Columbia River Project Water Use Plan

Transcription

1 Columbia River Project Water Use Plan KINBASKET AND ARROW LAKES REVEGETATION MANAGEMENT PLAN Reference: CLBMON-33 Arrow Lakes Reservoir Inventory of Vegetation Resources FINAL REPORT Study Period: 2010, Castlegar, BC

2 Original Report Cover CLBMON-33 ARROW LAKES RESERVOIR INVENTORY OF VEGETATION RESOURCES 2010 Final Report Submitted to: BC Hydro Water License Requirements Castlegar, B.C. by: Katherine Enns, H. Bruce Enns and A.Y. Omule Galena Bay, BC 602 Tamarack St. Castlegar, B.C. V1N 2J2 December 7, 2010 ii

3 Citation: Enns, K., H.B. Enns and A.Y. Omule CLBMON-33 Arrow Lakes Reservoir Inventory of Vegetation Resources: 2010 Final Report prepared by for BC Hydro, 86 pp. plus appendices Cover photo: Seepage enhancement of vegetation at Galena Bay in the Arrow portion of the Arrow Lakes Reservoir at middle elevation (type: PE) 2010 BC Hydro No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without prior permission from BC Hydro, Burnaby, B.C. iii

4 EXECUTIVE SUMMARY This report for the CLBMON-33 project (Arrow Lakes Reservoir Inventory of Vegetation Resources) presents the results of comparisons of vegetation spatial extent, structure and composition between 2007 and The project monitors landscape-level changes in vegetation communities in the reservoir drawdown zone between elevations 434m and 440m in response to the soft constraints operating regime. This report describes the status of the third year of monitoring under the CLBMON-33 project, including descriptions of the field work, aerial photography capture, as well as methods, analysis and interpretation of the field and mapped image data. Field work completed in 2010 included reassessment of previously established plots done for CLBMON-33 and a companion project, CLBMON-12, which addresses site level (1:1 scale) changes in existing vegetation and response to vegetation enhancement in the reservoir. Additional plots were established in previously unsampled study areas or incompletely sampled vegetation community types in The 2010 aerial photographic imagery, upon which this year s results are based, used digital image capture. Before this, imagery was captured on film. The methods of image capture and the resulting images are an improvement over previous film images. It is possible to make accurate comparisons to the previous imagery. The 2010 data analysis indicated that at the landscape level, there was no significant change in vegetation spatial extent or structure between 2007 and A landscapelevel change is defined as a 10 per cent incremental shift in the coverage of a vegetation community type from its previous condition. Of 398 polygons reviewed in the mapping only six polygons showed at least a 10 per cent change in vegetation community type cover between 2007 and 2010, and of those, two polygons showed a change greater than 10 per cent. However, coverage within vegetation community types at the landscape level had much greater resolution than 10 per cent increments, due to the quality of the 2010 imagery and the analysis of field data. At increased resolution, we found that covers and heights in some of the vegetation types were significantly lower in 2009 than in previous years. This was attributed to higher water levels in Near full pool conditions persisted from August 2008 to January 2009, and were followed by a decrease in covers of vegetation in some vegetation types, specifically between 434 and 436 metres. Recovery of the vegetation from this relatively high duration and frequency of inundation was also apparent in the data. The spatial extent, structure and composition (diversity, evenness, richness and distribution) of vegetation communities are clearly influenced by depth of inundation and have characteristic species structure and richness in relation to depth. However, species richness, diversity, evenness and distribution appear unchanged over time, whereas spatial extent (cover) and structure (heights) have responded to a single year shift in the soft constraints operating regime in Several of the vegetation variables are strongly influenced by environmental variables such as soil texture, and soil texture is in turn influenced by inundation in the reservoir. Comparisons between the high quality imagery in 2010 with future imagery of similar quality will show more precise vegetation responses and allow us to state more clearly how the vegetation is maintained by the soft constraints operating regime. Table 1 is a summary of the findings in relation to the management questions and the management hypothesis for this study, as well as a summary of additional work required to answer the management questions completely. iv

5 Table 1. CLBMON-33 Status of Objectives, Management Questions and Hypothesis after Year 3 for Monitoring Landscape-level Change in Vegetation Communities between 434m and 440m Objective Management Question Management Hypothesis Year 3 Status Complete the delineation of riparian and wetland vegetation communities within the drawdown zone between elevations 434m and 440m. What are the existing riparian and wetland vegetation communities in the Arrow Lakes Reservoir drawdown zone between 440m to 430m? The vegetation communities are grouped into 16 vegetation community types (VCTs) based on topography, parent materials (soils and non-soils), and vegetation. The dominant vegetation type is called PC: Reed Canary Grass mesic. It is not likely that any further examination of this question is required. Is the current distribution of vegetation communities in Revelstoke Reach representative of the remainder of the reservoir? The answer to this question is no, Revelstoke is not representative of the remainder of the reservoir (Enns et al. 2008). To ensure this is clear, in this year s work we determined that the proportions of vegetation community types in Revelstoke Reach are significantly different from those in the Arrow Lakes portion of the Arrow Lakes Reservoir (see Section 0). The Revelstoke Reach is dominated by PC. The reservoir becomes more diverse toward the south, with higher proportions of the other 15 vegetation community types. In 2012, either discriminant function analysis or a formal classification can be used to test the validity of the vegetation community types. Measure the changes in spatial extent, structure and composition (distribution and diversity) of the communities in the drawdown zone over What are the spatial extents, structure and composition (i.e., relative distribution and diversity) of these communities within the drawdown zone between 440m to 430m? H OA : There is no significant difference in the spatial extent of vegetation communities within the existing vegetated zones of There are distinct patterns in spatial extent (cover) and structure (heights) of the vegetation communities in the Arrow Lakes Reservoir. In general, vegetation cover is low at very low elevation and occasionally at high elevation. Heights also increase with elevation. Diversity, evenness, and richness are usually determined at v

6 time. Arrow Lakes Reservoir. H OB : There is no significant difference in the structure and composition (i.e., distribution and diversity) of vegetation communities within the existing vegetated zones of Arrow Lakes Reservoir. the species level and not at the landscape level. We have plotted these measures within years and compared them between years. Use of analysis of variance of year-wise differences in species diversity, evenness and richness within community types should be done with plot data to increase confidence in the answers given to this question. In particular, the question of change over time requires additional years of data. Assess whether the observed changes in the communities (as above) are attributable to the soft constraints operating regime of the reservoir. Does the soft constraints operating regime of Arrow Lakes Reservoir maintain vegetation spatial limits, structure and composition of existing vegetation communities in the drawdown zone? As above, with respect to H OA and H OB. Since the study began in 2007, the annual operating patterns have been similar, with the exception of 2008 where full pool persisted from the summer of 2008 to the early winter of 2009, which differed from the usual decline in water levels in the fall. Subsequently, some vegetation types showed a statistically significant decline in cover and height in the following year (2009). This was shown this year in map code data using confusion matrices and repeated measures analysis of variance. By 2010, a statistically significant recovery of the covers and heights had occurred in some of the vegetation communities. These results have been determined using mapping statistics in this paper, and in the analysis of variance tests for statistical differences in plant cover and heights within vegetation community types, from the field work, also included in this paper. Before 2010, we did not have adequate aerial photography coverage in 2008 to make comparisons over time. Comparisons between 2007, 2010 and 2012 imagery will be vi

7 necessary to show that the most typical of the soft constraint operating regime scenarios, which is to allow a late summer exposure period of low water, will allow vegetation to be maintained at the 2007 baseline level. Further, if there is a second, comparable high water period, we will be able to observe the same pattern in vegetation response. This will allow us to be more definitive regarding our conclusions about the role of the soft constraints operating regime on maintaining vegetation in the drawdown zone. Provide information on the effectiveness of the soft constraints operating regime at maintaining the existing spatial extent, structure and composition of existing vegetation communities. How do spatial limits, structure and composition of vegetation communities relate to reservoir elevation and topoedaphic site conditions such as aspect, slope, soil moisture, etc.? As above, with respect to H OA and H OB. The relationship between the spatial extent, structure and composition of the vegetation communities has been examined using RDA with forward selection. Both map data and field data were examined using RDA in 2008, 2009 and again in 2010, and similar results were obtained. Environmental conditions account for substantial variation on cover and heights, and these conditions are coupled with the soft constraints operating regime. Soil texture is a fundamentally important variable, and it is strongly influenced by depth and duration of inundation. Due to the nature of river flow, silts and clays accumulate at low elevations and gravels and sands accumulate at high elevation, and these influence the vegetation. The vegetation composition (diversity, evenness and richness) is examined at the species level and has been clearly linked with moisture and nutrients. Climate also is important. The climate regime is warmer in the Arrow Lakes portion than in the Revelstoke Reach and this influences the spatial extent and diversity of the vegetation types; specifically, the southern portion is more diverse than the northern portion. Trends in the field data indicate species richness has not changed over vii

8 time ( ). We will need to examine the data for a statistically significant change in vegetation community composition over time using more complete image and field data than are available this year. To test for a statistical difference in species richness over time, we would use repeated measures analysis of variance (Nagaike et al. 2006). Are there operating changes that can be implemented to maintain existing vegetation communities at the landscape scale more effectively? Based on the single occurrence of a sustained high water level from the summer of 2008 to the early winter of 2009, and the subsequent decline in cover and heights of vegetation in the low elevation bands of the reservoir, we recommend that water levels be allowed to decline for a period each midsummer to late fall for vegetation to recover and grow during the latter portion of the growing season. This recommendation is based on the response of vegetation to one sustained high water period. This observation is consistent with the patterns of higher covers and heights in years when the vegetation was exposed during the middle late summer period (i.e., the observational years of 2007, 2008 and 2010). Additional annual data analysis of trends will increase our confidence in this recommendation. viii

9 KEYWORDS Arrow Lakes Reservoir, Inventory of Vegetation Resources, landscape level, soft constraints operating regime, Vegetation Community Type, spatial extent, vegetation structure, vegetation composition, vegetation cover, vegetation height, vegetation diversity, vegetation distribution, vegetation evenness, species richness, aerial photographic interpretation elevation zone, water levels, vegetation response, scouring, vegetation recovery. ix

10 ACKNOWLEDGEMENTS We would like to acknowledge the reviews, advice and project management of Eva Boehringer of BC Hydro. Terrasaurus Aerial Photography Ltd. created the aerial photographic images, and 4DGIS did the georeferencing and aerial triangulation of the digital aerial photographs, and creation of the photomosaics. Susan K. Stevenson and Vivien Bowers provided technical reviews. Evan McKenzie and Audrey Ehman collected a large proportion of the field data and both have contributed to the document. Justin Overholt and Jane Enns provided GPS and plot layout in the field. Simone Runyan provided continuity with the last year of field work. Frank Lomer and Dr.Terry McIntosh provided verification and identification of unknown plants. Our thanks to them all. x

11 TABLE OF CONTENTS Executive Summary... iv KEYWORDS... ix ACKNOWLEDGEMENTS... x TABLE OF CONTENTS... xi List of FIGURES... xiii LIST OF TABLES... xx List of APPENDICES... xxii 1.0 INTRODUCTION Monitoring program management questions, objectives and scope Program hypothesis METHODS Field Work Field Work Post-field Work Imagery data collection and database extraction Post-flight Work Data Analysis RESULTS Question 1. What are the existing vegetation communities in the Arrow Lakes Reservoir drawdown zone between 430 and 440 metres? Question 2. Is the current distribution of vegetation communities in Revelstoke Reach representative of the conditions in the remainder of the reservoir? Question 3. What are the spatial extents, structure and composition (i.e., relative distribution and diversity) of these communities within the drawdown zone between 440 and 430 metres? Question 4. How do the spatial limits (cover), structure (height) and composition of the vegetation communities relate to elevation, the topoedaphic site conditions (aspect, slope and soil moisture, etc.)? Question 5. Does the soft constraints operating regime of Arrow Lakes Reservoir maintain spatial limits, structure and composition of existing vegetation communities in the drawdown zone? Review of the mapping Repeated Measures Analysis: Confusion Matrix Repeated Measures Analysis: ANOVA of map data - are vegetation cover and height maintained over time? Tukey s Boxplots and ANOVA: Field data comparisons - are vegetation cover and height maintained over time? Is the vegetation composition (distribution, diversity, evenness and richness) maintained? Examples of changes in vegetation spatial limits (cover), from the imagery of 2007 and Image analysis pilot study xi

12 3.6 Question 6. What potential changes to the soft constraints operating regime could be made in order to maintain the existing vegetation communities at the landscape scale? Hypothesis test summary DISCUSSION CONCLUSIONS SOURCES OF UNCERTAINTY AND POTENTIAL ERROR REFERENCES GLOSSARY APPENDIX 1. Data structures FOR CLBMON APPENDIX 2. Locations of polygons included in the Kappa statistic and the repeated measures analysis APPENDIX 3. Number of field plots and mapped polygons sampled APPENDIX 4. Vegetation cover and heights in ARROW LAKES RESERVOIR: APPENDIX 5. Analysis of VARIANCE RESULTS FOR Tables of Species Richness, Species Evenness, and Species Diversity APPENDIX 6. Definition of Exposure Regime Dummy Variables APPENDIX 7. RDA Analysis WITH ELEVATION APPENDIX 8. Image Analysis Pilot STUDY xii

13 LIST OF FIGURES Table 1. CLBMON-33 Status of Objectives, Management Questions and Hypothesis after Year 3 for Monitoring Landscape-level Change in Vegetation Communities between 434m and 440m... v Figure 1. Table 2. Table 3. Table 4. The location of the CLBMON-33 project, in the drawdown zone of the Arrow Lakes Reservoir, between Revelstoke and Castlegar B.C. Red dots are plot locations (stacked in the figure due to scale of the map) Management questions, objectives relating to them, and summary of findings to date and further work required to meet the objectives Vegetation Community Types that are thought to be mainly influenced by the Arrow Lakes Reservoir (i.e., Soft Constraints Operating Regime) Vegetation Community Types in the Arrow Lakes Reservoir that are included in the Mapping, but are thought to be less influenced by the Soft Constraints Operating Regime than by other Variables, or are uncommon Figure 2. Water levels in the Arrow Lakes Reservoir from 2005 to Table 5. Dates of Work undertaken to complete CLBMON-33 in Figure 4. Figure 5. Figure 6. Locations of the 398 polygons compared for changes in vegetation community type features (decile differences, spatial extent, structure and composition) between 2007 and Details of the polygon sample locations are provided in Appendix Illustrated example of a notched Tukey s boxplot. Boxplots show a statistically significant difference (p<0.05) between medians from comparably sampled populations when the notches or two or more boxplots do not overlap. The narrow centre of the box represents the median point in the data set. The top and bottom of the box are the 75 th and 25 th percentiles, respectively, and are called the hinges. The Hspread is the absolute value of the difference between the values of the two hinges. The whiskers show the range of values that falls within 1.5 Hspreads of the hinges. Inner fences (not plotted) are equal to the value of the hinge ± (1.5 x Hspread). Outer fences (not plotted) are equal to the value of the hinge ± (3.0 x Hspread). Asterisks and circles show values outside the inner and outer fences of the data distribution and are potential outliers. The notch in the box represents the 95 per cent confidence limits of the median. When the confidence limit exceeds the 75 th or 25 th percentile (i.e., the end of the box), the boxplot looks like it has been turned 'inside-out' (McGill et al. 1978) Typical elevation spread of Vegetation Community Types (VCTs: see Table 3 in the Arrow Lakes Reservoir. Only those VCTs with greater than three per cent cover of vegetation, sufficient sample size (Omule and Enns 2010) and primarily influenced by the soft constraints operating regime are shown; BE = Sandy beach, BG = Gravelly beach, CR = Cottonwood riparian, PA = Redtop upland, PC = Reed canary grass mesic, PE = Horsetail lowland, RR = Reed-rill xiii

14 Figure 7. Figure 8. Figure 9. Table 6. Total Area Coverage of each VCT, from Arrow Lakes (Arrow) and Revelstoke Reach (Revelstoke) Total Area Coverage of each VCT, from left (Northern Boundary) to right (Southern Boundary): 1-North Revelstoke Reach, 2-South Revelstoke Reach, 3-Beaton to Halfway River, 4-Nakusp to Arrow Ferry, 5-Arrow Ferry to Burton, 6-Edgewood to the Southern Boundary of the Map Area at Deer Park and Renata. (Appendix 2 identifies the locations for the start and endpoints of the areas illustrated above.) Total cover of vegetation in the three elevation bands (434 to 436 metres, 436 metres to 438 metres, and 438 to 440 metres) in the Arrow Lakes (A) and Revelstoke Reach (R) portions of the Arrow Lakes Reservoir Analysis of variance results for total vegetation cover and average maximum vegetation height by reach and elevation band. Significant results have a p-value < 0.05 and are shaded Figure 10. Total heights of vegetation in the three elevation bands (434 to 436 metres, 436 metres to 438 metres, and 438 to 440 metres) in the Arrow Lakes (A) and Revelstoke Reach (R) portions of the Arrow Lakes Reservoir Table 7. Figure 11. Figure 12. Environmental variables measured or estimated from the mapping for each polygon used in the RDA analysis for VCTs of the Arrow Lakes Reservoir in Sand, silt, and clay texture estimates are indicative of soil nutrient and moisture regimes RDA Biplot for the Arrow Lakes portion of the Arrow Lakes Reservoir. The red lines with the arrows are the vegetation variables (height and cover) and the blue lines show the environmental variables. The environmental variables explained approximately 28 per cent of the variation in vegetation cover and height. The first and second canonical variables explained about 20 per cent and 8 per cent, respectively, of the total variation in vegetation cover and height. The vegetation cover is negatively correlated with gravel and sand, (which is related to scouring of those materials) and vegetation height appears to increase with increase in coarse fragment content (sands and gravels; which is related to drainage and the development of shrub heights at the highest elevation), an increase in slope and with the highest elevation, warmest and driest weather regime N3 (Nakusp weather, 438 to 440 metres elevation) RDA Biplots for the Revelstoke Reach Portion of the Arrow Lakes Reservoir. The lines with the arrows are the vegetation variables (height and cover) and the remaining lines show the environmental variables. The environmental variables explained approximately 36 per cent of the variation in vegetation cover and height. The first and second canonical variables explained about 24 per cent and 12 per cent of the total variation in vegetation cover and height, respectively. The vegetation cover appears to be negatively related to estimates of sand and gravel fractions in materials and soils and the vegetation height appears to increase in areas with R2 and R3 exposure regime, high boulder content, and aspect. Height is negatively correlated with silts. Silts accumulate at low elevation and support shorter VCTs such as PE: Horsetail lowland. 39 xiv

15 Figure 13. Scouring of gravel and sand caused by high water wave action between 438 and 440 metres in the Arrow Lakes portion of the Arrow Lakes Reservoir. (Photo taken in 2007 at McDonald Park.) The 438 to 440 metre elevation band is visible in the left side of the photograph and shows wave scouring and exposure of gravels. The 438 to 436 metre elevation band is visible to the right in this photograph and shows the lack of scouring in the middle elevation band and development of high relative vegetation cover. These patterns in winter wave action influence vegetation cover and heights in the reservoir. High heights and low cover can occur at high elevation, and low heights with high cover in the middle elevation bands are common depending on water energy Table 8. Confusion matrix: summary of analyses for Vegetation Cover Table 9. Confusion Matrix: summary of analyses for Vegetation Height Table 10. Confusion Matrix: Summary of analyses for Vegetation Distribution Table 11. Table 12. Table 13. Table 14. Figure 14. Figure 15. Figure 16. Repeated Measures ANOVA p-values for the Arrow Lakes portion of the reservoir by Elevation Band, Year and Elevation Band x Year interaction for each VCT. Significant results have a p-value < 0.05 and are shaded Vegetation Community Type (VCT), Year and VCT x Year interaction for each Elevation Band in the Arrow portion of the Arrow Lakes Reservoir. Significant results have a p-value < 0.05 and are shaded Repeated Measures ANOVA p-values for Revelstoke Reach portion by Elevation Band, Year and Elevation Band x Year interaction effects for each VCT. Significant results have a p-value < 0.05 and are shaded Vegetation Community Type (VCT), year and VCT x Year interaction for each Elevation Band in the Revelstoke Reach portion of the Arrow Lakes Reservoir. Significant results have a p-value < 0.05 and are shaded Tukey s boxplots: change in total vegetation cover in BE: Beach (top) and CR: Cottonwood riparian (bottom) VCTs in the Arrow Lakes Reservoir drawdown zone between elevations m (Arr = Arrow portion, Rev = Revelstoke Reach portion of the Arrow Lakes Reservoir) between 2007 and See Figure 5 for a complete description of boxplots Tukey s boxplots: change in total vegetation cover in PA: Redtop upland (top) and PC: Reed canary grass mesic (bottom) VCTs in the Arrow Lakes Reservoir drawdown zone between elevations m (Arr = Arrow portion, Rev = Revelstoke Reach portion of the Arrow Lakes Reservoir) between 2007 and See Figure 5 for a complete description of boxplots Tukey s boxplots: change in total vegetation cover in PE: Horsetail lowland (top) and RR: Reed-rill (bottom) VCTs in the Arrow Lakes Reservoir drawdown zone between elevations m (Arr = Arrow portion, Rev = Revelstoke Reach portion of the Arrow Lakes Reservoir) between 2007 and See Figure 5 for a complete description of boxplots xv

16 Table 15. Figure 17. Figure 18. Figure 19. Table 16. Figure 20. Figure 21. ANOVA results for Total Vegetation Cover within selected VCTs by Reach (Arrow or Revelstoke) and Year (2007, 2008, 2009, 2010). Significant differences in mean total cover are shaded Tukey s boxplots: Maximum vegetation heights in BE: Beach (top) and CR: Cottonwood riparian (bottom) VCTs in the Arrow Lakes Reservoir drawdown zone between elevations m (Arr = Arrow portion, Rev = Revelstoke Reach portion of the Arrow Lakes Reservoir) between 2007 and See Figure 5 for a complete description of boxplots Tukey s boxplots: Maximum vegetation heights in PA: Redtop upland (top) and PC: Reed canary grass mesic (bottom) VCTs in the Arrow Lakes Reservoir drawdown zone between elevations m (Arr = Arrow portion, Rev = Revelstoke Reach portion of the Arrow Lakes Reservoir) between 2007 and See Figure 5 for a complete description of boxplots Tukey s boxplots: Maximum vegetation heights in PE: Horsetail lowland (top) and RR: Reed-rill (bottom) VCTs in the Arrow Lakes Reservoir drawdown zone between elevations m (Arr = Arrow portion, Rev = Revelstoke Reach portion of the Arrow Lakes Reservoir) between 2007 and See Figure 5 for a complete description of boxplots ANOVA results for Mean Maximum Vegetation Height within selected VCTs by Reach (Arrow or Revelstoke) and Year (2007, 2008, 2009, 2010). Significant differences in mean maximum height are shaded Shannon-Weiner Diversity Index (H ) for each Vegetation Community Type (VCT) from 2007 to 2010 in the Arrow Lakes Reservoir (both geographic areas combined). See Appendix 5 for the tabular values for this figure Pielou s species evenness (J) for each Vegetation Community Type (VCT) from 2007 to 2010 in the Arrow Lakes Reservoir (both geographic areas combined). See Appendix 5 for the tabular values for this figure.. 58 Figure 22. Species richness for all vegetation within the three elevation bands ( metres, metres and metres) from 2007 to 2010 in the Arrow Lakes Reservoir (both geographic areas combined). See Appendix 5 for the tabular values for this figure Figure 23. Figure 24. Figure 25. Species diversity (H) for all vegetation within the three elevation bands ( metres, metres and metres) from 2007 to 2010 in the Arrow Lakes Reservoir (both geographic areas combined). See Appendix 5 for the tabular values for this figure Pielou s species evenness (J) for all vegetation within the three elevation bands ( metres, metres and metres) from 2007 to 2010 in the Arrow Lakes Reservoir (both geographic areas combined). See Appendix 5 for the tabular values for this figure Species Richness for each elevation band from 2007 to 2010 in the Arrow Lakes Reservoir (both geographic areas combined). See Appendix 5 for the tabular values for this figure xvi

17 Figure 26. Phenological Differences in Polygon 10 between 2007 (left) and 2010 (right) in the Arrow Lakes Reservoir, from aerial photography. Note the greener and fluffier appearance of vegetation in the 2007 imagery indicating the three week difference in time of capture of the photographs. The 2007 photographs were taken on May 30; the 2010 photographs were taken on May Scale is approximately 1: Figure 27. Phenological Differences in Polygon 1777 between 2007 (left) and 2010 (right) in the Arrow Lakes Reservoir, from aerial photography. Growth of reed canary grass is more advanced in the 2007 photography, and shrubs have not released as much in 2010 as in This does not constitute a decile change. Scale is approximately 1: 1, Figure 28. Figure 29. Figure 30. Figure Figure 32. Figure 33. Figure 34. Example of a change in vegetation due to a combination of ORV use and phenology in Polygon 1763 in the Arrow Lakes Reservoir, from aerial photography. At middle to low elevation RR: reed-rill VCT between 2007 (left) and 2010 (right) constitutes a decile change in the vegetation mapping for this polygon. Scale is approximately 1: 1, Dispersal of logs from the LO: Blue wildrye log VCT between 2007 (left) and 2010 (right) in Polygon 2009 in the Arrow Lakes Reservoir, from aerial photography. Note difference in resolution at this scale. This constitutes a decile change and a vegetation community type change, from LO to RR, although the change is mainly in relation to woody debris. Scale is approximately 1: Example of a change in decile cover of PE: Horsetail lowland in Polygon 1710 in the Arrow Lakes Reservoir, from aerial photography. A low elevation PE: Horsetail lowland VCT between 2007 (left) and 2010 (right). This magnitude of change was seen in fewer than six out of 398 polygons. Scale is approximately 1: Differences in phenology in Polygon 1192 in the Arrow Lakes Reservoir, from aerial photography. Trees and shrubs appear to have shrunk between 2007 and 2010; the lack of leaves in the May 12, 2010 photo vs. the May 30, 2007 photo is notable. Scale is approximately 1: Differences in clump size in reed canary grass at Montana Slough, Revelstoke Reach 2007 vs. 2010, from aerial photography. Differences are possibly due to inundation-controlled colour expression and phenology. Scale 1: 1, Differences in cover and height of reed canary grass in the Arrow Narrows study area of the Arrow Lakes portion of the Arrow Lakes Reservoir, 2007 vs from aerial photography. Differences may be due to water level differences and phenological differences. Scale 1: 3, Recommended Change to the BC Hydro Soft Constraints Operating Regime based on Data from 2007 to Table 17. Summary of Statistical Results xvii

18 Figure A4-5. Table A7-1. Figure A7-1. Figure A7-2. Figure A7-3. Figure A7-4. Total maximum heights of vegetation in each of the main vegetated VCTs in the Arrow Lakes Reservoir drawdown zone. Data are plotted as Tukey s boxplots (TBP). TBP are a visual method of showing statistically significant differences (p<0.05) between data distributions from comparably sampled populations (Tukey, 1977; Velleman and Hoaglin, 1981; Wilkinson et al., 1992). The narrow centre of the box represents the median point in the data set for a variable. The upper portion of the TBP is the 75 th percentile of the data (population); the lower portion is the 25 percentile of the data (population). Circles, asterisks or vertical bars show the outliers Number of map observations in 2010 by Reach, VCT and Elevation Band (code 1, 2 or 3). Data records from some sample original polygons (290) are duplicated RDA biplot for the Arrow portion of the reservoir: soil and elevation variables standardized eigenvectors or scores. The lines with the arrows are the dependent variables (vegetation height and cover) and the remaining lines show the independent variables (soil, site and elevation). Vegetation cover is strongly negatively associated with gravel and sand; and vegetation height is positively associated with most soil variables, except silt, with which it is strongly negatively correlated. Elevation appears to be positively associated with vegetation height at high elevation ( m), but elevation is not associated with vegetation cover RDA biplot for the Arrow portion of the reservoir: wave action variables standardized eigenvectors or scores. The lines with the arrows are the dependent variables (vegetation height and cover) and the remaining lines show the independent variables (wave action). Vegetation cover is positively associated with SH (sheltered topography), S (seepage with some scouring,) and R (river entry); and vegetation height does not appear to be related to the wave action variables RDA biplot for Revelstoke Reach: soil and elevation standardized eigenvectors or scores. The lines with the arrows are the dependent variables (vegetation height and cover) and the remaining lines show the independent variables (soil, site and elevation). Vegetation cover is strongly negatively associated with gravel and sand; and vegetation height is positively associated with most environmental variables, except silt, with which it is strongly negatively associated. Elevation is associated positively with vegetation height at high elevation ( m) and negatively at low elevation ( m); elevation is positively associated with vegetation cover at low elevation ( m), but not at high elevation RDA biplot for Revelstoke Reach: wave action variables standardized eigenvectors or scores. The lines with the arrows are the dependent variables (vegetation height and cover) and the remaining lines show the independent variables (wave action). Vegetation cover is positively associated with R (river entry) and S (seepage with some scouring), and negatively associated with W (wave action) and E (high water effects). Vegetation height appears negatively associated with R (river entry) xviii

19 Figure A7-5. Figure A7-6. RDA biplot for the combined Arrow portion and Revelstoke Reach: soil and elevation standardized eigenvectors or scores. The lines with the arrows are the dependent variables (vegetation height and cover) and the remaining lines show the independent variables (soil, site and elevation). Vegetation cover is strongly negatively associated with gravel and sand; and vegetation height is positively associated with most soil and elevation variables, except silt, with which it is strongly negatively corrected. Elevation is positively associated with vegetation height at high elevation ( m), but elevation is not associated with vegetation cover RDA biplot for the combined Arrow portion and Revelstoke Reach: wave action variables standardized eigenvectors or scores. The lines with the arrows are the dependent variables (vegetation height and cover), and the remaining lines show the independent variables (wave action). Vegetation cover is positively associated with SH (sheltered); height appears negatively associated with EX (exposed promontory) xix

20 LIST OF TABLES Figure 3. Table A3-1. Table A3-2. Table A3-3. Table A3-4. General location of field plots completed in 2010 within study areas in the Arrow Lakes Reservoir for the CLBMON-33 project. Red dots are the plot locations (which tend to overlap at this scale) and yellow lines are the 43 study areas within the Arrow Lakes Reservoir. Details of the 559 plot locations are provided in the map database Numbers of field plots completed in each Vegetation Community Type in 2010 for CLBMON Number of mapped Polygons assessed in each Vegetation Community Type in 2010 for Repeated Measures Analysis. Data taken from the 2007 and 2010 mapping Number of aerial photographic observations made for the confusion matrix and RMA in 2007 and 2010 mapping: by Reach, VCT and Elevation Band (code 1, 2 or 3). Note that sample polygons may contain more than one VCT, and that if a sample polygon extended over two elevation bands, it was divided into two to three records. Therefore. some redundancy exists in the data set Numbers of field plots completed in each Vegetation Community Type in 2010 for CLBMON-33, including those that were also done in 2008 and 2009 for CLBMON-33 and CLBMON-12. Some plots done in 2010 were co-located in similar locations as the 2007 plots, but they were treated as new plots in the 2010 database Figure A4-1. Total Cover of Vegetation in the Three Elevation Bands (434 to 436 metres, 436 metres to 438 metres, and 438 to 440 metres) in the Arrow Lakes (A) and Revelstoke Reach (R) portions of the Arrow Lakes Reservoir. Data are plotted as Tukey s boxplots (TBP). TBP are a visual method of showing statistically significant differences (p<0.05) between data distributions from comparably sampled populations (Tukey, 1977; Velleman and Hoaglin, 1981). The narrow centre of the box represents the median point in the data set for a variable. The upper portion of the TBP is the 75 th percentile of the data (population); the lower portion is the 25 percentile of the data (population). Circles, asterisks or vertical bars show the outliers Figure A4-2. Total Heights of Vegetation in the three Elevation Bands (434 to 436 metres, 436 metres to 438 metres, and 438 to 440 metres) in the Arrow Lakes (A) and Revelstoke Reach (R) portions of the Arrow Lakes Reservoir. Data are plotted as Tukey s boxplots (TBP). TBP are a visual method of showing statistically significant differences (p<0.05) between data distributions from comparably sampled populations (Tukey, 1977; Velleman and Hoaglin, 1981; Wilkinson et al., 1992). The narrow centre of the box represents the median point in the data set for a variable. The upper portion of the TBP is the 75 th percentile of the data (population); the lower portion is the 25 th percentile of the data (population). Circles, asterisks or vertical bars show the outliers xx

21 Figure A4-3. Figure A4-4. Table A5-1. Table A5-2. Table A5-3. Total Cover of Vegetation by Community type for all combined data in both the Arrow Lakes and Revelstoke Reach portions of the Arrow Lakes Reservoir. BE = Beach, BG = Gravelly beach, CR = Cottonwood riparian forest, PA = Redtop upland, PC = Reed canary grass mesic, PE = Horsetail lowland, RR = Reed-rill. Data are plotted as Tukey s boxplots (TBP). TBP are a visual method of showing statistically significant differences (p<0.05) between data distributions from comparably sampled populations (Tukey, 1977; Velleman and Hoaglin, 1981; Wilkinson et al., 1992). The narrow centre of the box represents the median point in the data set for a variable. The upper portion of the TBP is the 75 th percentile of the data (population); the lower portion is the 25 percentile of the data (population). Circles, asterisks or vertical bars show the outliers Least Squares Means of total cover in BG: Gravelly beach, CR: Cottonwood riparian, PA: Redtop upland, PC: Reed canarygrass mesic, PE: Horsetail lowland and RR: Reed rill Shannon-Weiner Diversity Index (H ) for each Vegetation Community Type (VCT) from 2007 to 2010, in the Arrow Lakes Reservoir (both geographic areas combined) Pielou s species evenness (J) measures for dominant and codominant species of each Vegetation Community Type (VCT) in the Arrow Lakes Reservoir (both geographic areas combined) Species Richness measures for dominant and co-dominant species of each Vegetation Community Type (VCT) in the Arrow Lakes Reservoir (both geographic areas combined) Table A5-4. Shannon-Weiner Diversity Index (H ) for each Elevation Band from 2007 to 2010 in the Arrow Lakes Reservoir (both geographic areas combined) Table A5-6. Table A5-6. Table A6-1. Pielou s species evenness (J) measures for all vegetation within the three elevation bands ( metres, metres and metres) from 2007 to 2010 in the Arrow Lakes Reservoir (both geographic areas combined) Species Richness Elevation Band from 2007 to 2010 in the Arrow Lakes Reservoir (both geographic areas combined) Table of values describing the various aspects of the six exposure regimes constructed from weather data for the Arrow Lakes Reservoir Table A6-2. Table of variables from Table A6-1 with definitions xxi

22 LIST OF APPENDICES Appendix 1. Data structures FOR CLBMON Appendix 2. Locations of polygons included in the Kappa statistic and the repeated measures analysis Appendix 3. Number of field plots and mapped polygons sampled Appendix 4. Appendix 5. Vegetation cover and heights in ARROW LAKES RESERVOIR: Analysis of VARIANCE RESULTS FOR Tables of Species Richness, Species Evenness, and Species Diversity Appendix 6. Definition of Exposure Regime Dummy Variables Appendix 7. RDA Analysis WITH ELEVATION Appendix 8. Image Analysis Pilot xxii

23 1.0 INTRODUCTION The Water Licence Requirement project (CLBMON-33 Arrow Lakes Reservoir Inventory of Vegetation Resources) was initiated in This project is a 10- year monitoring program in the Arrow Lakes Reservoir, overseen by BC Hydro, to assess the impacts of the soft constraints operating regime on existing vegetation communities in the drawdown zone between elevations 434m and 440m (full pool). The Arrow Lakes Reservoir is situated on the Columbia River, between the Revelstoke Dam at Revelstoke in the north, and the Hugh Keenleyside Dam at Castlegar, British Columbia, in the south (Figure 1). Figure 1. The location of the CLBMON-33 project, in the drawdown zone of the Arrow Lakes Reservoir, between Revelstoke and Castlegar B.C. Red dots are plot locations (stacked in the figure due to scale of the map). 1

24 1.1 Monitoring program management questions, objectives and scope. The objectives of CLBMON-33 Arrow Lakes Reservoir Inventory of Vegetation Resources are to: Identify and spatially delineate existing riparian and wetland vegetation communities within the drawdown zone; Measure the spatial extent, structure and composition (i.e., distribution and diversity) of the communities in the drawdown zone at repeated time intervals over a 10 year period; Assess whether there are changes in the spatial extent, structure and composition of the communities in the drawdown zone over the monitoring period; Assess whether observed changes in the spatial extent, structure and composition are attributable to the soft constraints operating regime of the reservoir; and, Provide information on the effectiveness of the soft constraints operating regime at maintaining the existing spatial extent, structure and composition of the communities in the drawdown zone at the landscape level. The scope of the Arrow Lakes Reservoir Inventory of Vegetation Resources project is to identify and evaluate changes in vegetation communities at the landscape level, using aerial photographic interpretation and field data. The management questions, in relation to the objectives and findings to date, are summarized in Table 2. Table 2. Management Question What are the existing riparian and wetland vegetation communities in the Arrow Lakes Reservoir drawdown zone between 440m to 430mWhat? Is the current distribution of vegetation communities in Revelstoke Reach Management questions, objectives relating to them, and summary of findings to date and further work required to meet the objectives. Objectives Objective 1 applies. Objective 1 generally applies, but CLBMON-33 does not examine Summary of findings to date In 2007, a total of 16 vegetation community types (VCTs) were identified in the field (See Table 2). These were sampled and mapped at a scale of approximately 1: 10,000. Polygon delineation and classification were based primarily on topography, materials (soils and non-soils) and leading species. The characteristics of VCTs were further defined in All VCTs had plant species in common, but all VCTs also had characteristic plant species, especially in relation to topography and parent materials (Enns et al. 2007). No further work is required to define the existing vegetation community types, although changes in the vegetation within VCTs may occur, and will be monitored at the landscape and 1:1 scale. No, Revelstoke is not representative of the remainder of the reservoir (Enns et al. 2008). Revelstoke Reach is dominated by PC. The reservoir becomes more diverse toward the south, 2

25 representative of conditions in the remainder of the reservoir? What are the spatial extents, structure and composition (i.e., relative distribution and diversity) of these communities within the drawdown zone between 440 and 430 metres? How do the spatial limits (cover), structure (height) and composition of vegetation communities relate to elevation and the topo-edaphic site conditions (aspect, slope and soil moisture, etc.)? Does the soft constraints operating vegetation outside of the 43 study areas. Objective 2, 3 and 4 examine spatial extent (cover), structure (height) and composition (distribution and diversity) of types at the landscape scale. Environmental variables are not identified in the objectives for the study but are assumed to be variables that influence response of the vegetation along with the soft constraints operating regime (Objective 4). Objective 3 relates to this question, with higher proportions of the other 15 vegetation community types. The CLBMON-33 map areas include 43 discrete study areas covering 2,037 hectares. The zone between 434 and 440 metres is approximately 3,800 hectares. The combined study areas included in CLBMON-33 (Arrow Lakes proper and Revelstoke Reach) represent greater than half the total drawdown zone between elevations 434m and 440m, but they may not be representative of the proportions of types in the whole drawdown zone. The CLBMON-33 study areas are large, relatively flat and more vegetated in comparison to the rest of the reservoir. The vegetation being monitored in CLBMON-33 may still be characteristic of all the vegetation of the reservoir, however. Previous work indicates that some VCTs tend to only occur at low elevation (Enns et al.2007), whereas upland types tend to occur at higher elevation above 436m (Enns et al.2007). A relatively small number of VCTs occur at all elevations in the reservoir. Spatial extent (cover) and structure (heights) of vegetation within most VCTs increase with increasing elevation. In some cases, scouring by high water and waves results in lower cover of vegetation within VCTs at the highest elevations. However, average heights of high elevation VCTs are not affected by this. The distribution of vegetation within VCTs, as measured using codes for a range of distribution patterns, is not influenced by elevation. Diversity of the vegetation increases with elevation, but some low elevation VCTs, such as RR and PE, can have a high diversity of plants that are specifically adapted to the conditions of frequent, prolonged inundation (PE) or continuous seepage (RR). Diversity of VCTs was found to be associated with richer nutrient regimes and wetter hygrotopes (Enns et al. 2007). Drought tolerant VCTs (and species) had higher covers and heights where soils were sandy and fine textured (Enns et al. 2007), indicating soil nutrient regime and moisture availability is important for supporting vegetation in the reservoir. Previous to 2010, we examined within-year differences and similarities in the vegetation, 3

26 regime of Arrow Lakes Reservoir maintain spatial limits, structure and composition of existing vegetation communities in the drawdown zone? What potential changes to the soft constraints operating regime could be made in order to maintain the existing vegetation communities at the landscape scale? where maintenance of vegetation is considered over time. which could not be assumed to be time related (Enns et al. 2008). The imagery meant for comparison between 2007 and 2008 was insufficiently captured, and therefore did not allow us to determine if vegetation was being maintained by the soft constraints operating regime (Enns et al. 2008). Comparisons of vegetation over time the period 2007 to 2010 are reported in this paper. Additional years of data are required to address this management question conclusively. Objectives 4 and 5. The CLBMON-33 project had not been operational for long enough to answer this question. We need at least one more series of good imagery to assess the patterns in the soft constraints operating regime and to determine if vegetation can be maintained at its current level with the usual pattern in water levels, e.g to 2008 and 2009 to If a water regime similar to that of the summer to winter of does occurs, and the predicted decline in the vegetation occurs as predicted, then recommended changes will be based on two periods of high water to compare the effects of the soft constraints operating regime on vegetation over time. Also at that time the loss of vegetation, if it occurs, can be quantified. 1.2 Program hypothesis The interpretation of the aerial photographic images and field measurements must be analyzed using biometrics to address the following hypothesis and associated sub-hypotheses: H 0 : Under the soft constraints operating regime (or possibly a newly selected alternative after five years), there is no significant change in existing vegetation communities at the landscape scale. H 0A : There is no significant change in the spatial extent (cover, and number of hectares) of vegetation communities within the existing vegetated zones of Arrow Lakes Reservoir. H 0B : There is no significant change in the structure and composition (i.e., distribution and diversity) of vegetation communities within the existing vegetated zones of Arrow Lakes Reservoir. These hypotheses link back to the management questions in the following ways: a finding of no significant change 1 in existing vegetation communities at the landscape scale would be recorded as no significant difference in the existing vegetation communities in comparison to the 2007 baseline mapping. 1 Defined as a change of 10 per cent or greater in vegetation cover or height. 4

27 The vegetation community types (VCTs) occurring within the drawdown zone of the reservoir between 434 metres and 440 metres were defined in the field, and base mapping was created using delineation of polygons on aerial photographic mosaics at various scales for 43 study areas in the Arrow Lakes Reservoir. The VCTs are based on a combination of similar topography, parent materials or soils, and vegetation features. The VCTs are listed in Table 3 and Table 4. Table 3. Vegetation Community Types that are thought to be mainly influenced by the Arrow Lakes Reservoir (i.e., Soft Constraints Operating Regime). VCT Description Illustration BE BG Beach: mostly non-vegetated to sparsely vegetated with sand-tolerant herbaceous plants, seedlings and grasses, occasionally with sedges and mosses. BE occurs from 440 to below 434 metres, but is most abundant below 436 metres in the reservoir, is gently sloping and can be undulating. Materials are mineral soils only, composed of sands and small amounts of gravel, derived from sandy tills and sandy glaciofluvial materials, which was weathered in situ, then subjected to fluviation. BE contributes to downstream deposition. It is common throughout the Arrow Lakes Reservoir, but increases in abundance to the south. Gravelly beach: sparsely vegetated with grasses and herbaceous plants, especially drought tolerant weeds and occasionally small cottonwood and willows. BG occurs from 439 to 434 metres, but is most abundant between 437 and 435 metres elevation. It is often flat to gently sloping. Materials are non-soils, consisting of gravels and cobbles, but can have a sandy matrix. BG is derived from ablation till or glaciofluvial gravels, and is subjected to fluviation, but does not contribute much to downstream deposition of gravels, except in extreme events. It is fairly solid and resistant to scouring. BG is increasingly common and abundant toward the southern end of the Arrow Lakes Reservoir, and almost non-existent in the northern portion of the Revelstoke Reach. BE at Burton BG at Beaton 5

28 CR Cottonwood riparian: has a very high diversity of vegetation including trees, shrubs, herbs, grasses, mosses and lichens. This type includes early structural stages of previous cottonwood or coniferous dominated forests that were converted to pasture or lawn prior to CR is centred at the highest elevation in the reservoir, but on very sheltered well drained sites will extend down to 437 metres. Slopes range from very steep to flat, and are occasionally undulating. Soils range from undeveloped bouldery materials to well developed true soils, and are derived from virtually all possible terrain features, most commonly tills and glaciofluvial terraces, but also fluvial outwash and colluvial materials. CR is abundant throughout the Arrow Lakes Reservoir. CR at Illecillewaet PA Redtop upland: is dominated by shrubs and several species of grasses with a few drought tolerant herbs and several species of weeds. Lichens, mosses and liverworts are very common. Stunted trees often occur, and many of these predate the reservoir construction. PA is centred near the upper elevation of the drawdown zone, near 439 metres, but can extend down to 434 metres on very well drained materials. PA occurs almost always on convex shedding sites, but can have pockets of scoured out depositional areas, sometimes with pocket wetlands in them. Some fragments of developed true soils can persist. Almost all PA is derived from glaciofluvial gravels and gravelly till. It is subjected to high winter wave action and scouring, and both receives organic material and plant propagules from the drawdown zone, and contributes them. The vegetation in PA seems to provide some protection from scouring. Very common throughout the Arrow Lakes portion of the reservoir, but present in the Revelstoke Reach. PA in Revelstoke Reach PC Reed canary grass mesic: is comparatively species poor, and dominated by reed canary grass (Phalaris arundinacea), but can include a minor component of mint, horsetail and agronomic species. PC occurs in all slope positions, but is centred between 435 and 437 metres and is usually flat to undulating to gently sloping. Soils are close to true soils with some rudimentary layering, and are almost always sands with some silt and clay. They are derived from alluvial fans in Revelstoke Reach and derived from similar fans and glaciofluvial terraces in the Arrow Lakes. They are subjected PC in Revelstoke Reach 6

29 to scouring on raised topography and at the upper and lower elevation limits of their occurrence in the reservoir. PC is common and abundant throughout the reservoir, but is most common in Revelstoke Reach and declines toward the south. PE RR Horsetail lowland: has a very characteristics species assemblage consisting of horsetails, rushes, reeds and mosses. PE is centred between 433 and 435 metres elevation, but can extend upslope if moisture is available, and always occurs in concave, sloping topography. Soils are silty, often compacted, and occasionally include clays and organics and are poorly aerated. PE soils are possibly derived from organic materials or very silty glaciofluvial deposits. This type is very susceptible to scouring. PE occurs throughout the reservoir, but covers a small total area. It is almost absent from the southern end. Reed rill: can have comparatively rich vegetation, consisting predominantly of herbs, liverworts, wetland grasses, mosses, sedges, rushes and reeds. RRs originate from upwelling of underground water supplies and re-directed minor streams, usually at 436 metres and extending down slope. They can be serpentine, and occur on gently sloping to steeply sloping undulating topography. Soils often have a mix of coarse sands, enriched from groundwater transport of nutrients and moisture, and are weathering in situ in a variety of materials. They are not as common or abundant as some of the above VCTs and occur mainly in the midsection of the reservoir. This type is susceptible to damage from ORVs. Scouring does occur in sandy materials where water energy is high. RR is almost absent from the Revelstoke Reach portion of the reservoir. PE at Arrow Park RR at Arrow Park 7

30 Table 4. Vegetation Community Types in the Arrow Lakes Reservoir that are included in the Mapping, but are thought to be less influenced by the Soft Constraints Operating Regime than by other Variables, or are uncommon. VCT Description Illustration BB CL Boulders, steep: mostly with less than three per cent vegetation and steeply sloping, uncommon and increasing toward the south. BB is usually nonvegetated to very sparsely vegetated with less than three per cent vegetation cover and is not considered vegetated at the landscape scale. BB is usually derived from bouldery till. Cliffs and rock outcrops: often very steep, some gentle slopes but all are sparsely vegetated. This type occurs in fewer than 10 polygons in the base map. SF has insufficient frequency of occurrence to be considered for landscape scale analysis. CLs are derived from bedrock and colluvium. BB at Crawford Bay Ferry south terminal SF IN Failing slope: usually silty sands that have slumped in response to slope failure. Buried vegetation may occur. Approximately five polygons delineated. SF has insufficient frequency of occurrence to be considered for landscape scale analysis. SF appears to be derived from very sandy till and or glaciofluvial terrace edges and escarpments. Industrial/residential/recreational: on a variety of soils and materials with mixed slopes and slope positions, always influenced by recreational and industrial vehicle use. IN is most common in the northern end of Revelstoke Reach and in the central section of Arrow Lakes. Typically has weeddominated, drought and compaction-tolerant vegetation, and occasionally garden plants. IN is usually non-vegetated to very sparsely vegetated with less than three per cent vegetation cover and is not considered vegetated at the landscape scale. IN is of mixed origins. CL at Deer Park No Illustration available. IN at MacDonald Park boat launch 8

31 LO PO RS Blue Wild rye log zone: usually confined to high elevation, occasionally in sheltered coves and inlets, almost always at the top of the slope on convex to concave topography, dominated by logs and woody debris. LO was mostly in the Beaton to Nakusp portions of the reservoir in 2007, but its occurrence may have changed since then. LO can be ephemeral, and may be mapped within the PA type. LO is usually non-vegetated to very sparsely vegetated with less than three per cent vegetation cover and is not considered vegetated at the landscape scale. The LO type is not based on terrain; it is based on the presence of log debris. Pond: often in an upslope depressional area, resembling a cut-off oxbow or backchannel of the main stem of the reservoir. PO is restricted to the Revelstoke Reach and occurs in very few polygons. Mostly filled with still, brackish or slow moving water. Occasionally very sparsely vegetated with open water-tolerant wetland plants, such as ivy-leaved duckweed (Lemna trisulca) and pondweed (Potamogeton sp.). PO is usually non-vegetated to very sparsely vegetated with less than three per cent vegetation cover and is not considered vegetated at the landscape scale. PO overlays fluvial material. Willow stream entry: occurs from high to low elevation along incoming stream channels, usually gullied and undulating and almost always bouldery to gravelly with fine sand and silt deposits (i.e., mixed materials). RS is very gently sloping to moderately steeply sloped. The RS water supply is seasonal with a high flow in spring and fall freshet, and very low to completely dry during summer and winter. The effect of this water supply and its physical influence on the vegetation of RS is difficult to distinguish from the effects of the soft constraints operating regime. RS originated as minor, somewhat ephemeral, fluvial channels. LO in Revelstoke Reach PO at Montana slough SS Steep sand: sands, with slopes greater than 20 degrees, occurs at various slope positions in the drawdown zone and is relatively uncommon. SS is usually non-vegetated to very sparsely vegetated with less than three per cent vegetation cover and is not considered vegetated at the landscape scale. SS is very similar to BE, but is steep, whereas BE is gently sloping to flat. SS is mostly derived from very sandy glaciofluvial terrace escarpments or is weathering in situ as a new escarpment. RS at Deer Park SS in Revelstoke Reach 9

32 WR Silverberry river entry: occurs only in river entries with year-round water flow, from highest elevation locations to the lowest elevation, and is usually flat (although the sides of the river channels are included) Mainly bouldery and frequently inundated with river water. The effect of a continuous river entry water supply is dramatically greater than the influence of the soft constraints operating regime. WR is often non-vegetated to very sparsely vegetated with less than three per cent vegetation cover and is not considered vegetated at the landscape scale. WR persists as a major, active fluvial channel. WR at Halfway River The CLBMON-33 project uses mapping at the landscape level to define the vegetation communities introduced in the above table. In the mapping, these vegetation communities are expressed as deciles, i.e., 10 per cent increments of community types within polygons. A change in the vegetation communities at the landscape scale is therefore relatively coarse-grained, and may not detect change at the ground level. However, it could be argued that for a large area such as the Arrow Lakes Reservoir, only small scale changes are relevant to the overall character of vegetation in the drawdown zone. In their review of mapping and analysis of catchments at the landscape scale, Apan et al. (2002), identified the landscape perspective as the need to consider larger spatial and temporal scales in policies and guidelines for managing land. They suggested that a broad scale perspective is best for the incorporation of spatial relationships into land use planning, to allow for decisions about management actions that could result in a critical change in an existing condition (Apan, et al. 2002). A review of the literature applicable to this project indicates that few studies other than concurrent Water License Requirement projects (i.e., CLBMON-10 Kinbasket Reservoir Inventory of Vegetation Resources; Hawkes et al. 2007, 2008) compare to CLBMON-33. The soft constraints operating regime of the Arrow Lakes Reservoir has resulted in a general annual pattern of inundation of lowest water levels in the early spring, followed by a rapid midsummer inundation to full pool, and a more gradual decline in water levels over the winter months (Figure 2). 10

33 442 Water elevation (metres above mean sea level) Date Figure 2. Water levels in the Arrow Lakes Reservoir from 2005 to In all but 2008, the lowest elevation vegetation communities between 434 and 436 metres were exposed during the spring, followed by rapid reservoir level increases. A gradual decline in water levels followed the highest water periods, which tended to occur in July and August. Figure 2 shows that in most years, water levels dropped in August and a period of exposure occurred prior to the onset of winter. In 2008, however, the elevation bands from 434 to 438m remained inundated during the fall. Subsequently, the August to December low water periods resumed in 2009 and in This pattern appears to be important, and is discussed in the Results (Section 0). The management questions regarding the character of the vegetation (structure, spatial extent, diversity and distribution) assume that vegetation is influenced by water level patterns; however, the study also includes the consideration of environmental variables such as soil texture, precipitation and temperature, slope and aspect. The management questions regarding potential changes resulting from the soft constraints operating regime are being addressed with comparisons over time. In past years, when examining the current year s data, environmental variables such as soil moisture and textures have been found to have an influence on overall species diversity and vegetation heights (Enns et al. 2009; page 45) and these influences are as important as elevation (as a surrogate for inundation), but in fact, they are linked with elevation. The development of the 11

34 soil features are, in part, due to where they reside in the elevation profile. Silts accumulate at low to middle elevations and these materials only support shorter vegetation types that are adapted to the low elevation silt communities. Gravels and sands accumulate at high and low elevation; at high elevation they will support water intolerant, drought tolerant, tall shrubs. At low elevation, sands and gravels are mostly bare, due to the effects of scouring and duration of inundation. In previous years, the patterns in the vegetation communities in relation to the water levels indicate that cover and heights of vegetation have been influenced by duration of high water, that the low elevation vegetation communities in the reservoir are well adapted to frequent and prolonged inundation, and conversely, that the drought tolerant vegetation community types at higher elevation benefit from inundation (Enns et al. 2009). 12

35 2.0 METHODS 2.1 Field Work The methods used in CLBMON-33 include aerial photographic interpretation (API) and analysis, and the collection and analysis of field data, based on adaptation of methods developed by Keser (1976) and Luttmerding et al. (1980). In 2007, the aerial photographs were viewed in stereo on paper prints and on screen as mosaics. Polygons were delineated on the mosaics describing similar groupings of physiographic features and vegetation. For example, a headland would be delineated separately from a sheltered bay. Each polygon could include up to three vegetation community types (VCTs). As described in Table 3 and Table 1, the VCTs each have a characteristic set of topographic and soils features together with typical vegetation for those features. Some VCTs were based on physiographic features only, such as steep boulders (BB) as they had less than three per cent vegetation cover in the type, and were not visible as a vegetation feature at the landscape scale. To ensure consistency, the field data were collected in a similar fashion as in previous years (Enns et al. 2007, Enns et al. 2008, Enns et al. 2009). Plots were located within the VCTs in the polygons. Polygons were selected randomly for field sampling, but in some cases, specific VCTs were picked to increase the sample size for VCTs with a previously low sample size. Some of the plots that were done in previously sampled areas were resampled in Power analysis was used to arrive at more appropriate sample sizes than in previous years for both the map data and the field data collection and analysis (Omule and Enns, 2010). Pre-field planning included the identification of new plot locations in previously unsampled or undersampled study areas, using random generation of polygons. For each VCT, a list of polygon numbers was created, which included all polygons in which that community type occurred. Polygon numbers for field and aerial photographic sampling were randomly selected from all available polygons in the map through use of the Random function in Microsoft Excel. Field plots were randomly located using a random stick toss method. The GIS locations and field data for previous plots were recorded in the GPS (Nomad 800L with SX Blue Differential). Maps of each study area, showing the previous plot locations were prepared and placed in binders for use in the field. Pre-field orientation, field training, safety procedures and field standardization were completed by the field crew prior to the start of field data collection. Calibration and standardization of vegetation cover estimates and height measurements was done on May 13, 2010, and reviewed during every week of field work. Field data, including specimens from previous years work, were referred to in the field. A field audit by BC Hydro took place part way through the field season, on May 31, and the field audit document was received on September 29,

36 Table 5 shows the dates of activities in the field and after field work for the CLBMON-33 project. 14

37 Table 5. Dates of Work undertaken to complete CLBMON-33 in Action Date Aerial photographic capture (Terrasaurus Inc.) May 11 15, 2010 Field assessment of vegetation plots, field truthing of the 2007 mapping, collection of data for image comparison. May 13 June 17, 2010 Safety audit May 25, 2010 Field practices audit May 31, 2010 Orthorectification and aerial photograph mosaic creation (4D- GIS) Field data archiving, cleaning and data entry Photo review, field data records compilation, data review, data screening, plotting for non-linearity, etc. May 15 August 31, 2010 June 15 August 20, 2010 August 31 September 27, 2010 Data collection from the imagery, biometrics and writing August 31, 2010 present. Figure 3 shows the general locations of the field plots. Due to the size of the study area and the large number of plots that were done in the field, it is not possible to put all plot locations on a single map for display purposes. More detailed plot locations are available in the database for the project. Plots that were repeated over time in the same place were compared using data from both CLBMON-33 and CLBMON-12 (Arrow Lakes Reservoir Monitoring of Revegetation Efforts and Vegetation Composition Analysis - a companion project with similar field methods). VCTs with very low vegetation cover were not extensively sampled (or resampled) in Previously sampled low cover VCTs with only a few small plants in them were recorded and photographed, but were not included in the comparisons over time unless they had more than three per cent cover. Examples of community types with very low vegetation cover that is not visible at the landscape level (less than three per cent) include SS: Steep Sand, BB: Boulders and LO: log zone (Table 4). LO: Blue wildrye log zone and PO: Pond can also be ephemeral from year to year. VCTs such as CL: Cliff rock outcrop, PO: Pond, WR: Silverberry river entry and WS: Willow stream entry are either very uncommon or impacted more by incoming streams and rivers than by the soft constraints operating regime and were not included in the analysis of data. Although data were collected in repeated plots in 2010, these data were not extensively analyzed this year. In the final year of this project, these VCTs will be compared to previous plots established in these types in the field, and will be included in the data analysis and interpretation of the digital imagery. 2.2 Field Work The dates of the field work are important; fieldwork had to be done according to priority to complete the low elevation sites before they were inundated. Low elevation (434 to 436m) field work was completed in the first two weeks of the sampling period, from May 11 to May 20, Middle to high elevation field work (436 to 440m) was completed between the last week of May and the third 15

38 week of June. By June 17, 2010, all of the low elevation area was inundated, and the water level was at approximately 437.5m. Figure 3. General location of field plots completed in 2010 within study areas in the Arrow Lakes Reservoir for the CLBMON-33 project. Red dots are the plot locations (which tend to overlap at this scale) and yellow lines are the 43 study areas 2 within the Arrow Lakes Reservoir. Details of the 559 plot locations are provided in the map database. 2 A total of 43 areas were selected for re-vegetation and monitoring in Moody, A.I Mica - Revelstoke - Keenleyside Water Use Plan: Potential Areas for Vegetation Establishment in the Arrow Lakes Reservoir. BC Hydro Contract Report. 45 pages. 16

39 The consequence of this schedule was that higher elevation plots tended to be done last, and cover and heights of the vegetation may have been more advanced in high elevation plots as a result. Enns et al. (2009) noted that heights of reed canarygrass were often a full metre taller at the beginning of the sample period than at the end. The priorities for repeated measures analysis (RMA) sampling, as identified by Omule and Enns (2010), were to reassess all 152 of the plots established from 2007 to 2009 (including any established in CLBMON-12), and to sample an additional 571 plots for vegetation height and cover. The Shannon-Weaver diversity index analysis required a total of 375 plots (Omule and Enns 2010). The suggested target numbers for each vegetation community varied among types. All plots were completed within the 2010 field sample, with some exceptions. The exceptions were for those types that did not occur frequently enough to support the targeted number of samples. For example PO: pond only occurred in approximately four polygons in the mapped database and all of these were sampled. In previous years, plots consisted of three 0.5 m 2 subplots within a larger 5 x 10 m plot, with a list of all species made for each subplot, as well as for the larger plot. In 2010, the larger plot was completed without the use of the subplots, because the smaller plots did not add new species to the record for a site, but added considerably to the amount of time required to complete a plot. Also, data from the previous three 0.5 m 2 subplots were amalgamated in past reports to represent one large plot. Larger plots can be more accurately relocated over time. Previously sampled plots from both CLBMON-12 and CLBMON-33 were included in the 2010 sample. Additional polygons (in which new plots were to be located) were selected at random using the random function in Microsoft Excel to provide a list of potential polygons that could be sampled. A random stick toss method was used to identify the entry location for the sample polygon, as several potential representatives of a given VCT could be sampled within the same polygon. Therefore, the actual plot location was either in a previous year s plot (2007 to 2009), or nearest to that plot in the randomly located VCT identified for sampling by random numbers generation, followed by random stick toss using the mapping as the basis for the toss. During the field sampling, plots were laid out in the VCTs within the randomly located polygons. In previous years, plot squares were tossed into the VCT from its margin within the polygon. A tendency to toss into non-vegetated areas was noted when this method was used. The same method was used in 2010, but nonvegetated tosses were rejected. Occasionally, potential plots were rejected if they were disturbed by outdoor recreation vehicles (ORVs) or had been incorrectly labelled in the mapping. Vegetation plots were 5 by 10 metres and were oriented north to south or in the direction of Columbia River main stem water flow, to avoid crossing over an elevation band. Each plot was marked at the corners with a pin. The corners were temporarily flagged and GPS locations were recorded using the Nomad 800L GPS. GPS locations were also recorded on the plot forms for each plot. The location information on the plot form was photographed. A species list was made and all species recorded. Minimum and maximum heights of each species were measured to the nearest ten centimetres for the dominant individuals or clumps, 17

40 for heights up to 4 metres. Heights higher than 4 metres were estimated at 1.0 metre intervals, using a 2 metre measuring stick and standing at a distance from the subject vegetation. The vertical projection of each species cover value was estimated using the per cent cover guide from Describing Ecosystems in the Field (Luttmerding et al. 1990). Distribution codes were recorded for each species, also using the codes described in Luttmerding et al. (1990). Photographs were then taken of the plot from the south looking north, and of the surrounding polygon characteristics to the east, west and south. Photographs of the finished plot form were also taken at the plot. Plot location flags were removed, and a labelled wooded stake was installed at the downstream corner furthest from the water. The field data were downloaded weekly into a GIS database, and copies were made of each field form on the same day the data were collected, using photographic records. Checks of the GIS data accuracy were run each week, and plots were recorded on new versions of the maps for further use in the field. The photographs and field data were stored both on and off site. Changes to the mapping (i.e., any errors in original classification) were made on paper copies of the maps in the field. These changes to mapping are not to be confused with shifts in VCT classification due to plant successional change in response to the operating regime and other factors; they were simply errors due to lack of previous field verification on some sites with poor or limited access and poor resolution in the 2007 imagery. Map record changes were photocopied. Species identification was done daily, with reference to the plant collection for the project and use of taxonomic keys. Some specimens were sent to Frank Lomer or Dr. Terry McIntosh (private contracting botanists) for verification or identification. These data were used to complete the field records in the database. Field forms were corrected daily, and were ready for data entry during the field work. 2.3 Post-field Work Database entry was reviewed in-house prior to spreadsheet compilation. The structure of the database, including all tables and pivot tables, is shown in Appendix 1. Data were entered such that each species in each plot had a separate line in the database, and that line included all plot data, as well as the species cover value, the minimum and maximum heights, and any comments made in the field. The database includes life form codes for species and a classification for plots with very low cover (less than three per cent). There was an average of seven species per 5 by 10 metre oblong plot, in the 559 plots. Screening of the pre-existing database and coordination of the data to allow for comparisons were required to complete the data preparation, including the screening of field data from previous years. This year represents the first comparison to the previous years field data. Previously, only within-year data had been examined. The numbers of plots and polygons used in the analysis are shown in Appendix 3. Each field plot was assigned a climatic regime using the weather data supplied by BC Hydro, and an elevation band according to the digital elevation model (DEM) 18

41 based on the location of the plot corners acquired in the field with the Nomad 800L GPS latitude. There were three elevation ranges: 434 to 436 metres, 436 to 438 metres, and 438 to 440 metres (BC Hydro 2005, Grau and Walsh 2007). A small number of plots had one corner straying into an adjacent elevation band. These were repositioned (in the database) to the dominant elevation band for the plot. 2.4 Imagery data collection and database extraction The digital image data capture took place during the CLBMON-33 field work, when the water was below the minimum elevation required for data collection (434m). Digital colour aerial photography of the 43 study areas, at a scale of 1:5,000 was captured using a Hasselblad medium format digital camera mounted with a long lens to minimize water reflectance and tree lean. The camera was housed in a gyro-controlled mount, capturing imagery from a fixed-wing aircraft. The imagery was captured at a 10 cm scale during sunny conditions at a high sun angle. Study areas along the western side of the reservoir were captured during the morning hours to minimize tree shadow effects. Eastern side study areas were captured later in the day. The areas of capture were based on the vegetation polygon shapefile. The digital imagery was delivered in RAW and 16 bit TIFF formats. Airborne differential GPS coordinates of each photo centre were also delivered to aid in the photography processing. 2.5 Post-flight Work A total of 893 scanned digital images were triangulated. The previous BC Hydro aerial triangulation supplied 2017 horizontal and 3205 vertical controls. These adjusted coordinates were used as ground control data. All points were assigned weights of 0.1 metre. In addition, Terrasaurus s airborne GPS control data (893 positions) were assigned weights of 2.5 meter both horizontally and vertically. Photogrammetric preparation included point selection and coding using an automated point extraction and triangulation package, and photos were tied with a minimum of 15 tie and pass points. The photo ties were then bundled and tested for error. Both the BC Hydro control points and the airborne GPS control points were found to fit well in the aerial triangulation adjustment (4DGIS 2010). Data files were delivered at the end of September and beginning of October The resulting image files were used directly in the GIS without further processing. The relationship between stationary objects in the 2007 and 2010 images was very close, to the point where individual boulders appeared identically matched between the two images. There are residual problems that are common to all aerial photographic images; these were identified during the interpretation, and are discussed below. Once the 2007 and 2010 imagery was overlain with the 2007 vegetation community type map data, a series of clips of individual polygons were taken and compared to field data from both 2007 and 2010, as a means of informally comparing the pixelation of the 2007 imagery at high power (> 1: 500 scale) to the 2010 imagery. Then, using the sample size recommendations from the power analysis (Omule and Enns 2010), comparisons between the spatial extent, structure, and distribution of vegetation community types within polygons, as they appeared in 2007, were compared with 2010 (Figure 4). Appendix 2 shows details 19

42 of the locations of polygons included in the aerial photographic comparisons. Each randomly located polygon containing VCTs for comparison between 2007 and 2010 was observed at various scales, ranging from 1: 500 up to 1: 20,000, and an attribute table was created for the individual records. The distribution of sampled polygons was such that all study areas were included in the comparison. During the sampling, some polygons were rejected if parts of the VCT was obscured by shadows from leaning trees, or if the 2007 image was murky or too pixelated to make a reasonable comparison. Detailed data were collected from 398 polygons out of In addition, approximately 500 further polygons were quickly reviewed to determine if a change in decile coverage had occurred. 20

43 Figure 4. Locations of the 398 polygons compared for changes in vegetation community type features (decile differences, spatial extent, structure and composition) between 2007 and Details of the polygon sample locations are provided in Appendix 2. 21

44 The attributes included in the assessment were: Magnitude and direction of a decile change in VCT coverage in the polygon observed between 2007 and 2010, for the VCT represented in the polygon (for example, 10PA in 2007 to 9PA in 2010, or 6BG in 2007 to 5BG in 2010). Cover of vegetation within the areas classed as BE, BG, CR, PA, PC, PE and RR at 1: 1,000 scale in 2007 and Height of the leading vegetation of the VCT in question, within the polygon supplying the VCT (e.g., BG 2007 = 0.5 metres, BG 2010 = 0.1 metres). The error associated with this estimate is discussed below. Distribution of the vegetation within the areas classed as BE, BG, CR, PA, PC, PE and RR at 1: 1,000 scale in 2007 vs These data were combined with the environmental variables for the polygons, which included (for each polygon) latitude, slope, aspect, elevation (some polygons cross over elevation boundaries), numbers of days exposed, timing of days exposed, average temperature and accumulated precipitation, substrate drainage (estimated sand, silt, clay, gravel and cobble percentages), water energy and flow class for use in Repeated Measures Analysis (RMA). These variables are defined further in Appendix Data Analysis Using the map data on screen, a survey was conducted of the delineation and status of VCT deciles in 398 polygons (out of approximately 2000 polygon records), comparing deciles of VCTs as mapped in 2007 with status of the same VCT component in The polygons were selected at random, with sample sizes determined by power analysis (Omule and Enns 2010). Notes were taken as to the character and types of changes, where they occurred. The vegetation structure and composition from the field plot data were compared using notched Tukey s boxplots (Figure 5). Boxplots show the distribution of data within a set of samples more completely than comparisons of means, analysis of variance tables or use of standard error or standard deviations. Notched Tukey s boxplots are a visual method of showing statistically significant differences (p<0.05) among data distributions from comparably sampled populations (Tukey 1977; Velleman and Hoaglin 1981; Wilkinson et al. 1992). The notched boxplot appears as a narrow-waisted box. The narrowed centre is the median of the set of values plotted for a variable. The hinges, or narrow ends of the box, represent the 25th and 75th percentiles of the data. The position of the hinges relative to the median provides an indication of the skewness of the values in the given data. Circles or asterisks are often plotted above and below the main box. These data may be outliers. The vertical extent of the notch in the box represents the 95 per cent confidence limits about the median value. When several of these notched boxplots are shown in the same graph, the position of the notches relative to one another shows whether the medians of the datasets are significantly different at the 95 per cent confidence level. Notches that do not overlap indicate medians that are significantly different, while notches that do overlap indicate no significant difference among the medians. Sometimes the 95 per cent confidence level of the median exceeds either one or both of the 25th 22

45 and 75th percentiles (the hinges or ends of the box) and the graphical result of this is a box that appears 'inside-out' at one or both ends (Figure 5). Far outside value * Outside value Whisker Upper hinge (75 th percentile) 95% confidence interval on the median Median Lower hinge (25 th percentile) Whisker Figure 5. Illustrated example of a notched Tukey s boxplot. Boxplots show a statistically significant difference (p<0.05) between medians from comparably sampled populations when the notches or two or more boxplots do not overlap. The narrow centre of the box represents the median point in the data set. The top and bottom of the box are the 75 th and 25 th percentiles, respectively, and are called the hinges. The Hspread is the absolute value of the difference between the values of the two hinges. The whiskers show the range of values that falls within 1.5 Hspreads of the hinges. Inner fences (not plotted) are equal to the value of the hinge ± (1.5 x Hspread). Outer fences (not plotted) are equal to the value of the hinge ± (3.0 x Hspread). Asterisks and circles show values outside the inner and outer fences of the data distribution and are potential outliers. The notch in the box represents the 95 per cent confidence limits of the median. When the confidence limit exceeds the 75 th or 25 th percentile (i.e., the end of the box), the boxplot looks like it has been turned 'inside-out' (McGill et al. 1978). The total coverage of VCTs in Revelstoke Reach was compared with those in the Arrow Lakes (these areas of the Arrow Lakes Reservoir are shown above, in Figure 1). Tests for significant changes in average covers of leading species in VCTs and mean heights for leading species in VCTs in 2010 were completed. Tests for differences in overall VCTs site-wise over time (from 2007 to 2010) were also done. Post-hoc comparisons were used to identify differences among VCTs 23

46 by elevation band, as an indication of the effects of the soft constraints operating regime, based on depth and duration of inundation over time. For landscape-level analyses of diversity and evenness by VCT, the Shannon- Wiener Diversity Index was calculated as follows: H' = -Σ (p i ln p i ), where p i is the relative frequency or proportion of species i of the total number of occurrences of all species in each VCT. Pielou's Evenness Index (J) was then calculated as follows: J = H'/H' max = (-Σ (p i ln p i ))/ ln q, where q is species richness (or the total number of species in each VCT). The H' and J statistics were calculated in a similar fashion for each elevation band, based on total numbers of occurrences of all species in each elevation band and the species richness by elevation band, respectively. In addition to the decile change data, we also compiled data from the vegetation base map for use in map statistics. The map database contains polygon vegetation records, as well as environmental attributes for each sample polygon (Enns et al. 2007). Aerial photographic interpretation was used to compare the differences in polygons between the 2007 and 2010 imagery, at a greater intensity than the decile level (as above). Power analysis was used to determine the number of polygons required for an adequate sample size for each VCT (Omule and Enns 2010), which varied by VCT according to variability of cover and height within each VCT. The random numbers generation feature in MS Excel was used to select polygons representing the replicates of the VCTs, and an attribute table was created comparing the change in deciles and the change in cover and heights of the vegetation in the specific VCT between 2007 and The superior imagery of the 2010 photographs allowed for relatively accurate estimates of changes in cover and heights, but uncertainty in height measurements was occasionally high enough to cause the rejection of a polygon from the data collection process. The base map has complex polygons (each with up to three VCTs); each of these complex polygons was assigned a polygon ID. The VCTs were described from these original polygons. Any polygon straddling more than one elevation band was split into the relevant elevation bands. A given polygon that occurred in more than one elevation band may therefore have up to three records, in which all the data except the elevation band, including the polygon ID, are equal. This creates redundancy in the data, which is a source of uncertainty. The numbers of records (for each year) per VCT and elevation band are provided in 24

47 Table 5 5. Two types of analysis were used: coded data and the Kappa statistic (coarse grained analysis), and cover and height estimates and analysis of variance (ANOVA) (finer grained analysis). The map analysis was based on an adaptation of the confusion matrix approach (Congalton and Mead 1983): K P o 1 P P c c where P o is the proportion of observed agreements and P c is the proportion of agreements expected by chance (Sim and Wright 2005). Theoretically, the range of possible values of Kappa is -1 to 1. The value 1 represents perfect agreement in codes between 2007 and 2010 (i.e., no change). A value of 0 indicates agreement no better than that expected by chance. A value of -1 indicates agreement worse than that expected by chance. Negative kappa values are very rare in practice; thus, the range of K is between 0 and 1, and in most significance tests the lowest value K = 0 is assumed. The confusion matrix for each variable is square, and its rows and columns are constructed based on the codes of the variables, with the 2007 codes being the rows of the matrix and 2010 codes being the columns. The matrix cells are populated with polygon frequencies. For example, if the code in a polygon changed from a code 5 to a code 8, then it would go into the cell defined by row 5 and column 8. If there is no code change in any of the polygons, then all the frequencies would line up along the diagonal of the matrix. The Kappa statistic is estimated and a test of the null hypothesis of no agreement between 2007 and 2010 codes (K = 0) is performed. This null hypothesis in our application means that there have been changes in vegetation cover, height or distribution codes in a VCT between 2007 and Note that the distribution codes were based on the Daubenmire cover scale for cover and half metre increments for the height variables (Mueller-Dombois and Ellenberg 1974). The analysis was done by constructing matrices from the raw map cover and height data, and calculating the simple Kappa statistic, as well as a test for the statistical significance of the Kappa estimate. The significance tests were at the 95 per cent probability level (two-tail). The VCT codes for cover were used and compared between 2007 and We compared only those VCTs that were influenced primarily by the Arrow Lakes Reservoir water behaviour, and not by feeder streams and rivers, all terrain vehicles, or industrial activity. As coded data were considered coarse grained, and the image clarity from 2010 allowed a more detailed assessment of cover of vegetation within VCTs, a further analysis of change was used. This analysis was based on cover and height estimates for 2007 and 2010 of vegetation within VCTs, and used the same database as the Kappa statistic. A repeated measures analysis (RMA) ANOVA using a linear mixed-effects model (LMM) (Little et al. 2006) was used to test for significant differences in cover and height between 2007 and The potential fixed effects of interest were geographic area (Arrow or Revelstoke Reach), VCT, elevation band and year (2007 and 2010). The elevation bands and the water level data, by year, are surrogates for the differences in depth, timing and duration of inundation by year 25

48 in the reservoir. However, a single LMM incorporating all the four effects in a factorial arrangement could not be calculated. This is because of the duplication of the records for split polygons as described in Section 2, which would have resulted in pseudo-replication. Thus, for each geographic area, the RMA two-way ANOVA was used to test for the significance of the year effect for the elevation bands within each VCT, and for the VCTs within each elevation band. That is, we tested for a VCT effect, an elevation band effect, a year effect, a VCT x Year interaction effect, and an Elevation Band x Year interaction effect 3. However, it was not possible to test for significant differences among interactions between community types and elevation or between geographic areas. When the second high quality image is flown in 2012, it will be possible to test for differences between the geographic areas by aerial photographic interpretation of polygons split by elevation. The mathematical form of the fitted full LMM was: y ijk VCT Year VCT Year [1] i j i j ijk Or: y ijk Elevation Year Elevation Year [2] i j i j ijk where y ijk represents the observation of vegetation cover or height in the kth polygon in the ith VCT or Elevation Band; is the overall mean of the observations; and is the random error associated with the kth polygon. ijkm Random effects were assumed to be independent and normally distributed with mean zero, with a unique variance component. Model [1] was for comparison among VCTs for each elevation band and model [2] analysis by VCT. This allowed for tests for significant differences in changes of areas of VCTs within polygons over time. Change detection included an evaluation of error. Least squares means for each interaction term were calculated and their differences were tested for significance pairwise. These analyses were done using SAS (version 9.2). 4 RMA was also used to examine the influence of environmental variables on the VCT spatial extent, structure and composition. This step evaluated the differences in cover values and height values for VCTs to assess associations between dependant variables and elevation, slope, aspect, soil texture and climate variables from weather data supplied by BC Hydro. Canonical Redundancy Data Analysis (RDA) was used to relate the polygon vegetation attributes (Y, dependent variables) to the environmental attributes (X, independent variables). The RDA performs a principal components analysis on fitted values of Y, and produces canonical axes and eigenvectors. Each axis 3 This analysis is similar to the CLBMON-12 analysis by Enns and Gibeau 2009, pages SAS Institute Inc., Cary, NC, USA. 26

49 corresponds to a direction related to a linear combination of the X. The eigenvectors give the contributions of the X variables (and predicted Y) to the canonical axes. The eigenvectors are then used to produce RDA ordination biplots. The analyses were done separately for each geographic area, using SAS (version 9.2). The RDA analysis involved the following steps for each portion of the Arrow Lakes Reservoir (Arrow and Revelstoke Reach): Examine the scatter plots of the Y versus X to see if there were any relationships. Standardize the quantitative variables to mean zero and standard deviation 1. Code the six categories of the exposure regime qualitative variable into six dummy variables. There were 13 potential independent variables in the Arrow and 10 in Revelstoke Reach. Conduct forward selection regression analysis on the potential independent variables to attempt to reduce the number of the independent variables to be used in the RDA. Perform canonical RDA analysis using the standardized values in step 2, to relate Y to X. That is, assess how well the original Y variables could be predicted from the canonical variables (from the selected X independent variables). Create biplots from the RDA outputs of the normalized eigenvalues (scores of vegetation and environmental variables), for two principal axes. 27

50 3.0 RESULTS The 2010 results for CLBMON-33 are presented in relation to the management questions, and a summary is provided in the Discussion (Section 0). In some cases, we refer to both the field and map data to address a particular question, and in others, we use field data alone, specifically where diversity, evenness and richness of VCTs are considered. We have changed the order of the questions, slightly. 3.1 Question 1. What are the existing vegetation communities in the Arrow Lakes Reservoir drawdown zone between 430 and 440 metres? This management question was addressed in previous reports (Enns et al. 2007, Enns et al. 2008) by mapping and describing the existing vegetation at the beginning of the project. To verify and refine the vegetation community types, and to monitor the changes in the VCTs heights and covers over time, a total of 559 field plots were completed in See Appendix 3 for the numbers of plots completed in each VCT. The distribution of the VCTs with respect to their typical elevation is shown in Figure 6. Each VCT has a distinct distribution in the Arrow Lakes Reservoir, and the distribution is consistently aligned with physiography (e.g., the patterns in landforms), including slope, topography, and hydrology (Enns et al. 2007). For example, gentle alluvial fans have different VCT composition from steep or gentle exposed tills. The occurrence of bedrock or stream and river channel entries also determines the VCT composition. Small stream entries that have been cut off from their older patterns due to the creation of the reservoir form emergent underground streams (RR: Reed-rill), often percolating to the surface in the middle elevation band of the drawdown zone. Old serpentine fluvial channels and oxbows are the origin of PO types in Revelstoke Reach. Terrain texture seems to determine the prevalence of boulders, or sands or silt clay mixtures. If the exposed till is steep and bouldery, then BB: Boulders, steep is usually the prevalent VCT. The soft constraints operating regime tends to influence the deposition of fine textured materials, such as silts and clays. These often accumulate in the lower elevation VCTs, such as PE: Horsetail lowland. Sandy headlands and bays have characteristic sandy VCTs, such as BE: (sandy) Beach, and BG: Gravelly beach, which were formerly just sandy or gravelly textured till. The action of water in the reservoir has also exposed either sandy or gravelly material and re-deposited some of it downstream and in upslope positions, such as the PA: Redtop upland. Because deposition and fluviation are continual processes, changes in the VCTs are likely to occur over time, as materials develop and change in their distribution, similar to, but possibly more dynamic than an unmanaged river system. An exception to the importance of terrain and hydrology is IN: Industrial / Residential / Recreational, which occurs mainly where human populations cooccur. 28

51 Elevation (m) BE BG CR PA PC PE RR Vegetation Community Type Figure 6. Typical elevation spread of Vegetation Community Types (VCTs: see Table 3 in the Arrow Lakes Reservoir. Only those VCTs with greater than three per cent cover of vegetation, sufficient sample size (Omule and Enns 2010) and primarily influenced by the soft constraints operating regime are shown; BE = Sandy beach, BG = Gravelly beach, CR = Cottonwood riparian, PA = Redtop upland, PC = Reed canary grass mesic, PE = Horsetail lowland, RR = Reed-rill. There is a higher diversity of VCTs in the Arrow portion of the Arrow Lakes Reservoir (Figure 7) primarily due to the longer reach and higher variability. The Arrow portion spans approximately four climate gradients than the much shorter Revelstoke Reach portion, which occurs primarily within the wettest and coldest of the climatic zones 5. There is also a much higher diversity of parent materials and soils in the Arrow portion of the reservoir than in Revelstoke Reach. Some of the characteristics of BG: Gravelly beach and BE: Beach overlap can co-occur in both sections of the reservoir. PC: Reed canary grass mesic occurs at all elevations, but is most frequent at middle elevation and is a major component of Revelstoke Reach due to the very large, silty, sandy alluvial fans in this portion of the reservoir. CR: Cottonwood riparian and PA: Redtop upland are usually adjacent to each other at high elevations throughout the reservoir, whereas PE: Horsetail lowland median distribution is centred near the bottom of the elevation bands and is more common in the Arrow portion of the reservoir, in part because of steeper slopes in the Arrow portion. RR: Reed-rill is a middle elevation VCT that is influenced by underground stream entry into the reservoir and extends down, often perpendicular to the main flow of the Columbia River to the lowest elevation. It is not common in the Revelstoke Reach portion, although the 5 These approximate climate differences are loosely based on the biogeoclimatic gradients from Braumandl and Curran

52 vegetation in some of the gullied topography of the undulating alluvial fans of the Revelstoke Reach is similar to an RR VCT, but does not have a stream breaking through to the surface or any of the surficial changes that streams make to materials. That community is more common in the steeper Arrow portion of the reservoir, which is more commonly dominated by gravel and sand. No new vegetation communities were identified in either the fieldwork or from the map analysis following the fieldwork in Variation within and among species compositions of VCTs has been described in CLBMON-12 (Enns et al. 2009), and no departures from the previously described mapped classification were recorded in 2010, although additional data increased our understanding of cover, height and species diversity of the VCTs. Some VCTs were not sampled extensively in 2010, other than to record changes that may have taken place since 2009 in comparison to previous data and photographs. These were VCTs that have remained almost completely nonvegetated since 2007, or have been influenced by factors other than the soft constraints operating regime. These include the boulder dominated VCT (BB) and VCTs influenced by standing water or incoming rivers (PO - Ponds; SS - Steep sand; RS - Willow stream entry; WS Silverberry river entry), or moving debris (LO Blue wildrye log zone) or VCTs where industrial effects were most common (IN). The total cover and height results by VCT from field work done in 2010 were very similar to previous years. These results are presented in Appendix 4, including boxplots of total cover and total height, comparisons between VCTs, analysis of variance and post hoc comparisons. Management Question summary: Sixteen landscape-level vegetation communities in the Arrow Lakes Reservoir drawdown zone occur between 434 and 440 metres and are distributed according to characteristic patterns. CR: Cottonwood riparian occurs between 436 and 440 metres and is most common between 439 to 440 metres elevation in the Arrow Lakes Reservoir. PA: Redtop upland occurs between 435 and 439 metres and is very strongly aligned with the 439 metre elevation in the Arrow Lakes Reservoir. BE: Beach, BG: Gravelly beach, PC: Reed canary grass mesic, and RR: Reed-rill all occur in a wider range of elevations, from 434 to 438 metres elevation in the Arrow Lakes Reservoir. PE: Horsetail lowland is centred between 434 and 435 metres elevation in the Arrow Lakes Reservoir, but occasionally occurs at higher elevation if adequate moisture is available. The above VCTs are the main vegetation communities that respond mainly to the soft constraints operating regime. The remaining nine communities may also respond to the soft constraints operating regime, but they are either rare or largely impacted by industrial development, incoming streams, and other factors. 30

53 3.2 Question 2. Is the current distribution of vegetation communities in Revelstoke Reach representative of the conditions in the remainder of the reservoir? To illustrate the distribution of VCTs in Revelstoke Reach in comparison to the Arrow portion of the Arrow Lakes Reservoir, we plotted the total area covered by each VCT in the two reach portions (Figure 7). 100% 90% Proportion of Total Area (%) 80% 70% 60% 50% 40% 30% 20% 10% 0% PC PA BG BE Arrow PC PA BG BE Revelstoke Reach BB BE BG CL CR IN LO PA PC PE PO RR RS SF SS WR Figure 7. Total Area Coverage of each VCT, from Arrow Lakes (Arrow) and Revelstoke Reach (Revelstoke). This plot shows that Revelstoke Reach has less diversity of VCTs than the Arrow, and that the vegetation communities in Revelstoke Reach are not representative of the conditions in the remainder of the reservoir. The management question 2 could be stated in hypothesis form as: H 0 : The areal coverage of VCTs in Revelstoke Reach is the same as for the Arrow Lakes portion of the reservoir. H A : The areal coverage of VCTs in Revelstoke Reach is not the same as for the Arrow Lakes portion of the reservoir. A contingency table and chi-square test was used to test the null hypothesis 2 above. A chi-square value of 0.05,15 = 543.6, was significant, and therefore H 0 is rejected. The current distribution of vegetation communities in Revelstoke Reach is not representative of the conditions in the remainder of the reservoir. To further illustrate the distribution of VCTs by area, we plotted the total areal coverage of each VCT in six sections of the Arrow Lakes Reservoir (Figure 8). Revelstoke Reach has much higher total area of the PC: Reed canary grass mesic dominated VCT, and less area of the other VCTs in the reservoir. The steep sand (SS) and bouldery beach (BG) VCTs cover larger areas in the Arrow portion of the reservoir. The midslope RR: Reed-rill VCT is not common in the Revelstoke Reach, whereas RR covers a larger area in the middle portion of the 31

54 Arrow Lakes. These differences seem to align with differences in slope, parent materials and climate. The highest diversity of VCTs occurs from Edgewood south to Deer Park and Renata. 100% 90% Proportion of Total Area (%) 80% 70% 60% 50% 40% 30% 20% 10% 0% PC PC PC PC PC PC BG BG BG BG BE BG BE BE BE BE BG BE 1 - North Reach 2 - South Reach 3 - Beaton Nakusp - Ferry 5 - Ferry - Burton 6 - Edgewood - Halfway South Sub Reach BB BE BG CL CR IN LO PA PC PE PO RR RS SF SS WR Figure 8. Total Area Coverage of each VCT, from left (Northern Boundary) to right (Southern Boundary): 1-North Revelstoke Reach, 2-South Revelstoke Reach, 3-Beaton to Halfway River, 4-Nakusp to Arrow Ferry, 5-Arrow Ferry to Burton, 6-Edgewood to the Southern Boundary of the Map Area at Deer Park and Renata. (Appendix 2 identifies the locations for the start and endpoints of the areas illustrated above.) Management Question summary: The current distribution of vegetation communities in Revelstoke Reach is not representative of the conditions in the remainder of the reservoir; areal coverage of many of the VCTs differs significantly between the Arrow Lakes portion and Revelstoke Reach. 3.3 Question 3. What are the spatial extents, structure and composition (i.e., relative distribution and diversity) of these communities within the drawdown zone between 440 and 430 metres? The VCTs in the Arrow Lakes Reservoir show distinct patterns in average maximum vegetation heights and total cover (Appendix 4) in relation to elevation; this has been shown and discussed in previous reports (Enns et al. 2007, Enns et al. 2008, Enns et al 2009). Total vegetation cover by elevation band and reach is plotted in Figure 9. There are significant differences in vegetation cover among elevation bands and between reaches, but the progression of shorter to taller vegetation occurs with increasing elevation in both Arrow Lakes and Revelstoke 32

55 Reach. The results of a two-factor analysis of variance for total cover are shown in Table 6. Significant results have a p-value < 0.05 and are shaded (full analysis of variance tables and post hoc comparisons are provided in Appendix 4). When considering total cover by reach only, there was no significant difference between Revelstoke Reach and Arrow Lakes. When comparing within elevation bands for the entire reservoir, total cover increased significantly with increasing elevation. There was also a significant effect on the total cover due to an interaction between elevation band and reach. In the lowest elevation band ( metres), there was no difference in total vegetation cover between reaches. The middle elevation band ( metres) had significantly lower total vegetation cover in Revelstoke Reach than in the Arrow Lakes. Total cover in the upper elevation band ( metres) was significantly higher in Revelstoke Reach than in the Arrow Lakes. This latter difference is due to the greater presence of the cold- and wet-tolerant willow (primarily Salix sitchensis) in the highest elevation band of Revelstoke Reach. 300 Total Vegetation Cover (%) A m A m A m R m R m R m Reach and Elevation Band Figure 9. Total cover of vegetation in the three elevation bands (434 to 436 metres, 436 metres to 438 metres, and 438 to 440 metres) in the Arrow Lakes (A) and Revelstoke Reach (R) portions of the Arrow Lakes Reservoir. 33

56 Table 6. Analysis of variance results for total vegetation cover and average maximum vegetation height by reach and elevation band. Significant results have a p-value < 0.05 and are shaded. Effect Tested P-value Result Total Cover by Reach Not significant Total Cover by Elevation Band Significant Total Cover by Reach x Elevation Band Significant Average Maximum Height by Reach Significant Average Maximum Height by Elevation Band Significant Average Maximum Height by Reach x Elevation Band Significant Note: Full analysis of variance tables and post hoc comparisons are provided in Appendix 4. Average maximum height by elevation band and reach is plotted in Figure 10, illustrating significant differences in height between elevation bands and between reaches. A two-factor analysis of variance for height (Table 7) showed that average maximum vegetation height was significantly greater in Revelstoke Reach than the remainder of the reservoir. When comparing within elevation bands over the entire reservoir, average maximum vegetation heights were significantly higher with increasing elevation. There was also a significant interaction between elevation band and reach. The average maximum height was significantly greater in Revelstoke Reach within each elevation band than in the remainder of the reservoir. Average Maximum Vegetation Height (m) A m A m A m R m R m R m Reach and Elevation Band Figure 10. Total heights of vegetation in the three elevation bands (434 to 436 metres, 436 metres to 438 metres, and 438 to 440 metres) in the Arrow Lakes (A) and Revelstoke Reach (R) portions of the Arrow Lakes Reservoir. As with previous years observations, the VCTs that are confined to lower elevations, such as PE: Horsetail lowland, have lower vegetation heights than the 34

57 communities that are confined to higher elevations, such as PA: Redtop upland and CR: Cottonwood riparian. These are VCTs that are dominated and typified by the presence of trees and shrubs. The lower elevation PE: Horsetail lowland VCT often has high cover of inundation-tolerant or inundation-dependent vegetation. For those VCTs that extend across all elevations, heights and covers tend to increase with increasing slope, but can decline at high elevation if scouring from waves and log movement occurs during winter (Enns et al. 2009). Management Question summary: The spatial extents (cover of types and cover of vegetation within types), structure (heights) and composition within the drawdown zone between 440 and 434 metres have the following statistically significant patterns: Elevation: Mean Total Cover 438 to 440 m > 436 to 438 m > 434 to 436 m; Reach x Elevation: Mean Total Cover in Rev 438 to 440 m > Arrow 438 to 440 m > Arrow 436 to 438 m > Rev 436 to 438 m > Arrow 434 to 436 m = Rev 434 to 436 m Reach: Mean Max Height Revelstoke > Arrow; Elevation (whole reservoir): Mean Max Height 438 to 440 m > 436 to 438 m > 434 to 436 m; Reach x Elevation: Mean Max Height Rev 438 to 440 m > Arrow 438 to 440 m > Arrow 436 to 438 m = Rev 436 to 438 m = Arrow 434 to 436 m = Rev 434 to 436 m The vegetation composition within the drawdown zone between 440 and 434 metres is discussed in Question Question 4. How do the spatial limits (cover), structure (height) and composition of the vegetation communities relate to elevation, the topoedaphic site conditions (aspect, slope and soil moisture, etc.)? The map data were used to address the effects of elevation on spatial limits (cover), structure (height) and composition of the vegetation communities. Each VCT has a distinct median elevation, as reported above (Figure 6). The relationship between the VCTs and slope, aspect and soil moisture was examined in the field, including soil texture and nutritional measurements in 2007 (Enns et al. 2007). In 2008, Enns et al.(2009) explored these relationships further, demonstrating that drought-tolerant species co-occurred in drier soils and inundation-tolerant species occurred soils that were inundated for longer periods than soils at high elevation. The RDA analysis in 2009 seemed to verify the known autecological characteristics of the species in the reservoir as shown in eflora 6. Using map data, we examined relationships between spatial extent (mapped cover of vegetation in the VCT) and the set of environmental variables (Table 7). 6 accessed October

58 The values for a set of weather and water inundation variables were determined for the polygons based on their location in the reservoir. Considered together, these variables make up six exposure regimes based on the elevation band of the polygon and its proximity to the weather station that most closely reflects the weather experienced there (see Appendix 6). The exposure regime variables were limited by the fact that only two weather stations with records for the study areas, Nakusp and Revelstoke, had complete data. To simplify the analysis, we grouped the variables by elevation (three bands) and weather regime (two reaches). The exposure regime variables are further detailed in Appendix 6. The forward selection regression reduced the number of potential independent variables in the Arrow from 13 to 10; excluded were three exposure regime dummy variables representing N1 (Nakusp weather, low elevation), N2 (Nakusp weather, middle elevation) and R1 (Revelstoke weather, low elevation). The number of independent variables in Revelstoke was reduced from 10 to 9; excluded was the dummy variable representing R1 (Revelstoke weather, low elevation). Table 7. Environmental variables measured or estimated from the mapping for each polygon used in the RDA analysis for VCTs of the Arrow Lakes Reservoir in Sand, silt, and clay texture estimates are indicative of soil nutrient and moisture regimes. Environmental Variables Elevation Slope Aspect Terrain texture (estimates of moisture and nutrient classes) Temperature Precipitation Wave / Current action Metres above sea level Degrees Degrees Units/ Classes % boulders, gravels, sand, silt, clay Temperatures in degrees Celsius during exposure, by elevation zone Accumulated precipitation during exposure Classes: nil, low/current running over when flooded, moderate to high/waves/ripping/scouring causing surface erosion In the Arrow Lakes portion of the Arrow Lakes Reservoir, the environmental variables explained approximately 28 per cent of the variation in vegetation community type cover and height estimates from the aerial photography. The first and second canonical variables explained about 20 per cent and 8 per cent of the total variation in vegetation cover and height, respectively. The estimated heights of the vegetation in VCTs (from aerial photographic interpretation) appear taller in areas with more sand content, more slope and with exposure regime N3 (Nakusp weather, 438 to 440 metres elevation ) (Figure 11). The PC: Reed canary grass mesic VCT of the mid-slope mid-elevation portions of the drawdown zone probably contributed to this relationship. In the Revelstoke Reach portion of the Arrow Lakes Reservoir, the environmental variables explained approximately 36 per cent of the variation in vegetation cover and height (Figure 12).The first and second canonical variables explained about 36

59 24 per cent and 12 per cent of the total variation in vegetation cover and height, respectively. The vegetation cover in both Revelstoke Reach and Arrow Lakes portions of the reservoir was strongly negatively correlated with sand and gravel; coarse textured soils with few nutrients and/or a low water holding capacity together with scouring effects did not allow vegetation to develop high cover. Height was negatively correlated with silt content in soils. Silts were very common in the low elevation sites, and the plants that grow there are tolerant of fine textures as long as they have some exposure to water (eflora 7 ). The PE: Equisetum lowland is an example of this type. The plants in PE are adapted to being under water in anoxic soils with fine textures for long periods. 1.5 cover 1 silt 0.5 clay 0 boulder cover clay silt gravel height sand aspect slope N3 R2 R3 R2 R3 N slope boulder aspect -0.5 gravel sand height Figure 11. RDA Biplot for the Arrow Lakes portion of the Arrow Lakes Reservoir. The red lines with the arrows are the vegetation variables (height and cover) and the blue lines show the environmental variables. The environmental variables explained approximately 28 per cent of the variation in vegetation cover and height. The first and second canonical variables explained about 20 per cent and 8 per cent, respectively, of the total variation in vegetation cover and height. The vegetation cover is negatively correlated with gravel and sand, (which is related to scouring of those materials) and vegetation height appears to increase with increase in coarse fragment content (sands and gravels; which is related to drainage and the development of shrub heights at the highest elevation), an increase in slope and with the highest elevation, warmest and driest weather regime N3 (Nakusp weather, 438 to 440 metres elevation). 7 accessed May to September

60 In the Arrow Narrows portion of the reservoir, vegetation cover appeared to be weakly associated with silts and clays, which are common at low elevation and support the PE: Horsetail lowland VCT, which often had high cover. Vegetation height appeared to strongly increase with increase in coarse fragment content (sands and gravels), an increase in slope, and with the weather regime N3 (Nakusp weather, 438 to 440 metres elevation). This reflects the taller vegetation that occurs in the shrub dominated PA: Redtop upland, which is very common between 438 and 440 metres in the reservoir. Also, the ubiquitous reed canary grass (Phalaris arundinacea) was usually taller at higher elevation in the reservoir across all VCTs, unless scouring during the winter month from waves and log movement occurred. The patterns of diversity of VCTs in the Arrow Lakes portion of the reservoir from Edgewood to the Deer Park (Figure 8) tends to support the RDA findings presented in Figure 10; a higher variation in this part of the reservoir contributes to a greater diversity of shrubs, grasses and herbs at the upper elevations of the drawdown zone, even though cover may be low in these VCTs due to scouring. The Revelstoke Reach vegetation height appeared to increase with higher elevations, specifically with the climate classification of R2 and R3 (Revelstoke weather, 436 to 440 metres), high coarse fragment content, and slope (Figure 12). Coarse fragments in soils and non-soil parent materials were correlated with height in the Revelstoke Reach. Revelstoke Reach is cooler and wetter than the Arrow Lakes. Its climate and soils are more conducive to the presence of tall Sitka willow (Salix sitchensis). Sitka willow is abundant in Revelstoke Reach, but not in the warmer, drier Arrow Lakes. In Arrow Lakes, the drier climate is more conducive to the establishment of conifers and cottonwoods in the 438 to 440 metre elevation bands. 38

61 1.5 cover slope clay clay silt sand aspect slope aspect boulder 0 boulder cover gravel height R2 R3 height R2 R silt -0.5 sand -1 gravel -1.5 Figure 12. RDA Biplots for the Revelstoke Reach Portion of the Arrow Lakes Reservoir. The lines with the arrows are the vegetation variables (height and cover) and the remaining lines show the environmental variables. The environmental variables explained approximately 36 per cent of the variation in vegetation cover and height. The first and second canonical variables explained about 24 per cent and 12 per cent of the total variation in vegetation cover and height, respectively. The vegetation cover appears to be negatively related to estimates of sand and gravel fractions in materials and soils and the vegetation height appears to increase in areas with R2 and R3 exposure regime, high boulder content, and aspect. Height is negatively correlated with silts. Silts accumulate at low elevation and support shorter VCTs such as PE: Horsetail lowland. The cover of vegetation in each of the VCTs was also highly variable; this may be why vegetation cover is only negatively correlated with other variables. High covers can occur at all elevations and in all VCTs, depending on site specific factors. The PE: Horsetail lowland VCT can have 100 per cent cover, but be composed mainly of mosses and a few vascular plants. Scouring occurs at very low elevations and at high elevations, and is not as common at middle elevations (Figure 13), which reduces cover to a somewhat bimodal distribution in the data. 39

62 Figure 13. Scouring of gravel and sand caused by high water wave action between 438 and 440 metres in the Arrow Lakes portion of the Arrow Lakes Reservoir. (Photo taken in 2007 at McDonald Park.) The 438 to 440 metre elevation band is visible in the left side of the photograph and shows wave scouring and exposure of gravels. The 438 to 436 metre elevation band is visible to the right in this photograph and shows the lack of scouring in the middle elevation band and development of high relative vegetation cover. These patterns in winter wave action influence vegetation cover and heights in the reservoir. High heights and low cover can occur at high elevation, and low heights with high cover in the middle elevation bands are common depending on water energy. The effect of wave action in the RDA should be evaluated with caution. The wave and current action variables were derived from aerial photographic interpretation and were not based on actual measurements of water action (see Table 7 above for a list of the variables used in the RDA). Management Question summary: The spatial extents (cover of types and cover of vegetation within types), structure (heights) and composition relate to elevation and topo-edaphic site conditions in the following way: Arrow: Approx 28 per cent of variation in height and cover was explained by the environmental variables used in the RDA. Cover was negatively correlated with sands and gravels, which are influenced by scouring. Height was associated with higher elevation (more developed soils), increased coarse fragment content (better drainage) and increased slope. Revelstoke: Approx 36 per cent of variation in height and cover was explained by the environmental variables used in the RDA. Cover was 40

63 negatively correlated with sands and gravels, which are in turn influenced by scouring. Height was associated with higher elevation bands in Revelstoke Reach, where willows and other tall shrubs are adapted to the colder wetter climate, with increased presence of boulders (better drainage), and with aspects other than north- and east-facing ('increased aspect', i.e., with warmer sites). The data show that species diversity, richness and evenness tend to increase with elevation. These relationships are shown in Question 5, below. 3.5 Question 5. Does the soft constraints operating regime of Arrow Lakes Reservoir maintain spatial limits, structure and composition of existing vegetation communities in the drawdown zone? This management question implies a consideration of change over time. To address it, we examined data from field plots collected in 2007, 2008, 2009 and 2010, as well as the 2007 and 2010 map data. Seven categories of results were needed: Review of the mapping to record changes in deciles of VCTs within polygons between 2007 and This step also supplied the database for the following two categories of analysis. Repeated measures analysis of vegetation in mapped polygons using coded data (Confusion Matrix with the K-statistic), addressing differences in cover, height and distribution of the mapped data. Repeated measures analysis of the mapped vegetation data using ANOVA, addressing differences in cover and height. Comparisons of annual differences in cover and height, based on field data from 2007 through 2010, using Tukey s Boxplots and ANOVA. Annual trends in Shannon Weiner Index (H ), species evenness (J) and species richness. Examples of the mapping to illustrate the above findings. A pilot study to examine the use of image analysis as a means of defining change for the study areas as a whole (Appendix 8). The numbers of observations from the mapping and the field data for each of the subsections below are tabularized in Appendix Review of the mapping The comparison of decile coverage in VCTs between the 2007 and 2010 vegetation mapping showed that only six polygons out of a sample of 398 had a change in VCT decile coverage, and of these only two polygons had changes that were greater than one decile (i.e., greater than 10 per cent). Some examples of the changes in polygons from the mapping in 2007 and 2010 are shown in Section 0, following the presentation of the statistical analysis of the map and field data. 41

64 This result indicates that no significant change in vegetation has occurred between 2007 and 2010 in the Arrow Lakes Reservoir at the landscape scale. However, a 10 per cent difference in the vegetation may not be sufficiently conservative. Therefore, additional analysis was undertaken to examine change, using the field data and the mapping Repeated Measures Analysis: Confusion Matrix A confusion matrix analysis was performed to look further for a significant change in vegetation cover within VCTs in more detail than the decile level, between 2007 and The numbers of polygons in the Arrow Lakes and Revelstoke Reach portions of the Arrow Lakes Reservoir used in this analysis are shown in Appendix 3. The aerial photo-interpretative map data were first analyzed using coded variables for vegetation cover and height. The coded variables were used to develop a confusion matrix, which was then tested with the Kappa statistic. The results of the confusion matrix analysis are summarized in Table 8, Table 9, and Table 10. It is important to note that the null hypothesis, when using the Kappa statistic, is opposite to what one would normally expect, so a low probability and rejection of the null hypothesis means that no significant difference between coded values exists. Tables 8 to 10 provide: The vegetation community type (VCT) name (Column [1]). The dimensions of the confusion matrix. The number of rows and columns in the matrix correspond to the observed codes for each variable. The rows are the 2007 codes and columns are the 2010 codes (Column [2]). The sample size or number of sample polygons (Column [3]). The estimate of simple Kappa statistic (Column (4)), and its asymptotic estimate of the standard error (Column [5]). The range of possible values of Kappa is -1 to 1. The value 1 represents perfect agreement in codes between 2007 and 2010 (i.e., no change). A value of 0 indicates agreement no better than that expected by chance. A value of -1 indicates agreement worse than that expected by chance. The exact (or asymptotic) p-value of the test of the null hypothesis that the Kappa estimate is equal to zero, or K = 0 (two-tail, 95 per cent probability) (i.e., no agreement between 2007 and 2010 (Column [6]). This null hypothesis means that a change has occurred in a given variable (e.g., cover) between 2007 and The Kappa statistic test conclusion (Column [7]). If the null hypothesis is rejected it means that there is agreement in codes between 2007 and That is, the change in codes between 2007 and 2010 is not statistically significant (95 per cent probability). 42

65 Table 8. Confusion matrix: summary of analyses for Vegetation Cover. VCT Confusion matrix dim. No. sample polygons Kappa Estimate (K) 95% Conf. Int. of K Exact p- values Conclusion (*) (1) (2) (3) (4) (5) (6) (7) BE 3x Significant agreement BG 5x No significant agreement CR 6x << Significant agreement PA 6x < ** Significant agreement PC 7x << Significant agreement PE 6x << Significant agreement RR 7x Significant agreement * Significant agreement means that no significant change in VCT coding occurred between 2007 and 2010; no significant agreement means that a significant change in VCT coding occurred between 2007 and 2010 ** Asymptotic p-value; exact p-value could not be computed. Table 9. Confusion Matrix: summary of analyses for Vegetation Height. VCT Confusion matrix dim. No. sample polygons Kappa Estimate (K) 95% Conf. Int. of K Exact p- value Conclusion (*) (1) (2) (3) (4) (5) (6) (7) BE 1x1 10 No analysis since number of rows and columns is < 2. BG 2x2 30 No analysis since number of non-zero rows and columns is < 2. CR 5x << Significant agreement PA 5x < ** Significant agreement PC 2x No significant agreement PE 2x No significant agreement RR 2x No significant agreement * Significant agreement means that no significant change in VCT coding occurred between 2007 and 2010; No significant agreement means that a significant change in VCT coding occurred between 2007 and ** Asymptotic p-value; exact p-value could not be computed. 43

66 Table 10. Confusion Matrix: Summary of analyses for Vegetation Distribution. VCT Confusion matrix dim. No. sample polygons Kappa Estimate (K) 95% Conf. Int. of K (1) (2) (3) (4) (5) (6) (7) p-value Conclusion (*) BE 6x Significant agreement BG 8x << Significant agreement CR 6x << Significant agreement PA 9x << Significant agreement PC 6x << Significant agreement PE 5x << Significant agreement RR 7x << Significant agreement * Significant agreement means that no significant change in VCT coding occurred between 2007 and The confusion matrix did not take into account elevation; it indicated only if cover, height and distribution of the vegetation within VCTs had changed between 2007 and The confusion matrix analyses show significant changes only for cover in BG vegetation types and for height in PC, PE and RR vegetation types between 2007 and No significant changes in the distribution of VCTs were observed between 2007 and Repeated Measures Analysis: ANOVA of map data - are vegetation cover and height maintained over time? The results of the repeated measures ANOVAs are displayed in Table 11 and Table 12 for the Arrow portion of the reservoir, and in Table 13 and Table 14 for Revelstoke Reach. Table 11 shows that significant differences in mean cover exist between the years 2007 and 2010 for the BG, PE, and RR vegetation community types, when individual VCTs are considered. The RR and PE VCTs occurred at low elevation, while BG occurred over a wider elevation range, but was sparsely vegetated and often subjected to scouring. There were significant differences in mean height between 2007 and 2010 in the BE and BG community types, also. BE is similar to BG in being sparsely vegetated and subjected to scouring. The difference between the ANOVA results and the confusion matrix results occurred because ANOVA examines variation in actual cover values ranging from zero to 100 per cent, whereas the confusion matrix examines variation in coded values for cover, and is a broader level of analysis than ANOVA. The confusion matrix analysis examines change at the landscape level, whereas ANOVA examines change at a level that is much closer to the ground. This finer level of analysis is available to us due to the clarity in the aerial photographic imagery captured in There were no significant differences in mean covers or heights between elevation bands within any specific vegetation community type in the Arrow 44

67 Lakes portion of the reservoir. These results are based on cover and height differences estimated from aerial photography interpretation (and not coded as in the confusion matrix). The interaction effect (Elevation Band x Year) is not significant for any of the VCTs considered in this analysis. This indicates that the effects of Elevation Band and Year on the total cover and height of vegetation are independent of each other. Table 11. Repeated Measures ANOVA p-values for the Arrow Lakes portion of the reservoir by Elevation Band, Year and Elevation Band x Year interaction for each VCT. Significant results have a p-value < 0.05 and are shaded. VCT Elevation Band Cover Year Elevation Band x Year Elevation Band Height Year Elevation Band x Year BE BG < CR PA PC PE < RR < Table 12. Vegetation Community Type (VCT), Year and VCT x Year interaction for each Elevation Band in the Arrow portion of the Arrow Lakes Reservoir. Significant results have a p-value < 0.05 and are shaded. Veg. Attribute VCT Year VCT x Year Elevation Band m Cover < < Height < Elevation Band m Cover < < Height < Elevation Band m Cover < < < Height < Table 12 shows that there were significant differences in total cover and height among VCTs. This is an expected result, because these differences are observable in the field and on the aerial photographs. When the effect of the Year is considered independently of VCT (the Year column), there were significant 45

68 differences in mean vegetation covers and heights within the and metre elevation bands. In the metre elevation band, only total cover differed significantly between 2007 and The significant differences in height and cover from year to year illustrated by these results may be due to phenological differences resulting from the timing of air photo flights, to the differences in early season growing conditions between the two years, or to other effects, such as the soft constraints operating regime. The interaction effect shown in Table 12 (VCT x Year) is significant for the analysis of total cover in all three elevation bands. This indicates that the effects of the different VCTs and of the year are not independent from one another for the total cover analysis in the Arrow Lakes portion of the reservoir. Table 13. Repeated Measures ANOVA p-values for Revelstoke Reach portion by Elevation Band, Year and Elevation Band x Year interaction effects for each VCT. Significant results have a p-value < 0.05 and are shaded. VCT Elevation Band Cover Year Elevation Band x Year Elevation Band Height Year Elevation Band x Year BE BG CR PA PC PE RR The results of the ANOVAs summarized in Table 13 for Revelstoke Reach are similar to those presented in Table 11 for the Arrow Lakes. Significant differences in total cover and height were found for some VCTs due to the effect of Year. Mean vegetation cover was significantly different in 2007 than in 2010 only for the PA type. Mean heights were significantly different in 2007 than in 2010 for BE, CR and PC types. Within VCTs, there were no significant differences in total cover due to elevation. The effects of elevation and year are independent of one another in all cases for these analyses (i.e., no significant results in Elevation Band x Year columns). Table 14 shows that significant differences in mean covers and heights existed between vegetation community types within each elevation band for Revelstoke Reach, because each VCT has distinct patterns in heights and covers (see Appendix 4). However, when all VCTs were combined within an elevation band there was no significant difference in heights or covers between 2007 and This is an important result, because it implies that overall, at the landscape level, vegetation heights and covers have not changed significantly when comparing 2007 and 2010, although vegetation covers and heights may have changed within each of the intervening years. There was a significant result for the interaction effect of VCT x Year for the height analysis in the metre 46

69 elevation band. This indicates that the VCT and Year are not independent in the way that they affect mean height for that analysis. Table 14. Vegetation Community Type (VCT), year and VCT x Year interaction for each Elevation Band in the Revelstoke Reach portion of the Arrow Lakes Reservoir. Significant results have a p-value < 0.05 and are shaded. Veg. Attribute VCT Year VCT x Year Elevation Band m Cover < Height < Elevation Band m Cover < Height < Elevation Band m Cover < Height < Tukey s Boxplots and ANOVA: Field data comparisons - are vegetation cover and height maintained over time? We examined changes in vegetation within VCTs over time, using the field data. The numbers of plots that were repeated from previous fieldwork in the reservoir for CLBMON-12 and those done in 2010 for this project are shown in Appendix 3. Tukey s boxplots show the changes in covers and heights of vegetation within VCTs from plots repeated between 2007 and 2010 (Figure 14 through Figure 19). The data presented below are from Enns et al. (2007, 2008, 2009) and from the field work for this project completed in The data from 2010 include remeasurements of previous years plots, as well as new plots in Some of the previous years data were re-measured in 2010, but most are unique to that year of measurement. The data in Figure 14 through Figure 19 should be interpreted with caution, as some VCTs had limited data in 2007, and some VCTs were not extensively sampled in 2008 in Revelstoke Reach, due to the onset of high water during the sample period. Also, the 2007 measurements were taken in the fall, whereas all other measurements were collected in the spring. Measurements of heights in the CR: Cottonwood riparian plots between 2007 and 2009 did not include estimates of the taller tree canopy heights, whereas heights of crowns of trees were estimated in 2010, and are therefore much higher than previous years. 47

70 BE 120 Total Vegetation Cover (%) Arr 07Rev 08Arr 08Rev Year and Reach 09Arr 09Rev 10Arr 10Rev CR 300 Total Vegetation Cover (%) Arr 07Rev 08Arr 08Rev Year and Reach 09Arr 09Rev 10Arr 10Rev Figure 14. Tukey s boxplots: change in total vegetation cover in BE: Beach (top) and CR: Cottonwood riparian (bottom) VCTs in the Arrow Lakes Reservoir drawdown zone between elevations m (Arr = Arrow portion, Rev = Revelstoke Reach portion of the Arrow Lakes Reservoir) between 2007 and See Figure 5 for a complete description of boxplots. 48

71 PA R Total Vegetation Cover (%) Arr 07Rev 08Arr 08Rev Year and Reach 09Arr 09Rev 10Arr 10Rev PC 150 Total Vegetation Cover (%) Arr 07Rev 08Arr 08Rev Year and Reach 09Arr 09Rev 10Arr 10Rev Figure 15. Tukey s boxplots: change in total vegetation cover in PA: Redtop upland (top) and PC: Reed canary grass mesic (bottom) VCTs in the Arrow Lakes Reservoir drawdown zone between elevations m (Arr = Arrow portion, Rev = Revelstoke Reach portion of the Arrow Lakes Reservoir) between 2007 and See Figure 5 for a complete description of boxplots. 49

72 PE 200 Total Vegetation Cover (%) Arr 07Rev 08Arr 09Arr 09Rev 10Arr 10Rev Year and Reach RR 200 Total Vegetation Cover (%) Arr 07Rev 08Arr 09Arr 10Arr Year and Reach 10Rev Figure 16. Tukey s boxplots: change in total vegetation cover in PE: Horsetail lowland (top) and RR: Reed-rill (bottom) VCTs in the Arrow Lakes Reservoir drawdown zone between elevations m (Arr = Arrow portion, Rev = Revelstoke Reach portion of the Arrow Lakes Reservoir) between 2007 and See Figure 5 for a complete description of boxplots. 50

73 Table 15 provides the results of analyses of variance conducted on these field data for total cover in selected VCTs by Reach and Year. Mean total covers were significantly lower in 2009 than in any other year since the beginning of the CLBMON-33 and CLBMON-12 studies for CR, PA, PC and PE VCTs. Only the PC vegetation community type showed significant differences in mean total cover between reaches when considered together with Year (Reach x Year). The Reach and Year effects were independent for all other VCTs, and there were no significant differences in mean total cover due to Reach alone. Table 15. ANOVA results for Total Vegetation Cover within selected VCTs by Reach (Arrow or Revelstoke) and Year (2007, 2008, 2009, 2010). Significant differences in mean total cover are shaded. Type Factor P Conclusion Post hoc comparison results BE* Reach No sig. diff. N/A BE* Year No sig. diff. N/A BE* Reach x Year No sig. diff. N/A CR* Reach No sig. diff. N/A CR* Year Significant difference CR* Reach x Year No sig. diff. N/A PA Reach No sig. diff. N/A PA Year Significant difference PA Reach x Year No sig. diff. N/A PC Reach No sig. diff. N/A PC Year Significant difference PC Reach x Year Significant difference PE** Reach No sig. diff. N/A PE** Year Significant difference PE** Reach x Year No sig. diff. N/A Mean total cover in 2010 greater than in 2008, which in turn was greater than in (i.e > 2008 > 2009) 2007 > 2008 > 2010 > > 2008 > 2010 > 2009 Rev2007 > Arrow2007 > Arrow 2008 = Rev2008 = Arrow2010 > Rev2010 > Arrow2009 = Rev > 2010 > 2009 * Insufficient data to include 2007; Year effect had three levels: 2008, 2009, 2010 **Insufficient data to include 2008; Year effect had three levels: 2007, 2009,

74 BE Average Vegetation Height (m) Arr 07Rev 08Arr 08Rev Year and Reach 09Arr 09Rev 10Arr 10Rev 7 CR Average Vegetation Height (m) Arr 07Rev 08Arr 08Rev Year and Reach 09Arr 09Rev 10Arr 10Rev Figure 17. Tukey s boxplots: Maximum vegetation heights in BE: Beach (top) and CR: Cottonwood riparian (bottom) VCTs in the Arrow Lakes Reservoir drawdown zone between elevations m (Arr = Arrow portion, Rev = Revelstoke Reach portion of the Arrow Lakes Reservoir) between 2007 and See Figure 5 for a complete description of boxplots. 52

75 PA Average Vegetation Height (m) Arr 07Rev 08Arr 08Rev Year and Reach 09Arr 09Rev 10Arr 10Rev PC Average Vegetation Height (m) Arr 07Rev 08Arr 08Rev Year and Reach 09Arr 09Rev 10Arr 10Rev Figure 18. Tukey s boxplots: Maximum vegetation heights in PA: Redtop upland (top) and PC: Reed canary grass mesic (bottom) VCTs in the Arrow Lakes Reservoir drawdown zone between elevations m (Arr = Arrow portion, Rev = Revelstoke Reach portion of the Arrow Lakes Reservoir) between 2007 and See Figure 5 for a complete description of boxplots. 53

76 PE RR Average Vegetation Height (m) Average Vegetation Height (m) Arr 07Rev 08Arr 09Arr 09Rev 10Arr 10Rev Year and Reach 0 07Arr 07Rev 08Arr 09Arr 10Arr Year and Reach 10Rev Figure 19. Tukey s boxplots: Maximum vegetation heights in PE: Horsetail lowland (top) and RR: Reed-rill (bottom) VCTs in the Arrow Lakes Reservoir drawdown zone between elevations m (Arr = Arrow portion, Rev = Revelstoke Reach portion of the Arrow Lakes Reservoir) between 2007 and See Figure 5 for a complete description of boxplots. Table 16 provides the results of analyses of variance conducted on these field data for maximum height in selected VCTs by Reach and Year. 54

77 Table 16. ANOVA results for Mean Maximum Vegetation Height within selected VCTs by Reach (Arrow or Revelstoke) and Year (2007, 2008, 2009, 2010). Significant differences in mean maximum height are shaded. Type Factor P Conclusion Post hoc comparison results BE Reach No sign.diff. N/A BE Year No sign.diff. N/A BE Reach x Year No sign.diff. N/A CR* Reach No sign.diff. N/A CR* Year Significant difference Mean max height was significantly greater in 2010 than in 2008, which in turn was not significantly different than mean max height in 2009 (i.e > 2008 = 2009) CR* Reach x Year No sign.diff. N/A PA Reach No sign.diff. N/A PA Year Significant difference 2009 > 2010 > 2008 = 2007 PA Reach x Year No sign.diff. N/A PC Reach No sign.diff. N/A PC Year No sign.diff. N/A PC Reach x Year No sign.diff. N/A PE** Reach Significant difference Mean max heights in the PE type were significantly greater in Revelstoke Reach than in the Arrow Lakes. (i.e. Rev > Arrow) PE** Year Significant difference 2007 > 2010 > 2009 PE** Reach x Year No sign.diff. N/A * Insufficient data to include 2007, Year effect had three levels: 2008, 2009, 2010 **Insufficient data to include 2008, Year effect had three levels: 2007, 2009, 2010 Mean maximum heights did not show the same consistent trends that were shown by the analyses of mean total covers. Heights in CR were significantly greater in 2010 than in previous years due to the inclusion of trees in the height measurements, as discussed above. Heights in PA were significantly greater in 2009 than in 2010, 2008 or Heights in PA in 2008 and 2007 were not significantly different, but were shorter than those in PE was the only VCT to show significant differences in heights due to the reach in which the vegetation occurred, where mean heights in Revelstoke were greater than in the Arrow Lakes. Heights in PE were also significantly different from year to year, with 2007 being the tallest, followed by 2010 and The Reach and Year effects on mean maximum height were independent for all VCTs. Several CR plots that were repeated showed a huge increase in total heights between previous years and 2010, which is a function of measuring heights of only shrubs in CR in previous years and including trees in However, even with these limitations, some trends in the data appear evident. Cover of vegetation of BE: Beach VCT did not 55

78 vary over time, but did appear to be slightly higher in the Arrow portion of the Arrow Lakes Reservoir in 2010 compared to other years. Cover of vegetation in the CR: Cottonwood riparian VCT was variable between years, but the higher number of samples collected in this type in 2010 showed a high variation in total covers in this VCT, with slightly depressed cover values in Total cover values in the PA: Redtop upland and the PC: Reed canary grass mesic VCTs were slightly depressed in 2009 compared to previous years. A similar pattern of lower total vegetation cover in 2009 in the PE: Horsetail lowland VCT is shown in Figure 16. The RR: Reed-rill VCT had low sample sizes prior to 2010, but total covers of vegetation in the VCT appear to have increased since The average heights of vegetation in BE: Beach VCT were similar to the total cover over time, i.e., no real differences occurred in this low-profile low-cover sanddominated community type. Average heights in the CR: Cottonwood riparian VCT did not vary over time, probably because of the typical range of elevation of this VCT. The CR VCT rarely occurs below 440 to 438 metres and appears to be little affected by variation in the soft constraints operating regime, compared to VCTs that range below 438 m. Total vegetation heights in the PA: Redtop upland and the PC: Reed canary grass mesic VCTs either remained the same over time (PC), or increased slightly (PA) since There appears to have been a depression in heights in This is most evident in the PE: Horsetail lowland VCT and in the RR: Reed-rill VCT, although the results for the RR VCT prior to 2010 are based on small sample sizes Is the vegetation composition (distribution, diversity, evenness and richness) maintained? We used the confusion matrix (Section above) to test for differences in distribution patterns of vegetation at the landscape level. The map data were used to determine if distribution of vegetation in VCTs had changed between 2007 and The confusion matrix (Table 10) indicated there was no significant change in the distribution of mapped VCTs between 2007 and Distribution was described by using nine codes for various patterns of abundance and aggregation of vegetation (Luttmerding et al. 1990). These codes are usually used at the field plot level, but patterns in vegetation at the landscape level also resemble the distribution codes described in Luttmerding et al. (1990). The landscape-level comparison of the aerial photography did not show a consistent pattern of change in distribution between 2007 and 2010 (data not shown). In almost all cases, mesic, mid-elevation vegetation (PC: Reed canary grass mesic) had an even distribution of plants (data are not shown, but the pattern is illustrated in Figure 13). Vegetation in VCTs at the high and low elevation zones in the reservoir tended to have a clumped and/or random distribution. Although some vegetation appeared to fill in or become slightly less clumped in 2010 in comparison to 2007, scouring of materials also caused some vegetation to become slightly more scattered with smaller clumps in Mesic vegetation appeared to maintain its even distribution between 2007 and 2010, except on the natural boundaries of communities where some scouring or increases in vegetation cover occurred. The maintenance of diversity and richness over time in VCTs was examined using diversity and evenness indices and measures. These elements of the 56

79 management questions were examined at the taxon level (genus and species where known) using field data. Taxon diversity, richness and evenness were compared over time for all previous years in comparison to 2010 (Figure 20 through Figure 25). The tables showing the taxon diversity, richness and evenness values are in Appendix 5. Species Diversity (H') BE BG CR PA PC PE RR Vegetation Community Type Figure 20. Shannon-Weiner Diversity Index (H ) for each Vegetation Community Type (VCT) from 2007 to 2010 in the Arrow Lakes Reservoir (both geographic areas combined). See Appendix 5 for the tabular values for this figure. Shannon-Weiner H indices combine species richness with the evenness of distribution of the number of individuals among species. However, this calculation uses cover values instead of a count of individuals. There was not much change within VCT Diversity Indices over time, but relatively high species richness and evenness exist within some of the VCTs, such as CR and PA, as a result of their position on the slope. The CR and PA VCTs receive the shortest duration of inundation and the shallowest depths of inundation, have remnants of complete soil or complete soil, and have higher species richness as a result. Evenness is lower in PC: Reed canary grass mesic VCT than other VCTs, because of the dominance of reed canarygrass and the scarcity of other species (Figure 20). 57

80 Species Evenness (J) BE BG CR PA PC PE RR Vegetation Community Type Figure 21. Pielou s species evenness (J) for each Vegetation Community Type (VCT) from 2007 to 2010 in the Arrow Lakes Reservoir (both geographic areas combined). See Appendix 5 for the tabular values for this figure. Species Richness BE BG CR PA PC PE RR Vegetation Community Type Figure 22. Species richness for all vegetation within the three elevation bands ( metres, metres and metres) from 2007 to 2010 in the 58

81 Arrow Lakes Reservoir (both geographic areas combined). See Appendix 5 for the tabular values for this figure Shannon-Wiener Diversity Index (H') Year m m m Figure 23. Species diversity (H) for all vegetation within the three elevation bands ( metres, metres and metres) from 2007 to 2010 in the Arrow Lakes Reservoir (both geographic areas combined). See Appendix 5 for the tabular values for this figure Pielou's Evenness (J) Year m m m 59

82 Figure 24. Pielou s species evenness (J) for all vegetation within the three elevation bands ( metres, metres and metres) from 2007 to 2010 in the Arrow Lakes Reservoir (both geographic areas combined). See Appendix 5 for the tabular values for this figure Species Richness Year m m m Figure 25. Species Richness for each elevation band from 2007 to 2010 in the Arrow Lakes Reservoir (both geographic areas combined). See Appendix 5 for the tabular values for this figure. Patterns of species richness in four yearly measurements (Figure 22) probably reflected the much higher number of plots in 2010 than in previous years. More than double the number of plots were completed in 2010: 150 to 230 per year from 2007 to 2009, compared to 559 in Additional sampling undoubtedly contributed to higher species richness in previously undersampled, but relatively species-rich VCTs, such as CR: Cottonwood riparian and PA: Redtop upland. In comparison, additional sampling in species-poor VCTs, such as BE: Beach and BG: Gravelly beach did not result in much of an increase in species richness, despite the higher number of plots in The mesic PA: Reed canary grass VCT was assumed to be rather species poor, but in fact, more sampling to meet the power analysis requirements resulted in an increase in species richness in The trends in species richness are such that there are few differences between years, and that within years, the patterns are very strongly related to the depth and duration of inundation. This pattern does not appear to have been influenced by the change in duration of inundation in 2008, although the relationship has not been tested statistically. 60

83 3.5.6 Examples of changes in vegetation spatial limits (cover), from the imagery of 2007 and 2010 The aerial photographic imagery shows several differences in appearance between 2007 and Some of these differences are due to different media being used (film vs. digital imagery) 8. The aerial photography was sampled for the repeated measures analysis, and in that process, time comparisons between 398 polygons were made. Change in spatial limits, i.e., cover, was difficult to assess, as the phenology (stage of growth) of the vegetation in 2007 was approximately two to three weeks ahead of vegetation in Also, prior to CLBMON-33, the vegetation experienced years of relatively low water. The comparisons of field data from 2007 to 2010 show that the lowest vegetation covers occurred in 2009, and if aerial photographic imagery had been available for 2009, these lower covers would likely have been detectable. Comparisons in the appearance of vegetation between the same polygons, shot in 2007 and shot in 2010, are shown in Figure 26 to Figure 33. Figure 29 and Figure 30 show pixelation of the 2007 film image at approximately 1: 300 scale that was not present in the 2010 digital image. It also shows that individual shrubs were smaller in crown size and height in 2010 than in When observed in the field, some of the individual shrubs that were smaller in the 2010 image had evidence of dieback; crowns were partially dead and suckering was occurring. This was particularly evident at Burton and in Revelstoke Reach. Figure 26. Phenological Differences in Polygon 10 between 2007 (left) and 2010 (right) in the Arrow Lakes Reservoir, from aerial photography. Note the greener and fluffier appearance of vegetation in the 2007 imagery indicating the three week difference in time of capture of the photographs. The 2007 photographs were taken on May 30; the 2010 photographs were taken on May Scale is approximately 1: 500. The cover of standing dead crowns on reed canary grass is also higher in the 2007 image than in 2010, and this has an influence on the interpretation. Dead 8 E.g., different pixel sizes, color intensity differences, blurred mosaic lines and poor resolution in Due to the better resolution in the 2010 imagery, it was possible to see more clearly where errors in the base mapping were made; these can be easily corrected. 61

84 crowns cover other vegetation and appear flat and brown, as in Figure 26 above. These colour and size difference are a function of the time of capture. For example, Figure 32, taken at Montana Slough (shown at a scale of 1: 1,000), shows the effect of moisture on the colour of the vegetation and slightly smaller clump sizes of reed canary grass in 2010 in comparison to In Figure 33, taken in the Arrow Narrows (shown at a scale of 1: 3,000), the higher water level in the 2010 imagery is evident (even though the imagery was taken two weeks earlier), and the vegetation cover appears to be taller in 2007 than in Figure 27. Phenological Differences in Polygon 1777 between 2007 (left) and 2010 (right) in the Arrow Lakes Reservoir, from aerial photography. Growth of reed canary grass is more advanced in the 2007 photography, and shrubs have not released as much in 2010 as in This does not constitute a decile change. Scale is approximately 1: 1,500. The differences shown in Figure 26, Figure 27, Figure 31, Figure 32 and Figure 33 are not enough to cause a difference in cover at the decile level in the illustrated VCTs; the aerial photographic interpretation attributed the differences in these comparisons to phenology and moisture. In contrast to these, Figure 28, Figure 29 and Figure 30 illustrate actual decile changes in VCT cover in polygons, primarily due to scouring. The conditions illustrated in Figures 28, 29 and 30 were exceptional: almost all polygons showed no removal or addition of vegetation to the VCTs. 62

85 Figure 28. Example of a change in vegetation due to a combination of ORV use and phenology in Polygon 1763 in the Arrow Lakes Reservoir, from aerial photography. At middle to low elevation RR: reed-rill VCT between 2007 (left) and 2010 (right) constitutes a decile change in the vegetation mapping for this polygon. Scale is approximately 1: 1,500. Figure 29. Dispersal of logs from the LO: Blue wildrye log VCT between 2007 (left) and 2010 (right) in Polygon 2009 in the Arrow Lakes Reservoir, from aerial photography. Note difference in resolution at this scale. This constitutes a decile change and a vegetation community type change, from LO to RR, although the change is mainly in relation to woody debris. Scale is approximately 1:

86 Figure 30. Example of a change in decile cover of PE: Horsetail lowland in Polygon 1710 in the Arrow Lakes Reservoir, from aerial photography. A low elevation PE: Horsetail lowland VCT between 2007 (left) and 2010 (right). This magnitude of change was seen in fewer than six out of 398 polygons. Scale is approximately 1: 300. Figure 31. Differences in phenology in Polygon 1192 in the Arrow Lakes Reservoir, from aerial photography. Trees and shrubs appear to have shrunk between 2007 and 2010; the lack of leaves in the May 12, 2010 photo vs. the May 30, 2007 photo is notable. Scale is approximately 1:

87 Figure 32. Differences in clump size in reed canary grass at Montana Slough, Revelstoke Reach 2007 vs. 2010, from aerial photography. Differences are possibly due to inundation-controlled colour expression and phenology. Scale 1: 1,000. Figure 33. Differences in cover and height of reed canary grass in the Arrow Narrows study area of the Arrow Lakes portion of the Arrow Lakes Reservoir, 2007 vs from aerial photography. Differences may be due to water level differences and phenological differences. Scale 1: 3,000. Differences in colour expression between the two sets of imagery are consistent: the white balance of the 2007 imagery is not correct, and the images have a blue tint. It may be possible to correct these discrepancies by using photo development software to change the white balance. The results of the image comparisons done so far indicate that although the quality of the 2007 imagery is 65

88 not as good as the 2010 imagery, the differences in the cover and heights of the vegetation are observable with a high degree of certainty Image analysis pilot study The use of image analysis to record changes in the whole population was proposed as a pilot study (Enns 2010). This was discussed in Enns and Tomlins (2010). The results of the pilot study indicate that image comparisons can be used to show changes in vegetation for the study areas, but interpretation of change in the imagery will still be necessary (Appendix 8). Management Question summary: The vegetation of the Arrow Lakes Reservoir appears to have been maintained in most VCTs up to the winter of 2008 when a decline in cover of some VCTs subsequently occurred, and was recorded in the field season of Some VCTs appear to have recovered in The soft constraints operating regime varied in 2008 from previous years, by having an exceptionally high water period that extended from June 2008 to January This coincided with the decline in cover observed in Several individual tests and examinations of the field data show this relationship. Heights of vegetation were not as affected by the summer to winter period of 2008, and some of the vegetation appears to have increased in cover in 2010 in response to the return to the pre-2008 pattern in the soft constraints operating regime. The following results were recorded: Significant cover change for BG type only; significant height change for PC, PE and RR only; no significant changes in distribution code. Significant cover change for BG, PE and RR from 2007 to 2010; significant height change for BE and BG from 2007 to Significant differences in cover and height between VCTs for all elevation bands; significant differences in cover and height between 2007 and 2010 in all elevation bands except height in 438 to 440 m; some significant differences in cover due to interaction effect of VCT x Year. Significant cover change for PA from 2007 to 2010; significant height change for BE, CR, and PC from 2007 to Significant differences in heights between VCTs for all elevation bands; no significant differences in cover and height between 2007 and 2010 due to year effect alone; a significant difference in height for the lowest elevation band due to interaction effect of VCT x Year. BE: (2007 excluded) no significant differences; CR: (2007 excluded) significant effect due to year, 2010 cover > 2008 cover > 2009 cover; PA: significant effect due to year, 2007 > 2008 > 2010 > 2009; PC: significant effect due to year, 2007 > 2008 > 2010 > 2009; significant interaction effect of Reach x Year, Rev2007 > Arrow2007 > Arrow2008 = Rev2008 = Arrow2010 > Rev2010 > Arrow2009 = Rev2009; PE: (2008 excluded) significant effect due to year, 2007 > 2010 > BE: (2007 excluded) no significant differences; CR: (2007 excluded) significant effect due to year, 2010 heights > 2008 heights = 2009 heights; PA: significant effect due to year, 2009 > 2010 > 2008 = 2007; PC: no 66

89 significant effects due to year or reach; PE: (2008 excluded) significant effect due to year, 2007 > 2010 > The above VCTs are the main vegetation communities that respond mainly to the soft constraints operating regime. The remaining nine communities may also respond to the soft constraints operating regime, but they are either rare or largely impacted by industrial development, incoming streams, or other factors. Vegetation community composition is a function of the duration and depth of inundation, which is a condition brought about by the soft constraints operating regime. At the landscape level, cover and heights of deciles of VCTs have not changed between 2007 and At a closer level, cover and heights of some VCTs declined significantly between 2008 and In general, cover and heights of some VCTs appear to have been impacted by the soft constraints operating regime of Those VCTs are in areas of the drawdown zone where scouring takes place, and where the impact of prolonged inundation may be detrimental to the maintenance of vegetation community structure, extent and composition. The cover and height of vegetation in some of these VCTs appears to have recovered to previous years values. 3.6 Question 6. What potential changes to the soft constraints operating regime could be made in order to maintain the existing vegetation communities at the landscape scale? The investigation of change in vegetation using aerial photography interpretation and deciles (10 per cent increments) of cover of vegetation within VCTs showed almost no change in the vegetation of the Arrow Lakes Reservoir, when 2007 was compared to 2010 at the landscape scale using the aerial photographs. Only six out of 398 polygons showed a decile change, and only two of these were greater than 10 per cent. This is an overwhelmingly low number of records of change in a large sample. However, we suspect that changes of less than 10 per cent, which were not detected with these methods, may be ecologically significant. When we examined coded cover values (Kappa statistic and the confusion matrix) and aerial photographic interpreted cover estimate comparisons (analysis of variance), some changes in covers and heights of the vegetation between 2007 to 2010 became more apparent. A decline in cover in low elevation VCTs was recorded in the field in 2009, after the reservoir had stayed near full pool for 218 days of the previous year. Vegetation grows under the water between May and September (Enns et al. 2007), but it requires an exposure period, including fall and winter, in order to maintain cover. Therefore, it is likely that the losses of vegetation cover noted in 2009 could have been reduced by allowing the reservoir levels to decline to below 434 metres in the fall of Figure 34 shows the period of prolonged inundation in Plant diversity at low and high elevations is influenced by inundation. Prolonged inundation may reduce diversity and increase evenness at lower elevations. Moisture availability (due to inundation) at high elevations results in higher diversity. 67

90 Management Question summary: We recommend that the soft constraints operating regime be maintained in the pattern seen between 2005 and 2008 and from 2009 to 2010, to reduce the risk of loss of vegetation cover or height in some of the VCTs of the Arrow Lakes Reservoir. Figure 34. Recommended Change to the BC Hydro Soft Constraints Operating Regime based on Data from 2007 to Hypothesis test summary Some of the results presented in this paper clearly address the management questions with statistically significant or non-significant results. Table 17 is a review of those results as they relate to the following hypothesis: H 0 : Under the soft constraints operating regime (or possibly a newly selected alternative after five years), there is no significant change in existing vegetation communities at the landscape scale. H 0A : There is no significant change in the spatial extent (cover, and number of hectares) of vegetation communities within the existing vegetated zones of Arrow Lakes Reservoir. H 0B : There is no significant change in the structure and composition (i.e., distribution and diversity) of vegetation communities within the existing vegetated zones of Arrow Lakes Reservoir. 68

91 Table 17. Summary of Statistical Results. MQ Object of Tests Tests Significant Results Q1: Vegetation Cover by Reach and Elevation Band (Appendix 4) Vegetation Height by Reach and Elevation Band (Appendix 4) ANOVA plus post hoc comparisons ANOVA plus post hoc comparisons Elevation: Cover 438 to 440 m > 436 to 438 m > 434 to 436 m; Reach x Elevation: Rev 438 to 440 m > Arrow 438 to 440 m > Arrow 436 to 438 m > Rev 436 to 438 m > Arrow 434 to 436 m = Rev 434 to 436 m. Reach: Height Revelstoke > Arrow; Elevation: Height 438 to 440 m > 436 to 438 m > 434 to 436 m; Reach x Elevation: Rev 438 to 440 m > Arrow 438 to 440 m > Arrow 436 to 438 m = Rev 436 to 438 m = Arrow 434 to 436 m = Rev 434 to 436 m. Q2: Areal coverage by VCTs in Rev Reach vs. Arrow Lakes Q3: same as Q1 Q4: Heights and Cover as they relate to topoedaphic site conditions Contigency table with 2 statistic RDA The coverage of VCTs in Revelstoke Reach is not the same as in the Arrow Lakes portion of the reservoir. Arrow: Approx 28 per cent of variation in height and cover was explained by the environmental variables used in the RDA; Cover not associated with environmental variables; Height associated with higher elevation, increased coarse fragment content and increased slope Revelstoke: Approx 36 per cent of variation in height and cover was explained by the environmental variables used in the RDA; Cover not associated with environmental variables; Height associated with 2 higher elevation bands (R2 and R3), increased presence of boulders and aspects other than north and east facing ('increased aspect'). Q5: Repeated measures analysis of vegetation in mapped data Repeated measures analysis of vegetation in mapped data Confusion matrix with K statistic RMA ANOVA: Arrow - Selected VCTs individually by Elevation band and Year Significant cover change for BG type only; Significant height change for PC, PE and RR only; No significant changes in distribution code. Significant cover change for BG, PE and RR from 2007 to 2010; Significant height change for BE and BG from 2007 to

92 MQ Object of Tests Tests Significant Results RMA ANOVA: Arrow - Elevation Bands separately for change in cover and height due to VCT and Year Significant differences in cover and height between VCTs for all elevation bands; Significant differences in cover and height between 2007 and 2010 in all elevation bands except height in 438 to 440 m; Some significant differences in cover due to interaction effect of VCT x Year. Comparison of field data from 2007 through 2010 for changes in cover for selected VCTs (BE, CR, PA, PC, PE) Comparison of field data from 2007 through 2010 for changes in heights for selected VCTs (BE, CR, PA, PC, PE) RMA ANOVA: Revelstoke - Selected VCTs individually by Elevation band and Year RMA ANOVA: Revelstoke - Elevation Bands separately for change in cover and height due to VCT and Year ANOVA plus post hoc comparisons ANOVA plus post hoc comparisons Significant cover change for PA from 2007 to 2010; Significant height change for BE, CR, and PC from 2007 to Significant differences in heights between VCTs for all elevation bands; No significant differences in cover and height between 2007 and 2010 due to year effect alone; A significant difference in height for the lowest elevation band due to interaction effect of VCT x Year. BE: (2007 excluded) No significant differences; CR: (2007 excluded) Significant effect due to year, 2010 cover > 2008 cover > 2009 cover; PA: Significant effect due to year, 2007 > 2008 > 2010 > 2009; PC: Significant effect due to year, 2007 > 2008 > 2010 > 2009; Significant interaction effect of Reach x Year, Rev2007 > Arrow2007 > Arrow2008 = Rev2008 = Arrow2010 > Rev2010 > Arrow2009 = Rev2009; PE: (2008 excluded) Significant effect due to year, 2007 > 2010 > BE: (2007 excluded) No significant differences; CR: (2007 excluded) Significant effect due to year, 2010 heights > 2008 heights = 2009 heights; PA: Significant effect due to year, 2009 > 2010 > 2008 = 2007; PC: No significant effects due to year or reach; PE: (2008 excluded) Significant effect due to Reach, Rev > Arrow; Significant effect due to year, 2007 > 2010 >

93 MQ Object of Tests Tests Significant Results Comparison of species diversity (H') for each VCT from 2007, 2008, 2009 and 2010 from field data. Shannon Weiner Index (H') plotting of trends over time. Slight increases in diversity over time in some VCTs which is probably due to higher numbers of plots done in each consecutive year. No clear effect of 2008 high water scenario in the soft constraints operating regime on the diversity of plants in each VCT over time. Comparisons between species evenness (J) for each VCT from 2007, 2008, 2009 and 2010 from field data. Comparisons between species richness for each VCT from 2007, 2008, 2009 and 2010 from field data. Comparisons between species diversity for all vegetation within three elevation bands , and m from 2007, 2008, 2009 and 2010 from field data. Pielou's Species evenness (J), plotting of trends over time. Species Richness, plotting of trends over time. Shannon Weiner Index (H'); plotting of trends by elevation over time, as an indication of the effects of depth and duration of inundation. Slight decrease in species evenness over time in some VCTs which is probably due to higher numbers of plots done in each consecutive year. No clear effect of 2008 high water scenario in the soft constraints operating regime on the diversity of plants in each VCT over time. These are observations only and not statistically significant. Increases in species richness over time in 2010 in most some VCTs which is probably due to higher numbers of plots done in No clear effect of 2008 high water scenario in the soft constraints operating regime on the diversity of plants in each VCT over time. These are observations only and not statistically significant. Very little change in the patterns in species diversity X elevation band from 2007 to Species diversity by elevation: m < < for all years except 2009 where m = m. These are observations only and not statistically significant. 71

94 MQ Object of Tests Tests Significant Results Comparisons between species evenness for all vegetation within three elevation bands , and m from 2007, 2008, 2009 and 2010 from field data. Pielou's (J); plotting of trends by elevation over time, as an indication of the effects of depth and duration of inundation. Very little change in the patterns in species evenness X elevation band from 2008 to In 2007 species evenness was lower at low elevation and higher at high elevation (Species diversity by elevation: m < < ); possibly due to small number of field plots in In 2008 and 2009 species evenness appeared to be similar across all elevation bands, and species evenness increased slightly by elevation in 2010: m < < These are observations only and not statistically significant. Comparisons between species richness for all vegetation within three elevation bands , and m from 2007, 2008, 2009 and 2010 from field data. Distribution of vegetation in mapped VCTs Species richness; plotting of trends by elevation over time, as an indication of the effects of depth and duration of inundation. Species richness consistently increased with elevation for all measurement years metres < metres < metres with very similar patterns in species richness by elevation band for each year. These are observations only and not statistically significant. See confusion matrix results. 72

95 4.0 DISCUSSION In summary, at the landscape level, based on a sample of 398 polygons out of approximately 2000, a total of six polygons showed a change in vegetation community type and of those only three were decisive changes in the decile cover values for VCTs in the polygons. No changes were noted in the classification of the vegetation community types in the field or in the interpretation of the 2010 imagery, although additional clarity in the image and higher numbers of samples in the field added to our certainty regarding the VCT classification and interpretation. There are differences between the 2007 and 2010 imagery that have to be accounted for in the interpretation. For example, vegetation growth stages of some sites were behind in 2010 compared to 2007, and as a result vegetation was less abundant and shorter in stature in many places, but not all. The distribution of vegetation communities in Revelstoke Reach was not representative of the remainder of the reservoir, although vegetation communities were very similar in composition and structure and shared many features. The vegetation in the six portions of the reservoir, ranging from Revelstoke to Deer Park is likely to be representative of the vegetation of the whole reservoir. No new vegetation types were identified using improved imagery in 2010 or in the course of the fieldwork in 2010, which was roughly three times more extensive than in previous years. Still, there is always a possibility new VCTs may occur as the study area is very large and spans several bedrock and climatic regimes (Enns et al. 2007). Cover of vegetation in the three elevation ranges of 434 to 436 metres, 436 to 438 metres and 438 to 440 metres typically progressed from low to high with increasing elevation. Plant communities that were inundation-tolerant, however, usually had high cover at low elevation (434 to 436 metres), but heights of vegetation were usually low at low elevation. Heights of vegetation generally increased with elevation. Environmental variables such as soil texture and nutrition (which are related to the fractions of sand, silt, clay, gravels and cobbles) are both positively and negatively correlated with vegetation cover and heights. Silt and clay (rich nutrient regimes) have been positively correlated with cover (Enns et al. 2009), and sands and gravels are correlated with low cover where scouring of materials and exposure of sand and gravel has occurred. As explained in previous years, the autecology and adaptability of the vegetation itself explains a lot of the patterns in the vegetation of the reservoir at the landscape scale. It is possible, perhaps even likely, that some environmental effects outweigh the influences of the soft constraints operating regime, at the landscape scale. However, observations of the trends in the vegetation from 2007 to 2010 suggest that cover of vegetation will decline if duration of inundation extends over most of the growing season into the fall, and it is possible that if this were continue over some unknown period of time, a subsequent further loss of vegetation could occur. The results illustrate what is relatively obvious in the field. Beaches with high portions of both gravel and sand materials in the texture fraction tend to have very low cover of vegetation in general and are susceptible to scouring from 73

96 wave action at very high elevation and very low elevation. They can also have low overall heights, except above 438 metres in the CR: Cottonwood riparian and PA: Redtop upland vegetation community types, where vegetation is inundated less frequently and for much shorter durations than vegetation community types below 438 metres. The invasion of taller species, such as willows, contributes to outliers in the height distributions of the uplands (CR, PA) and both beach vegetation community types (BE, BG). Changes in vegetation at the landscape level were examined using the mapped data and two types of repeated measures analysis: confusion matrices and analysis of variance. The results of the mapped comparisons (confusion matrix using coded data) between 2007 and 2010 recognized a significant change in heights over time in those VCTs that occur at all three elevations, and a significant difference in cover in the bouldery beach community type between the two years. The results of the ANOVA from continuous data were similar, but in addition, significant differences were recognized in the vegetation cover of the BE: Beach VCT between 2007 and Cover and heights in all vegetation types were significantly different between 2007 and 2010, when elevation of the type was included in the analysis. Field data from 2007 to 2010 indicated a depression in vegetation cover and heights in some VCTs in 2009, especially at low elevations, which we attributed to the high water levels from summer 2008 to early spring A review of the 2007 vs imagery shows the effect of a 2.5 week difference in the timing of image capture. Examples of aerial photographic comparisons between polygons in 2007 vs (including some of the six polygons with decile changes) show the effects of log redistribution, die-back of shrub cover (Salix sitchensis), ORV-induced reduction in cover and exposure of materials (e.g., scouring) and the appearance of shrubs in very early spring versus mid to late spring. Species richness increased with elevation, except in water-requiring or watertolerant vegetation at the lowest elevation range under study ( m). Species diversity and evenness varied slightly between vegetation community types, and increased within VCTs with increasing elevation. These patterns have not changed over time since the beginning of the observation period. The total number of species (species richness) has increased over time in comparison to previous years in the more complex, upland shrub and forest vegetation communities. This is due to additional sampling effort in In fact, it is very likely that increasing the sample size even further in CR: Cottonwood riparian would cause another increase in known species for this VCT. However, the emphasis on sampling should be on those VCTs at risk to loss of vegetation spatial extent, structure and composition in response to the soft constraints operating regime. Graphs of the soft constraints operating regime show that the elevations above where most of the CR occurs are not inundated for long periods. CR may in fact benefit from inundation in many respects, due to its typically brief increased water supply to roots during the growing season. Despite additional sampling, species richness in sandy and gravelly vegetation community types at low elevations has remained similar over the years, indicating that this type is sufficiently sampled (Mueller-Dombois and Ellenberg, 1974). 74

97 Growth resumes in the fall for most species in the drawdown zone, if the water recedes in August or September (Enns et al. 2007). This was evident in a comparison between the appearance of the vegetation in the imagery collected in May, and the plot data collected in September-October of 2007, which showed that the vegetation had clearly grown under the water. The effect of the sustained high water levels in the fall of 2008 on vegetation is evident in the data, with field data showing that recovery seems to be occurring in some VCTs. An examination of changes in the imagery over time using a spectral classification will potentially provide a simple overview of the population. For example, change in total area of vegetation in all study areas in comparison to change in gravel and sand coverage in all study areas could be used to indicate areas of change (see Appendix 8). The first four management questions addressed in this paper are descriptive: What is the vegetation composed of, how is the vegetation distributed, is one part of it any different from the rest, and how is it affected by environmental variables other than the soft constraints operating regime? The remaining two questions address how the vegetation is maintained by the soft constraints operating regime, and what changes might be made to maintain vegetation. When addressing these questions, we are assuming that vegetation changes naturally over time, that the heights of vegetation should increase over the period of the observations, and that in the absence of the soft constraints operating regime, the effects of the environment would not be any different from the effects we observed. We have confidence in answering the first four questions because we have four years of analysis with similar results every year. New species were added to species richness of under-sampled VCTs in 2010, but there are few differences in the statistics from previous years. The vegetation of the Arrow Lakes Reservoir drawdown zone, when considered at the landscape level using sampling and statistics, does not appear to have experienced an overall significant change in cover or height since 2007, nor does composition of the vegetation appear to have changed. This is based on the observation that only six of 398 polygons underwent a decile change in the coverage of the vegetation community type. When coded variables describing cover and heights of vegetation were considered, a significant difference was observed for low elevation VCTs between 2008 and 2009, and we postulate that the higher water levels in 2008 to 2009 are responsible. Phenological differences have to be considered in the interpretation of these data. The imagery in 2007 was captured between May 30 and May 31, and the weather had been relatively warm prior to the capture. In 2010, the imagery was captured between May 11 and the 15, and the weather had been cool and wet, although the winter weather was very mild. Differences were observed that could be attributed to phenological states in the two photo capture periods but they were not great enough to cause a shift in decile coverage within VCTs. 75

98 5.0 CONCLUSIONS When considering the reservoir as a whole, there is no significant difference between total vegetation cover in Revelstoke Reach and the Arrow portion of the Arrow Lakes Reservoir for all elevation bands. The ANOVA test for significant differences in vegetation cover between elevation and geographic areas of the reservoir (Arrow vs Revelstoke Reach) (Appendix 4) shows no differences between the two parts of the reservoir overall, but significant differences were found between total vegetation cover by elevation band. The post-hoc comparisons based on field data (Appendix 4) show that when elevation and geographic areas of the reservoir are considered together, total cover in the low and middle elevation bands in the Arrow Lakes proper and Revelstoke Reach are not significantly different from each other, but Arrow and Revelstoke Reach total covers are significantly different in the high elevation range ( m). Further, total vegetation cover based on field plot data in Revelstoke Reach at high elevations was higher than high elevation plots in the Arrow portion of the Arrow Lakes Reservoir. This was also noted in Enns et al. (2008) and Enns et al. (2009), and was attributed to the steeper slopes in the drawdown zone of the Arrow portion, which caused some scouring at high elevation due to higher winter wave action. In previous years, the greater species richness and diversity in the Arrow Lakes portion of the Arrow Lakes Reservoir was attributed to higher variation in parent materials and soils in the Arrow Lakes portion than in the Revelstoke Reach portion (Enns et al. 2007). The higher numbers of sites sampled in 2010 vs and 2008 (and the CLBMON-12 sites completed in 2009) support this observation. Stream entries (RS) and disturbed sites (IN) had higher species richness in Revelstoke Reach than in the Arrow Lakes portions of the Arrow Lakes Reservoir in These VCTs are strongly influenced by immigration of new species and the introduction of weeds. We have only seen one divergence from the main pattern of the soft constraints operating regime, and there is evidence in the data to link this divergence to a loss of cover in low elevation VCTs. We believe the following evaluations of the hypotheses are reasonable but tentative assumptions: H 0 : Under the soft constraints operating regime (or possibly a newly selected alternative after five years), there is no significant change in existing vegetation communities at the landscape scale. H 0A : There is no significant change in the spatial extent (cover and number of hectares) of vegetation communities within the existing vegetated zones of Arrow Lakes Reservoir. Accepted for the reservoir as a whole at a 10 per cent level of accuracy. Rejected for some VCTs at a high resolution. H 0B : There is no significant change in the structure and composition (i.e., distribution and diversity) of vegetation communities within the existing vegetated zones of Arrow Lakes Reservoir. Rejected for some VCTs at a high resolution with respect to structure. Accepted for the reservoir as a whole at a 10 per cent level of accuracy, with respect to distribution. No distinct conclusion is available regarding diversity until more comparisons in the aerial photographic capture over time are available. Data available so far suggest acceptance of the null hypothesis with respect to diversity, evenness and richness at the landscape scale. 76

99 These statements are based on a single change in the soft constraints operating regime, with four years of observation in the field, and a comparison between two somewhat dissimilar sets of images. A second example of a variance from the usual operating regime, or conversely, the maintenance of the operating regime in its usual form over the next three to four years should show a measureable response in the vegetation. A third capture of good quality imagery will allow for better resolution of the hypotheses. 77

100 6.0 SOURCES OF UNCERTAINTY AND POTENTIAL ERROR There are several potential sources of error in field and mapping studies. The following sources of error and uncertainty apply to this study. Redundancy in the polygon data used in the RMA occurs for those polygons that cross elevation zone boundaries and may be rectified by fragmenting the polygon and re-assigning different cover and height estimates from the aerial photographic interpretation. However, redundancy applies to relatively few polygons, and there appear to be few differences in cover and heights between the fragments of a polygon due to variation within the polygon. The use of aerial photographic interpretation to supply data for RDA is limited because estimates of scouring, wave action and water energy are based on the appearance of the vegetation and materials and soils, and not on actual measures of scouring or water behaviour. The low vegetation cover observations in 2009 could be a function of low overall cover estimates made by the team members (i.e., an aberration); however, they were consistently low, and some team members were consistently in the field over the four years of estimates. Further, calibration of cover occurred every year, and especially in Comparisons between images are always problematic, due to various observational and image quality issues. However, the clarity and detail available in the 2010 images allowed for calibration of the 2007 images and verification of features and conditions that were previously not available. Statistical tests of samples of the populations, in this case vegetation types within a large reservoir, have high uncertainty when variation is also high. The use of repeated measures helps define changes in certain locations, but a large sample size is needed to make sure these locations are representative. 78

101 7.0 REFERENCES Apan, A Mapping and analysis of change in the riparian landscape structure of the Lockyer Valley Catchment, Queensland Australia. Landscape and Urban Planning. 59: (1) BC Hydro Consultative Committee report: Columbia River Water Use Plan, Volumes 1 and 2. Report prepared for the Columbia River Water Use Plan Consultative Committee by BC Hydro, Burnaby, BC. 924 pp. Braumandl, T.F. and M. P. Curran A field guide for site identification and interpretation for the Nelson Forest Region. BC Ministry of Forests. Research Branch. Victoria, B.C. Land Management Handbook No pp. Congalton, R.G. and R.A. Mead A quantitative method to test for accuracy and correctness in photo-interpretation. Photogrammetric Engineering and Remote Sensing 49:69-74 Enns, K.A., R. Durand, P. Gibeau and B. Enns Vegetation monitoring of the Arrow Lakes Reservoir. CLBMON-33 Arrow Lakes Reservoir Inventory of Vegetation Resources (2007). Report prepared by for BC Hydro. Castlegar, B.C. 116 pages including appendices. Enns, K., P. Gibeau and B. Enns CLBMON-33 Arrow Lakes Reservoir Inventory of Vegetation: 2008 Final Report. Report prepared by for BC Hydro. Castlegar, B.C. 78 pages. Enns, K., P. Gibeau and B. Enns CLBMON-12 Monitoring of revegetation efforts and vegetation composition analysis. Report prepared by for BC Hydro. Castlegar, B.C. 94 pages. Enns, K.A CLBMON-33 Arrow Lakes Reservoir Inventory of Vegetation Resources: Revised Study Design. Castlegar, B.C. 57 pages Enns, K. and G. Tomlins Draft image comparison. Letter to BC Hydro. October 31, Appended. Grau, M. and S. Walsh Mid Columbia River and Arrow Lakes Reservoir: overview summary for the revegetation program physical works (CLBWORKS-2) Consultative Draft. Version 2.0 Unpublished paper for BC Hydro. Castlegar,B.C. 28 pages. Hawkes, V.C., C. Houwers, J.D. Fenneman, and J.E. Muir Kinbasket and Arrow Lakes reservoir revegetation management plan: Kinbasket Reservoir inventory of vegetation resources. Annual Report LGL Report EA1986. Unpublished report by LGL Limited environmental research associates, Sidney, BC, for BC Hydro Generations, Water Licence Requirements, Burnaby, BC. 82 pp. Hawkes, V.C. and J.E. Muir CLBMON-10 Kinbasket Reservoir Inventory of Vegetation Resources. Annual Report LGL Report EA1986. Unpublished report by LGL Limited environmental research associates, Sidney, BC, for BC Hydro Generations, Water License Requirements, Castlegar, BC. 78 pp. 79

102 Keser, N Interpretation of landforms from aerial photographs. With Illustrations from British Columbia. British Columbia Forest Service. Government of British Columbia. Victoria, B.C. 215 pages. Little, R.C., G.A. Milliken, W.W. Stroup and R.D. Wolfinger SAS system for mixed models. Second Edition. SAS Institute, Inc., Cary, NC, USA. 249 pp. Luttmerding, H.A., D.A. Demarchi, E. C. Lea, D. V. Meidinger and T. Vold Describing Ecosystems in the Field. Second Edition. BC Ministry of Environment, Lands and Parks and Ministry of Forests. Victoria, BC. 213 pages. McGill, R., Tukey, J.W., and Lasen, W. A Variations of box plots. The American Statistician, 32, pages Mueller-Dombois, D. and H. Ellenberg Aims and methods of vegetation ecology. John Wiley and Sons. Toronto. 547 pages. Nagaike, T., A. Hayashi, M. Kubo, K. Takahashi, M. Abe and N. Arai Changes in plant species diversity over 5 years in Larix kaempferi plantations and abandoned coppice forests in central Japan. Forest Ecology and Management 236 (2006) Omule, A.Y. and K.A. Enns CLBMON-33: determination of optimal sample size for repeated measures analysis. For B.C. Hydro. Delphinium Holdings, Inc. 22 pages. Sim, J. and C.C. Wright The Kappa statistic in reliability studies: use, interpretation, and sample size requirements. Physical Therapy, Volume 5, Number 3, pages Tukey, J.W Exploratory data analysis. Addison Wesley, Reading, Massachusetts. 688 pages. Velleman, P.F. and D. C. Hoaglin Applications, basics and computing of exploratory data analysis. Duxbury Press, Boston. 237 pages. Wilkinson, L., M. Hill, J.P. Welna, S. Miceli, G. Birkeneuel and E. Vang Statistics: SYSTAT for Windows, Version 5 Edition. Evanston Ill. 636 pages 80

103 8.0 GLOSSARY error - the probability of rejecting a true null hypothesis in statistical testing ß error - the probability of accepting a false null hypothesis in statistical testing % cover - see per cent cover a priori - in statistics, a priori knowledge is prior knowledge about a population, rather than that estimated by recent observation abundant - large numbers of individuals adjusted-r2 - a modification of R2 that accounts for the number of explanatory (independent) variables included in multiple regression or canonical statistical analyses aerial photographic image capture - the process of recording photographic images on film or in a digital format from the air aerial photographic imagery - refers to a collection of aerial photographic images or pictures aerial photographic mosaic - an aerial photograph of a large area, made by carefully fitting together aerial photographs of smaller areas so that the edges match in location, and the whole provides a continuous image of the larger area aerial photographic interpretation - the process of studying and gathering information on aerial photographs required to identify the various natural and cultural features on the ground aerial photography - taking photographs of the ground from the air aerial triangulation - using the process of triangulation to correct for distortion on aerial photographic imagery; see triangulation air photo mosaic - see aerial photographic mosaic anoxic - lacking oxygen ArcGIS - a group of software products (Geographic Information System GIS) used to manage, view and manipulate spatial data, create maps and perform basic spatial analysis aspect - the direction of a slope gradient or the horizontal direction to which a mountain slope faces asymptotic - a variation in the distribution of a variable in relation to the population. autecological - pertaining to autecology autecology - a subfield of ecology that deals with the biological relationship between an individual species and its environment autocorrelation - the internal correlation between members of series of observations ordered in time or space biometrics - the study of measurable biological characteristics 81

104 boxplots - visually display the differences between groups of data by showing the dispersion and skewness of data without making any assumptions about their underlying statistical distributions canonical analysis - a multivariate technique in statistics used with multiple regression models to assess the linear relationships between groups of variables in a data set canonical correspondence analysis (CCA) - a multivariate, direct gradient analysis method that is derived from correspondence analysis but has been modified to allow environmental data to be incorporated into the analysis change detection - detecting differences over time in vegetation characteristics including cover, height, vigour, distribution, and composition (species richness, abundance, diversity and evenness) chlorosis - a condition in which leaves produce insufficient chlorophyll and have a yellow appearance central tendency - tendency for values to lie close to the median or mean. climatic regime - the conditions of the climate including temperature, humidity, rainfall, snowfall and wind in a given region over a long period of time colluviation - a terrain process whereby materials are developed by downslope movement common - widely distributed community - see vegetation community complex polygon - a polygon that contains more than one unit or type composition - see vegetation composition confounding variable - an extraneous variable in a statistical model that correlates (positively or negatively) with both the dependent variable and the independent variable; a confounding variable can alter the outcome of a study by showing a false correlation between the dependent and independent variables, leading to an incorrect rejection of the null hypothesis confusion matrix - a table that contains information about actual and predicted classes done by a classification system; the performance of such classification systems is commonly evaluated using the data in the table or matrix concave - terrain that is bowl-shaped, receiving water convex - terrain that is upside-down bowl-shaped, and shedding water correspondence analysis - a statistical technique designed to analyze simple twoway and multi-way tables containing some measure of correspondence between the rows and columns; the analysis is an ordination technique that provides a graphic method for exploring the systematic relationships between categorical variables in the tables when there are incomplete a priori expectations as to the nature of those relationships - 82

105 covariance - a measure of the linear relationship between two random variables or how much the two variables change together cover - cover estimate of vegetation in a polygon vegetation community type, see also vegetation cover (estimated in the field) cover value - see vegetation cover value database - an organized collection of data for one or more uses, typically in digital form data cleaning - the act of detecting and correcting (or removing) incomplete, incorrect or inaccurate records from a record set, table, or database data screening - similar to data cleaning; the process of checking for completeness, validity and quality of input data Daubenmire cover classes - classes used to group % cover value data; 1 = zero to five, 2 = five to 10, 3 = 10 to 25, 4 = 25 to 50, 4 = 50 to 75, 6 = 95 to 100. decile - one of nine values that divide data or area into 10 equal parts or 10 per cent increments dieback - the loss of upper crown structure in vegetation, usually seasonal, but may be pathological and often results in regrowth degrees of freedom - the number of independent units of information in a sample used in the estimation of a parameter or calculation of a statistic DEM - digital elevation model dependent variables - the variables that are being measured in an experiment and that are being affected by or responding to the independent or explanatory variables digital image analysis - see image analysis digital image capture - the process of recording photographic images in a digital format digital imagery - images created in a digital format that contain a fixed number of rows and columns of digital values called picture elements or pixels; each pixel in the digital image represents a brightness of a given colour direct gradient analysis - a statistical method that examines the response of individual species or vegetation community types to gradients of environmental factors including climate, parent material, topography, other organisms and time distribution - see plant species distribution distribution codes - see plant distribution codes diversity - the number of different classes (i.e., community types, genera, species) in a specified area; also see species diversity drawdown zone - the shoreline area along the Arrow Lakes Reservoir located between low and high water levels 83

106 dummy variable - a binary variable that is used to represent a given level of a categorical variable (i.e., a variable that is either present or absent); the most common choices as dummy variables are 1 and 0 edaphic - related to or caused by particular soil conditions rather than by physiographic or climatic factors e-flora - an electronic atlas of the plants of British Columbia eigenvector - an eigenvector of a matrix is a vector such that when we multiply the matrix by the vector we get the vector back again except that it has been multiplied by a particular constant, called the eigenvalue ephemeral - lasting only a short period of time (e.g., an ephemeral stream that only exists for a short period following precipitation or snowmelt) error matrix - see confusion matrix evenness - see species evenness field truthing - see field verification field verification - confirming the accuracy and reliability of mapping by on-theground field surveys and sampling fixed effects - observed values of explanatory (independent) variables that are all treated as if they were non-random (i.e., the effects are deliberately controlled by the experimenter) forward selection - a procedure used to select only the independent (explanatory) variables that contribute significantly to explaining the variance in dependent (response) variables within multiple regressions or canonical models fluviation - a terrain process whereby materials are developed by water movement geographic information systems (GIS) - a set of tools that captures, stores, analyzes, manages, and presents data that are linked to location(s); the systems digitally create and manipulate spatial areas glaciofluvial - a terrain process whereby materials are developed by a combination of glacial movement and melting glacier water movement ground truthing - confirming the accuracy and reliability of mapping by on the ground field surveys and sampling height - see vegetation height herb - non-woody vascular plants including forbs and graminoids (plants with grass-like growth form including grasses, sedges, rushes and spikerushes) herbaceous - descriptor for plants that are classified as herbs heterogeneity - the quality of being different in kind or nature heterogeneous - differing in kind or type; consisting of dissimilar elements or parts Ho - see null hypothesis 84

107 homogeneity - the quality of being similar or comparable in kind or nature homogeneous - of the same or similar kind or nature hygrotope - unit in a classification of soil moisture regimes hypotheses - the plural of hypothesis hypothesis - a tentative statement that proposes an explanation for an observable phenomenon; the hypothesis can be tested by further investigation and statistical methods image analysis - the extraction of meaningful information from images including the extraction of data from aerial photography imagery using air photo interpretation methods and digital image processing techniques image data capture - see aerial photographic image data capture image training sites - sites selected in the field that are homogeneous, typical and representative of information classes (e.g., vegetation community types) that are used in conjunction with image processing software to develop characterizations of the spectral properties (reflectance s) of each class included in an image classification scheme; the unique spectral properties of classes are used to "train" the classification decision rules used in developing the classification imagery - see aerial photographic imagery incidental species - species that occur as a minor part of the species composition in the study area independent variables - variables that are presumed to affect or explain a dependent variable; typically the variables in an experiment representing the values that are being manipulated or changing intrinsic - an essential or inherent property of a system or thing inundated - flooded by standing or slow-moving water inundation - surface flooding by standing or slow-moving water Kappa (K) statistic - a statistical measure of agreement used in assessing the degree to which two or more observers, examining the same qualitative (categorical) data, agree when it comes to assigning the data to categories kurtosis - the degree of "peakedness" of a probability distribution; a measure of whether the data are peaked or flat relative to a normal distribution leading species - the dominant species with the greatest abundance within a specified area leading vegetation - the dominant vegetation community type with the largest spatial extent within a polygon or other specified area least square means - within-group means that are appropriately adjusted for the other effects in the model; they estimate the marginal means for a balanced population (as opposed to the unbalanced design) life form codes - codes that indicate the growth form of plant species (i.e., tree, shrub, graminoid, forb, moss, lichen) 85

108 linear mixed effects ANOVA model - see linear mixed effects model (LMM) linear mixed effects model (LMM) - a statistical model that investigates responses from a subject (dependent variable) that are thought to be the sum (linear additions) of fixed and random effects; the model shows the amounts that fixed and random factors contribute to the overall variability in the dependent variable loam - soil containing a relatively equal mixture of sand and silt and a smaller proportion of clay (about % concentration respectively) lowland - land lying below the area where water usually flows or that is influenced by flooding and poor drainage massive catastrophic failure - structural failure in a landform or feature causing debris torrent or slide. median - a numerical value that divides a sample, population or probability distribution in half when all data values are listed in order; it is the preferred measure of central tendency for a skewed distribution (in which the mean would be biased) mesic - medium soil moisture regime where a site has neither excess soil moisture or a moisture deficit microtopography - small scale variations in the land surface shape (i.e., hummocks and hollows) mixed effects model - a statistical model that investigates the responses from a dependent variable due to both fixed effects (factors with values that are deliberately controlled by the experimenter) and random effects (factors with values that represent a random sample of a larger population) moisture regime - a parameter of soils that corresponds to a soil's ability to receive and retain moisture; eight classes represent the relative amounts of soil moisture available for plant growth in a biogeoclimatic subzone monomials - a monomial is a constant (number) or a product of non-negative powers of variables mosaic - see aerial photographic mosaic multiple regression - a statistical method used to quantify the relationship between several independent (explanatory) variables and a dependent (response) variable multivariate - multivariate statistics encompasses the simultaneous observation and analysis of more than one statistical variable at a time; multivariate analysis involves analyzing more than one related outcome measure per dependent variable simultaneously with adjustment for multiple confounding variables (covariates) necrosis - premature death of cells and living tissue non-linearity - refers to a situation or process in which there is no simple proportional relation between cause and effect non-soil - parent materials with no soil development 86

109 normal distribution - a probability distribution of a random variable which is bellshaped, symmetrical, and single peaked and where the mean, median and mode coincide and lie at the center of the distribution; a normal distribution is fully specified by two parameters - mean and the standard deviation normality - a property of a random variable that is distributed according to a normal distribution null hypothesis - a type of hypothesis used in statistics that proposes that no statistical significance exists in a set of given observations; it is presumed to be true until statistical evidence rejects it for an alternative hypothesis nutrient regime - a parameter of soils that indicates a soil's ability to supply major nutrients required for plant growth; five classes of soil nutrient regimes are recognized in a biogeoclimatic subzone ocular estimates - subjective visual estimations of a value (e.g., % cover) ordination - a statistical method in multivariate analysis that is complementary to data clustering and used mainly in exploratory data analysis; ordination orders multivariate objects so that similar objects are near each other and dissimilar objects are farther from each other and the relationships between objects (i.e., floristic and environmental gradient parameters) can be characterized numerically and/or graphically ordination diagram - a graph used to display the relationships between multivariate objects resulting from an ordination analysis orthogonal axes - a set of two or more mutually perpendicular axes orthophoto mosaic - see orthorectified aerial photographic mosaic orthophotographs, orthophotos - see orthorectified photographs, photos orthorectified photographs, photos - aerial photographs that have been corrected for distortion after image capture orthorectified aerial photographic mosaic - an large aerial photograph created by fitting together a series of smaller air photos that have been corrected for distortion orthorectification - the process of correcting for errors and distortion in aerial photographic imagery; for accurate removal of image distortions, a digital elevation model and adequate GPS-derived ground control points are required orthorectified - having gone through the process of orthorectification ORV - outdoor recreational vehicle outliers - observations that deviate markedly from other members of the sample in which they occur p-value - in statistical significance testing, the p-value is the probability that the observed data or a more extreme outcome would have occurred by chance; the lower the p-value, the less likely the result is if the null hypothesis is true 87

110 parent material - the underlying geological material (i.e. bedrock, glacial till, colluvium, lacustrine silts, and fluvial sand, silt and gravel deposits) in which soil horizons form pathogens - biological agents that cause disease to their hosts pathology - the study and diagnosis of disease per cent cover (% cover) - the area of the foliage of a plant species projected onto the ground within a specified area expressed as a percentage of the area perennial plants - a plant that lives for more than two years, usually flowering each year phenological differences - differences in the development of plants or animals related to the effects of climatic conditions on growth and life cycle stages phenology - the study of periodic plant and animal life cycle events such as flowering, breeding, and migration, in relation to climatic conditions photogrammetric - pertaining to photogrammetry, which is the practice of determining the geometric properties of objects from photographic images photomosaic - see aerial photographic mosaic physiographic features - the natural shapes, patterns and landforms of the environment that are associated with bedrock relief, glaciation, tectonic activity, volcanism, mass wasting, erosion and water movement physiography - the systematic classification and description of the physical and geographic patterns, processes and phenomena of the natural environment Pielou's evenness index - see species evenness index pivot table - a data summarization tool found in data visualization programs such as spreadsheets; pivot-table tools can automatically sort, count, and total the data stored in one table or spreadsheet and create a second table (the "pivot table") that displays the summarized data pixel - a single point in a raster image; the pixel is the smallest display element of a digital image that can be controlled pixelation - an effect caused by displaying a digital image at such a large size that individual pixels are visible to the eye plant distribution codes - number codes used to represent dispersion patterns of individuals of a plant species within a specified area plant species distribution - the spatial arrangement or dispersion pattern of individuals of a plant species within a specified area plant succession - the replacement of one plant community by another; a site may often progress to a stable terminal community called the climax polygon - a discrete, topographically defined unit in space and time, surrounded by a line delineating it from all other polygons; the two-dimensional shape delineated on a map or air photo can be modelled and stored within a 88

111 digital database where its spatial position in the database is defined by the co-ordinates of its vertices polynomials - a polynomial is an expression of finite length constructed from variables and constants, using only the operations of addition, subtraction, multiplication, and non-negative, integer exponents polynomial equation - any mathematical equation in which one or both sides are in the form of a polynomial; polynomial equations can be used to adequately model the complex spatial nature of an area pooled data - data from two or more groups of data or two or more data sets that are combined for statistical analysis post-hoc comparisons - comparing the results of an experiment or analysis after it has concluded for patterns that were not specified based on prior knowledge about a population power - see statistical power power analysis - a statistical method used to calculate the minimum sample size required to accept the outcome of a statistical test with a particular level of confidence principal components analysis (PCA) - a statistical method involving a mathematical procedure that transforms a number of possibly correlated variables into a smaller number of uncorrelated variables called principal components; the first principal component accounts for as much of the variability in the data as possible, and each succeeding component accounts for as much of the remaining variability as possible; PCA aims at reducing a large set of variables to a small set that still contains most of the information in the large set qualitative - refers to data that are described in terms of some quality or categorization and includes virtually any information that can be captured that is not numerical in nature quantitative - refers to data that are numeric values representing amounts that are measured R2 (R-squared) - the coefficient of determination of multiple regression statistical analyses that measures the fraction of variance in Y (dependent or response variable) that is explained by a linear combination of the independent (explanatory) variables in X random error - errors in measurement that lead to measured values being inconsistent when repeated measures of a constant attribute or quantity are taken; the effect of random errors may be reduced by repetition of the measurements and averaging the results redundancy analysis (RDA) - a multivariate direct gradient analysis method in which species are presumed to have linear relationships to environmental gradients - repeated measures analysis (RMA) - a statistical method in which the same measures are collected multiple times for each subject but under different conditions 89

112 (RMA) ANOVA - a repeated measures analysis of variance that can be used to compare values for the same variable from the same sample population that has been measured at different times and/or under different conditions replicates - Samples taken in a population, e.g., VCT polygons, or vegetation plots. rill - an underground water supply, usually a buried stream, often comes to ground in a reservoir riparian - along the bank of a river, lake or wetland serpentine - bending from side to side as in the movement of a serpent, in reference to river channel shape. Statistical Analysis System (SAS) - a comprehensive computer software system for data processing and analysis scouring - the removal of soil and substrate materials by moving water and/or wave action seepage - subsurface or surface groundwater discharge having less flow than a spring serpentine creek or river - a moving water body with sinuous channel formation shading effect - an artefact in a photograph where tree and shrub canopies cause shadows to appear in imagery; is both an advantage (used in interpretation of aerial photography for estimating height; provides a signature pixel colour for year-wise comparisons) and a disadvantage (obscures details) Shannon index of diversity (H') - a measure of species diversity that considers both the total number of species and their relative abundances (proportions) within a specified area Shannon-Weaver diversity index (H') - also known as the Shannon-Weiner diversity index; see Shannon index of diversity shape file - a digital vector storage format for storing geometric data types (points, lines and polygons) and their associated attribute information in tables of records to specify what they represent significance - see statistical significance significant difference - see statistical significant difference skewness - the degree of asymmetry about a central value of a distribution slope - see slope gradient slope gradient - the steepness of a slope slope position - refers to the position of a site on the gradient from high elevation to low elevation soft constraints operating regime - the operating system used to allocate water of the Arrow Lakes Reservoir under competing demands, where a set of soft constraints that represent desirable conditions but not inviolable ones, are used to guide operational decisions each year to balance the objectives of 90

113 different stakeholders including those concerned with recreation, fish, wildlife, vegetation, culture & heritage and erosion soil - a true soil has organic material incorporated with mineral material, and indications of development, i.e., layering over time. soil/substrate texture - relative proportions of the three particle sizes - sand, silt and clay in the fine fraction portion of the soil or substrate spatial extent - the distribution of plant community cover at the site or landscapelevel species composition - see vegetation composition species diversity - the number of different species and their relative abundances in a given area or habitat; species diversity can be measured and quantified using an index of diversity species evenness - the degree of equitability in the distribution of individuals among a group of species; the relative abundance with which each species is represented in an area species evenness index (J) - a mathematical expression that provides a measure of species evenness or equability within a specified area spectral data - The range in expression of colour in an image species richness - the total number of different species in a specified area statistical population - total of all the experimental units (i.e., polygons) in the study area statistical power - the probability that a statistical test will produce a significant difference at a given significance level; the capacity of a statistical test to provide the smallest ß error for the designed a error statistical significance - the statistical significance of a result is the probability that a difference (i.e. between means) in a sample occurred by pure chance and that no such difference exists in the population from which the sample was drawn statistical significant difference - in statistical testing, a difference between two statistics is significant when the results show that the difference is too great to have occurred purely by chance (i.e. the results of the test are strong enough to prove that the null hypothesis needs to be rejected) statistical variance - the main measure of variability for a data set; variance is one of several descriptors of a population frequency distribution where it describes how far values lie from the mean statistical variation - the range of differences observed for a variable in a sample population statistically significant - see statistical significance structure - see vegetation structure sub-hypotheses - secondary hypotheses that are related to the main hypotheses of the experiment or study 91

114 substrate - the surface layer of the ground where soils develop and that plants and animals may live upon; substates can include both biotic and abiotic materials surface expression - terrain texture, usually determined in aerial photographic interpretation, followed by field truthing surficial changes modification of the soil or parent material surfaces, usually caused by water, grazing, ice, downslope movement, etc. taxa - groups of living organisms, at various levels of classification texture - see soil/substrate texture till - materials (rock) dropped from a retreating glacier or worked by a moving glacier topo-edaphic - pertaining to soil and topographic features that influence the distribution, composition and structure of plant species and communities topography - the surface shape of the land in terms of elevation, slope and orientation trained classification - a grouping of pixels, often in relation to field expression triangulated - see triangulation triangulation - the process of determining the location of a point by measuring angles to it from known points at either end of a fixed baseline rather than measuring distances to the point directly; the point can then be fixed as a third point of a triangle with one known side and two known angles two-tailed statistical test - the test uses a two-tailed probability to determine significance where the null hypothesis will be rejected for values of the test statistic that are either sufficiently small or sufficiently large upland - land lying above the area where water usually flows or that is influenced by flooding and poor drainage UTM coordinates - x and y coordinates, called eastings and northings, respectively, that locate a point in a universal transverse mercator (UTM) projection of the earth's surface variance - see statistical variance variation - see statistical variation vectors - a quantity possessing both a magnitude and a direction vegetation or plant community - a group of interacting plants species inhabiting a given area; the components of each plant community are influenced by soil type, topography, climate, organisms and disturbance vegetation structure - vegetation structure is characterized primarily by the horizontal and vertical distributions of plant biomass, particularly foliage biomass vegetation community type - the basic map unit in the reservoir draw down zone based primarily on physiographic (terrain, soil/substrate) features and vegetation characteristics; VCT 92

115 vegetation composition - includes the different plant species, their abundances and distributions within a vegetation community vegetation cover - the area of the foliage of a plant species projected onto the ground within a specifies area vegetation cover value - the area of ground covered by vegetation or a plant species expressed as a percentage of the sample area covered vegetation height - the height of a plant species or overall vegetation in a community type measured from the ground level to the top of the plants or canopy water regime - soft constraints operating regime; how the water levels are managed in the reservoir wetland - sites dominated by hydrophytic vegetation where soils are watersaturated for a sufficient length of time such that excess water and resulting low soil oxygen levels are principal determinants of vegetation and soil development 93

116 APPENDIX 1. DATA STRUCTURES FOR CLBMON Table 1-1. Database description for all data from all plots completed in each Vegetation Community Type in 2010 for CLBMON-33. Spreadsheet Name Data entry Validation List Species list by life form Species list by Latin name Species list by common name Most common species 2010 Vegetated plot totals Non-vegetated plot totals All 2010 plot totals Average cover - pivot table Average cover of most common sp Arrow average cover table Arrow average cover graph Revelstoke average cover table Revelstoke average cover graph BE - pivot table BE - freq. and incidental species Incidental sp total Species richness BC species list Description Entered from 2010 field forms as laid out by Bruce Enns. Connected to Data Entry spreadsheet by data validation tool to ensure proper codes are entered; updated for Life form codes obtained from BC Ministry of Forests and Range species list. From validation list. From validation list. Uses Excel pivot table tool to tally number of occurrences of each plant species, divided by Revelstoke Reach and Arrow plots. Totals plot numbers by original year measured, type and area. Obtained by copying data entry list and using remove duplicates tool to get unique plot names from first column. Totals number of non-vegetated plot numbers by type. Combines above two totals. Average, maximum and minimum cover of species occurring in 20 or more plots. Created by pivot table tool. Displays data from above spreadsheet as graph comparing Revelstoke and Arrow average cover of species. Average, maximum and minimum cover of Arrow species occurring in 10 and 20 or more plots. Created by pivot table tool. Displays data from above spreadsheet as graph. Average, maximum and minimum cover of Revelstoke species occurring in 10 and 20 or more plots. Created by pivot table tool. Displays data from above spreadsheet as graph. Uses pivot table tool to tally number of occurrences of each species in BE plots (same method for every type). Takes above pivot table data and divides occurrences by number of vegetated plots in each type and area to obtain frequencies. Incidental species list contains species with <90, >10% frequencies (same method for every type). Counts number of incidental species in each type and compares them in graph form. Compares number of species occurring in each type. Downloaded from e-flora 9, connected to species lists to assure species names and codes are up-to-date. 9 Accessed June to September

117 APPENDIX 2. LOCATIONS OF POLYGONS INCLUDED IN THE KAPPA STATISTIC AND THE REPEATED MEASURES ANALYSIS. Cyan-coloured areas represent polygons chosen for analysis 95

118 Cyan-coloured areas represent polygons chosen for analysis 96

119 Cyan-coloured areas represent polygons chosen for analysis 97

120 Cyan-coloured areas represent polygons chosen for analysis 98

121 Cyan-coloured areas represent polygons chosen for analysis 99

122 APPENDIX 3. NUMBER OF FIELD PLOTS AND MAPPED POLYGONS SAMPLED. Table A3-1. Numbers of field plots completed in each Vegetation Community Type in 2010 for CLBMON-33. Type Arrow 434 to 436 m 436 to 438 m 438 to 440 m Arrow Total Revelstoke Reach 434 to 436 m 436 to 438 m 438 to 440 m Revelstoke Total Grand Total BB BE BG CR IN LO PA PC PE PO RR RS SS Grand Total Table A3-2. Number of mapped Polygons assessed in each Vegetation Community Type in 2010 for Repeated Measures Analysis. Data taken from the 2007 and 2010 mapping. Vegetation Community Type BE BG CR PA PC PE RR Total Number of polygons Table A3-3 shows the number of polygon - elevation zone samples used in the confusion matrix analysis. 100

123 Table A3-3. Vegetation Community Type (VCT) Number of aerial photographic observations made for the confusion matrix and RMA in 2007 and 2010 mapping: by Reach, VCT and Elevation Band (code 1, 2 or 3). Note that sample polygons may contain more than one VCT, and that if a sample polygon extended over two elevation bands, it was divided into two to three records. Therefore. some redundancy exists in the data set. Arrow Lakes Total Revelstoke Reach Total Elevation Bands 434 to 436 m 436 to 438 m 438 to 440 m Elevation Bands 434 to 436 m 436 to 438 m 438 to 440 m BE BG CR PA PC PE RR Total Table A3-4. Numbers of field plots completed in each Vegetation Community Type in 2010 for CLBMON-33, including those that were also done in 2008 and 2009 for CLBMON-33 and CLBMON-12. Some plots done in 2010 were co-located in similar locations as the 2007 plots, but they were treated as new plots in the 2010 database. Type Grand Total BB BE BG CR IN LO PA PC PE PO

124 RR RS SS Grand Total APPENDIX 4. VEGETATION COVER AND HEIGHTS IN ARROW LAKES RESERVOIR: ANALYSIS OF VARIANCE RESULTS FOR 2010 Vegetation cover in the Revelstoke Reach and Arrow Lakes portions of the Arrow Lakes Reservoir. Management Question: Is there a difference between the Revelstoke Reach and the Arrow Lakes portion of the reservoir in vegetation cover, when considered by elevation band? 300 Total Vegetation Cover (%) A m A m A m R m R m R m Reach and Elevation Band Figure A4-1. Total Cover of Vegetation in the Three Elevation Bands (434 to 436 metres, 436 metres to 438 metres, and 438 to 440 metres) in the Arrow Lakes (A) and Revelstoke Reach (R) portions of the Arrow Lakes Reservoir. Data are plotted as Tukey s boxplots (TBP). TBP are a visual method of showing statistically significant differences (p<0.05) between data distributions from comparably sampled populations (Tukey, 1977; Velleman and Hoaglin, 1981). The narrow centre of the box represents the median point in the data set for a variable. The upper portion of the TBP is the 75 th percentile of the data (population); the lower portion is the 25 percentile of the data (population). Circles, asterisks or vertical bars show the outliers. 102

125 Table A4-1. Analysis of variance of total vegetation cover in Revelstoke Reach and Arrow Lakes portions of the Arrow Lakes Reservoir, within three consecutive elevation bands: 434 to 436 metres, 436 to 438 metres, and 438 to 440 metres. Categorical values encountered during processing are: REACH$ (2 levels) Arrow, Revelstoke ELEVBAND$ (3 levels) , , Dep Var: TOTCOVER N: 549 Multiple R: Squared multiple R: Analysis of Variance Source Sum-of-Squares DF Mean-Square F-Ratio P REACH$ ELEVBAND$ REACH$*ELEVBAND$ Error Durbin-Watson D Statistic First Order Autocorrelation

126 Post hoc comparisons: Least Squares Means 63 TOTCOVER Arrow Revelstoke REACH$ No differences can be detected when both sections of the reservoir are considered, if elevation bands are not distinguished. Least Squares Means TOTCOVER ELEVBAND$ 104

127 Least Squares Means TOTCOVER TOTCOVER Arrow Revelstoke REACH$ 24 Arrow Revelstoke REACH$ TOTCOVER Arrow Revelstoke REACH$ Least Squares Means, however, indicates a significant difference among all three elevation bands. Cover increases with higher elevations. 105

128 Vegetation heights in Arrow Lakes vs. Revelstoke Reach Portions of the Arrow Lakes Reservoir Average Maximum Vegetation Height (m) A m A m A m R m R m R m Reach and Elevation Band Figure A4-2. Total Heights of Vegetation in the three Elevation Bands (434 to 436 metres, 436 metres to 438 metres, and 438 to 440 metres) in the Arrow Lakes (A) and Revelstoke Reach (R) portions of the Arrow Lakes Reservoir. Data are plotted as Tukey s boxplots (TBP). TBP are a visual method of showing statistically significant differences (p<0.05) between data distributions from comparably sampled populations (Tukey, 1977; Velleman and Hoaglin, 1981; Wilkinson et al., 1992). The narrow centre of the box represents the median point in the data set for a variable. The upper portion of the TBP is the 75 th percentile of the data (population); the lower portion is the 25 th percentile of the data (population). Circles, asterisks or vertical bars show the outliers. 106

129 Table A4-2. Analysis of variance of total vegetation heights in Revelstoke Reach and Arrow Lakes portions of the Arrow Lakes Reservoir, within three consecutive elevation bands: 434 to 436 metres, 436 to 438 metres, and 438 to 440 metres. Categorical values encountered during processing are: REACH$ (2 levels) Arrow, Revelstoke ELEVBAND$ (3 levels) , , Dep Var: AVGMAXHT N: 553 Multiple R: Squared multiple R: Analysis of Variance Source Sum-of-Squares DF Mean-Square F-Ratio P REACH$ ELEVBAND$ REACH$*ELEVBAND$ Error Least squares means. LS Mean SE N REACH$ =Arrow REACH$ =Revelstoke ELEVBAND$ = ELEVBAND$ = ELEVBAND$ = REACH$ =Arrow ELEVBAND$ = REACH$ =Arrow ELEVBAND$ = REACH$ =Arrow ELEVBAND$ = REACH$ =Revelstoke ELEVBAND$ = REACH$ =Revelstoke ELEVBAND$ = REACH$ =Revelstoke ELEVBAND$ =

130 Durbin-Watson D Statistic First Order Autocorrelation Least Squares Means AVGMAXHT Arrow Revelstoke REACH$ Least Squares Means AVGMAXHT ELEVBAND$ 108

131 Vegetation cover by vegetation community type (combined data for all plots, by VCT) 300 Total Vegetation Cover (%) BE BG CR PA PC PE RR Vegetation Community Type Figure A4-3. Total Cover of Vegetation by Community type for all combined data in both the Arrow Lakes and Revelstoke Reach portions of the Arrow Lakes Reservoir. BE = Beach, BG = Gravelly beach, CR = Cottonwood riparian forest, PA = Redtop upland, PC = Reed canary grass mesic, PE = Horsetail lowland, RR = Reed-rill. Data are plotted as Tukey s boxplots (TBP). TBP are a visual method of showing statistically significant differences (p<0.05) between data distributions from comparably sampled populations (Tukey, 1977; Velleman and Hoaglin, 1981; Wilkinson et al., 1992). The narrow centre of the box represents the median point in the data set for a variable. The upper portion of the TBP is the 75 th percentile of the data (population); the lower portion is the 25 percentile of the data (population). Circles, asterisks or vertical bars show the outliers. The Tukey s Boxplot indicates significant differences among VCTs. 109

132 Table A4-3. Analysis of variance of total vegetation cover of seven Vegetation Community Types in the combined Revelstoke Reach and Arrow Lakes portions of the Arrow Lakes Reservoir. Categorical values encountered during processing are: VCT$ (7 levels) BE, BG, CR, PA, PC, PE, RR Dep Var: TOTCOVER N: 506 Multiple R: Squared multiple R: Analysis of Variance Source Sum-of-Squares DF Mean-Square F-Ratio P VCT$ Error *** WARNING *** Case 341 is an outlier (Studentized Residual = 4.178) Case 477 is an outlier (Studentized Residual = 4.445) Durbin-Watson D Statistic First Order Autocorrelation ANOVA indicates a significant difference between total vegetation covers in seven VCTs Post-hoc comparisons among VCTs COL/ ROW VCT$ 1 BE 2 BG 3 CR 4 PA 5 PC 6 PE 7 RR Using least squares means. Post Hoc test of TOTCOVER

133 Using model MSE of with 499 DF. Matrix of pairwise mean differences: Bonferroni Adjustment. Matrix of pairwise comparison probabilities: BE differs from all except BG. BG differs from all other except BE. CR differs from all others. PA differs from BE, BG, CR, PC. PC differs from all except PE. PE differs from all except PA and PC. RR differs from all except PA. 111

134 Post Hoc Comparisons: Least Squares Means 116 TOTCOVER BG CR PA PC PE RR VCT$ Figure A4-4. Least Squares Means of total cover in BG: Gravelly beach, CR: Cottonwood riparian, PA: Redtop upland, PC: Reed canarygrass mesic, PE: Horsetail lowland and RR: Reed rill. 112

135 Average Maximum Vegetation by Height by vegetation community type (combined data for all plots, by VCT) Average Maximum Vegetation Height (m) BE BG CR PA PC PE RR Vegetation Community Type Figure A4-5. Total maximum heights of vegetation in each of the main vegetated VCTs in the Arrow Lakes Reservoir drawdown zone. Data are plotted as Tukey s boxplots (TBP). TBP are a visual method of showing statistically significant differences (p<0.05) between data distributions from comparably sampled populations (Tukey, 1977; Velleman and Hoaglin, 1981; Wilkinson et al., 1992). The narrow centre of the box represents the median point in the data set for a variable. The upper portion of the TBP is the 75 th percentile of the data (population); the lower portion is the 25 percentile of the data (population). Circles, asterisks or vertical bars show the outliers. 113

136 Table A4-4. Analysis of variance of total vegetation heights of seven vegetation community types in the Arrow Lakes Reservoir. Categorical values encountered during processing are: VCT$ (7 levels) BE, BG, CR, PA, PC, PE, RR Dep Var: AVGMAXHT N: 506 Multiple R: Squared multiple R: Analysis of Variance Source Sum-of-Squares DF Mean-Square F-Ratio P VCT$ Error Clear, significant differences in height between VCTs. Post-hoc comparisons: Durbin-Watson D Statistic First Order Autocorrelation COL/ ROW VCT$ 1 BE 2 BG 3 CR 4 PA 5 PC 6 PE 7 RR Using least squares means. Post Hoc test of AVGMAXHT Using model MSE of with 499 DF. Matrix of pairwise mean differences:

137 Bonferroni Adjustment. Matrix of pairwise comparison probabilities: BE differs from CR and PA. BG differs from CR and PA. CR differs from all others. PA differs from all others. PC differs from CR and PA. PE differs from CR and PA. RR differs from CR and PA. 115

138 Least Squares Means 3 2 AVGMAXHT BG CR PA PC PE RR VCT$ CR is significantly taller than all others. PA is significantly taller than all others except CR. No other differences in average maximum height exist. Least Squares Means 3 2 AVGMAXHT BG CR PA PC PE RR VCT$ CR is significantly taller than all others; PA is significantly taller than all others. Except for CR, no other differences in average maximum height exist. 116

139 ANOVAs for Cover in Type by Reach and Year For BE, all data for 2007 were excluded because of only a single measured cover value in this type in Revelstoke in Source Sum-of-Squares DF Mean-Square F-Ratio P REACH$ YEAR REACH$*YEAR Error No significant differences were found for Total Vegetation Cover between Reaches or Years. There was no interaction effect as the effects of Reach and Year were independent. For CR, all data for 2007 were excluded because of only a single measured cover value in this type in Revelstoke in Source Sum-of-Squares DF Mean-Square F-Ratio P REACH$ YEAR REACH$*YEAR Error Significant differences in total cover by year were indicated. Post hoc comparison showed total cover in 2010 was significantly greater than in 2008, which was significantly greater than in 2009 (i.e > 2008 > 2009). There were no significant differences in total cover due to the effects of Reach or Reach x Year. For PA, all data in all years were used. Source Sum-of-Squares DF Mean-Square F-Ratio P REACH$ YEAR REACH$*YEAR Error Significant differences in total cover by year were indicated. Post hoc comparison showed total cover ranked from highest to lowest: 2007 > 2008 > 2010 > There were no significant differences in total cover due to the effects of Reach or Reach x Year. For PC, all data in all years were used. Source Sum-of-Squares DF Mean-Square F-Ratio P REACH$ YEAR REACH$*YEAR

140 Error Significant differences in total cover by year were indicated. Post hoc comparison showed significantly higher cover in 2007 > 2008 > 2010 > There was a significant interaction effect of Reach and Year on total cover. Post hoc comparison of interaction least square means showed that total cover was greatest in Rev2007 > Arrow2007 > Arrow2008 = Rev2008 = Arrow2010 > Rev2010 > Arrow2009 = Rev2009. There were no significant differences in total cover due to the effects of Reach alone. For PE, all data for 2008 were excluded because of no measured cover values for this type in Revelstoke in Source Sum-of-Squares DF Mean-Square F-Ratio P REACH$ YEAR REACH$*YEAR Error Significant differences in total cover by year were indicated. Post hoc comparison showed total cover ranked from highest to lowest: 2007 > 2008 > 2010 > Thereh were no significant differences in total cover due to the effects of Reach or Reach x Year. ANOVAs for Height in Type by Reach and Year For BE, all data in all years were used. Source Sum-of-Squares DF Mean-Square F-Ratio P REACH$ YEAR REACH$*YEAR Error No significant differences were found for Average Heights between Reaches or Years. There was no interaction effect as the effects of Reach and Year were independent. For CR, all data for 2007 were excluded because of only a single measured height value in this type in Revelstoke in Source Sum-of-Squares DF Mean-Square F-Ratio P REACH$ YEAR REACH$*YEAR Error Significant differences in average height by year were indicated. Post hoc comparison showed average heights significantly greater in 2010 than in 2008 or 2009, i.e., 2010 > 2008 = There were no significant differences in total cover due to the effects of Reach or Reach x Year. 118

141 For PA, all data in all years were used. Source Sum-of-Squares DF Mean-Square F-Ratio P REACH$ YEAR REACH$*YEAR Error Significant differences in average height by year were indicated. Post hoc comparison showed height ranked from highest to lowest: 2009 > 2010 > 2008 = There were no significant differences in total cover due to the effects of Reach or Reach x Year. For PC, all data in all years were used. Source Sum-of-Squares DF Mean-Square F-Ratio P REACH$ YEAR REACH$*YEAR Error No significant differences in average height were indicated due to Year, Reach or Reach x Year interactions. For PE, all data for 2008 were excluded because of no measured cover values for this type in Revelstoke in Source Sum-of-Squares DF Mean-Square F-Ratio P REACH$ YEAR REACH$*YEAR Error Significant differences in average height by reach and by year were indicated. Post hoc comparison showed average heights were significantly greater in Revelstoke Reach than in Arrow Lakes. Post hoc comparison of average height by year resulted in the following rankings: 2007 > 2010 > There were no significant differences in total cover due to the effects of Reach x Year. 119

142 APPENDIX 5. TABLES OF SPECIES RICHNESS, SPECIES EVENNESS, AND SPECIES DIVERSITY. Table A5-1. Shannon-Weiner Diversity Index (H ) for each Vegetation Community Type (VCT) from 2007 to 2010, in the Arrow Lakes Reservoir (both geographic areas combined). Vegetation Community Shannon-Weiner Diversity Index (H') BE BG CR PA PC PE RR Table A5-2. Pielou s species evenness (J) measures for dominant and codominant species of each Vegetation Community Type (VCT) in the Arrow Lakes Reservoir (both geographic areas combined). Vegetation Community Pielou's Evenness (J) BE BG CR PA PC PE RR

143 Table A5-3. Species Richness measures for dominant and co-dominant species of each Vegetation Community Type (VCT) in the Arrow Lakes Reservoir (both geographic areas combined). Vegetation Community Species Richness BE BG CR PA PC PE RR Table A5-4. Shannon-Weiner Diversity Index (H ) for each Elevation Band from 2007 to 2010 in the Arrow Lakes Reservoir (both geographic areas combined). Elevation Band Shannon-Wiener Diversity Index (H') to 436 m to 438 m to 440 m Table A5-6. Pielou s species evenness (J) measures for all vegetation within the three elevation bands ( metres, metres and metres) from 2007 to 2010 in the Arrow Lakes Reservoir (both geographic areas combined). Elevation Band Pielou's Evenness (J) to 436 m to 438 m to 440 m Table A5-6. Species Richness Elevation Band from 2007 to 2010 in the Arrow Lakes Reservoir (both geographic areas combined). Elevation Band Species Richness

144

145 APPENDIX 6. DEFINITION OF EXPOSURE REGIME DUMMY VARIABLES Table A6-1 describes the actual values of the various exposure regime variables, which were replaced by simple coded values (ExposureRegime) to eliminate the 16 additional variables which would otherwise have unnecessarily complicated the RDA and RMA analyses. These 16 variables did not have any variation within a given combination of weather and elevation band, which is what allowed them to be replaced by a single value for ExposureRegime. Table A6-1. Table of values describing the various aspects of the six exposure regimes constructed from weather data for the Arrow Lakes Reservoir. Weather Revelstoke Nakusp ElevBand TotSub TotExp SubDate1_ Jun Jun Jun Jun Jun-06 SubDate2_ Jun Jun Jun Jun-07 AvgDTemp Jun-06 MaxDTemp MinDTemp TotPrecip TotSub TotExp SubDate1_ Jun Jun Jun Jun-09 SubDate2_ Jun Jun Jun Jun Jun-10 AvgDTemp Jun-10 MaxDTemp MinDTemp TotPrecip ExposureRegime R1 R2 R3 N1 N2 N3 123

146 Table A6-2. Table of variables from Table A6-1 with definitions. Variable Weather Definition Weather is the name of the weather station that applies to the polygon TotSub07 Total number of submersions from May 1, 2006 to Jun 30, 2007 TotExp07 Total number of days exposed (not submerged) from May 1, 2006 to Jun 30, 2007 SubDate107 Date of first submersion after May 1, 2006 SubDate207 Date of second submersion (if exists) AvgDTemp07 Average daily temperature during period(s) of exposure MaxDTemp07 Maximum daily temperature during period(s) of exposure MinDTemp07 Minimum daily temperature during period(s) of exposure TotPrecip07 Total mm of precipitation (rain and snow) during period(s) of exposure TotSub10 Total number of submersions from May 1, 2009 to Jun 30, 2010 TotExp10 Total number of days exposed (not submerged) from May 1, 2009 to Jun 30, 2010 SubDate110 Date of first submersion after May 1, 2009 SubDate210 Date of second submersion (if exists) AvgDTemp10 Average daily temperature during period(s) of exposure MaxDTemp10 Maximum daily temperature during period(s) of exposure MinDTemp10 Minimum daily temperature during period(s) of exposure TotPrecip10 Total mm of precipitation (rain and snow) during period(s) of exposure

147 APPENDIX 7. RDA ANALYSIS WITH ELEVATION Data The dataset consisted of 609 map observations or the year There were 466 observations taken from mapped polygons in the Arrow Reach and 143 in the Revestoke Reach (Table 1). Table A7-1. Vegetation Community Type (VCT) Number of map observations in 2010 by Reach, VCT and Elevation Band (code 1, 2 or 3). Data records from some sample original polygons (290) are duplicated. 10 Arrow Reach Elevation Bands Total Revelstoke Reach Elevation Bands BE BG CR PA PC PE RR Total Total There were two vegetation variables in the dataset: height and cover; these were the dependent variables (Y). There were nine environmental variables: aspect, slope, % clay, % silt, % sand, % gravel, % boulders, elevation and wave action. 11 They were quantitative, except for elevation and wave action which were qualitative. Elevation had three categories: m, m and m (coded 1, 2 and 3, respectively). Influence from water movement or water behavior had seven categories coded as W = extreme wave action, SH = sheltered from wave action, S = seepage or ground water nearby, R = influenced by incoming river, L = leeside of current, EX = exposed to ripping and E = small stream influence. Dummy variables were created for the elevation and wave-action categories. Thus, there were altogether 17 potential independent variables (X), including the dummy variables, for further analysis. Analysis 10 Each original complex polygon was assigned a Vegetation Community Type (VCT), and vegetation and environmental attributes. Approximately one-third of the polygons were then split into up to three elevation bands. Thus, in the database, any split complex polygon has two or three duplicate records of all the data, except the elevation band code. The polygon data may also be repeated, in a small number of cases, where the same complex polygon was sampled for change in a second different VCT (in the same original polygon). 11 BC Hydro requested that elevation be explicitly included in the RDA analysis, as it is an important variable in addressing the management questions. The Exposure Regime variable (elevation level) used earlier was dropped from the revised analysis, as this variable is virtually the same as elevation band, if the data are analyzed separately by reach. In addition, K. Enns requested that wave action (described in the text above) also be included in the analysis. 125

148 Canonical Redundancy Data Analysis (RDA) was used to relate the polygon vegetation attributes (Y: height and cover) to the environmental attributes (X: soil, site and wave action). The RDA analyses were done separately for each reach, and combined reaches, using the following steps implemented in the SAS computer program (version 9.2): Examined the scatter plots of Y versus X to examine the variation and any relationships among the variables. 2. Standardized the quantitative variables to have mean zero and standard deviation one. 3. Conducted a multiple regression with forward selection of Y versus X. The purpose of this regression analysis was to attempt to reduce the potential number of the independent variables X to be used in the RDA. 4. Performed canonical RDA analysis using the standardized values in step 2, to relate Y to X, that is, to examine how well the original Y variables could be predicted from the canonical variables (from the selected X independent variables). The outputs from this step included normalized eigenvectors, amount of variation explained by the canonical variables corresponding to X. 5. Created biplots from the RDA outputs of the normalized eigenvectors (scores of vegetation and environmental variables), for the two principal axes. Results & Interpretation Scatter plots The scatter graphs relating vegetation cover and height to the environmental variables showed high variation in the data (not shown here). Forward selection The forward selection process reduced the number of potential independent variables in the Arrow from 17 to 13, in Revelstoke from 17 to 14, and in the combined reaches from 17 to 14. Excluded by the forward selection process were the dummy variables representing the low and middle elevation bands and the wave action categories; all the soil variables were kept. RDA analysis The results of the RDA analyses were similar in the Arrow and Revelstoke for the soil and elevation variables, and not much different from the results of the earlier analysis (in the body of this report) that used the Exposure Regime variable. The results of the analyses are summarized in biplots in Figures 1 and 2 (Arrow), Figures 3 and 4 (Revelstoke) and Figures 5 and 6 (combined reaches). Arrow Lakes portion of the reservoir In the Arrow, the selected environmental variables (soil, site, high elevation and wave action variables) explained approximately 35 per cent of the variation in vegetation cover and height. Vegetation cover was strongly negatively associated with sand and gravel, and vegetation height was positively associated with most soil and high elevation variables except silt, with which it is negatively correlated (Figure 1). This is because silt dominated communities tend to support species that are short in stature, such horsetails, 12 SAS Institute Inc., Cary, NC, USA. 126

149 rushes and reeds. Elevation was positively associated with vegetation height at the high elevation ( m), but not at the low elevation. Elevation was not associated with vegetation cover at any elevation level, as cover is very variable, and can be high at low elevation if conditions are right for vegetation, but also low, of scouring has occurred or if sands are common. Sand and gravel deposits are common at both high and low elevation. Vegetation cover was positively associated with S (seepage) and R (river entry), which is an indication of the importance of water supplies to plants. The vegetation height did not appear to be related to the wave action variables (Figure 2). Revelstoke Reach portion of the reservoir In Revelstoke, the selected environmental variables explained approximately 44 per cent of the variation in vegetation cover and height. Vegetation cover was strongly negatively associated with sand and gravel, similar to the Arrow Lakes portion of the reservoir and vegetation height was positively associated with most environmental variables except silt, which was negatively correlated with height (Figure 3). As in the Arrow Lakes portion of the reservoir, silts tend to accumulate at low elevation, and support species with low stature. Elevation was positively associated with vegetation height at the high elevation ( m) and negatively at the low elevation ( m). Elevation was positively associated with vegetation cover at the low elevation, but not at the high elevation. Vegetation cover was positively associated with R (seepage and moisture availability, including from streams and rivers) and S (seepage with some scouring), and negatively associated with W (wave action) and E (high water effects only). Vegetation height appears negatively associated with R (river entry) (Figure 4). Combined Arrow Lakes and Revelstoke Reach For the combined data, the environmental variables explained approximately 34 per cent of the variation in vegetation cover and height. The relationships of the vegetation and environmental variables (Figures 5 and 6) were similar to those observed in the separate Arrow Lakes and Revelstoke Reach analyses. Vegetation cover was strongly negatively associated with sand and gravel, and vegetation height was positively associated with most environmental variables except silt, with which it is negatively correlated (Figure 5). Elevation was positively associated with vegetation height at the high elevation ( m) and negatively at the low elevation ( m). Elevation was positively associated with vegetation cover at the low elevation, but not at the high elevation. Vegetation cover was positively associated with SH (sheltered), and vegetation height appeared negatively associated with EX (exposed promontory) (Figure 6). 127

150 cover silt clay 0 High E slope boulder aspect gravel sand height -1 Figure A7-1. RDA biplot for the Arrow portion of the reservoir: soil and elevation variables standardized eigenvectors or scores. The lines with the arrows are the dependent variables (vegetation height and cover) and the remaining lines show the independent variables (soil, site and elevation). Vegetation cover is strongly negatively associated with gravel and sand; and vegetation height is positively associated with most soil variables, except silt, with which it is strongly negatively correlated. Elevation appears to be positively associated with vegetation height at high elevation ( m), but elevation is not associated with vegetation cover. 128

151 cover SH S L 0 R EX height -1 Figure A7-2. RDA biplot for the Arrow portion of the reservoir: wave action variables standardized eigenvectors or scores. The lines with the arrows are the dependent variables (vegetation height and cover) and the remaining lines show the independent variables (wave action). Vegetation cover is positively associated with SH (sheltered topography), S (seepage with some scouring,) and R (river entry); and vegetation height does not appear to be related to the wave action variables. 129

152 cover Low E slope clay 0 silt aspect height boulder sand -0.8 gravel -1 Figure A7-3. RDA biplot for Revelstoke Reach: soil and elevation standardized eigenvectors or scores. The lines with the arrows are the dependent variables (vegetation height and cover) and the remaining lines show the independent variables (soil, site and elevation). Vegetation cover is strongly negatively associated with gravel and sand; and vegetation height is positively associated with most environmental variables, except silt, with which it is strongly negatively associated. Elevation is associated positively with vegetation height at high elevation ( m) and negatively at low elevation ( m); elevation is positively associated with vegetation cover at low elevation ( m), but not at high elevation. 130

153 cover L R S 0 height E W Figure A7-4. RDA biplot for Revelstoke Reach: wave action variables standardized eigenvectors or scores. The lines with the arrows are the dependent variables (vegetation height and cover) and the remaining lines show the independent variables (wave action). Vegetation cover is positively associated with R (river entry) and S (seepage with some scouring), and negatively associated with W (wave action) and E (high water effects). Vegetation height appears negatively associated with R (river entry). 131

154 cover silt clay High E 0 slope boulder aspect gravel sand height Figure A7-5. RDA biplot for the combined Arrow portion and Revelstoke Reach: soil and elevation standardized eigenvectors or scores. The lines with the arrows are the dependent variables (vegetation height and cover) and the remaining lines show the independent variables (soil, site and elevation). Vegetation cover is strongly negatively associated with gravel and sand; and vegetation height is positively associated with most soil and elevation variables, except silt, with which it is strongly negatively corrected. Elevation is positively associated with vegetation height at high elevation ( m), but elevation is not associated with vegetation cover. This negative association between height and silt is because silts tend to occur at low elevations where heights are almost always lower (e.g., PE, BE, BG). Silts accumulate at low to middle elevations, and these materials only support shorter vegetation types that are adapted to the low elevation silt communities. Gravels and sands accumulate at high elevation, and at low elevation; at high elevation they will support water intolerant, drought tolerant tall shrubs. At low elevation, sands and gravels are mostly bare, due to the effects of scouring and duration of inundation. 132

155 cover SH L 0.2 S R E 0 EX height Figure A7-6. RDA biplot for the combined Arrow portion and Revelstoke Reach: wave action variables standardized eigenvectors or scores. The lines with the arrows are the dependent variables (vegetation height and cover), and the remaining lines show the independent variables (wave action). Vegetation cover is positively associated with SH (sheltered); height appears negatively associated with EX (exposed promontory). 133

156 APPENDIX Tamarack Street Castlegar, B.C. Canada V1N 2J2 Phone: Fax: IMAGE ANALYSIS PILOT STUDY DATE: October Eva-Maria Boehringer, M.Sc., MRM, RPBio Natural Resource Specialist BC Hydro Generation Water License Requirements th Street Castlegar, B.C. V1N 2N1 Phone: (PAX 73416) Fax: (PAX 78390) Cell: RE: Image analysis pilot study for CLBMON-33 Dear Ms. Boehringer, A pilot study of the use of digital image analysis (DIA) to change in the vegetation of the Arrow Lakes Reservoir was proposed in It was not possible to use DIA to compare the 2007 to the 2008 film imagery, due to differences in color and lack of complete coverage between these two capture years. It is possible to use the 2007 baseline imagery to compare to the 2010 image and to pixel data for future years, however. In 2010, the digital aerial photographic files were collected and a trial change detection exercise was completed for the imagery for Burton, north of the Needles ferry in the central portion of the Arrow Lakes portion of the Arrow Lakes Reservoir. The Burton study area is a large, undulating alluvial fan with patches of both homogenous and variable vegetation. The interpretation of this site is supported by ample field work and ground photography. An example of the 2007 and 2010 image with a specific principle components analysis enhancement of the two images is shown in Figure1. 134

157 Figure 1. Burton image dated circa May 30 th, 2007 (top), May 12, 2010, and principle component analysis enhancement of the two captures (bottom). Arrow denotes a change in the vegetation. Circle shows gravels and sands. A mosaic line appears as a dark green artifact in the lower left side of the bottom photograph. 135