A Multi-Country Assessment of Vegetation Dynamics, Soil Erosion, and Watershed Degradation After Wildfires

A Multi-Country Assessment of Vegetation Dynamics, Soil Erosion, and Watershed Degradation After Wildfires Item Type text; Proceedings Authors Neary, Daniel G.; Orr, Barron; Leeuwen, Wim; Aguilar, Susana Bautista; Wittenberg, Lea; Carmel, Yohay Publisher Arizona-Nevada Academy of Science Journal Hydrology and Water Resources in Arizona and the Southwest Rights Copyright, where appropriate, is held by the author. Download date 01/09/2018 00:15:13 Link to Item http://hdl.handle.net/10150/296612

A MULTI -COUNTRY ASSESSMENT OF VEGETATION DYNAMICS, SOIL EROSION, AND WATERSHED DEGRADATION AFTER WILDFIRES Daniel G. Neary,l Barron Orr,2 Wim Leeuwen,2 Susana Bautista Aguilar,3 Lea Wittenberg,4 and Yohay Carmel5 Abstract. Wildfires in semi -arid regions can induce vegetation changes that alter fire frequency, increase erosion, and lead to general watershed degradation (Allen et al. 1998; DeBano et al. 1998). This paper describes an International Arid Lands Consortium project in Arizona, Spain, and Israel that is examining the use of MODIS satellite imagery and ground -based verifications to develop a tool to evaluate vegetation changes, erosion, and landscape degradation after wildfires (Justice et al. 1998). Site -specific GIS data, erosion rates, and vegetation changes are being used to validate MODIS -based models on the Indian and Rodeo - Chediski wildfire sites in Arizona, the Guadalest, Calderona, and Millares wildfire sites in Spain, and several unnamed wildfires on Mount Carmel, in Israel. LANDSCAPE DEGRADATION PROJECT Drought, wildfire, and precipitation events can have a devastating impact on the sustainable use of dryland resources. The primary objective of the research project described in this paper is to integrate seasonal and geo -spatial vegetation and climate data sources (e.g. remotely sensed data) and erosion plot data in a land -degradation assessment model to determine rates of land degradation and recovery after wildfire events. A geo -spatial land - degradation information system is being used to develop seasonal assessment maps of land degradation potential for sites in the USA, Spain, and Israel. Soil erosion data from field plots at the different sites will be used to evaluate the framework of land degradation that is represented by the Revised Universal Soil Loss Equation model (Inbar 1USDA Forest Service, Rocky Mtn. Research Station, Flagstaff, Arizona 2Office of Arid Lands Studies, Univ. of Arizona, Tucson 3Department of Biology, University of Alicante, Spain 4Department of Geography, Haifa University, Israel 5Technion Israel Institute of Technology, Haifa et al. 1998; Gottfried et al. 2003; Soler and Sala 1992). The system incorporates relief and soil data in combination with geo- spatial and temporal precipitation and remotely sensed vegetation cover data for sites that were affected by recent fire events (2000-2003). The landscape degradation model will function as a conceptual framework that will be evaluated in terms of regeneration capacity after fire events (Naveh and Dan 1973; Swanson 1981). The model will also allow us to generate seasonally representative maps of land degradation and erosion risk at the regional scale, which will make it possible to focus on regions at greatest risk. These sensitive or vulnerable sites should have priority for both fire prevention and post -fire mitigation actions in case of fire. The regional soil loss potential maps provide improvements in four areas: (1) delineation of the regions at greatest risk, (2) assessment of the relative importance of factors contributing to land degradation, (3) assessment of vulnerability to further degradation, and (4) a new perspective for the development of options for more sustainable land use. Both scientists and decision makers can use the resulting framework to assess the ability of landscapes to support natural and human land use. GOAL AND OBJECTIVES The primary goal of this research is to enhance the ability of fire managers, land managers, and natural resource staff to assess the temporal vegetation dynamics that influence landscape vulnerability to precipitation and wildfire in fire -prone dryland ecosystems and predict post -fire soil erosion risk. The project objectives are the following: (1) Test the hypothesis that enhancing temporal resolution (dynamic versus static, or biweekly instead of annual) will improve the accuracy -and thus the utility -of current erosion models. These models include the Universal Soil Loss Equation

38 Neary, Orr, Leeuwen, Bautista Aguilar, Wittenberg, and Cannel (USLE) and the Revised Universal Soil Loss Equation (RUSLE), as well as other versions. (2) Extend and evaluate the model of choice, which was successfully applied in the U.S. Southwest, to other semi -arid regions. This comparative part will elucidate similarities and dissimilarities between remote regions, which in turn may culminate in (a) a better understanding of the determinants of post - fire erosion and vegetation regeneration, and (b) better mapping of predicted high- erosion -risk areas at a large scale. METHODS This International Arid Lands Project plans to achieve its objectives through the following steps: 1. Assessing the temporal dynamics of vegetation on selected erosion and wildfire research sites in fire -prone arid and semi -arid regions of the United States, Spain, and Israel through biweekly vegetation indices derived from NASA MODIS satellite imagery (250 m spatial resolution). 2. Understanding how precipitation patterns (through time and space) influence the vegetation dynamics, through spatially interpolated precipitation data from a rain gage network associated with each study area and the multi- temporal vegetation indices. 3. Assessing vegetation regeneration capacity (by vegetation type) through time on burned areas, through a combination of ground -based vegetation assessments and analysis of the vegetation indices relative to a baseline. 4. Relating these vegetation dynamics in both burned and unburned areas to soil loss estimates obtained through the erosion /vegetation regeneration model developed. 5. Integrating these elements and addressing plot to regional scaling issues to map landscape vulnerability to precipitation and wildfire. APPROACH The following steps are being taken to meet the objectives outlined above: Logistics for a Tri- National Project Involving Five Research Teams The project team has implemented a number of measures to ensure efficient communication and to facilitate data sharing. Subcontracts have been established with three institutions: the University of Alicante in Spain, the University of Haifa in Israel, and the Technion Israel Institute of Technology, also in Haifa. Periodic international teleconferences will keep team members updated. The project has also established a website ( http : / /wildfire.arid.arizona.edu /) and a dedicated project data server with a Secure Shell File Transfer client. Site visits have been completed to the Indian Fire, Prescott National Forest, Arizona; the Rodeo -Chediski Fire, Apache -Sitgreaves National Forest, Arizona; the Guadalest Fire, Alicante Region, Spain; the Millares Fire, Alicante Region, Spain; the Calderona Fire, Valencia Region, Spain, and two unnamed fires on Mt. Carmel, Haifa, Israel. Data set acquisition meetings have been set up to obtain information (i.e. Prescott National Forest GIS Coordinator meeting), and an annual principal investigators workshop will further project progress (December 15-19, 2004, Alicante, Spain; and March 14-19, Haifa, Israel). Data format protocols have been developed to maintain continuity between countries. Study Site Delineation and the Compilation of Existing Data Sets The Arizona, Israel, and Spain teams have refined their selection of study sites to include areas that have burned since the original proposal was written. Sites in Arizona include the Indian Fire on the Prescott National Forest southwest of Prescott and the Rodeo -Chediski Fire on the Apache -Sitgreaves National Forest south of Heber and Show - low. In Spain, the wildfire sites selected for use in the study include Guadalest (Alicante Region), Millares (Alicante Region), and Calderona (Valencia Region). In Israel the study sites are in Mt. Carmel National Park (Haifa, Israel) Temporal Vegetation Dynamics Assessed Through Multi- Temporal Satellite Imagery The project team has obtained multi- temporal moderate resolution imaging spectroradiometer ( MODIS) data from 2000 to the present for tiles covering northern Arizona, southeastern Spain, and the Mt. Carmel region of Israel (Justice et al. 1998). The primary MODIS product of interest for this study is the biweekly Normalized Difference Vegetation Index (NDVI) at 250 m resolution. NDVI is successful as a vegetation measure in that it is sufficiently stable to permit meaningful comparisons of seasonal and inter -annual changes in vegetation growth and activity. Across the three countries this amounts to 200 gigabytes of satellite data needing modification into formats that can be analyzed in a geographic information system (GIS; Desmet and Grovers 1996). Project geo- spatial teams have tested an automated conversion of these data into formats that

Multi -Country Assessment of Watershed Degradation After Wildfire 39 can be used for the research. A team meeting in Alicante, Spain revealed the specific geographic and cartographic parameters necessary for each country and study sites. When these scripts have been modified, we will then automate the acquisition of new data as it becomes available throughout the life of the project. During the Alicante meeting the project teams agreed to analyze MODIS data to help understand vegetation dynamics before and after wildfire. This involves the creation of scripts to aggregate data for explicit spatial extents (e.g., the perimeter of a vegetation type or a sub -catchment). It also involves exploratory data analysis of the temporal trends in vegetation greenness to identify useful indicators (e.g., onset of greenness, peak greenness, length of growing season). This approach also involves the identification of suitable methods to assess inter -annual trends. These analyses will assist in the development of the modified, spatially explicit RUSLE model, recognition of vegetation regeneration patterns after wildfire, and post -fire vulnerability assessment. All Other Land Degradation and Post -Fire Vulnerability Data During the Alicante meeting, data sets for all key model parameters were identified by each project team member. These include precipitation (and associated rainfall erosivity parameters), elevation, soil erodibility parameters, vegetation measures (e.g. cover, biomass, species composition), and management parameters (e.g. terraces, experimental post -fire treatments). Individual studies of soil erosion, runoff, soil water repellency, and vegetation cover are being conducted at all sites independently of this project. Model Development Team members introduced and discussed the basic land degradation model during the 2004 team meeting in Alicante, Spain, as well as the modified, spatially explicit version of RUSLE. This revision of the USLE model incorporates temporally dynamic land cover and precipitation parameters, accumulated slope length and slope steepness parameters, soil erodibility parameters, and, potentially, a parameter to address management (e.g. terracing). A manual has been written for the team members to explain how the model functions and is run in a GIS environment (Desmet and Grovers 1996). AML scripts are included in this manual to allow all members to run the model at their GIS labs. Post -Fire Vulnerability Assessment: Field Methods and Conceptual Framework A rapid field -based post -fire vulnerability (to land degradation) assessment method was also discussed at the team meeting in Alicante. The method is based on a soil condition assessment methodology developed by David Tongway in Australia. Project members are working to adapt these methods to assess vegetation recovery after wildfire events. The adapted methods will initially be tested by the Spain team on the Calderona site and by the Israel team on data collected during related research on wildfires in Mt. Carmel National Park. The Spain team has been working with management agencies to develop a broader, spatially explicit, conceptual framework for mapping post - fire land degradation vulnerability. The conceptual framework developed by CEAM for Valencia will be evaluated for the potential inclusion of the modified, spatially explicit RUSLE being developed in this research. The Spain team also hopes to provide assistance in assessing vegetation regeneration. Ultimately, the project plans to use the modified, spatially explicit version of RUSLE to assess the factors of vulnerability identified in a broader conceptual framework being used for management in southeastern Spain. Validation and Verification: Erosion Factor Sensitivity Analysis Another outcome of the Alicante meeting in 2004 was a more detailed methodology for model validation and sensitivity analysis. Specific data sets were discussed that could be used for model validation and verification. Differences in the spatial and temporal patterns of model results will be validated in relative terms. The project team decided to conduct a sensitivity analysis of all erosion model factors to be used for each study area in order to examine the relative importance of each erosion factor under different conditions. The analysis will begin with an assessment of the literature on RUSLE as applied to natural vegetation. Next, individual parameters will be tested under the conditions of each study area. Erosion data for parameter testing will be drawn from field studies (e.g., erosion pins on the Rodeo -Chediski Fire site, silt fences installed at the Indian Fire site for different post -wildfire treatments, and small plot assessments of erosion on burned watersheds in Spain and Israel). The Spain team installed some silt fences at the Calderona

40 Neary, Orr, Leeuwen, Bautista Aguilar, Wittenberg, and Cannel site in 2005. Finally, model results are currently being compared to benchmark erosion data (i.e., what management agencies currently use to make decisions, such as the Terrestrial Ecosystem Survey "current erosion rate" and "soil loss tolerance" factors). EXAMPLE CASE STUDY ANALYSIS Remote sensing is a valuable tool for data visualization as well as a powerful analytical tool. The following case study demonstrates the use of satellite images for the visualization of post -wildfire vegetation dynamics coupled with graphical representations of quantitative data extracted from the imagery to illustrate a more analytical evaluation of change in vegetation dynamics after wildfire. On May 15, 2002, a wildfire started that eventually burned approximately 530 ha in the Prescott National Forest, Arizona, named the Indian Fire. This case study shows the region of the fire just prior to the burn, shortly after the burn, and for some time after the event. A series of images (Figures 1, 3, 5, 7, and 9) illustrate the impact of the fire. Each image is accompanied by a graph showing the entire time series of NDVI values for each of two polygons in the image (Figures 2, 4, 6, 8, and 10). The fire polygon (light gray in this figure and yellow in the original images) shows the burn perimeter, and the darker (blue in the original images) polygon shows original vegetation that was not burned. To calculate the NDVI value for each polygon the values of all pixels within each polygon were averaged for each time step, resulting in a single value that represents the average NDVI value for that polygon. The specific time step for each image shown is indicated on the graph by a dotted line, which includes the end date of the composite period shown in the image. The image and the phenology graph for the period ending on January 16, 2002 show that both the reference area and the area that will burn later in the season have more actively photosynthesizing vegetation than at any time for the period of record shown here (April 2000 to February 2005). The amounts of vegetation for both areas is similar, as demonstrated by examination of the image and as summarized on the chart as the average NDVI of each area. The figures for May 24, 2002 illustrate the consistent decline in actively photosynthesizing vegetation over the nearly 5 -month period immediately before the fire. Note the decline in NDVI values in the graph and the low vegetation illustrated in the image as compared with the January 16 image. This most likely indicates that the large amount of vegetation present in January had dried up just prior to the fire. Also notice that the areas are still fairly close in NDVI levels, with the area to be burned having slightly higher levels of NDVI than the reference area. It is important to note that the Indian Fire started on May 15th, but because the 16-day NDVI composites choose the highest NDVI values to include in the composite, the fire cannot be detected in these data until the next composite. Figure 5 is the first image where the wildfire is visible. The composite covers the period from May 25 through June 9, 2002. In the image, one can see that the western portion of the burned region was more severely burned than the eastern portion. This illustrates a limitation in the phenology graph, in that it gives only an average value for the entire polygon, with no spatially explicit information included. In Figure 7, one year after the initial image, the difference between the peak vegetation of the burned and unburned areas can be clearly seen. The vegetation difference is still clear 2 years after the fire (Figure 9), although the phenology curves indicate that the gap between the two areas may be beginning to close (Figure 10). Investigation on the ground has shown, however, that although the burned region is being revegetated, it is not returning quickly to its prior state of pine forests; rather, it is moving to a chaparral vegetation type. It may be a long time before the vegetation curves in the two polygons resemble one another, if ever. SUMMARY AND CONCLUSIONS This paper summarizes an International Arid Lands Consortium study in the USA, Spain, and Israel that is using MODIS satellite imagery and ground -based verifications to develop a tool to evaluate vegetation changes, erosion, and landscape degradation after wildfires. Wildfires have become major forces influencing Mediterranean and western USA landscapes. Preliminary results indicate that this GIS -based tool has potential to give land managers a new tool for assessing post - wildfire changes in erosion and vegetation type over large landscapes. ACKNOWLEDGMENTS The authors would like to acknowledge assistance from graduate students Grant Casady, University of Arizona, USA; Angeles Garcia Mayor, University of Alicante, Spain; and Naama Tesler and Ronel Barzilai, University of Haifa, Israel.

Multi-Country Assessment of Watershed Degradation After Wildfire 41 Figure 1. January 16, 2002, MODIS satellite image, Indian Fire, Prescott National Forest (left fire area; right control area). January 16. 20( 2 7000 o z 5000 3000 R0 0 4 Reference Fire Figure 2. Phenology graph for January 16, 2002, based on MODIS satellite images.

1 Reference 42 Neary, Orr, Leeuwen, Bautista Aguilar, Wittenberg, and Carmel Figure 3. May 24, 2002, MODIS satellite image, Indian Fire, Prescott National Forest (left fire area; right control area). -f- Fire Figure 4. Phenology graph for May 24, 2002, based on MODIS satellite images.

Multi -Country Assessment of Watershed Degradation After Wildfire 43 Figure 5. June 9, 2002, MODIS satellite image, Indian Fire, Prescott National Forest (left fire area; right control area). 9000 8000 7000 ôo _8 x 6000 > o z 5000 4000 3000 O o o o o o N N N N N N V) CY) C) CY) O) C ).4. 0 O O O O O O O O O O O O O O O O Cn in Ú) i) U) U7 i) L-75 Ln Cif ia ia in- 7 tá Cn to in in-.4 6 6 6 T T m 6 N à co 6 4 co -O T T T T T T Ó Cn Ó Ó Ó Ó at Ta a -0- Reference -a- Fire Figure 6. Phenology graph for June 9, 2002, based on MODIS satellite images.

44 Neary, Orr, Leeuwen, Bautista Aguilar, Wittenberg, and Carmel Figure 7. January 16, 2003, MODIS satellite image, Indian Fire, Prescott National Forest (left fire area; right control area). 8 O o o N N N N N N C9 C'7 M M co co 0 ó 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 o i"ri a r Ta at Ta a a a T aaaa T r r -. Reference ---0- Fire T T Ó á Figure 8. Phenology graph for January 16, 2003, based on MODIS satellite images.

Multi -Country Assessment of Watershed Degradation After Wildfire 45 Figure 9. June 8, 2003, MODIS satellite image, Indian Fire, Prescott National Forest (left fire area; right control area). 9000 8000 7000 8 o _o x 6000 > O Z 5000 4000 3000 ó ó ó ó ó ó ó ó ó 0 0 0 0 0 0 ó ó ó ó ó ó ó ó ó CO C') iñ u in ñ ñ ñ ñ in n ñ n n n n n ir) in n ñ ñ n Ln ñ í a m a a a a a a a a a a a a a a d a a a a a4 a a a - -- Reference -6- Fire Figure 10. Phenology graph for June 8, 2003, based on MODIS satellite images.

46 Neary, Orr, Leeuwen, Bautista Aguilar, Wittenberg, and Carmel LITERATURE CITED Allen, C. D., J. L. Betancourt, and T. W. Swetnam. 1998. Landscape changes in the southwestern United States: Techniques, long -term data sets and trends. In Perspectives on the Land Use History of North America: A Context for Understanding Our Changing Environment, edited by T. D. Sisk, pp. 71-84. Biological Science Report USGS/BRD/BSR-1998-0003. U.S. Geological Survey, Biological Resources Division, Reston, VA. DeBano, L. F. D. G. Neary, and P. F. Ffolliott. 1998. Fire's Effects on Ecosystems. John Wiley & Sons, New York. 333 p. Desmet, P. J.J., and G. Grovers. 1996. A GIS procedure for automatically calculating the USLE LS factor on topographically complex landscape units. J. Soil and Water Cons. 51(5): 427-433. Gottfried, G. J., D. G. Neary, M. B. Baker, Jr., and P. F. Ffolliott. 2003. Impacts of wildfires on hydrologic processes in forest ecosystems: Two case studies. In USDA Agricultural Research Service, The First Inter - agency Conference on Research in the Watersheds, 27-30 October 2003, Benson, AZ. Inbar, M., L. Wittenberg, and M. Tamir. 1998. Runoff and erosion processes after forest fire in Mount Carmel, a Mediterranean area. Geomorphology 24: 17-33. Justice, C., D. Hall, V. Salomonson, J. Privette, G. Riggs, A. Strahler, W. Lucht, R. Myneni, Y. Knjazihhin, S. Running, R. Nemani, E. Vermote, J. Townshend, R. Defries, D. Roy, Z. Wan, A. Huete, W. van Leeuwen, R. Wolfe, L. Giglio, J -P. Muller, P. Lewis, and M. Barnsley. 1998. The moderate resolution imaging spectroradiometer (MODIS): Land remote sensing for global change research. IEEE Transactions on Geoscience and Remote Sensing 36(4): 1228-1249. Naveh, Z., and J. Dan. 1973. The human degradation of Mediterranean landscapes in Israel. In Mediterranean Type Ecosystems, edited by F. D. Castri and H. A. Moony, pp. 373-390. Springer Verlag, New York. Soler, M., and M. Sala. 1992. Effects of fire and clearing in a Mediterranean Quercus ilex woodland: An example approach. Catena 19: 321-332. Swanson, F.. 1981. Fires and geomorphic processes. In Proceedings of Fire Regimes and Ecosystems Conference, Honolulu, pp. 401-420. General Technical Report WO-26 USDA, Washington, D.C.