Rangeland Carbon Fluxes in the Northern Great Plains

Transcription

1 Rangeland Carbon Fluxes in the Northern Great Plains Wylie, B.K., T.G. Gilmanov, A.B. Frank, J.A. Morgan, M.R. Haferkamp, T.P. Meyers, E.A. Fosnight, L. Zhang US Geological Survey National Center for Earth Resources Observation Science (EROS) USGS: Earth Surface Dynamics, Land Remote Sensing, and Geographic Analysis and Monitoring National Science Foundation/EPSCoR Grant No through the South Dakota Center for Biocomplexity Studies and by the State of South Dakota USAID GL-CRSP project Grant No. PCE-G to Univ. of California, Davis USDA Agricultural Research Service Agriflux network US Agency for International Development Grant No. PCE-G , Global Bureau, Office of Agriculture and Food Security U.S. Department of the Interior U.S. Geological Survey

2 Objectives Exhaustively model and map Carbon fluxes in rangelands to quantify the fluxes and Investigate temporal and patterns to understand climate and environmental influences.

3 METHODS: Develop Empirical Model Tree Algorithms to Predict Flux Tower Carbon Observations From Spatial Data Localized Detailed Measurements Spatial Variables Through Time C Flux Mapping Model Tree If NDVI <.4 C fluxes = f(x1, X2, X4) If NDVI >=.4 C fluxes = f(x1, X4, X6) C Flux Maps Northern Great Plains

4 Flux Tower Data is used to Develop Models for the Mapping of 1-day Carbon Dynamics on Rangelands Lethbridge (LE), Canada: Larry Flanagan Ft. Peck (FP), MT; Tilden Meyers Mandan (MA), ND; Al Frank New Towers Cottonwood, SD Miles City (MC), MT: Marshall Haferkamp Brookings, SD Cheyenne (CN), WY: Jack Morgan

5 METHODS: Tower C Fluxes Associated with Basic Ecosystem Functions of Gross Primary Production (Pg) and Total Ecosystem Respiration (Re) are Derived From Light Response Curves on 3 min. Datasets. Tagir Gilmanov, SDSU, coordinator of the FLUXNET WORLDGRASSFLUX working group N E E P(Q,T s ; α,β,γ,θ,κ )= 1 2θ αq + β (αq + β)2 4αβθQ γ eκt s, Derives estimates of daytime respiration independent of nighttime flux data (others often use relationships derived from nighttime fluxes and apply them to daytime observations) PROVIDES: 1) Tower Pg, Re, and NEE 2) Facilitates flux tower data gap filling 3) Quantifies ecophysiological parameters (quantum yield (α), maximum gross photosynthesis (β), & daytime respiration (γ))

6 4-May 15-Mar 25-Jan 6-Dec 23-Jun 9-Jan 2-Nov 1-Oct 12-Aug Ecosystem Process Carbon Partitioning: Added Value Re Bkwd_finalGPP_RE_NEEmodel4c.xls NEE Pg Source Sink g C m-2 day-1

7 8 14 Flux Tower Pg Group: Significant Early Rain, High Pg Pg (g C m-2/day-1) day precipitation (mm) CH 1998 Pg MA 1999 Pg CH 1998 ppt MA 1999 ppt 1-Jan 2-Feb 1-Apr 3-May 19-Jul 7-Sep 27-Oct 16-Dec 8 14 Flux Tower Pg Group: Moderate Rain, Moderate Pg Pg (g C m -2 /day -1 ) day precipitation (mm) FP 2 Pg MA 2 Pg MA 21 Pg MC 21 Pg FP 2 ppt MA 2 ppt MA 21 ppt MC 21 ppt 1-Jan 2-Feb 1-Apr 3-May 19-Jul 7-Sep 27-Oct 16-Dec Day of Year 8 14 Flux Tower Pg Group: Low Rain, Low Pg Pg (g C m-2/day-1) day precipitation (mm) LE 2 Pg LE 21 Pg MC 2 Pg MC 2 ppt LE 2 ppt LE 21 ppt 1-Jan 2-Feb 1-Apr 3-May 19-Jul 7-Sep 27-Oct 16-Dec Day of Year

8 Model Development Data Set Consists of Flux Tower Pg, Re, and NEE Combined With Spatial Variables from Remote Sensing and GIS (1-day Time Step) Variable name Seasonally Dynamic Solar Radiation nd ppt temp par Definition function of DOY and latitude Smoothed NDVI precipitation temperature PAR Seasonally Dynamic but dependent on time sincesost Days since Start of Season (NDVI metric) DOY day of year = photoperiodism solstice days prior or post summer solstice Annually Dynamic sosn sost tin year Site Dynamic sitecat c4pct co3surf pctclaysurf pctgrass omernick Elevation Aspect NDVI at Start of Season Start of Season DOY Time Integrated NDVI Categorical variable for testing year effects Categorical variable for testing site effects % C4 from STATSGO Surface calcium carbonates from STATSGO % clay in surface layer from STATSGO % grass from STATSGO Categorical variable for testing ecoregion effects

9 Tower versus Spatial Data - Takes into account the accuracy of spatial data sets when developing model. - Process-based models typically rely heavily on precipitation - Precipitation can have high spatial and temporal variability that is difficult to map accurately. Precipitation Temperature y = 1.331x R 2 =.4247 y = 1.171x R 2 =.9852 Tower Precipitation (mm) Tower Temperature ( o C) Mapped Precipitation (mm) Mapped Temperature ( o C)

10 Distinct Model Trees Models that predict: Net Ecosystem Exchange (NEE) Gross Primary Production (Pg) Respiration (Re)

11 Empirical Piecewise Regression Model Cubist Parsimonious and transparent selection of variables Stratification of data into homogeneous information spaces Multiple regression defined for each information space

12 NEE Model example (g CO 2 m -2 day -1 ) Rule 1: [98 cases, mean , range to , est err ] if nd <= 14 and sincesost > 85 then NEE = nd -.91 ppt -.19 sosn -.9 par -.1 temp -.15 sincesost Rule 2: [21 cases, mean , range to , est err ] if nd > 14 and par <= 1 then NEE = par -.4 temp +.33 nd -.3 sosn -.8 ppt -.14 sincesost Rule 3: [6 cases, mean , range to , est err ] if nd <= 14 and sincesost <= 85 then NEE = par -.46 temp -.54 tin +.43 nd -.4 sosn -.12 ppt -.19 sincesost

13 Rules or respective information spaces Flux Tower Training Data Set and Respective NEE Rules 4 3 Tower NEE (g C m -2 day -1 ) rule 1, Rel. Err..11 rule 2, Rel. Err..24 rule 3 Rel. Err..16 rule 4 Rel. Err..25 rule 5 Rel. Err..22 April - Oct May 15-Mar 25-Jan 6-Dec 23-Jun 28-Feb 9-Jan 2-Nov 1-Oct 12-Aug Date

14 Variables used to predict NEE Variables used to predict P g Variables used to predict R e Theme Count Pct. Weight Pct par nd sincesost temp ppt tin sosn Theme Count Pct. Weight Pct nd par doy temp sincesost c4pct sost tin Theme Count Pct. Weight Pct temp nd ppt par sincesost doy pctgrass tin

15 Model Accuracy Assessment MAD = 1/n observed predicted Range = 95th percentile - 5th percentile Relative Error = MAD / Range

16 Winter fluxes difficult to estimate. NGP winter NEE NEE (g C m-2 day-1) Temperature (C) 3 min. data sets indicate snow depth and wind speed important

17 Winter fluxes (Nov. - Mar.) are need to convert model tree maps (Apr. - Oct.) to Annual Fluxes -.1 Mandan 1-2 Mandan 96- Miles City Lacombe Average Average NEE (g C m -2 day -1 ) Winter C flux Grassland Studies Mandan 1-2, Gilmanov 22; Mandan 96-, Frank et al. 22, Miles City, Gilmanov 22; Lacombe, Baron et al. 24

18 Cross validation Results Methodology to quantify robustness of small training data sets Random cross-validation produces realistic error estimates Cross-validation by sites evaluates influence of individual sites Cross-validation by year evaluates influence of individual years Five-fold NEE (g C m -2 day -1 ) cross-validation (random) Random sample as test Fold 1 Fold 2 Fold 3 Fold 4 Fold 5 Hold out data sample size Mean Absolute Difference Relative Error Average MAD =.48

19 Cross-validation NEE (g C m -2 day -1 ) by site Site withheld as test LE MA FP CN MC None Sample size (train) Mean Absolute Difference (train) Relative Error (train) Sample size (test) Mean Absolute Difference Relative Error Cross-validation NEE (g C m -2 day -1 ) by years Year withheld as test None Sample size (train) Mean Absolute Difference (train) Relative Error (train) Sample size (test) Mean Absolute Difference (test) Relative Error (test)

20 Pg: Jack_site_summary3.xls Robustness of Pg model tree: Cross Validation using Sites Site Withheld CN FP LE MA MC None MAD (g C m -2 day -1 ) Relative Error N Cheyenne and Mandan are influential sites (high Pg in CH 1998 and MA 1999) (MA is only high precipitation, eastern site) Relative Errors on the other sites was from 7 to 12 %

21 Results How Well do 1-day Model Tree Predictions Averaged over the Growing Season Compare with Flux Tower Observation? NGP Modeled April -- Oct. average Re NEE (g C (g mc -2 mday -2-1 day ) versus -1 ) versus Observed Re NEE NGP Modeled April - Oct. average Pg (g C m -2 day -1 ) versus Observed Pg, MAE =.15 Tower Tower NEE Re.82.5 y = 1.274x y =.8382x R 2 R = =.639 MC 2 2 MC 2 MC 21.4 MC 21 MA 21 MA 21 MA 2 MA 2.2 FP FP 2 MA 1999 MA :1 Line 1:1 Line Regression Regression Miles City 2 flux data begins 4/ Modeled Re Modeled NEE Twrlocalbxy_gpp_nee1_Re.xls Tower Pg y =.962x R 2 =.949 Miles City 2 flux data begins 4/ Modeled Pg MC 2 MC 21 MA 21 MA 2 FP 2 MA :1 Line Regression

22 Model Tree Predictions Match 21 Flux Tower Observations Better than MODIS Net Photosynthesis Captures General Tend in Tower Pg Predictions seem site specific Model Tree (trained from all towers) (withheld 21 MAD was.48) MODIS GPP 6 5 MAD =.32 Relative Error = MAD =.56 Relative Error =.16 Estimated Pg (g C m -2 day -1 ) LE MA MC 1:1 Line Estimated Pg (g C m -2 day -1 ) LE MA MC 1:1 Line Tower Pg (g C m -2 day -1 ) Tower Pg (g C m -2 day -1 )

23 Regional Comparison of 21 MODIS GPP and Model Tree Pg on Rangelands General Agreement MODIS had outlier values > 6 g C m -2 day -1 Pg (g C m-2 day-1) MODIS Model Tree

24

25

26

27

28 NEE (g C m 2 yr -1 ) 1998 = = = = -6. Average = -16.

29 Why was 2 NEE so low? 9 Precipitation (mm) Ppt_temp_monthy_rangelands.xls Month Ppt_temp_monthly_rangelands.xls NEE (g C m 2 yr -1 ) 1998 = = = = -6. Average = -16. Low June precipitation Low previous fall precipitation

30 Investigating Causes of Sinks and Sources: 1999 High NEE 1 and NEE (gc m -2 day -1 ) 1 High 1 High 2 Low 1 Low 2 Low NEE 1 and /9 3/31 5/2 7/9 8/28 1/17 12/6 4 Date Percent Clay 2 NDVI High 1 High 2 Low 1 Low High 1 High 2 Low 1 Low 2 2/9 3/31 5/2 7/9 8/28 1/17 12/6 Date

31 Environmental Drivers of Source Areas Average Temperature STATSGO April-June Range (Apr. percent - Production Oct., clay Precipitation ) Normal Year (C) Percent mm13.7 clay Kg 15.5 ha <= >=

32 Conclusions Model Tree predictions of Pg and NEE agreed well with tower observations Heavy reliance of Model Tree predictions on NDVI insures tracking of temporal and spatial dynamics not evident in climatic data sets Winter flux estimates need improvement N. Great Plains Rangelands from were a WEAK source of C.