Location: Range Management ResearchTitle: Field validation of biomass retrieved from Landsat for rangeland assessment and monitoring Author
Submitted to: International Geoscience and Remote Sensing Symposium Proceedings
Publication Type: Proceedings
Publication Acceptance Date: 7/15/2010
Publication Date: 7/25/2010
Citation: Browning, D.M., Peters, D.C., Steele, C., Rango, A. 2010. Field validation of biomass retrieved from Landsat for rangeland assessment and monitoring [abstract]. 2010 International Symposium on Geoscience and Remote Sensing, July 25-30, 2010, Honolulu, HI. TH3.L05.1. Interpretive Summary:
Technical Abstract: Measures of terrestrial biomass that are accurate and repeatable are important to monitoring and modeling productivity in the world’s grassland and savanna (i.e., rangeland) ecosystems. Remote sensing offers potential to characterize seasonal changes in biomass in a consistent non-destructive manner. We capitalized on long-term measurements of aboveground biomass at the Jornada Basin (JRN) Long-Term Ecological Research site in the northern Chihuahuan desert. We sought to answer two questions: (1) Does Landsat 5 Thematic Mapper (TM) imagery serve as a reliable and accurate proxy for vegetation biomass in this highly heterogeneous arid ecosystem? and (2) How do differences in vegetation structure influence biomass/SVI relationships? For this pilot study, seven images corresponding to biomass sampling dates from Nov 2006 to Nov 2008 were selected from the Landsat image archive to relate biomass to image-based indicators of vegetative vigor. TM images were processed to yield top of atmosphere radiance values corrected for atmospheric effects using the image-based COST method. We evaluated the performance of four SVIs: normalized difference vegetation index (NDVI), enhanced vegetation index (EVI), soil-adjusted vegetation index (SAVI), and modified soil-adjusted vegetation index (MSAVI2). To facilitate comparison of results from Landsat and the MODIS sensor in this arid ecosystem, we report values for EVI. EVI values are reported as the mean of 25 TM pixel values corresponding to the 150-m X 150-m area centered on the fifteen 0.5-ha study sites. Relationships between EVI and biomass were linear and consistent for sites within each of five vegetation communities (i.e. creosote, mesquite, tarbush, grassland, and playas). Field measurements of biomass ranged widely over time and across sites (0 to 422.2 g/m2). EVI tracked changes in biomass reasonably well with R2 values ranging from 0.238 (for playa sites) to 0.789 (for mesquite sites). EVI exhibited a low dynamic range in values across large differences in plant biomass. Grassland EVI responses do not always clearly reflect changes in biomass which may be due to the prominence of senescent vegetation at grassland and playa sites or confounded by the prominence of bare soil in arid ecosystems. On-going analyses will evaluate the performance of other SVIs. The range in physiognomy and productivity at the JRN is well-suited for field validation exercises for biophysical parameters retrieved via remote sensing. Accurate and robust relationships between spectral vegetation indices derived from freely available satellite imagery and field measurements of biomass are powerful tools for monitoring landscape condition in arid and semi-arid rangelands.