Location: Range Management ResearchTitle: Integrating phenology and patterns in spatial autocorrelation to delineate perennial grass cover using Landsat: A case study in a southwestern U.S. arid ecosystem Author
Submitted to: Meeting Abstract
Publication Type: Abstract Only
Publication Acceptance Date: 3/15/2013
Publication Date: N/A
Citation: N/A Interpretive Summary:
Technical Abstract: The presence and persistence of perennial grasses are important indicators for ecosystem function, biodiversity, and sustainable land management practices worldwide. Reliably distinguishing the perennial grass cover is challenging; this is especially true in arid ecosystems in the southwestern U.S. due to the prominence of bright exposed soils and modest vegetation cover. We integrate an understanding of plant phenology and measures of spatial autocorrelation in growing season Normalized Difference Vegetation Index (NDVI) values derived from Landsat 5 Thematic Mapper (TM) imagery from 2000 to 2010. Using time series TM data we combine knowledge of phenology for perennial grasses and shrubs and multi-scale patterns of spatial autocorrelation to distinguish between shrub- and grass-dominated plant communities. We differenced NDVI values between May when shrubs are green and September images when grasses are green and computed Moran’s I and Local Indicators of Spatial Autocorrelation (LISA). We conducted analyses of spatial autocorrelation by soil type because soils influence vegetation state changes in the region. We hypothesized that areas dominated by perennial grasses would exhibit high positive spatial autocorrelation at broad spatial scales while the patterns for shrub-dominated areas would be autocorrelated at local spatial scales. For years of above-average rainfall, large patches of positive and high autocorrelation in growing season NDVI corresponded to grass-dominated sites with a mean patch size of 14.5 ha and a growing season NDVI 0.083. Smaller patches of positively correlated NDVI difference corresponded to shrub-dominated sites with a mean patch size of 0.25 ha and a growing season NDVI of 0.07. Next steps will involve comparison of spatially autocorrelated patches with long-term estimates of biomass and percent cover. The integration of spatial autocorrelation metrics of growing season NDVI based on phenology could facilitate the application of remote sensing technologies to enhance broad-scale monitoring efforts and improve land management decision-making.