Evaluating plant community responses to rainfall extremes in southwest U.S. ecosystems using radiometric change detection

Background/Question/Methods

Spatially-explicit depictions of plant productivity and biomass over large areas are critical to monitoring landscapes in highly heterogeneous arid and semi-arid ecosystems. Radiometric change detection techniques applied to Landsat Thematic Mapper (TM) imagery provide an opportunity to characterize changes in land surface conditions that are relevant to land managers. Importantly, these techniques are straightforward and readily transferable to those without expertise in remote sensing. We sought to answer three questions: (1) can we use differences between pre- and post-growing season spectral vegetation index values (i.e., normalized difference vegetation index (NDVI)) to locate areas of extreme change in vegetation? (2) in what ecological sites do these areas of extreme change occur? and (3) can extreme changes in vegetation be related to vegetation dynamics conceptualized in state-and-transition models. We calculated the difference between NDVI values representing pre- (May) and post- (September/October) growing season conditions, standardized NDVI difference values to Z-scores to identify areas of extreme high and low change over the growing season, and determined the plant and soil features associated with significantly high or low NDVI difference values. Calculations were made for two years selected to represent below- (2003) and above-(2009) average rainfall conditions.

Results/Conclusions  

Two percent of the 5696 –km² study area experienced significant reduction in NDVI values in 2003. The extreme reductions in NDVI values in 2003 were most commonly observed in Rockland and Hills ecological sites in a Historical Savanna state, but were also observed on an area of Clay Loam Upland in a Historical Grassland state. Interestingly, part of this Clay Loam Upland area also demonstrated an extreme increase in NDVI over the wet 2009 growing season. Such responses to wet and dry years detected by NDVI could portend community phase changes within the Historical Grassland state or a transition to an alternative state. The NDVI difference in 2009 following above-average rainfall in 2006 and 2008 resulted in 3.7% of the area experiencing a significant increase between May and Sep. The largest most contiguous patches of significant increase in NDVI occurred on the Clay Loam Upland ecological site and correspond to the Historical Grassland state. The vegetation index differencing method and translation of growing season differences to Z-scores permit examination of changes over large areas and can be applied by non-experts. Use of additional image dates could enhance interpretations of vegetation dynamics, focus field reconnaissance efforts, and become a routine operational work flow for natural resource managers.