What does the 2003 SouthWest USA vegetation dieback event tell us about vegetation resilience to climate change? Results from a high-resolution land surface modeling exercise

In mountainous landscapes, topographic variability and the resulting biotic heterogeneity occur at scales substantially smaller than the grid-cell size of global climate simulations. In the Southwest USA, vegetation typically occupies smaller upland regions. Traditional land-surface modeling spatial scales thus fail to simulate the existence of these ecosystems for the present day. Lack of detailed representation of the land surface is a common problem, but in this region in particular, recent climate events have shown that the upland evergreen woodlands are sensitive to climate change-type drought events. This infrequently observed type of mortality event can potentially provide important constraints on the sensitivity of contemporary ecosystems to climate extremes, but only if the vegetation can be simulated at a resolution that captures the basic topographically driven patterns of existing ecotones. To represent these upland vegetated regions and investigate the impacts of future climate on vegetation mortality, we generated observed climate data sets at 32km resolution using the North American Regional Reanalysis (NARR), to drive the Community Land Model (CLM3.5). We calibrated the parameters controlling stress-induced mortality thresholds and severity in CLM3.5, to match the persistent NDVI reductions after the 2003 mortality event derived from long-term AVHRR images, data derived from rainfall manipulation experiments and site-specific observations of plant mortality. These empirical constraints on the likely thresholds for vegetation mortality allow us to constrain the likely impact of future climate change scenarios on other upland vegetation in the South West USA. .