The relative contribution of climatic, edaphic, and biotic drivers to risk of tree mortality from drought

Abstract:

In recorded history, the 2011 Texas Drought was comparable in severity only to a drought that occurred 300 years ago. By mid-September, 88% of the state experienced 'exceptional' conditions, with the rest experiencing 'extreme' or 'severe' drought. By recent estimates, the 2011 Texas Drought killed 6.2% of all the state's trees, at a rate nearly 9 times greater than average. The vast spatial scale and relatively uniform intensity of this drought has provided an opportunity to examine the comparative interactions among forest types, terrain, and edaphic factors across major climate gradients which in 2011 were subjected to extreme drought conditions that ultimately caused massive tree mortality. We used maximum entropy modeling (Maxent) to rank environmental landscape factors with the potential to drive drought-related tree mortality and test the assumption that the relative importance of these factors are scale-dependent. Occurrence data of dead trees were collected during the summer of 2012 from 599 field plots distributed across Texas with 30% used for model evaluation. Bioclimatic variables, ecoregions, soils characteristics, and topographic variables were modeled with drought-killed tree occurrence. Their relative contribution to the model was seen as their relative importance in driving mortality. To test determinants at a more local scale, we examined Landsat 7 scenes in East and West Texas with moderate-resolution data for the same variables above with the exception of climate. All models were significantly better than random in binomial tests of omission and receiver operating characteristic analyses. The modeled spatial distribution of probability of occurrence showed high probability of mortality in the east-central oak woodlands and the mixed pine-hardwood forest region in northeast Texas. Both regional and local models were dominated by biotic factors (ecoregion and forest type, respectively). Forest density and precipitation of driest month also contributed highly to the regional model. The local models gave more importance to available water storage at root zone and hillshade. Understanding how environmental factors drive drought-related mortality can help predict vulnerable landscapes and aid in preparing for future drought events.