Biotic determinants of ecosystem sensitivity of semi-arid grasslands to extreme drought

Semi-arid grasslands are predicted to be either highly sensitive or insensitive to extreme drought when compared to other ecosystems. Low sensitivity is predicted because plants in semi-arid ecosystems have evolved under high water stress, and thus should be less sensitive to drought than plants in more mesic ecosystems. In contrast, high sensitivity is expected based on the Huxman-Smith model which predicts that ecosystems with the lowest mean annual precipitation should be most responsive to variation in precipitation. We tested these predictions with a four-year extreme drought experiment imposed simultaneously in four semi-arid grasslands of the US: desert grassland in NM (SBK), shortgrass steppe in NM and CO (SBL and SGS, respectively), and northern mixed grass prairie in WY (CHY). We found that as predicted by the Huxman-Smith model these grasslands exhibited relatively high sensitivity to extreme drought, but that sensitivity varied five-fold: 90% reduction in total aboveground net primary production (ANPP) for SBK vs. 50% SBL, 15% SGS and 5% for CHY. We assessed the potential biotic mechanisms that could be underlying variable sensitivity: 1) differences in dominant photosynthetic pathway (C4 vs C3), 2) differences in physiological drought adaptation of congeneric species, and 3) differences in physiological adaptation of conspecific populations caused by genotypic variation. For sites with different dominant physiological pathways (CHY vs. SGS), we found that the CHY was least sensitive to extreme drought due to a greater abundance of C3 grasses that were less responsive to drought when compared to C4 grasses at SGS. For sites with congeneric species (SBK and SBL), we found that Bouteloua eripoda was more sensitive to drought than its congener, B. gracilis, due to a lower hydraulic safety margin. Finally, for sites with different conspecific populations, we found that the B. gracilis population at SBL had a lower hydraulic safety margin than the population at SGS, resulting in greater sensitivity of the SBL population to extreme drought. Overall, our results suggest that differential sensitivity of semi-arid grasslands can result from a hierarchy of biotic mechanisms, ranging from the community to the genetic level.