Assessing Forest Carbon Response to Climate Change and Disturbances Using Long-term Hydro-climatic Observations and Simulations

Long-term climatic and hydrologic observations on the Santee Experimental Forest in the lower coastal plain of South Carolina were used to estimate long-term changes in hydrology and forest carbon dynamics for a pair of first-order watersheds. Over 70 years of climate data indicated that warming in this forest area in the last decades was faster than the global mean; 35+ years of hydrologic records showed that forest ecosystem succession three years following Hurricane Hugo caused a substantial change in the ratio of runoff to precipitation. The change in this relationship between the paired watersheds was attributed to altered evapotranspiration processes caused by greater abundance of pine in the treatment watershed and regeneration of the mixed hardwood-pine forest on the reference watershed. The long-term records and anomalous observations are highly valuable for reliable calibration and validation of hydrological and biogeochemical models capturing the effects of climate variability. We applied the hydrological model MIKESHE that showed that runoff and water table level are sensitive to global warming, and that the sustained warming trends can be expected to decrease stream discharge and lower the mean water table depth. The spatially-explicit biogeochemical model Forest-DNDC, validated using biomass measurements from the watersheds, was used to assess carbon dynamics in response to high resolution hydrologic observation data and simulation results. The simulations showed that the long-term spatiotemporal carbon dynamics, including biomass and fluxes of soil carbon dioxide and methane were highly regulated by disturbance regimes, climatic conditions and water table depth. The utility of linked-modeling framework demonstrated here to assess biogeochemical responses at the watershed scale suggests applications for assessing the consequences of climate change within an urbanizing forested landscape. The approach may also be applicable for validating large-scale models.