H53G-0933:
Uncertainty Analysis Using BMA for Hydrologic Projections under Future Climate Change

Friday, 19 December 2014
Ehsan Beigi, Louisiana State University, Baton Rouge, LA, United States and Frank T-C Tsai, LSU, Baton Rouge, LA, United States
Abstract:
This study conducts uncertainty analysis on future region-scale hydrologic projections under the uncertain climate change projections of the Intergovernmental Panel on Climate Change (IPCC) Fifth Assessment Report. The choice of the global climate models (GCMs), the greenhouse gas concentration trajectories and the GCM initial conditions are considered to be three major sources of uncertainty in the projected precipitation and temperature. This study uses the 133 sets of downscaled precipitation and temperature of the 1/8 degree BCCA projections for uncertainty analysis, which are derived from 22 CMIP5 GCMs, four emissions paths (RCP2.6, RCP4.5, RCP6.0, and RCP8.5), and different number of GCM initial conditions. The downscaled precipitation and temperature are used in the hydrologic model HELP3 to derive high-resolution spatiotemporal distributions of surface runoff, evapotranspiration and groundwater recharge from 2010 to 2099. The hierarchical Bayesian model averaging (HBMA) method is adopted to segregate and prioritize the three sources of climate projection uncertainty, obtain the ensemble mean of hydrologic projections, and quantify the hydrologic projection uncertainty. Posterior model probabilities in the BMA are calculated based on a performance criterion and a convergence criterion. The performance criterion is the GCM performance in reproducing the historical climate. The convergence criterion is the closeness of GCM simulation to the ensemble mean of future projections. Different likelihood functions are used to investigate their impacts on the posterior model probabilities. The methodology is applied to the study of hydrologic projections and uncertainty for the area of the Southern Hills aquifer system, southwestern Mississippi and southeastern Louisiana. The study area is divided into more than 2.6 million subdivisions by intersecting various datasets through the ArcGIS. The analysis is computationally intensive. Parallel computation is used to obtain the results. It is found that the HBMA method is able to quantify the contributions of individual sources of uncertainty in the climate projections to the total uncertainty in the hydrologic projections.