Uncertainties, Relationships, and Optimal Blends of Ensemble-Mean NLDAS Drought Indices

We analyze uncertainties and relationships among four ensemble-mean NLDAS drought indices (percentiles) derived from monthly mean evapotranspiration (ET), total runoff (Q), top 1m soil moisture (SM1) and total column soil moisture (SMT). The results show that the uncertainty is the smallest for SM1, the largest for SMT, and is in-between for ET and Q. The strongest relationship is between SM1 and SMT, and the weakest relationship between ET and Q. The strong relationship between ET and SM1 (SMT) appears in arid/semi-arid region, and that between Q and SM1 (SMT) appears in east and west wet regions. Drought frequency analysis shows that SM1 has most frequent drought occurrence, followed by SMT, Q and ET. In this study, we firstly established the linkage between NLDAS drought indices (a research product) and US Drought Monitor (USDM, an operational product) by comparing drought area percentage (100 x drought area/total area in Continental United States) derived from NLDAS and USDM. We suggest an optimization approach to search for optimal weights for an optimal blend of NLDAS drought indices by minimizing the root mean square error between NLDAS and USDM drought area percentage for a 10-year period (2000-2009). After we used 2-year (2010-2011) data to validate the optimal blend of NLDAS drought indices, we reconstruct a 30-year (1980-2009) optimal blended NLDAS drought index and monthly drought percentage. Overall results show that optimal blended NLDAS drought index performs the best, followed by SM1 and SMT.