Application of USDM statistics in NLDAS-2: Optimal blended NLDAS drought index over the continental United States

Abstract

This study performs three experiments to calibrate the drought area percentages in the continental United States (CONUS), six U.S. Drought Monitor (USDM) regions, and 48 states downloaded from the USDM archive website. The corresponding three experiments are named CONUS, Region, and State, respectively. The data sets used in these experiments are from the North American Land Data Assimilation System Phase 2 (NLDAS-2). The main purpose is to develop an automated USDM-based approach to objectively generate and reconstruct USDM-style drought maps using NLDAS-2 data by mimicking 10 year (2000–2009) USDM statistics. The results show that State and Region have larger correlation coefficients and smaller root-mean-square error (RMSE) and bias than CONUS when compared to the drought area percentages derived from the USDM, indicating that State and Region perform better than CONUS. In general, State marginally outperforms Region in terms of RMSE, bias, and correlation. Analysis of normalized optimal weight coefficients shows that soil moisture percentiles (top 1 m and total column) play the dominant role in most of the 48 states. The optimal blended NLDAS drought index (OBNDI) has higher simulation skills (correlation coefficient and Nash-Sutcliffe efficiency) in the South, Southeast, High Plains, and Midwest regions when compared to those in the West and Northeast. The highest simulation skills appear in TX and OK. By using optimal equations, we can reconstruct the long-term drought area percentages and OBNDI over the continental United States for the entire period of the NLDAS-2 data sets (January 1979 to present).