Impact of Spatial and Temporal Index Weight Variation on Estimates of Drought Extents

The purpose of this work is to evaluate objectively blended drought indices, with constant, equal and unequal, as well as spatially and temporally varying weights. Spatially and temporally varying weights are currently not part of current practice in the USDM (US Drought Monitor). The three indices employed in this work are the 6-month standard precipitation index (SPI6), total column soil moisture percentile (SMT), and the 3-month standard runoff index (SRI3). Four cases (case 1-equal weight, case 2-unequal weight, case 3-spatially varied unequal weight, and case 4 - spatially and seasonally varied unequal weight) are tested and verified against the drought extent for the six USDM drought regions. The results show that the weights are unequal with spatial and seasonal variation. Overall, SPI6 is the most important factor and plays an important role in all USDM regions, SMT is the second most important factor and plays an important role in all six USDM regions except for Southeast region, and SRI3 is the third most important factor, especially in the eastern part (i.e., Northeast, Southeast, Midwest) of the continental United States. In the South and West regions, the weights have little seasonal variation. In the Northeast and Midwest, the weights have large seasonal variation. In the High Plains and Southeast regions, the seasonal variation of the weights is moderate. In Northeast and Midwest regions, as complexity of the weights increases, estimated drought extents have smaller RMSE (root-mean-square-error) and bias, and larger correlation and NSE (Nash-Sutcliffe Efficiency), indicating improved simulations for all four drought categories. In High Plains, South, Southeast, and West regions, the results are mixed. At the same time, we compare NSE and RMSE calculated from a new joint drought index (JDI) based on SPI6 and SMT/SRI3 with case 1 and case 4 in High Plains, South, Southeast, and West region when USDM drought extents are used as a benchmark. The results show that JDI outperforms the linearly combined drought index for Case 1and Case 4 for all drought categories in four regions except for less severe drought categories (D0-D4, D1-D4) in South and Southeast. This suggests that JDI is a promising approach when a linear combination of multiple drought indices does not work well.