A new drought monitoring approach: Vector Projection Analysis (VPA)

Abstract

In this research, a new drought monitoring approach with an adaptive index-Vector Projection Analysis (VPA) and Vector Projection Index of Drought (VPID)- was developed that considers multiple drought indicators in various climate zones across the Contiguous United States (CONUS) and East Asia. A major advantage of VPA is that it uses multiple dependent variables (i.e., surface-based drought indices) and multiple independent variables (i.e., satellite-derived drought factors) to capture varied climate and environmental characteristics. Therefore, the VPA-based indices can be adopted for different drought types (i.e., meteorological, agricultural and hydrological droughts) depending on the user's selection of drought factors and indices. In VPA, the weights of each drought factor are generated through correlations (r) between each dependent variable and a satellite-derived drought factor that is obtained from the new generation sensor systems, Visible Infrared Imaging Radiometer Suite (VIIRS) and Global Precipitation Mission (GPM). Three schemes of VPID with different combinations of variables focused on integrated-, short-, and long-term drought (VPIDinte, VPIDshort, and VPIDlong, respectively) were evaluated over the CONUS and East Asia. All three schemes showed good agreement with surface-based drought indices resulting in averaged r (from 0.42 to 0.68) across both study areas. While VPIDshort or VPIDlong were more correlated with short (1-3 months) or long (9-12 months) term surface-based drought indices, VPIDinte provided a more generalized (i.e., integrated from 1 to 12 months) index that is suitable for varied drought conditions. The spatial distributions of drought from VPIDs agreed with United States Drought Monitor (USDM) data and Emergency Events Database across East Asia for both extreme drought and normal conditions. In particular, the drought areas of VPIDinte corresponded with the USDM results with r = 0.83 and root mean square error = 9.4%. Crop yield trends were also consistent with the VPIDs results in most years for both study areas. The proposed approach, VPA and VPID, can be adopted for any region with different combinations of surface-based drought indices and satellite-derived drought factors.