Multiscale hydrological drought analysis: Role of climate, catchment and morphological variables and associated thresholds

Identification of thresholds associated with key climate, catchment and morphological variables for hydrological droughts can further improve our understanding of evolution and propagation of droughts in a complex water resource system. These thresholds are associated with complex interaction between climate and catchment variables and they are often connected through hierarchical as well as non-linear relationships. The advantage of selecting a multi-factor predictor domain can detect multiple thresholds that may not be observed by analyses limited to single predictors. In the present study, we developed a conceptual modeling framework by integrating a hydrological model developed based on the Soil and Water Assessment Tool (SWAT) and statistical models to quantify the potential influence of climate, catchment, and morphological variables and their thresholds on hydrological drought duration and severity for the watersheds located in Savannah River Basin (SRB). The concept of standardized runoff index (SRI) was used to derive the multiscale hydrological drought time series (i.e., SRI 1, SRI 6, and SRI 12) to investigate short term, medium term, and long term drought events based on their duration and severity. It was observed that the linear models developed based on the climate variables may not be capable for predicting the duration of multiscale hydrological droughts, whereas, the performance of statistical models can be significantly improved by the addition of catchment and morphological variables. In addition, among the morphological variables stream order seems to have a significant control over short, medium and long term drought duration across the study area. In the second phase of our analysis, we employed classification and regression tree (CART) algorithm for quantifying the thresholds associated with climate, catchment, and morphological variables that have potential influence on the hydrological drought. The result indicates that the variables and its associated threshold vary for short, medium, and long term drought. The proposed modeling framework can be extended for ungauged basins to improve the drought management.