DROUGHT CHARACTERIZATION AND PROPAGATION OVER INDIAN SUB-CONTINENT: A COMPLEX NETWORK APPROACH

Abstract

Drought is a natural disaster that affects water resources, agriculture, and social and economic development due to its long-term and frequent occurrence. However, spatiotemporal assessment of drought characteristics over India at the sub-basin scale based on terrestrial water storage is unexplored. In the first part of this study, the terrestrial water storage anomalies (TWSA) obtained from a Gravity Recovery and Climate Experiment and precipitation data are used to compute Combined Climatological Deviation Index (CCDI) and GRACE-Drought Severity Index (GRACE-DSI). The trend characteristics of GRACE-DSI and CCDI are investigated using the non-parametric Mann-Kendall test, and its slope is estimated using the Theil-Sen slope estimator. Our results showed that GRACE-DSI exhibits significant negative trends over most of the Indian sub-basins compared to CCDI, indicating that most of the drought events are due to depletion of TWS. The sub-basin showing significant negative GRACE-DSI trends and significant positive CCDI trends conclude that precipitation is available, preventing TWS from depletion. The number of sub-basins showing significant negative trends for GRACE-DSI is more than that for CCDI. Hence TWS is depleting for most of the subbasins in India. The maximum drought duration obtained by GRACE-DSI and CCDI is 26months (2002-2004) and 17 months (2013-2015), respectively. The maximum drought severity computed by GRACE-DSI and CCDI is -44.2835 and -13.4392, respectively. In the second part of this study, we have utilized a complex network technique to study the drought propagation over the Indian subcontinent. In Methodology, we first computed drought event series based on GRACE-DSI. Then by the event synchronization method, we computed the strength matrix. By applying the threshold value, we have converted the strength matrix into an adjacency matrix and then constructed the network. After network construction, we have computed network matrices like degree, indegree, outdegree, betweenness centrality, and closeness centrality. Out results show that Krishna upper, Krishna middle, Manjra, the Godavari lower, and Mahanadi lower subbasins are more critical subbasins because they have high degree values between 33 to 42. The Godavari lower, Krishna upper, and Mahanadi lower subbasins are more important pathways for drought propagation because they have a high value of betweenness centrality.