Multivariate analysis of compound hydroclimatic extremes in snow-fed sub-catchments: an application for climate-resilient water management of Narayani Basin, Nepal

Compound hydroclimatic extremes challenge water management worldwide due to their complex and nonlinear hydrological impacts. Drought-to-flood (DtF) events occur when prolonged dry periods are followed by rapid wet anomalies with distinct impacts from regular floods. Such events occur across diverse hydroclimatic settings; in heterogeneous snow-fed basins, DtF behaviour is governed by precipitation variability, antecedent soil-moisture conditions, snow processes, and, where present, reservoir regulation. This study investigates DtF transitions across six sub-catchments of the Narayani River Basin, Nepal, using monthly precipitation (17 stations), streamflow (6 gauges), remote sensing soil moisture, and historical drought–flood records. Historical DtF events were fitted with monthly Standardized Precipitation Index (SPI) signals, and sensitivity analysis of three ΔSPI thresholds (1.0, 1.25, and 1.5) identified 1.25 as the optimal criterion for distinguishing meaningful DtF transitions. The calibrated ΔSPI framework identifies DtF events and integrates ΔSPI–ΔSSI regression with multivariate Principal Component Analysis (PCA) of ΔSPI, ΔSSI, and ΔSWDI to investigate coupled hydroclimatic controls and process-level DtF dynamics. Results reveal pronounced spatial contrasts across hydroclimatic regimes. Highly snow-dominated Budhi Gandaki exhibits weak or opposing ΔSPI–ΔSSI responses during snow storage, followed by strengthened rainfall–runoff coupling during snowmelt (R2 up to 0.96). Moderately snow-influenced sub- catchments display mixed rainfall–runoff responses (R2 = 0.02–0.79), reflecting varying contributions of snowmelt and rainfall, whereas rainfed reservoir- regulated Tadi Khola and Rapti show weak pre-monsoon responses during reservoir filling (R2 = 0.001–0.96) and shift to strong monsoon coupling (R2 = 0.75–1.00) during reservoir release. PCA shows that PC1 explains 40%–100% of the joint ΔSPI–ΔSSI–ΔSWDI variance, with contrasting loading directions between precipitation, streamflow, and soil-moisture during decoupled periods, but increasingly aligned ΔSPI–ΔSSI loadings during stronger hydrological coupling. These patterns indicate that elevation and snow storage-melt timing regulate snow-influenced transitions, while antecedent soil-moisture conditions and reservoir operation control DtF behaviour in regulated systems, highlighting that precipitation anomalies alone are insufficient to explain observed transitions. The proposed framework provides a robust and transferable tool for detecting compound hydroclimatic extremes and supports early warning systems, reservoir management, and climate-resilient water planning in snow-affected, data-scarce mountain basins.