Abstract Droughts alter the precipitation‐runoff (PR) relationship, thereby influencing water resources planning and management. Previous studies have mainly focused on the influence of multi‐year droughts on PR relationship. The influence of different drought types across various timescales (monthly, seasonal, and annual) remains understudied. In this study, we applied a Cumulative Distribution Function‐Linear Regression‐Random Forest (CDF‐LR‐RF) algorithm to evaluate, bias correct, and integrate multi‐source precipitation data sets for drought assessment in a data‐poor basin. Then, meteorological drought (MD) and hydrological drought (HD) were identified using the Standardized Precipitation‐Evapotranspiration Index (SPEI) and Standardized Runoff Index (SRI), respectively, with compound drought (CD) events defined where MD and HD overlapped. The PR relationship was characterized using Slope (conversion speed) and R 2 (relationship strength). The approach was tested in the source region of the Yellow River basin (SRYB) in northwest China. The key findings include: (a) The CDF‐LR‐RF method efficiently integrates multi‐source precipitation data, significantly improving accuracy for drought assessment. (b) In the SRYB, MD, HD, and CD exhibit clear time‐scale effects, with CD being more severe and longer‐lasting than MD, but less so than HD. (c) Different drought types impact the PR relationship differently across timescales, with short‐term droughts (identified through monthly and seasonal SPEI/SRI) showing a stronger effect than long‐term droughts (e.g., annual scale). (d) CD influences both the strength and conversion speed of the PR relationship, with its impact generally stronger than MD but weaker than HD. These insights help managers predict water availability and target drought responses for specific drought types.
An et al. (2026) studied this question.