Multi-scale assessment and entropy-MCDM framework for evaluating reanalysis precipitation datasets over Indian basins

• Novel Entropy-MCDM-based ranking of reanalysis rainfall over all Indian river basins. • ERA5 excels across overall statistical metrics but underestimates extremes. • IMDAA best captures extremes, especially in orographic regions. • MERRA improves at basin scale despite grid issues; CFSR shows limited consistency. • Grid and basin-scale analysis reveals spatial and scale-sensitive patterns. Accurate estimates of precipitation time-series data are critical for governing multiple dimensions of water security, including water availability, disaster management and climate impact assessments. Reanalysis datasets have gained attention as promising alternatives to sparse gauge networks and satellite products, offering consistent and continuous spatio-temporal coverage. However, their fidelity over large, monsoon-influenced, and hydro-climatically diverse regions like India, remain insufficiently explored. Where such evaluations have been conducted, they are often limited to specific events, lacking the integration of decision-support systems for gaining deeper insights into their effectiveness. This study presents a systematic multi-scale evaluation of well-known state-of-the-art reanalysis precipitation datasets, ERA5, IMDAA, MERRA-2, and CFSR/CFS v 2 across entire India over each grid as well as 25 basins. Using 34 years of data (1990–2024), we evaluate their performance rigorously through a set of 11 statistical metrics at daily, monthly, and seasonal scales. Key metrics include the linear correlation coefficient (CC), Kling–Gupta Efficiency (KGE; integrating correlation, variability, and bias), Nash–Sutcliffe Efficiency (NSE; predictive skill), and Percent Bias (PBIAS; mean deviation). In addition, distributional differences are analyzed using Kernel Density Estimation (KDE), a nonparametric approach for quantifying distributional shape. We introduce a spatially adaptive Multi-Criteria Decision-Making framework, combining Shannon Entropy and Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) to develop a hierarchical ranking of these datasets at the grid-scale and basin-scale. Our results reveal ERA5 as the most consistent dataset across all timescales (mean monthly CC: ∼0.81, KGE: ∼ 0.56, NSE ∼ 0.53, PBIAS: +6%). IMDAA aces in monsoon and mountainous basins (seasonal CC: 0.74, KGE: 0.49, and KDE shows widest IQR among all). Both CFSR/CFSv2 and MERRA-2 display low performance (KGE: 0.01 and 0.09) and significant bias (PBIAS: –29 % and –22 %). The study proposes a scalable evaluation framework transferable to other regions, providing actionable insights for hydrological applications and climate-resilient water management.