<p>Standardised Precipitation–Evapotranspiration Index (SPEI) is widely used for drought monitoring because it integrates precipitation and atmospheric evaporative demand, offering advantages over precipitation-only indices by capturing the impact of rising temperatures under global warming. However, existing global SPEI products are typically provided at coarse spatial resolutions and often contain biases that limit their applicability for local-scale drought hazard assessment. This study addresses these limitations by developing a high-resolution (0.1°) SPEI dataset for Vietnam covering the period 1980–2022 using a Geographically Weighted Regression (GWR) model. By integrating observations from 130 meteorological stations with ERA5-Land predictors, including precipitation, temperature, solar radiation and wind speed, the GWR model explicitly accounts for spatial non-stationarity, enabling simultaneous downscaling and bias correction. The validity of the GWR-generated results was evaluated using multiple complementary approaches. Advanced spatial diagnostics using Local Bivariate Relationships (LBR) and Lee’s L statistics confirmed strong spatial consistency between the GWR-derived SPEI and the ERA5-Drought SPEI product, while highlighting the model’s enhanced ability to capture localised drought patterns that are often obscured in coarse-resolution global datasets. In addition, validation against independent station observations demonstrated robust predictive performance, with low Root Mean Square Error (RMSE) values of 0.60 ± 0.12 for 2015 and 0.56 ± 0.12 for 2020, and a near-zero mean bias (0.01 ± 0.02). Beyond conventional drought intensity mapping, this study introduces the Standardized Precipitation–Evaporation Spatial Inhomogeneity Index (PESI), which combines SPEI intensity with spatial inhomogeneity derived from Local Indicators of Spatial Association (LISA). The proposed PESI characterizes the differentiation between spatially fragmented and concentrated drought clusters, providing a more robust basis for drought hazard assessment in particular countries like Vietnam, where drought patterns are uneven.</p> Graphical Abstract <p></p> <p>This study addresses the limitations of the existing global Standardised Precipitation Evapotranspiration Index (SPEI), which suffers from coarse spatial resolution and biases when applied to local drought hazard assessment in Vietnam. The primary objective was to downscale the SPEI to enhance its resolution and reliability for the local area. The methodology first involved computing a ground-truth SPEI for Vietnam using 43 years (1980–2022) of temperature and rainfall data from local meteorological stations. Next, the Geographically Weighted Regression (GWR) model was applied, using the ground-truth SPEI as the target variable and explanatory climate variables derived from the ERA5-land reanalysis dataset. The GWR-derived non-stationary coefficients were then used to predict a high-resolution SPEI at 0.1 degrees spatial resolution. The predicted SPEI was validated through multiple complementary methods, including Standard error quantification, spatial-temporal pattern inspection, Local Bivariate Relationship (LBR) analysis, and Bivariate Spatial Association (Lee’s L), with a specific focus on critical drought years (2015 and 2020). Key findings indicate that the GWR-derived SPEI achieved higher resolution and improved reliability. It demonstrated strong visual agreement with historical drought patterns and better captured local climate conditions than the existing global ERA5-Drought product. Finally, the research focused on conducting a Local Indicators of Spatial Association (LISA) analysis to generate an SPEI Spatial Inhomogeneity Index (SPEI-SI). By modelling the relative importance of SPEI intensity and SPEI-SI, the study proposes a framework for integrating both components into a new Drought Hazard Index (DHI) for enhanced and comprehensive drought risk assessment in Vietnam.</p>