BD-AgroClim: Bangladesh Agro-Climatic Dataset for Reanalysis Validation and Bias Correction

BD-AgroClim is a research dataset and accompanying analysis resource developed for the validation, evaluation, and bias correction of gridded climate reanalysis and satellite-derived climate products over Bangladesh. The collection integrates station-based observations with ERA5-Land and NASA POWER climate data and includes station metadata, terrain information, yield-related data, paired validation datasets, statistical bias-correction outputs, machine-learning correction results, and supporting analysis products. The resource was developed to investigate spatial and terrain-related controls on climate-product bias and to evaluate statistical and machine-learning approaches for improving gridded climate estimates. The analysis includes pooled validation, rainfall detection assessment, terrain–bias relationships, statistical bias correction, machine-learning residual correction, leave-one-station-out cross-validation, island holdout evaluation, and SHAP-based feature-importance analysis. The repository also includes the analysis notebooks used to generate the validation results, correction models, figures, and tables associated with the study. This repository is intended to support reproducibility, transparent methodological evaluation, and further research on climate data validation and bias correction over Bangladesh.