Global crop suitability datasets for 17 crops under present (2024) and future climate scenarios (2041-2100)

Accurately assessing crop suitability is essential for ensuring food security and guiding agricultural adaptation to climate change. However, existing suitability analyses are mainly based on empirical rules and coarse-resolution predictors, which limits their representativeness and prevents them from capturing contemporary global agricultural patterns with sufficient spatial detail. Here we propose a transferable novel framework that integrates crop samples from remote sensing-based crop layers and high-resolution environmental predictors to model the global suitability of 17 major crops, using the random forest model under present (2024) and future climate scenarios with a 1 km × 1 km resolution. The models achieve an overall accuracy exceeding 0.9 based on the independent test set. Compared with the Global Agro-edaphic suitability dataset from Global Agro-Ecological Zones-Model (GAEZ), our present suitability results show an improved accuracy, with a global pixel-level correlation coefficient (r) of 0.57, and strong agreement in classified suitability area comparisons (r = 0.99, R² = 0.97). This study offers valuable support for optimizing cropping patterns, enhancing agricultural resilience, and informing climate adaptation strategies.