coffee distribution map of main production region of Yunnan in 2023 based on DNN and GEE

Coffee, as a perennial commodity crop, plays a crucial role in global agricultural markets, regional livelihoods, and poverty alleviation. Yunnan Province of China, located at latitudes of 21°8’~29°15’N, represents the northernmost coffee-growing region worldwide, and its coffee products have attracted increasing attention from the global market. However, the lack of a high-resolution coffee distribution map has limited environmental research and policy-making related to coffee cultivation, particularly in this subtropical and marginal growing area. In this study, we utilized Sentinel-2 optical data and Shuttle Radar Topographic Mission (SRTM) topographic data from the Google Earth Engine (GEE) platform to generate the first 10-m resolution coffee distribution map of Yunnan Province for 2023. An object-based Simple Non-Iterative Clustering (SNIC) segmentation was first applied to reduce within-class spectral variability, followed by a Deep Neural Network (DNN) classifier for coffee identification. The proposed approach achieved an overall accuracy (OA) of 0.87, with user accuracy (UA), producer accuracy (PA), and F1 score of 0.90, 0.96, and 0.93, respectively. Furthermore, SHapley Additive exPlanations (SHAP) were employed to interpret the DNN predictions and assess feature importance. The results indicate that shortwave infrared (SWIR) and red-edge features derived from dry-season Sentinel-2 composites play a dominant role in coffee discrimination, highlighting the importance of moisture-related and vegetation structural information. These findings not only improve the interpretability of the proposed mapping framework but also provide valuable insights into the spectral characteristics of coffee plantations. Overall, the generated coffee map offers a solid foundation for future studies on deforestation, sustainable development, and livelihood protection related to coffee cultivation in Yunnan Province. We provided two files. "2023_yunnan_coffee_10m.tif " is a 10 meter coffee distribution data directly calculated based on remote sensing images provided by the GEE platform. A value of 1 indicates the presence of coffee at that location, while a value of 0 indicates the absence of coffee. 2023_yunnan_coffee_1km. tif "is the proportion of coffee area in a 1km grid obtained by aggregating the previous 10m coffee distribution data. The numerical range is 0~1, where 0 indicates that there is no coffee in the 1km grid and 1 indicates that all coffee is present in the 1km grid.