Mapping soil copper availability in a mining-affected city using UAV hyperspectral estimates and multi-temporal Sentinel-1/2 data

Introduction Copper is an essential micronutrient for plants; however, elevated concentrations of soil available copper (ACu) may adversely affect plant growth in mining‐affected environments. Large‐area surveys of ACu based on conventional soil sampling are time‐ and labour‐intensive and rarely provide sufficient coverage for continuous municipality-wide mapping. Methods This study developed a cross‐scale framework integrating UAV hyperspectral estimates with multi‐temporal Sentinel‐1/2 data to map ACu in exposed soils across Daye City, China. UAV‐derived ACu estimates from two survey sites were aggregated to the Sentinel 10 m grid and used as response labels. Four model‐specific predictor‐selection and regression pipelines were independently optimised and evaluated using leave‐one‐site‐out validation. Results The ensemble combining competitive adaptive reweighted sampling with a deep neural network (CARS‐DNN) showed the strongest agreement with held‐out UAV‐derived labels ( R 2 = 0.70; RMSE = 0.28 mg kg −1 ). The resulting 10 m map showed predominantly lower predicted ACu values, with localised higher‐value patches concentrated in mining‐disturbed areas and distinct spatial patterns among representative mining districts. Discussion By extending detailed UAV‐derived ACu information beyond individual survey areas, the framework provides a practical basis for relatively low‐cost municipality‐wide screening of soil Cu availability and for prioritising subsequent soil and plant investigations in mining‐affected regions.