Spatiotemporal cotton yield variations in relation to time-series satellite imagery: Implications for within-season crop management

Understanding the spatial and temporal variability of crop growth and yield is fundamental to optimizing site-specific, time-sensitive crop management to improve productivity and resource use efficiency. This study assessed the spatiotemporal cotton yield patterns and their relationships with satellite-derived vegetation indices, topography, and apparent soil electrical conductivity (EC_a) across multiple sites and seasons.

Landsat-based NDVI and its temporal derivatives were calculated to estimate cotton spatiotemporal growth patterns. Management zones were delineated using K-means clustering based on topography and EC_a. Correlation analysis was conducted to assess relationships among yield, NDVI, NDVI derivatives, topography, and EC_a, and to identify variables strongly associated with cotton yield variability. Random Forest was employed to predict yield and evaluate the contributions and interactions of environmental factors.

NDVI and its derivatives are strong predictors of cotton yield, with topography dynamically influencing these patterns in response to seasonal moisture. Lower elevations and slopes exhibited higher NDVI and yields due to improved water availability and lower erosion, whereas summits displayed moderate but unstable crop growth and yield. Random Forest models predicted yield robustly in dry years (R² = 0.75–0.83) but moderately in wet years (R² = 0.53–0.65).

NDVI and NDVI derivatives are good predictors of spatiotemporal cotton growth and yield. Integrating NDVI and NDVI derivatives with zones delineated using EC_a and topography provides robust yield prediction for precision crop management. Furthermore, applying machine learning techniques enhances the interpretability of complex environmental interactions, providing deeper insights to facilitate site- and time-specific management.