Evaluation of cotton planting suitability in Xinjiang based on climate change and soil fertility factors simulated by coupled machine learning model
Annotatsiya
Cotton is the world’s most widely cultivated fiber crop and holds great significance in Xinjiang. However, unsuitable planting environments can hinder farmer income and result in a substantial waste of agricultural resources.This study explores suitability of cotton planting areas in Xinjiang to reduce agricultural inputs and pollution. The goal is to promote sustainable agricultural development by considering both climate change and soil fertility, factors often overlooked in previous research. We analyzed climate change trends in Xinjiang and used machine learning-transfer component analysis to build a transferable coupling model for total nitrogen (TN) and soil organic carbon (SOC) indicators, resulting in a cotton suitability zoning that accounts for climate and soil fertility factors. Xinjiang has seen an overall increase in cumulative temperature and rainfall, with southern Xinjiang showing the most significant rise (4.02% in temperature and 16.26% in rainfall). The random forest model (RF) outperformed multivariate linear regression (MLR) and support vector machines (SVM) in predicting soil fertility indicators (TN: R 2 = 0 . 80 , SOC: R 2 = 0 . 77 ). The RF-TCA coupling model enhanced adaptability, with better performance in TN prediction compared to SOC. The Xinjiang cotton suitability zoning, based on meteorological and soil data, indicates a northward shift in suitable cotton planting areas in northern Xinjiang, while southern Xinjiang continues to maintain a substantial number of suitable planting zones. Notably, the disparity in suitability between the two regions has been narrowing over time. The research offers valuable insights for optimizing cotton planting locations, enhancing resource efficiency, and promoting sustainable development in Xinjiang. • Gray relation and full subset analysis of remote sensing indices improved model prediction accuracy. • An RF-TCA coupled machine learning model was developed to enhance transferability. • A comprehensive system was established to identify suitable cotton planting areas in Xinjiang, using meteorology and soil fertility factors.