Research on soil spotted degradation prediction in qinling tea-producing area of China based on deep learning

Soil Spotted Degradation (SSD) is a critical manifestation of land degradation that poses substantial constraints on agricultural development. However, the underlying mechanisms driving SSD and the methodologies for accurately predicting its occurrence remain poorly understood. In recent years, SSD has become increasingly prevalent in the tea-producing regions of the Qinling Mountains, China, intensifying the urgency of addressing soil-related challenges in the area. Consequently, developing accurate SSD prediction models has emerged as a pressing priority. This study focuses on Shangnan County, Shaanxi Province, China. Authentic SSD samples were collected through field investigations supported by remote sensing data and processed using the QGIS platform. To overcome limitations in sample compilation and the underutilization of feature information in SSD prediction, we propose a novel deep learning framework that integrates Stacked Autoencoders (SAE) with Dense Residual Networks (DRN). The performance of the proposed SAE-DRN model was benchmarked against conventional Support Vector Machine (SVM), hybrid Convolutional Neural Network—Random Forest (CPCNN-RF), and U-Net models. Experimental results demonstrate that the proposed SAE-DRN model achieved superior performance, with an overall accuracy of 0.87, an F1 score of 0.89, and an area under the receiver operating characteristic curve (AUC) of 0.92. Compared with the baseline models, SAE-DRN exhibited greater robustness and adaptability under small-sample conditions, producing more precise and reliable predictions of SSD occurrence. The findings underscore the potential of the SAE-DRN approach to guide tea plantation site selection and inform soil disease prevention strategies, thereby addressing key agricultural and environmental challenges in the Qinling region. Moreover, this method holds promise for broader application beyond tea cultivation systems, offering valuable insights for sustainable agricultural development, ecological restoration, and resource management in diverse agroecosystems.