BP-CNN-Based Study on the Hierarchical Classification of Soil Nematode Communities in Tea Gardens of the Qinba Mountains

Abstract

To achieve efficient and precise classification of nematode community distribution in tea garden soils of the Qinba Mountains, this study integrated a backpropagation neural network (BP-NN) with a convolutional neural network (CNN) to construct a BP-CNN model, which was applied to assess soil health in tea gardens within this region. Multiple gradient sampling points were established in typical tea gardens across the Qinba Mountains. Samples were collected at varying altitudes and soil types to obtain nematode community species composition, abundance, and environmental factor data. Using nematode community characteristics as input and soil health grades as output, the BP-CNN model was constructed. CNN extracted deep features from the data, while BP-NN performed classification predictions. The model was validated against standalone BP-NN and CNN models. Results indicate the BP-CNN model achieved a grading accuracy of [XX%], significantly outperforming single models. Tea garden soil nematode communities in the Qinba Mountains can be classified into [X] grades, each corresponding to specific soil physicochemical properties and elevation ranges. The abundance of dominant groups such as bacteriophagous nematodes and fungivorous nematodes serves as a core indicator for grading. This study demonstrates that the BP-CNN model can effectively classify nematode communities in tea garden soils within this region, providing a tool for intelligent soil health assessment. Its classification results also offer scientific basis for tea garden soil management.