Predicting the anthropogenic impacts on vegetation diversity of protected rangelands: an application of artificial intelligence

This study delves into anthropogenic impacts on vegetation diversity within mountainous protected rangelands, exploring habitat weakening and biodiversity loss. Employing artificial intelligence, specifically multilayer perceptron (MLP), radial basis neural network (RBFNN), and support vector regression (SVR), we predict vegetation diversity responses to ecological conditions, livestock grazing, and tourism. Assessing 305 sample plots with 21 variables, the MLP model demonstrated superior accuracy (R2 = 0.93 in training, R2 = 0.81 in the test dataset) compared to RBFNN and SVR. Sensitivity analyses highlighted anthropogenic factors like distance to tourist destinations, roads, pastures, and animal husbandries as significant influencers of vegetation diversity in mountainous protected rangelands. To enhance practical application, a user-friendly graphical interface was developed, enabling rangeland managers to utilize the MLP model. This tool facilitates estimation of livestock grazing and tourism impacts on vegetation diversity, empowering informed decision-making for the preservation and sustainability of mountainous protected rangeland ecosystems.