Introduction The karst peak-cluster depression ecosystem is ecologically fragile, and the relationship between vegetation diversity and soil nutrients is a key scientific issue for ecological restoration. However, the relationships among soil chemical composition, nutrient elements, physicochemical properties, and plant diversity indices across different vegetation types, as well as the pathways through which key factors influence each other, remain unclear.Methods This study investigated grasslands, shrublands, and forests in Pingtang County, Qiannan Prefecture, Guizhou Province. Using correlation analysis and partial least squares structural equation modeling (PLS-SEM), we examined the associations and driving factors linking soil physicochemical properties, nutrient elements, mass ratios, and plant diversity.Results and discussion We found that: (1) Shrubland vegetation exhibits significantly higher soil organic carbon (SOC) content than grasslands and forests, indicating greater carbon sequestration potential; total nitrogen (TN) was the primary limiting nutrient. (2) Shrubland communities reached peak values for Margalef, Shannon-Wiener, and Pielou indices, with SOC showing a significant positive correlation with the Pielou index. (3) Soil organic carbon (SOC), total nitrogen (TN), and total phosphorus (TP) content were core factors regulating plant species diversity. The soil nitrogen-to-phosphorus (N:P) ratio was a key regulator, significantly influencing species composition and spatial distribution. (4) The PLS-SEM model further quantified the interaction pathways among these factors, providing a theoretical basis for optimizing vegetation diversity and managing soil nutrients in regional rock desertification control.