A Functional Diversity Approach to Assessing Rangeland Health and Condition

Rangelands are critical ecosystems facing degradation from climate change and anthropogenic pressures. Traditional rangeland health assessments, while valuable, are often qualitative, labour-intensive, and difficult to scale. This study introduces and validates a novel, quantitative framework that integrates key plant functional traits with machine learning to support rangeland health assessment. We conducted extensive field surveys across five bioclimatic regions in Iran, collecting data on 17 standard rangeland health indicators, vegetation cover, soil organic matter, and key plant functional traits. We calculated nine taxonomic and functional diversity metrics and scaled them relative to ungrazed reference sites. Using machine learning algorithms (random forest, support vector regression, and extreme gradient boosting), we modeled the relationships between these functional diversity metrics and the core rangeland health attributes: soil/site stability, hydrologic function, and biotic integrity. We found that a minimal set of functional diversity metrics—specifically functional richness (FRic) and Rao's quadratic entropy (RaoQ)—coupled with simple-to-measure variables like canopy cover, were robust predictors of core rangeland health attributes (soil/site stability, hydrologic function, and biotic integrity), outperforming traditional metrics like species richness alone. Model performance was highest for random forest models, although results varied among regions and attributes. We finally operationalized our findings into an open-source R package, Rangeland Health and Condition (RHC), which automates health assessment using easily measurable functional trait data. This tool allows managers to accurately predict rangeland health scores without extensive qualitative evaluation. Our approach represents a shift toward a predictive, data-driven framework for rangeland management, demonstrated here within montane rangelands across a precipitation gradient, and provides a foundation for future testing and application across broader rangeland systems. The framework provides a foundation for scalable rangeland monitoring and early detection of degradation, contingent on the availability of spatially explicit input data.