Mapping and conservation analysis of upper montane evergreen broad-leaved forest using machine learning and geodata

The upper montane evergreen broad-leaved forest (UMEBLF) provides unique ecological functions but exhibits a highly fragmented spatial distribution. Mapping the suitable habitat of UMEBLF and assessing its conservation status are therefore crucial for regional protection and management of this vegetation. In this study, Yunnan Province, China, was used as a case study and employed 11 topographic and climatic variables, including mountain relative height, to identify an optimal Random Forest model for simulating and mapping the suitable distribution of UMEBLF under current and future climate scenarios. In addition, the montane cloud-fog belt within the monsoon core region was incorporated to evaluate the reliability of the mapping results under current climatic conditions. By comparing the spatial distribution of nature reserves across the province, the current protection status of UMEBLF and its potential future changes were clarified. Results indicate that under current climatic conditions, areas with high habitat suitability for UMEBLF cover approximately 4.2 × 10⁴ km², primarily distributed across the mid-montane zones of western and central Yunnan. The lower boundary of suitable habitat closely coincides with the elevation of the cloud-fog belt and is primarily located on middle to upper slopes, at approximately two-thirds of the mountain's relative height above the base surface. Under future warming scenarios, the spatial pattern of suitable habitat exhibits relatively stable core areas and highly dynamic marginal zones. Under the SSP5-8.5 scenario, the area of highly suitable habitat is projected to decline to approximately 3.0 × 10⁴ km² by the 2090s. Currently, 32 nature reserves contain areas with high habitat suitability, with 75% concentrated on the southwestern side of the Cangshan-Ailao Mountain region. In the future, the proportion of highly suitable habitat areas within protected zones is expected to decrease by 15.9%-18.7%, with reductions occurring mainly within experimental zones of nature reserves. This study demonstrates an integrated framework combining geospatial big data and machine learning to map and analyze UMEBLF distribution, providing a scientific basis for mountain forest conservation and management under climate change.