Protected areas (PAs) within biodiversity hotspots play a crucial role in climate change mitigation, yet evidence-based insights into factors governing their carbon storage potential remains limited. This study bridges this gap by employing an integrated, machine-learning approach to quantify aboveground carbon (AGC) and soil organic carbon (SOC) stocks and their drivers among three major physiographic regions of Nepal, namely, the Dhorpatan Hunting Reserve (DHR) in the Himalayas, the Shivapuri Nagarjun National Park (SNNP) in the mid-hills, and the Chitwan National Park (CNP) in the Terai. A total of 288 soil samples were collected using stratified random sampling based on predefined environmental parameters. Despite comparable conservation practices, these PAs exhibited different patterns of carbon storage, with PA in the mid-hills containing the most AGC and the Himalayan PA retaining the most SOC stocks. Based on principal component analysis, AGC was governed largely by tree structural traits while SOC was influenced by soil factors. By employing a machine-learning approach, random forest modelling identified the relative importance and non-linear interactions between tree traits and soil properties. These findings provide one of the first comparative assessment across Nepal’s major physiographic gradients integrating vegetation, soil and carbon pools. The results emphasize that effective strategies for maximizing carbon sequestration and biodiversity conservation requires strategies tailoring to distinct ecological processes operating within each physiographic region. • Mid-hill protected areas had highest aboveground carbon stock • Primarily tree traits predicted aboveground carbon (AGC) stock • Mainly soil properties predicted soil carbon stock • PCA and random forest identified biotic/abiotic factors carbon storage • Protected areas play a critical role in climate mitigation.
Khanal et al. (Wed,) studied this question.