Mapping bamboo distribution in Eastern Nepal using random forest and SNIC classification with sentinel 2 imagery

Bamboo is an ecologically and economically significant resource that contributes to biodiversity conservation, soil stabilization, and the livelihoods of rural communities. Despite its importance, spatial information on bamboo distribution in Nepal remains limited, therefore compromising effective management and conservation efforts. This work maps eastern Nepal’s bamboo-growing areas in 2023 using Sentinel-2 satellite images with the Google Earth Engine (GEE) platform. Two classification techniques-Simple Non-Iterative Clustering (SNIC) and Random Forest (RF) were applied to evaluate their performance in bamboo mapping. Independent training and validation samples, supported by participatory field verification using QField, were used to improve classification reliability. The RF model was developed using 48 spectral and terrain variables, including elevation, slope, and aspect, while RF robustness reduced the influence of variable redundancy. Segmentation quality and the 0.95 probability threshold were visually evaluated using high-resolution reference imagery to minimize false positives. The results revealed notable differences between the two approaches. The RF approach identified 148 km² of bamboo-growing area, whereas the SNIC approach mapped 132 km2. Accuracy assessment indicated that RF achieved a higher overall accuracy (85%) than SNIC (68.9%), demonstrating that the RF model is robust for bamboo classification. Although seasonal variability and explicit topographic correction were not fully accounted for, future studies using multi-season imagery could further improve classification accuracy. This study presents the bamboo distribution map for eastern Nepal, providing valuable spatial information to help bamboo entrepreneurs, policymakers, environmentalists, and researchers develop better strategies for bamboo resource management in the country.