Predictive mapping of bamboo cultivation suitability in Thailand using cloud-based modeling

Bamboo is increasingly recognized for its ecological and industrial value. In Thailand, 69 bamboo species have been recorded, including 50 native species. Bamboo has long supported local livelihoods and material practices, but the country still lacks a quantified and spatially continuous national assessment of the environmental conditions associated with bamboo cultivation suitability. We implemented a cloud-based species distribution modeling workflow using mapped bamboo occurrence and eleven topographic, climatic and soil predictors. Complementary Random Forest workflows generated a continuous probability surface and a binary suitability map at 1 km resolution. The probability-based model achieved a test accuracy of 85.6%, a Cohen’s Kappa value of 0.670, and an area under the receiver operating characteristic curve of 0.914, while the binary model achieved a test accuracy of 85.3% and a Kappa value of 0.665. At the national scale, the probability output identified approximately 43,952 km² (8.57% of Thailand) above the 0.5 threshold, while the binary classification identified 41,625 km² (8.11%) as environmentally suitable. The northern and eastern regions had the highest proportional suitability, whereas the northeastern region contained the largest absolute suitable area. Environmental conditions in mapped and predicted suitable areas broadly overlapped with Thailand’s national baseline, although selected topographic and soil variables showed clearer differentiation. This indicates that bamboo is broadly compatible with Thailand’s environmental setting, while suitability remains spatially selective. The probability and binary outputs overlapped with 92.66% and 92.21% of the mapped bamboo reference area, respectively, and identified favourable conditions beyond currently mapped occurrence. Our findings provide a national environmental baseline for cultivation planning and resource assessment rather than an estimate of immediately available plantation land. They also demonstrate that accessible cloud-based computing and globally available datasets can support national-scale suitability assessment, with finer-resolution shared data offering scope for more locally specific applications.