Prediction of faunal distribution: Comparative evaluation of common spatial models using Himalayan bumblebee

Species distribution models (SDMs) are computational tools utilized to predict the geographical range of species by analyzing occurrence data and environmental conditions. While multiple algorithms are available for implementing SDMs, studies focusing on pollinators like bumblebees (genus Bombus ) in the Himalayas remain scarce, despite their ecological importance in forest and agroecosystems. We performed this work to identify the most effective SDM algorithms for modeling the distribution of genus Bombus in the Himalayas, thereby improving conservation strategies. We compared eight SDM algorithms, including Artificial Neural Network (ANN), Classification Tree Analysis (CTA), Flexible Discriminant Analysis (FDA), Generalized Additive Model (GAM), Generalized Boosting Model (GBM), Generalized Linear Model (GLM), Maximum Entropy Model (MAXENT), and Random Forest (RF), along with an ensemble method. Evaluations were based on metrics such as the area under the curve (AUC) for receiver operating characteristic (ROC), true skill statistic (TSS), and statistical properties like skewness and kurtosis. Among single models, RF stood out for its strong ROC, TSS, and statistical properties. The ensemble approach, however, was overall the best option across these metrics. Researchers are encouraged to select SDM algorithms after careful evaluation of study-specific factors such as geographic area, sample size, species diversity, and spatial variables. The scientific community should develop and authenticate optimal standards that integrate biodiversity concepts, ecological traits, environmental factors, ecosystem-based approaches, and field data to ensure effective model assessment, informed decision-making, and policy formulation for conserving biodiversity and advancing sustainable development goals in the Himalayas. • A comparison between eight SDM algorithms was performed using Himalayan bumblebees. • Among single models, Random Forest stood out for strong evaluation and statistics. • The ensemble approach was overall the best option across all of the models. • SDMs of pollinators in the Himalayas aid in achieving sustainable development goals.