Forest fire pattern and vulnerability mapping using deep learning in Nepal

Abstract Background In the last two decades, Nepal has experienced an increase in both forest fire frequency and area, but very little is known about its spatiotemporal dimension. A limited number of studies have researched extent, timing, causative parameters, vulnerability factors regarding Nepal. Our study analyzed trends patterns for decades fire-vulnerability risk based on historical incidents across country. Results We spatial temporal fires extent burned area using Mann-Kendall trend test machine-learning approaches maximum entropy (MaxEnt), deep neural network (DNN). More than 78% was recorded between March May. The total increased over years since 2001 by 0.6% annually. obtained from categorized into four classes—very high, low, low. Conclusions Although models comparable, DNN slightly outperformed MaxEnt model. uses a complex structure algorithms modeled human brain that enables processing relationship input output dataset, making DNN-based recommended MaxEnt. These findings can be useful initiating implementing most suitable management intervention.