Integrating Multi-Sensors Data for Species Distribution Mapping Using Deep Learning and Envelope Models

The integration of ecological and atmospheric characteristics for biodiversity management is fundamental long-term ecosystem conservation drafting forest strategies, especially in the current era climate change. explicit modelling regional responses their impact on individual species a significant prerequisite any adaptation strategy. present study focuses predicting distribution Rhododendron arboreum, medicinal plant found Himalayan region. Advanced Species Distribution Models (SDM) based principle predefined hypothesis, namely BIOCLIM, was used to model potential arboreum. This hypothesis tends vary with change locations, thus, robust models are required establish nonlinear complex relations between input parameters. To address this relation, class deep neural networks, Convolutional Neural Network (CNN) architecture proposed, designed, tested, which eventually gave much better accuracy than BIOCLIM model. Both were given 16 parameters, including variables, statistically resampled then utilized establishing linear relationship fit occurrence scenarios species. parameters mostly acquired from recent satellite missions, MODIS, Sentinel-2, Sentinel-5p, Shuttle Radar Topography Mission (SRTM), ECOSTRESS. performance across all thresholds evaluated using value Area Under Curve (AUC) evaluation metrics. AUC be 0.917 CNN, whereas it 0.68 respectively. metrics indicate superiority CNN over BIOCLIM.