Integrating machine learning-based habitat suitability modeling with land use analysis for the conservation and rehabilitation of Elaeocarpus prunifolius in Meghalaya, India

Mismatch between broad spatial scales at which niche models operate vis-& agrave;-vis the finer localized scales required for conservation actions often hinder the effective translation of model outputs into actionable strategies. This study addresses this issue by integrating ecological niche modeling (ENM) with land use and land cover (LULC) analysis to improve the conservation status of a threatened tree species Elaeocarpus prunifolius in Northeast India. Using the Maximum Entropy (Maxent) model, we predicted the potential distribution of E. prunifolius using MODIS-based Enhanced Vegetation Index (EVI) and ASTER-based elevation data. The predicted distribution, covering 8.97% (approximate to 201,321 ha) of Meghalaya's total geographical area, was further refined through the overlay of LULC maps, identifying approximate to 17,033 ha as highly suitable habitat. The species' distribution was predominantly concentrated in the Khasi and Jaintia Hills, with a narrow potential distribution zone extending into the Garo Hills. The species was observed at elevations ranging from 700 to 1500 m above sea level (asl). Field validation of the predictive distribution maps resulted in the discovery of previously undocumented populations, primarily concentrated in the East Khasi Hills (EKH) and West Jaintia Hills (WJH), thereby confirming the accuracy of the habitat suitability model. This integrated approach demonstrates the effectiveness of combining ENM with LULC analysis for the precise identification of conservation sites, thereby improving the potential for successful conservation and reintroduction efforts for E. prunifolius. The study highlights the critical role of integrating predictive models with ground validation in developing informed and effective conservation strategies.