Behera M.D.Krishna J.S.R.Paramanik S.Kumar S.Behera S.K.Anto S.Singh S.N.Verma A.K.Barik S.K.Mohanta M.R.Sahu S.C.Jeganathan C.Srivastava P.K.Pradhan B.2025-01-132025-01-1320245643295https://dl.bhu.ac.in/ir/handle/123456789/1845The leaf area index (LAI) provides valuable input for modeling climate and ecosystem processes. However, ground-based observations are necessitated across various phenophases from dense tropical forests for a better understanding in terms of their contribution to carbon fixation. In this study, Digital Hemispherical Photography (DHP) was used for LAI observation from Similipal Biosphere Reserve, and to predict high-resolution LAI using Random Forest Machine Learning approach. Observations were taken from ninety-three Elementary sampling units (ESUs) corresponding to the beginning and end of leaf fall seasons across moist deciduous, dry deciduous, and semi-evergreen forests. LAI demonstrated high values for dry deciduous, followed by semi-evergreen and moist deciduous forests for the start of the leaf fall season, whereas moist deciduous forests demonstrated high values during the end of the leaf fall season. Satellite-based spectral reflectance bands of Sentinel-2 and vegetation indices (VIs) were used as predictor variables, wherein the band-7, band-8, band-12, enhanced vegetation index (EVI), and Red-edge based EVI were evaluated as the most dominant responsive variables for LAI estimation. Random Forest (RF) model provided good accuracy (R2 = 0.64, RMSE = 0.62) with observed DHP-based LAI. However, a comparison of RF model-based predicted LAI with global LAI products (MOD15A2H and VNP15A2H) provided a moderate correlation. Such studies demonstrate the potential of site or region-specific case studies to evaluate coarser-resolution global LAI products for possible improvement. � International Society for Tropical Ecology 2024.enDigital hemispherical photographyLeaf area indexRandom forestSentinel-2Similipal biosphere reserveArticle10.1007/s42965-024-00327-y