Improving Land Information Through Integrating Remote Sensing and Field Surveys: Evidence from the Bangladesh National Forest Inventory

Reliable land cover (LC) information is essential for scaling plot-based measurements in National Forest Inventories (NFIs). This study assessed the impact of using remote sensing (RS)-derived LC compared with field-assigned LC on the precision of key forest indicators in Bangladesh’s NFI (2015–2019). Field data from 1,781 plots were integrated with a 2015 national LC map produced from SPOT-6/7, Landsat, and Sentinel-2 imagery. Sampling errors of forest indicator estimates were evaluated across LC classes and ecological zones. Results show that RS derived LC reduced sampling errors for most biomass related indicators, including above- and below-ground biomass, tree volume, basal area, and carbon pools, by 15–20% on average, with some reductions exceeding 50%. Improvements were less consistent for regeneration related indicators (saplings, seedlings). These findings highlight the advantages of RS derived LC for improving NFI precision, while underscoring the continued need for advancing ontology driven approaches with necessary strengthening of field crew capacity to ensure consistent application of land cover standards.