Enhancing GEDI above ground biomass density estimates in contrasting forests of Pakistan

Forests are considered a critical resource for climate change mitigation because they capture and store carbon dioxide in forest biomass. Large-scale assessments of forest biomass have benefited recently from the increasing availability of ground-based forest inventory data and detailed vegetation structural data from airborne and terrestrial light detection and ranging (LiDAR). However, both terrestrial and airborne LiDAR data are costly, and not feasible for many nations. In contrast, since 2019, the spaceborne Global Ecosystem Dynamics Investigation (GEDI) laser instrument has collected publicly available data for global temperate and tropical forests. GEDI waveforms in conjunction with models calibrated using a worldwide network of field data to produce biomass products. However, relatively few field plots are available in the forests of South Asia in GEDI's current calibration dataset. We show here that GEDI's statistical estimators of Above Ground Biomass Density (AGBD) can be enhanced using local models calibrated strategically with GEDI data and field plots (594) in two contrasting forest ecosystems (i.e., mountain forests with evergreen broadleaf and needle-leaf trees, and coastal mangroves) in Pakistan. On-orbit GEDI AGBD estimations in mountainous areas have an RMSE of 95 Mg/ha / 87 % whereas estimates in mangrove areas have an RMSE of 8.4 Mg/ha / 89 %. GEDI AGBD estimates in mountain regions consistently underestimate field measurements in regions with higher AGBD (>300 Mg/ha), and overestimate in regions with low AGBD (<50 Mg/ha). In contrast, the locally recalibrated biomass models using the field measured data return RMSE of 39.6 Mg/ha / 36 % in mountain regions and 4.1 Mg/ha / 39 % in mangroves. GEDI is a valuable auxiliary resource for AGBD estimates in the absence of updated NFI measurements or over jurisdictions lacking enough reference data. However, local calibration using field data can increase accuracy and precision, particularly in areas where the base models have sparse local calibration data. Improved estimates from GEDI could then be used with active and passive satellite data for regional AGBD mapping and to lower uncertainties in regions with low NFI sampling coverage.