Assessment of spaceborne and airborne lidar metrics using Fay-Herriot models to support forest biomass estimation

Abstract

Accurate estimation of Aboveground Biomass Density (AGBD) is essential for understanding carbon cycling and informing forest management and climate mitigation strategies. This study evaluates the use of Fay-Herriot (FH) models to estimate AGBD by integrating metrics from spaceborne LiDAR (GEDI), airborne LiDAR (ALS), and their combination. We assessed predictive performance across two contrasting forest environments: eucalyptus plantations in Hawai‘i and Mediterranean pine forests in Spain. Four estimation methods were compared at each site: FH models using only ALS data, only GEDI data, both data sources combined, and direct estimation using only field data. A model selection process was employed to identify candidate predictors, and all models were rigorously evaluated. To assess the performance of each estimator, Root Mean Square Error (RMSE) and relative efficiency—compared to direct estimation—were used as indicators. The results demonstrate that FH models, regardless of the auxiliary variables used, consistently outperformed direct estimation methods, as evidenced by lower RMSE values. Relative improvements over direct estimations were 18 %, 19 %, and 21 % for ALS, GEDI, and their combination in Hawai‘i; and 31 %, 29 %, and 31 % for the respective auxiliary datasets in Spain. Combining ALS and GEDI yielded only marginal improvements over using each set individually. Furthermore, both datasets exhibited comparable performance. Regarding the predictors, structural metrics related to vertical complexity emerged as key drivers of performance. Together, these results demonstrate that both ALS and GEDI data substantially enhance AGBD estimation within FH frameworks, with GEDI providing a cost-effective alternative at operational scales where ALS data are unavailable.