Impacts of Sample Design on Estimation of Aboveground Biomass: Implications for the Assimilation of Lidar and Forest Inventory Data

The availability of lidar data that can be used to characterize forest structure and estimate aboveground biomass (AGB) is rapidly increasing. When lidar data are considered in conjunction with forest inventory data to estimate AGB, the order of acquisition for these data products may impact the quality of the resulting estimates. In this work, we address this question in the context of uncertainty reduction with respect to estimation of AGB in a degraded forest in Paragominas, Brazil. We have developed a simulation framework that quantitatively assesses the uncertainty associated with estimation of AGB for different sampling strategies that combine forest inventory and lidar data. We utilize a Bayesian hierarchical modeling (BHM) data assimilation framework to combine information from the forest inventory and lidar data products into a higher order data product of AGB. Spatially explicit realizations of AGB are generated under different sampling strategies. Sampling strategies are assessed using the distributional properties of the assimilated higher order data product in the context of uncertainty reduction. We consider both spatially explicit maps of uncertainty as well as the standard deviation of the posterior predictive distributions of AGB as endpoints for the quantification of uncertainty. This framework allows for the explicit characterization of important sources of uncertainty. Our results show that a significant reduction in the uncertainty associated with estimation of AGB can be realized when design optimization is utilized in this context.