Performance of GEDI footprint aboveground biomass models

Abstract:

The Global Ecosystem Dynamics Investigation (GEDI) is a multibeam waveform lidar on the International Space Station. GEDI data are producing globally representative measurements of vertical height profiles (waveforms) and estimates of aboveground carbon stocks. Here we document on-orbit performance of algorithms used to produce the first release of the GEDI04_A data product. The GEDI04_A data product is derived from statistical models that relate waveform height metrics to field-estimated aboveground biomass. We used a waveform simulator to generate simulated GEDI waveforms from discrete-return airborne laser scanning (ALS) data, and associated height metrics from simulated waveforms with field-estimated aboveground biomass. Here we apply these models to recorded GEDI data at locations worldwide using a representative sample within 7 plant functional types on 6 continents and GEDI-ALS crossovers. We compare model performance on recorded GEDI data to coincident simulated waveforms, and evaluate the dependence of model performance to GEDI beam sensitivity and other measurement conditions using the percentage root mean squared error (RMSE) and bias. We also document the degree to which training data are representative of the global distribution of waveform relative height metrics in recorded GEDI data. Our analysis benchmarks performance of the first version of the GEDI04_A data product and the degree to which currently-selected models are representative of the domains in which they are applied.