Using Landsat Time-Series and LiDAR to Inform Aboveground Forest Biomass Baselines in Northern Minnesota, USA

Abstract

The publicly accessible archive of Landsat imagery and increasing regional-scale LiDAR acquisitions offer an opportunity to periodically estimate aboveground forest biomass (AGB) from 1990 to the present to align with the reporting needs of National Greenhouse Gas Inventories (NGHGIs). This study integrated Landsat time-series data, a state-wide LiDAR dataset, and a recent cycle of the national forest inventory (NFI) records in Minnesota, USA, to obtain a spatially explicit inventory of AGB across a large region of space and time back to the 1990 baseline used by the US NGHGI. Pixel-level polynomial models were fit to 6 time-series metrics of Landsat data to obtain fitted predictors that were ultimately coupled with the NFI data in a nonparametric modeling framework to map temporal AGB baselines. Eighteen candidate models, formulated using different combinations of LiDAR and Landsat metrics, revealed that the model using both Landsat and LiDAR metrics consistently performed better than the alternative models. The RMSE of the model using both Landsat and LiDAR was 27.2 Mg ha−1, against 31.39 Mg ha−1 for the model using only LiDAR metrics. We conclude that the fitted Landsat-based model (RMSE = 47.64 Mg ha−1) provides acceptable accuracy for the 1990-baseline mapping of AGB.