Mapping and scaling of in situ above ground biomass to regional extent with SAR in the Great Slave region

Authors: Kraatz, Simon; BOURGEAU-CHAVEZ, L. - Michigan Technological University; BATTAGLIA, M. - Michigan Technological University; POOLEY, A. - Michigan Technological University; SIQUEIRA, P. - University Of Massachusetts, Amherst

Submitted to: Earth and Space Science
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 12/6/2022
Publication Date: 12/13/2022
Citation: Kraatz, S.G., Bourgeau-Chavez, L., Battaglia, M., Pooley, A., Siqueira, P. 2022. Mapping and scaling of in situ above ground biomass to regional extent with SAR in the Great Slave region. Earth and Space Science. 9:Article e2022EA002431. https://doi.org/10.1029/2022EA002431.
DOI: https://doi.org/10.1029/2022EA002431

Interpretive Summary: Synthetic Aperture Radar (SAR) is a proven technology for estimating the amount of vegetation biomass from space. Vegetation interacts differently with radar depending on many factors such as radar frequency, type of vegetation and soils and hydrology. The upcoming NASAISRO SAR (NISAR) radar mission will be the first to provide plentiful and freely available L-band frequency data. The main purpose of this study is to scale in situ biomass data to map biomass over a wider region by using air and spaceborne L-band radar data. A secondary goal is to report on the level of accuracy achieved by this approach, and whether it falls within the NISAR mission requirements. We found that the NISAR mission requirements could be met, but only after accounting for scaling differences between the in situ biomass estimates (~0.2 hectares) and SAR data, meaning that the 1 hectare SAR data needed first to be averaged over about 4 hectare regions. This is a new finding, as other researchers generally recommend 1-hectare grids. This research will be of value to natural resource managers, modelers, and remote sensors, who are hoping to calibrate and validate products.

Technical Abstract: The North American boreal region (NABR) is increasingly threatened by climate change, with wildfire being the most prevalent disturbance factor. Upcoming spaceborne Synthetic Aperture Radar (SAR) missions at L- (or P-) band, such as the upcoming NASA ISRO SAR (NISAR), have great potential to advance our knowledge on the distribution and amount of above-ground biomass (AGB) in the NABR. AGB estimation with SAR is challenging due to a lack of suitable data. Also, different forest settings (e.g., hydrology, tree densities, species) have different confounding impacts on radar cross sections (RCS), necessitating models to be regionally calibrated. Here we combine commercial and freely available L-band SAR data to obtain denser time-series and use logarithmic regression to relate RCS values to AGB. The model is calibrated using AGB at 14 AGB validation plots (AGBVs). We calibrate and evaluate the RCS to AGB model using different areas of aggregation (AOA) consisting of small (~0.1 ha. ~subpixel), medium (~3.5 ha, ~2x2 pixels) and large (~14 ha, 4x4 pixels) AOAs. Root-mean-square errors (RMSEs) were about 60 Mg/ha, irrespective of AOA, attributed to poor model performance at ~two AGBVs. However, when instead using NISAR assessment criteria the small, medium, and large ROI respectively had RSME’s of 32, 15 and 21 MG/ha. The AOA methodology was shown to be able to yield greatly improved AGB, estimates even when limited to a pre-defined data gridding scheme such as NISAR’s future AGB products (1 ha grids).