Abstract
This article uses the two-level model (TLM) to predict above-ground biomass (AGB) from TanDEM-X synthetic aperture radar (SAR) data for Sweden. The SAR data were acquired between October 2015 and January 2016 and consisted of 420 scenes. The AGB was estimated from forest height and canopy density estimates obtained from TLM inversion with a power law model. The model parameters were estimated separately for each satellite scene. The prediction accuracy at stand-level was evaluated using field inventoried references from entire Sweden 2017, provided by a forestry company. AGB estimation performance varied throughout the country, with smaller errors in the north and larger in the south, but when the errors were expressed in relative terms, this pattern vanished. The error in terms of root mean square error (RMSE) was 45.6 and 27.2 t/ha at the plot- and stand-level, respectively, and the corresponding biases were -8.80 and 11.2 t/ha. When the random errors related to using sampled field references were removed, the RMSE decreased about 24% to 20.7 t/ha at the stand-level. Overall, the RMSE was of similar order to that obtained in a previous study (27-30 t/ha), where one linear regression model was used for all scenes in Sweden. It is concluded that, using the power law model with parameters estimated for each scene, the scene-wise variations decreased.
Highlights
I N SWEDEN, remote sensing has long been used for wallto-wall forest mapping, demanded both by the wood industry and for more general mapping of natural resources
The models were validated on the National Forest Inventory (NFI) plots used as training data, and evaluated on the stand-level inventory data provided by Sveaskog
The prediction error of above-ground biomass (AGB) in terms of root mean square error (RMSE) was larger for the training data (45.6 t/ha) than the evaluation data (27.2 t/ha), see Table III
Summary
I N SWEDEN, remote sensing has long been used for wallto-wall forest mapping, demanded both by the wood industry and for more general mapping of natural resources. Satellite-based images with 25 m resolution were a valuable resource. The entire country was laser scanned, which resulted in forest mapping products [including forest height and above-ground biomass (AGB)] with 12.5 m resolution, with the majority of the forest covered between 2009 and 2015 [1]–[3]. Due to the active management of Manuscript received February 28, 2020; revised May 11, 2020 and September 4, 2020; accepted October 5, 2020. Date of publication October 13, 2020; date of current version November 11, 2020.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callMore From: IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
