Assessment of above-ground forest biomassby radar methods
I. A. BABIY1, S. T. IM1,2,3,4, V. I. KHARUK2,3
1Reshetnev Siberian State University of Science and Technology, Krasnoyarsk, Ruissia
2Sukachev Institute of Forest (department of FIC KSC SB RAS) Siberian Branch Russian Academy of Sciences, Krasnoyarsk, Ruissia
3Siberian Federal University, Krasnoyarsk, Ruissia
4Katanov Khakassian State University, Abakan, Ruissia
Keywords: biomass, remote sensing, allometry, carbon, monitoring, SAR
Abstract
The dynamics of forest biomass in boreal forests have a significant impact on global carbon cycles. Biomass assessments contribute to understanding the carbon balance of forest vegetation in Siberia. This paper discusses methods for estimating above-ground forest biomass based on radar remote sensing data used in modern research (2010-2021). Methodologies used for biomass assessments are described, including stages of field research, data pre-processing, and modelling of the relationship of remote sensing (RS) data with biomass. Radar sensing has limited capabilities for assessing forest biomass related to characteristics of the survey equipment and parameters of stands. In modern research, a combination of optical and radar data of RS is carried out, which allows to obtain more accurate assessment of biomass using regression models, machine learning, and special techniques (BIOMASAR, SWCM, MaxEnt). The use of data on the optical depth of vegetation cover, estimated from microwave survey data, makes it possible to solve the saturation problem when estimating large amounts of biomass. Comparison of the accuracy of biomass estimation methods is difficult due to the lack of uniform approaches to conducting experiments and calculating errors. Biomass assessment errors based on optical and radar data vary considerably, averaging ~25 %. The assessment of the biomass of boreal forests of Siberia is difficult due to the small amount of supporting field materials. Nowadays, to assess the biomass of boreal forests with a high spatial resolution, it is promising to develop methods based on machine learning algorithms for radar remote sensing data from the Sentinel-1 and ALOS-PALSAR satellites.
|
