Modelling Forest Aboveground Biomass Based on GF-3 Dual-Polarized and WorldView-3 Data: A Case Study in Datong National Wetland Park, China

Abstract

Aboveground biomass (AGB) models based on field-measured and remote-sensed data can help to understand and monitor ecosystems and evaluate the impacts of human activity. To create improved forest AGB models for use in an ecological rehabilitation area of Huainan, China, a suite of methods was used to evaluate a combination of the Chinese GaoFen-3 (GF-3) satellite’s synthetic aperture radar (SAR) data and vegetation indexes derived from the WorldView-3 satellite. Using vegetation indices and radar backscatter coefficients, a total of six modelling methods were applied to generate three AGB models, which included multivariable linear regression, linear, exponential, power, logarithmic, and growth functions. The results indicate that the observed root mean square errors (RMSE) of the best models, which included exponential functions based on the variables NDVI and HV, as well as their combination in a multivariable linear regression, were 43.74 Mg/ha, 30.87 Mg/ha, and 26.72 Mg/ha, respectively. The best model used multivariable linear regression with combined SAR and NDVI data (R2 = 0.861). The RMSEs were lowest for mixed forest, moderate for coniferous forest, and highest for broad-leaved forest. The results indicate that a combination of optical and microwave remote-sensing images can be used to effectively improve AGB estimation accuracy.