Chapter 8 - Estimating aboveground stand carbon by combining Sentinel-1 and Sentinel-2 satellite data: a case study from Turkey

Abstract

Accurate and consistent estimation of carbon captured in forest ecosystems has become very important for sustainable management and planning of forest resources in recent years. The objective of this study is to determine the success of the relationships between variables obtained from Sentinel-2 and Sentinel-1 satellite data and aboveground stand carbon (AGSC) values generated from ground measurements with different modeling techniques in pure Scotch pine stands. A total of 86 sample plots are used as ground data. The AGSC value of each sample plot is calculated using the regional allometric single-entry equation. The relationships between the vegetation indices and band reflectance values obtained from Sentinel-2 satellite data for each sample plot and the backscattering values (dB) obtained from Sentinel-1 VV and VH polarization with AGSC amounts are modeled by multiple linear regression (MLR) analysis, deep learning algorithms (DLAs), and support vector machine (SVM) modeling techniques. A combination of Sentinel-2 and Sentinel-1 satellite data could increase the accuracy of AGSC estimation by using the SVM [fit index (FI) = 0.877] modeling technique better than DLAs (FI = 0.857) and MLR (FI = 0.748) modeling techniques. The results demonstrated that AGSC in pure Scotch pine stands can most accurately be predicted depending on the combination of Sentinel-2 and Sentinel-1 satellite data.