Estimation of Aboveground Carbon Using Different Remote Sensing Data and Modelling Techniques

Aim of study: Forests contribute significantly to the global climate by acting as carbon sinks and controlling energy and water flows. This study aimed to model the aboveground carbon (AGC) of pure Scots pine stands within the boundaries of the Sinop Regional Directorate of Forestry in Turkey, using data obtained from various sensor images, including Sentinel-1 (S1), Sentinel-2 (S2), Landsat 8 OLI (L8) and Unmanned Aerial Vehicle (UAV) images, with artificial neural network (ANN) and multiple linear regression (MLR) modeling techniques. Area of study: The study was carried out within pure Scots pine stands located in Sinop Regional Directorate of Forestry. Material and method: a total of 184 sample plots were taken and field measurements were made in these sample plots. 80% of the sample plots (150) were used to fit the models and 20% (34) were used to test the models. The AGC values of each sample plot were estimated with the allometric equation. Brightness values and backscatter values from S1, vegetation indices, reflectance and texture values obtained for different window sizes (3x3, 5x5, 7x7 and 11x11) and different orientations (0°, 45°, 90° and 135°) from L8 and S2, and vegetation indices, band reflectance and digital band obtained from UAV were used in the study. Main results: The results indicated that the texture variables obtained for the 15x15 of the Sentinel-2 image for AGC estimation, together with the MLR modeling technique, were the most successful technique compared to other images and ANN analysis (R2=0.86). Research highlights: The results have shown that AGC can be predicted at high success levels with ANN modeling technique with remote sensing data sets.