Estimating the Spatial Distribution of Above-ground Carbon of Zagros Forests using Regression Kriging, Geographically Weighted Regression Kriging and Landsat 8 imagery

Estimating aboveground carbon (AGC) of forest is a fundamental task for sustainable management of forest ecosystems; therefore, there is a critical need for appropriate approaches for quantifying of AGC. The most commonly used approaches for estimating include global regression models that estimate the target variable over a wide range using cost-effective auxiliary data. Traditional regression models with fixed regression coefficients at all locations do not consider heterogeneity and spatial structure in modeling. The objective of this study is estimating the AGC using Regression Kriging, Geographically Weighted Regression Kriging and Landsat 8 data and compare methods.

The study was carried out in part of Zagros Forest, in Kohgiluyeh and Boyer-Ahmad Province. Totally, 184 plots (30×30 meters) surveyed and AGC were calculated by allometric equations. 32 variables were extracted from Landsat 8 as auxiliary data in the modeling process. The assessment of accuracies of methods was evaluated by K-fold cross validation via criteria such as coefficient of variation (R2), root mean square error (RMSE).

The results showed that Geographically Weighted Regression Kriging (R 2 = 0.66, RMSE= 21) had a better performance compared to Regression Kriging.

Hybrid methods with heterogeneity and spatial correlation can be a good alternative to early regression methods for estimating aboveground carbon (AGC).