Spatial distribution of forest carbon storage in Maoershan region, Northeast China based on geographically weighted regression kriging model.

Abstract

Forest carbon storage has important impacts on climate change. The previous models do not take into consideration of the inherent spatial correlation structure of residual and non-stationary of forest carbon storage which limits the prediction accuracy. Based on ETM+ remotely sensed imagery and 193 fixed plots of Maoershan Experimental Forest Farm of Northeast Forestry University, we established the geographically weighted regression kriging (GWRK) model between forest carbon storage and extracted factors from remotely sensed imagery and topographic factors. The prediction accuracy of GWRK, ordinary least square (OLS) model and geographically weighted regression (GWR) were compared. The results showed that the mean absolute error (MAE) and root mean square error (RMSE) of GWRK were lower than those of OLS and GWR models, while the mean error (ME) of GWRK model was lower than that of GWR model and was close to that of OLS model. The prediction accuracy of GWRK model was 83.2%, which was 6% and 10% higher than that of OLS model (73.7%) and GWR model (77.3%). Therefore, the GWRK model was more effective in estimating forest carbon storage than the others. The mean value of forest carbon storage predicted by GWRK model was 70.31 t·hm-2. The relatively high values presented in high altitude area, indicating that altitude had a great impact on forest carbon storage.