A new geographically weighted stacked regression method for forest aboveground carbon storage estimation: A case study of bamboo forest
Jingyi Wang, Xuejian Li, Huaqiang Du, Guomo Zhou, Fangjie Mao, Mingshi Li, Enbin Liu, Weiliang Fan, Ning Han, Yanxin Xu, Zihao Huang
Abstract
Forest aboveground carbon storage (AGC) is a key criterion reflects the carbon sequestration capacity, health and productivity of forest ecosystem. Bamboo forests are potential to sequester carbon thereby are significant in climate change investigations. Currently, monitoring forest AGC is an essential mission when pursuing Chinese Double-Carbon Goals Policy. However, the spatial heterogeneity of the forest AGC may impact model fitting, necessitating specially designed models to capture and reflect this heterogeneity. Moreover, neighboring pixels in remote sensing imagery are often highly correlated, yet few studies have explored how this spatial correlation affects AGC estimating. In this study, a geographically weighted stacked regression strategy was proposed which added the geographical information to model integration and provided a highly-accurate predictions with R 2 of 0.83, and RMSE at 1.84 Mg ha −1 . Its R 2 increased by 19 % compared to the least accurate model, while the RMSE was decreased by 40 %. The predicted AGC distribution is similar to the actual, which representing the practical value of the proposed method. The resulting AGC maps can targeted develop the bamboo forests in response to ongoing climate change.