Establishment of Forest Dynamic Aboveground Carbon Stock Estimation Model Combining Remote Sensing Factors and Forest Stand Factors

Taking the Pinus densata in Shangri-La City as the research object, based on the 1987-2017 national forest inventory data and Landsat time-series data, we calculated the 5-year and 10-year periodical changes and annual average changes of carbon stock, remote sensing factor and stand factor over a period of 30 years, and extracted remote sensing factor and stand factor that had higher correlation with the aboveground carbon stock of Pinus densata through correlation analysis, with the aim of improving the accuracy of forest aboveground carbon stock estimation. Through correlation analysis, we extracted remote sensing factors and stand factors with higher correlation with aboveground carbon stock of alpine pine, and constructed an aboveground carbon stock estimation model by using the random forest method, and combined the stand factors with the modeling of changes in remote sensing factors, in order to improve the accuracy of aboveground carbon stock estimation in two forests of canopy density and mean age. The results showed that the modeling effect of remote sensing factor combined with 10 year change in mean age was the best, with R² of 0.882, RMSE of 0.398 t/(hm²∙a), and P of 0.755. Three types of changes were used to construct a model for estimating the dynamics of aboveground forest carbon stock, and the model based on 10 year regular change had the best fitting effect and prediction accuracy. The modeling effect of remote sensing factor combined with average age was better than that combined with depression, and its R² enhancement effect was 5.16%, 2.86%, and 1.39% for the modeling of 5year, 10 year, and annual average amount of change, respectively. This study shows that the texture factor can effectively reflect the changes of aboveground carbon stock in forests, and the carbon stock estimation model constructed by combining the remote sensing factor with the changes of stand factor can effectively improve the accuracy of the dynamic estimation of aboveground carbon stock in forests.