Remote Sensing Estimation of Aboveground Carbon Storage in Picea Schrenkiana var. Tianschanica Forest Based on GF-7 Satellite Data

Objective Forest ecosystems are an important component of the global carbon cycle, and accurate estimation of forest carbon storage is of great significance for understanding and responding to global climate change.

Methods Based on GF-7 remote sensing imagery, this study used the K-means clustering method to divide the study area into five land cover types, and employed Support Vector Machine (SVM), Random Forest (RF), BP Neural Network (BP), and multiple linear regression models to invert and estimate the aboveground carbon storage. By comparing the key indicators such as R², RMSE, and EA among different models, the SVM model was finally determined as the optimal model.

Results The results showed that the SVM model performed excellently across multiple evaluation metrics, including R² (0.841 98), RMSE (1.405 635 t C), and EA (0.841 354), effectively capturing the spatial heterogeneity of carbon storage. K-means clustering analysis revealed that Class 2 contained the most abundant aboveground information and was ultimately selected as the main category for carbon storage inversion.

Conclusion Based on the aboveground carbon storage distribution map of Compartment 3, areas with higher carbon storage were mainly concentrated in the elevation range of 2 111.6–2 330.5 m within the study area, with values ranging from 7.73 to 9 t C. Elevation, sufficient water availability, and mild human disturbance were the main factors contributing to the higher carbon storage in this region. This study provides a scientific basis for carbon storage management and monitoring of natural forests in northern mountain areas, and also offers a reference for future research on more refined carbon storage estimation by integrating multi-source remote sensing data.