Abstract: To explore the potential of Sentinel-2 for estimating aboveground carbon storage in arboreal forests on the North Slope of Tianshan Mountains, a pilot study was conducted using Sentinel-2 remote sensing imagery from 2021 and quadrat survey data. By integrating spectral information, vegetation indices, textural features, and topographic factors, optimal variables were screened using mean residual sum of squares (RMS), akaike information criterion (AIC), and adjusted coefficient of determination (R²_adjust). Both the partial least squares regression model and the robust estimation model were developed, with their accuracy evaluated through the coefficient of determination (R²), root mean square error (E_RMS), relative root mean square error (E_RRMS), and bias (b_bias). The results showed that 22 highly significant remote sensing factors (P<0.01) were extracted from Sentinel-2 remote sensing imagery data, with 7 modeling factors ultimately selected through variable optimization. These factors covered three categories: Spectral information (band11, band12, band4, band5), vegetation indices (NDVI, RVI), and textural features (b11-Mean). Both the partial least squares regression model and robust estimation model demonstrated high predictive accuracy and reliability, with the former outperforming the latter. The results indicated the strong applicability of Sentinel-2 for aboveground carbon storage estimation in arboreal forests on the North Slope of Tianshan Mountains.

Key words: Sentinel-2, remote sensing imagery, arboreal forests, aboveground carbon storage, partial least squares regression, robust estimation, the North Slope of Tianshan Mountains