Prediction of Urban Forest Aboveground Carbon Using Machine Learning Based on Landsat 8 and Sentinel-2: A Case Study of Shanghai, China

The aboveground carbon storage (AGC) of urban forests is an important indicator reflecting the ecological function forests. It essential to monitor AGC and analyze their spatiotemporal distributions. Remote sensing a technical tool that can be leveraged accurately forest AGC, whereas machine learning algorithm for accurate prediction AGC. Therefore, in this study, single Landsat 8 (L) remote data, Sentinel-2 (S) combined (L + S) data are used as sources. Four methods, support vector regression (SVR), random (RF), XGBoost (extreme gradient boosting), CatBoost (categorical predict based on two phases sample plots Shanghai. We chose optimal model simulate distribution. study shows both models separate OLI satellite distribution Shanghai forest. Nevertheless, accuracy CatBoost-integrated higher than others, with fitting validation R2 values 0.99 0.70, respectively. RMSE was also smaller at 0.67 6.29 Mg/ha, uncertainty spatial derived from 2016–2019 small consistent actual situation. Furthermore, statistics showed increased 24.90 Mg/ha 2016 25.61 2019.