Changes in the waxing and waning of carbon pools in forest ecosystems have a profound impact on the global carbon cycle. The study takes forests in a city as an example, uses sample plot survey data and Sentinel-2 and GF-1 satellite remote sensing data as data sources, and after multiple feature extraction, the RBF neural network optimized by the PSO algorithm is used for modeling the dynamic estimation of carbon storage, assessing the accuracy of the model and inverting the estimation of carbon storage in the sample area. The results show that the PSO-RBF model in this paper has the best fit and prediction accuracy for carbon storage estimation, with the highest R² of 0.644 and the smallest RRMSE and MAE of 0.287 and 15.368 in the combination of the two data sources, and the estimated value of forest carbon storage in 2022 in the sample area is 6809451.80t, which differs from the actual value by 0.04% and the difference of predicted and actual values in each subarea is 0.04%. The predicted and measured values in subregions had a difference of less than 0.34%. The forest carbon storage model that utilizes PSO-RBF has a good estimation effect overall. The technical support provided by this study enables accurate monitoring of carbon storage in forest ecosystems.