Urban forests play a vital functional role in urban ecosystems, serving as crucial carbon sinks for absorbing urban carbon emissions. However, current quantitative assessments of urban forests’ carbon sequestration capacity remain incomplete. Based on the ground-measured forest data, this paper integrates Sentinel-1 radar satellite data and Sentinel-2 optical data. Four linear regression models, namely multiple stepwise regression, ridge regression, elastic net regression, and partial least squares regression, are employed to construct the fitting models for the above-ground forest biomass in Guilin City. Moreover, the fitting effect of the models is tested and their accuracy is verified. The research shows that the partial least squares regression model performs the best in estimating forest biomass, with a determination coefficient R2 reaching 0.733 and a root mean square error RMSE of 38.922 t/hm2. The elastic net regression model ranks second, with R2 = 0.659 and RMSE = 39.477 t/hm2. For the ridge regression, R2 = 0.543 t/hm2 and RMSE = 42.413 t/hm2. The multiple stepwise regression has the worst performance, with R2 = 0.512 and RMSE = 41.895 t/hm2. The vegetation biomass in Guilin City exhibits obvious regional distribution characteristics. The biomass in the main urban area generally concentrates within the range of 120–160 t/hm2. The biomass values in areas such as city center parks are relatively high, and the main high values are distributed in the mountain forest areas on the northwest and southeast sides of the city.