In this study, we used radar data from the ALOS-2 PALSAR-2 satellite to build biomass estimation models and then create a biomass map in Can Gio Mangrove Biosphere Reserve. We used the single variable regression and multivariate regression method, in which 30 sample plots for training model and 15 sample plots for validation model, the coefficient of determination (R 2 ) and root mean square error (RMSE) were used as metrics for evaluating the biomass estimates. The regression analyses showed that the HV polarization was highly related to the biomass, linear model (R 2 = 0.74; RMSE = 28.16), exponential model (R 2 = 0.69; RMSE = 28.73), and polynomial model (R 2 = 0.76; RMSE = 28.03). However, the HH polarization did not show a high relationship with the above-ground biomass, linear model (R 2 = 0.42), exponential model (R 2 = 0.46), or polynomial model (R 2 = 0.42). We also tried multiple linear regression between the parameters extracted from radar image (HH, HV, and textures) and field biomass. The coefficient of determination (R 2 ) between the biomass and two independent variables (HH and HV) was 0.79, and RMSE was 29.78. However, the model with the combination of HV variable and eight texture variables provided a better result (R 2 = 0.81; RMSE = 27.76), in other word the model could explain 81% variation of forest biomass. This model was used to produce the aboveground biomass map in Can Gio Biosphere Reserve in South of Vietnam.