Lakes play an important role in the global carbon cycle; however, there are still large uncertainties in the estimation of global lake carbon emission due to the limitations in conducting field surveys at large geographic scales. Using long-term Moderate-Resolution Imaging Spectroradiometer (MODIS) imagery and field observation data in eutrophic Lake Taihu, we developed a novel approach to estimate the concentration of dissolved carbon dioxide (cCO2) in lakes. Based on the MODIS-derived chlorophyll-a concentration, lake surface temperature, diffuse attenuation coefficient of photosynthetically active radiation, and photosynthetically active radiation, a spatially explicit cCO2 model was developed using multivariate quadratic polynomial regression (coefficient of determination (R2) = 0.84, root-mean-square error (RMSE) = 11.81 μmol L-1, unbiased percent difference (UPD) = 22.46%). Monte Carlo simulations indicated that the model is stable with relatively small deviations in cCO2 estimates caused by input variables (UPD = 26.14%). MODIS data from 2003 to 2018 showed a significant declining trend (0.42 μmol L-1 yr-1, p < 0.05) in the annual mean cCO2. This was associated with a complex balance between the increasing algae biomass and decreasing external inputs of inorganic carbon, nutrients, and organic matter. The high spatiotemporal variabilities in cCO2 were attributed to river inputs and seasonal changes in temperature and algae biomass. The study shows that satellite remote sensing can play an important role in the field of inland water carbon cycling, providing timely much-needed insights into the drivers of the spatial and temporal changes in dissolved CO2 concentrations in inland waters.