The CO2 concentration is an important type of climate forcing and CO2 flux over the earth’s surface is a key component of atmospheric CO2 budget. Since observed CO2 flux is rarely available over extensive areas, most CO2 flux values were estimated only by climate models. The carbon sequestration by the crops only crudely represented in the models. For example, most climate models use climatological or static crop growth and development that do not change from year to year, indistinguishable between flood and drought years. To improve the CO2 flux (i.e., photosynthesis) from crops we coupled crop models (CERES for corn and CropGro for soybean) with the regional model (MM5) along with the land surface model (LSM). This crop-climate coupled model with interactive crop phenology can simulate interannual variations in CO2 and water fluxes from the surface. The coupled model was used to simulate CO2 fluxes in the past couple of growing seasons in the central U.S. Results were compared with available CO2 flux observations at some AmeriFlux sites. It is found that the coupled model gives more realistic seasonal accumulation of CO2 fluxes and that the dynamic crop development in the coupled model has a strong feedback on regional climate, most noticeably precipitation and evapotranspiration, the other two types of key climate forcing. The typical climate models using static crop phenology significantly overestimate CO2 fluxes during early growing season because of positive biases in specifying leaf area index.