The interaction of groundwater (GW) and surface water (SW) not only sustains the runoff in dry seasons, but also plays an important role in regulating aquatic ecosystems. Hydrological engineers proposed the idea of modeling flood routing using the Muskingum-Cunge method, by ignoring GW-SW interaction during flooding which may however contribute to the outflow. This study proposes an improved nonlinear Muskingum-Cunge flood routing model considering lateral inflow, which is denoted as NMCL1 and NMCL2 that can simulate the flood routing and calculate GW-SW exchange. In addition, both the linear and nonlinear lateral inflow (with the channel inflow) are discussed, and both the stable lateral inflow due to GW-SW exchange and the transient/conventional lateral inflow changing with the river inflow are considered for the first time. Sensitivity analysis has shown that different parameters have different effects on the simulation results. Three different flood cases documented in literature with one measured from Zhongtian River, China, were selected to compare the classical and the updated Muskingum-Cunge methods. Two different floods of the River Wye are selected to verify the accuracy of the calibrated model. Comparison has shown that, for several cases, the proposed method is capable of obtaining the optimal simulation results. For the case of Zhongtian River, the proposed method can estimate the GW-SW interaction and lateral inflow reliably. The proposed method inherits the ability of Maskingum-Cunge in flood routing. Moreover, the new Muskingum-Cunge method can quantify GW-SW exchange, and the estimation has reliably owned to the nonlinearity and sign flexibility of the calculated exchange process.