Rice is traditionally grown in the flooded paddies, which makes anaerobic soil condition and contributes to significant methane (CH4) emissions. Thus, understanding characteristics of CH4 emission from irrigated rice paddy is important for climate smart agriculture. Land surface models (LSMs) are widely used to estimate CH4 fluxes and help to find the major controlling processes for CH4 production. However, misrepresentation of agricultural management practices like irrigation and difficulties in model parametrization can cause large discrepancies between observation and model estimation. In this study, we estimated CH4 flux from an irrigated rice paddy field using Joint UK Land Environment Simulator (JULES) LSM, and compared with eddy flux measurement data. Observed CH4 flux showed bi-peak distribution due to reduction of CH4 production during mid-season drainage (MSD) periods. The JULES LSM reproduced the CH4 production well after MSD, but failed to estimate CH4 peak before MSD. This is partly caused by improper representation for flooded irrigation condition in JULES LSM. To simulate irrigation effects, sub-model related to CH4 process was applied with observed soil water and water table depth data independently. Sub-model result was better than previous results, but failed to reproduce first CH4 peak also. These results indicate the importance of proper representation of irrigation practice and existence of unrepresented processes related with early peak of CH4 production.