A model-based 72-hour maximum precipitation (MP) was estimated for American River Watershed (ARW) in California by means of a physically-based numerical atmospheric model (Ohara et al. 2011). First, comparing the simulated basin-averaged precipitation against the PRISM data, a regional atmospheric model, MM5, was calibrated. Second, using the calibrated regional atmospheric model, the model-simulated precipitation field in ARW was successfully validated at nine individual rain gauge stations in the watershed. Then, the 1997 storm event, one of the severest storm events over ARW, was maximized by the modification of the model boundary and initial conditions. The initial and boundary conditions in the outer domain of the atmospheric model were modified by three methods: 1) the atmospheric moisture, 2) the storm event duration with maintaining equilibrium atmospheric condition, and 3) spatially shifting the atmospheric conditions. It was found that these modifications of the model boundary conditions significantly increased the precipitation over ARW. These results clearly indicate the importance of wind and moisture conditions at the boundary of the atmospheric modeling domain. The artificially maximized storm yielded 549 mm of 72-hour precipitation depth by the combination of the humidity and storm duration maximization, and 541 mm depth by shifting the atmospheric conditions to the ones toward south by 5.0 degrees. Consequently, the 72-hour maximum precipitation of the 1997 event over ARW can be maximized to as high as 550 mm. Although this study presents only a demonstrative maximization work, it shows that the presented modeling approach can be a potential alternative to the standard Probable Maximum Precipitation (PMP) estimation without depending upon the linear relationships required in the standard PMP method.