An estimation of the nonpoint sources of fecal coliform (FC) has been formulated as an inverse parameter estimation problem, using FC observation data from the receiving waters. An inverse algorithm was integrated into the three-dimensional transport model, HEM-3D. The modified Gauss-Newton method was used for the optimal estimation of the loads. The approach is straightforward and requires less effort in modification of the existing hydrodynamic model. Model experiments were conducted to investigate the feasibility of load estimation in a tidal river. The parameter uncertainty and accuracy associated with load estimation are discussed using data sets with different sampling sizes. The inverse model was applied to the tidal Wye River, Maryland, USA, to estimate FC nonpoint sources and to develop an allowable load for the river to attain water quality standards. The model results suggest that the inverse modeling approach is suitable for estimating FC nonpoint sources. The inverse method not only provides an efficient approach to assess the contribution of nonpoint sources, but has the additional advantage of addressing the problems of the uncertainty and error associated with FC simulation in the estuary using hydrodynamic models.