Inverse empirical models can inform and improve more complex process-based models by quantifying the principal factors that control water quality variation. Here we developed a multiple regression model that explains 81% of the variation in filtered methylmercury (FMeHg) concentrations in Fishing Brook, a fourth-order stream in the Adirondack Mountains, New York, a known “hot spot” of Hg bioaccumulation. This model builds on previous observations that wetland-dominated riparian areas are the principal source of MeHg to this stream and were based on 43 samples collected during a 33 month period in 2007–2009. Explanatory variables include those that represent the effects of water temperature, streamflow, and modeled riparian water table depth on seasonal and annual patterns of FMeHg concentrations. An additional variable represents the effects of an upstream pond on decreasing FMeHg concentrations. Model results suggest that temperature-driven effects on net Hg methylation rates are the principal control on annual FMeHg concentration patterns. Additionally, streamflow dilutes FMeHg concentrations during the cold dormant season. The model further indicates that depth and persistence of the riparian water table as simulated by TOPMODEL are dominant controls on FMeHg concentration patterns during the warm growing season, especially evident when concentrations during the dry summer of 2007 were less than half of those in the wetter summers of 2008 and 2009. This modeling approach may help identify the principal factors that control variation in surface water FMeHg concentrations in other settings, which can guide the appropriate application of process-based models.