This article presents an empirical approach to correcting for spatial interactions in stated preference data when valuing large-scale, spatially variable environmental improvements. This approach is presented in the context of a contingent valuation study estimating the benefits of reduced non-point source pollution in Green Bay, Wisconsin. The significant spatial variation of water clarity conditions in this large water body was captured using satellite-derived GIS data. This article focuses on two significant challenges: first, ensuring respondents are adequately informed of how the proposed change will impact their individual utility stream; second, dealing with the spatial effects within the estimation model. The GIS water clarity data were used to measure the initial conditions faced by each individual parcel. Including this information in the analysis significantly increased the estimated expected WTP of some individuals but decreased that of others. Some of the difference in aggregated benefits is likely due to issues of spatial correlation between properties that is unaccounted for in the simpler models.