Terrestrial and aquatic ecosystems are intimately linked by the export of elements from watersheds. Although export is influenced by land cover within watersheds, few models evaluate how the spatial configuration of land cover influences loading. In this study we examined spatial variation of land cover at a 10 ϫ 10 m resolution by developing a mass balance, maximum likelihood model of lake iron (Fe) concentrations in 93 watersheds. The model estimated lake iron concentrations based on loading, within-lake processes and losses. Two models were developed. One considered loading from eight land cover types, whereas the second model included the distance of each grid cell to account for Fe losses along flow paths to the lake. In-lake production and losses were accounted for as a function of lake area, water color, and discharge. If we treated watersheds as homogeneous source areas, export was estimated as 450 mg Fe m Ϫ2 yr Ϫ1 ; however, in spatial models export varied from negligible to 5,400 mg Fe m Ϫ2 yr Ϫ1 based on differential loadings from eight cover types. Accounting for losses of Fe based on distance from the lake did not improve the model. Although areal export of Fe was greater from wetlands, upland forests dominate the landscape and thus accounted for on average 75% of the total Fe load. Fe losses from lakes were primarily regulated by discharge; however, water color and lake depth were also important. Overall, the analysis revealed that lake Fe concentrations are related to land cover based on strong differential Fe loadings.