Diatom-based bioassessment in wetlands requires quantitative characterization of spatial and temporal variability of the diatom assemblages within each wetland. The purpose of this study was to examine surface-sediment diatom distributional patterns in a wetland to determine how best to sample these systems to capture spatial variability in the assemblage. Diatoms and environmental conditions were characterized from 29 sampling points within a wetland in the floodplain of the Columbia River, Oregon, USA. A total of 159 diatom taxa were identified in the surface-sediment samples. Species richness was high at each sampling point (median: 42, range: 23–57), and relative abundances of common taxa varied between 15 and 39% throughout the wetland. Assemblages contained taxa with both benthic (e.g., Staurosira construens, Nitzschia palea, Fragilaria capucina, Achnanthidium minutissimum) and planktonic (e.g., Aulacoseira granulata and Tabellaria spp.) preferences. Non-metric multidimensional scaling (NMDS) techniques detected differences in the low marsh and upper marsh sediment diatom assemblages. Geostatistical analysis showed spatial autocorrelation of the diatom assemblage in the wetland, measured as semivariance and Moran’s I. A simulation procedure indicated that changes in diatom species richness stabilized after approximately five samples were composited. Our results suggest that the wetland surface sediment diatom assemblage is heterogeneous and that hydrologic gradients may be an important structuring force. Diatom-based bioassessment has the potential to be a useful tool in assessing wetland environmental conditions; however, the shallow nature and complex hydrology of these systems require careful sampling design to adequately characterize the diatom assemblage.