Lakes are conduits of greenhouse gases to the atmosphere; however, most efflux estimates for individual lakes are based on extrapolations from a limited number of locations. Within‐lake variability in carbon dioxide (CO2) and methane (CH4) arises from differences in water sources, mixing, atmospheric exchange, and biogeochemical transformations, all of which vary across multiple temporal and spatial scales. We asked, how variable are CO2 and CH4 across the surface of a single lake, how do spatial patterns change seasonally, and how well does the typical sampling location represent the entire lake surface? During the 2016 ice‐free period, we mapped surface water concentrations of CO2 and CH4 approximately weekly in Lake Mendota (USA) and modeled diffusive gas exchange. During stratification, CO2 was generally lower than atmospheric saturation (mean 19.81 μM) and relatively homogenous (mean coefficient of variation 0.12), whereas CH4 was routinely extremely supersaturated (mean 0.29 μM) with greater spatial heterogeneity (mean coefficient of variation 0.65). During fall mixis, concentrations of both gases increased and became more spatially variable, but their spatial arrangements differed. In this system, samples collected from the lake center reasonably well represented the spatially weighted mean CO2 concentration but overestimated annual CO2 efflux by 21%. For CH4, the lake center underestimated annual diffusive efflux by only 8.6% but poorly represented lakewide concentrations and fluxes on any given day. Upscaling from a single site to the whole lake requires consideration of spatial variation to assess lakewide carbon dynamics due to heterogeneity in within‐lake processing, transport to the lake surface, and exchange with the atmosphere.