As hydraulic, hydrologic, and biogeochemical models evolve toward greater spatial resolution and larger extent, robust morphometric data sets are essential to constrain their results. Here we present the Landsat-derived North American River Width (NARWidth) data set, the first fine-resolution, continental scale river centerline and width database. NARWidth contains measurements of >2.4x10(5)km of rivers wider than 30m at mean annual discharge. We find that conventional digital elevation model-derived width data sets underestimate the abundance of wide rivers. To calculate the total surface area of North American rivers, we extrapolate the strong observed relationship between river width and total surface area at different river widths (r(2)>0.99 for 100-2000m widths) to narrower rivers and streams. We conservatively estimate the total surface area of North American rivers as <mml:msubsup>1.24-0.15+0.39</mml:msubsup>x10(5)km(2) (1 sigma confidence intervals), values <mml:msubsup>20-15+38</mml:msubsup>% greater than previous estimates used to evaluate greenhouse gas efflux from rivers to the atmosphere.