In this paper, we estimated impervious surface from United States census housing density data sets for the conterminous United States to examine the distribution and extent of impaired watersheds, and to estimate the risk to watersheds from development in the near future. We used regression tree methods to relate estimates of current housing density to the 2001 National Land Cover Database (NLCD) percent urban imperviousness. As of 2000, we estimate 83,749 km2 of impervious surface (IS) cover across the United States (about 9.6% lower than the NLCD). We estimate that IS cover will expand to 114,070 km2 by 2030. About 7% of eight-digit Hydrologic Unit Code HUC watersheds 3.6% of the conterminous United States were stressed or degraded 5% IS in 2000, and we estimated that this will increase to nearly double to 8.5% of watersheds by 2030 6.3% of area . We explored the subtle differences of fine-grain pattern for different urban land use types by comparing our national estimates of IS to those developed for the Chesapeake Bay watershed. We also found important nonlinear affects of watershed scale and aggregation, whereby estimates of IS could differ by roughly ten-fold.