Improving estimates of simulated runoff quality and quantity using road-enhanced land cover data

Hydrologic model estimates of runoff quantity and quality are appropriately sensitive to the extent and distribution of impervious area (IA) inputs. This research enhances land use maps of IA extracted from road networks and demonstrates its impact on hydrologic model parameters for 704 watersheds in New York. National Land Cover Database (NLCD) 1992 and 2001 data at 30 m pixel resolution provided baseline maps, and enhancement involved reclassifying pixels with overlapping high accuracy vector roads. NLCD 1992 is limited by its lack of explicit IA estimates, instead relying on wide ranges of IA for developed classes, and by the failure of its transportation class to capture most roads. NLCD 2001 has no transportation class, and its explicit IA estimates were found to underestimate IA. Road enhancement caused significant increases in hydrologic parameters such as the curve number, runoff coefficient, and event mean concentration based pollutant loads, with a p<1E-3 for a paired, two-tailed t-test. Magnitude of hydrological parameter changes increased with increasing watershed road density. Roads in the enhanced NLCD were treated as a directly connected impervious area, which caused large differences in pollutant loads generated by explicit runoff simulation. While NLCD allowed runoff to filter through wetlands and forests, enhanced NLCD directly discharged polluted road runoff and doubled phosphorus export coefficient estimated loads. This research suggests that watershed models use road enhanced NLCD to represent IA.