Impervious cover is an important metric for watershed health because it directly interferes with natural hydrology and is correlated with human activities that pollute waterways. It is a particularly convenient metric because national estimates of impervious cover are provided at 30 m resolution by the National Land Cover Database (NLCD). However, NLCD estimates of impervious cover systemically underestimate impervious cover. In the present study, we quantify bias in the NLCD impervious cover map for New Hampshire's coastal watershed by comparing the NLCD raster with a high resolution (1 foot) raster of impervious surfaces developed by NH GRANIT for the Piscataqua Region Estuaries Partnership (PREP). The two rasters were each aggregated to 120 m to smooth fine-scale noise, and cell-level values were compared. Zonal statistics were then used to compute the percent impervious cover using both datasets for subwatersheds identified by USGS HUC12 hydrological units. Finally, we compare the results of projecting future change in impervious cover using the NLCD and PREP rasters as alternative base maps for changing impervious cover in a rapid development scenario for the years 2010 to 2100.
Results/Conclusions
On average, the NLCD map underestimated impervious cover for the coastal watershed by 0.95% compared to the PREP dataset. The NLCD map tended to overestimate impervious cover for 120 m cells in the range 20% to 70% impervious cover, and to underestimate impervious cover for cells less than 10% or more than 90%. For the HUC12 watersheds, NLCD underestimated impervious cover by 1.0 to 1.5% for watersheds below 9% impervious cover, and overestimated impervious cover by 0 to 0.9% for three watersheds with more than 9% impervious cover. This pattern is notable because it has been established for a variety of metrics that watershed health begins to degrade rapidly as impervious cover exceeds 10%. Planners often adapt thresholds in the range of 8 to 10% in identifying acceptable levels of development. If a threshold of 10% is adopted for our results, the NLCD dataset categorizes one coastal watershed as “impacted” that would be classified as “protected” according to the PREP dataset. However, as impervious cover is projected to increase, watersheds reach the “impacted” threshold later in the century with NLCD than with PREP as a base map, potentially reducing the urgency of watershed protection as perceived by planners.