Assessment of Impervious Surface Estimation Techniques

Abstract

Impervious surfaces have been identified as a reliable indicator of the impacts of development on water resources. Research has found that imperviousness has a direct effect on local surface water as well as indirect effects on downstream receiving waters. Some of the affected characteristics of a given watershed due to the increase of impervious surfaces include hydrological impacts (the amount of runoff, peak discharge rates, and baseflow are altered), physical impacts (stream morphology and temperature are changed), water quality impacts (nutrient and pollutant loads increase), and biological impacts (stream biodiversity decreases). While attention has been focused on quantifying these relationships, little work has been done to assess the efficacy of various methods for estimating and mapping impervious surfaces. This paper assesses the results from six techniques for estimating the percent of impervious surface compared to photogrammetrically derived calibration and validation data from high spatial resolution digital planimetric datasets for 53 towns in Connecticut and New York. Impervious surface estimation layers and techniques examined include: (1) National Land Cover Dataset (NLCD) 2001 impervious surface layer, derived through regression tree classification of Landsat Enhanced Thermatic Mapper (ETM) data; (2) Connecticut’s Changing Landscape (CCL) 2002 impervious surface layer, derived through Landsat ETM subpixel classification and quantification of percent imperviousness; (3) and (4) land use-specific coefficients for two different land cover datasets, NLCD 2001 and CCL 2002, as modeled with the Impervious Surface Analysis Tool (ISAT); (5) and (6) a population density and a land cover-based regression model—Estimation Tool for Impervious Surfaces (ETIS)—using US Census Bureau population and CCL 2002 and NLCD 2001 land cover. In comparing results with the reference data, it has been found that estimates of census tract-wide imperviousness based on the direct Landsat-derived spectral methods tested yielded the lowest accuracy with an RMSE of 7.2% (NLCD) and 10.2% (CCL), respectively. The ISAT and ETIS, which leveraged readily available land cover information, and in the case of ETIS, population data, demonstrated the highest accuracy of the techniques examined with an RMSE of 4.4% (NLCD ISAT), 6.3% (CCL ISAT), 4.2% (NLCD ETIS), and 4.7% (CCL ETIS). An assessment of impervious surface estimates will provide decision makers and planners with useful information to guide them in selecting the optimal method of mapping imperviousness given their programmatic needs, available data, and technical resources.