![[logo] US EPA](http://www.epa.gov/epafiles/images/logo_epaseal.gif){kind=logo}
A research plan for the detection and mapping of impervious surfaces: a multi-scale, multi-date approach in mid-Atlantic sub-watersheds
RESEARCH SUMMARY
Investigators: S. Taylor Jarnagin
EPA Landscape Ecology Branch,
Environmental Sciences Division,
USEPA/ORD National Exposure Research Laboratory
Mail Drop E243-05
109 T.W. Alexander Drive
Research Triangle Park, NC 27711
E-mail: jarnagin.taylor@epa.gov
Work Office Telephone: 919-541-1987
Work Fax: 919-541-4329
With the advent of the first comprehensive clean water initiatives some twenty‑five years ago (Federal Water Pollution Control Act, 1972), many of the major end‑of‑pipe sources of water pollution such as sewage and industrial waste have been addressed. The predominant cause of remaining water pollution problems may be traced to runoff from NPS (USEPA, 1994). In urban basins these sources are related to human habitation and the subsequent buildup of impervious surfaces. These surfaces impede the infiltration of water into soil and thereby change the flow dynamics, sedimentation load and pollution profile of stormwater runoff (Slonecker, et al., 2001). In populated suburban areas these would include features such as roads, parking lots, roofs, driveways and sidewalks. Due to their ability to alter discharge rates and facilitate transport of pollutants, these land-use features have taken on a particular importance since the post World War II suburban expansion. Most early studies of impervious surfaces and streamflow focused on modeling stormwater flow as an instrument to predict flood frequency (Carter, 1961). In the last twenty years however, assessments have shifted towards an ecological perspective centered on the role of impervious surfaces as a stressor of aquatic environments (Klein, 1979) such that current literature generally recognizes runoff from impervious surfaces as the leading contributor to NPS pollution in urban watersheds (CWP, 2003).
Problem: Although the effect of urban development on water quality has been generally known for almost forty years (Carter, 1961), a basic problem remains in determining the explicit amount and spatial distribution of impervious surface cover on the landscape. The percent imperviousness of a watershed is recognized as a key landscape indicator of ecological condition, and yet timely and accurate quantitative measurements of impervious surface area remain problematic. This is particularly true for local scale mapping efforts over broad regional areas. Most often, in remote sensing investigations, a factor for imperviousness is either estimated from general land use classification without any assessment of accuracy or it is mapped from aerial photographs utilizing labor-intensive methods that cannot be cost-effectively implemented at the regional scale. An automated multi-scale strategy that can accurately map impervious surfaces would, therefore, add significantly to the level of detail and efficiency of current efforts attempting to understand and model ecosystem dynamics. This would then allow for multi-scale ecological impact assessments of impervious surfaces across large drainage areas in a community-based environmental strategy. Of importance to an automated extraction strategy is an accuracy assessment evaluation based upon explicit impervious surface mapping from high spatial resolution data.
While high spatial resolution data, such as aerial photography and IKONOS 1-meter imagery, provide accurate mapping results, they cannot be cost-effectively applied over large areas. Conversely, while medium spatial resolution Landsat imagery may be cost-effectively applied over large areas, there is a lack of research focused on automated strategies using regional or national level data with documented assessments of accuracy.
Hypotheses: Using a multi-scale approach, we propose to determine the uncertainty in quantity and spatial characterization of impervious surface area as derived from aerial photography, National Land Cover Data 92 (NLCD92) and NLCD00 across the mid-Atlantic region.There will be three phases to this investigation, to include:
Phase 1:
Vegetation leaf-off sequential aerial imagery data sets (1949, 1963, 1971, 1979, 1988 and 1994) enabled the interpretive mapping of impervious surfaces in the upper Accotink Creek watershed in Fairfax, County, Virginia. This allowed a determination of the percent of impervious surface for each year. This data, along with concurrent peak streamflow and daily precipitation data was used to investigate the historical relationship between the growth of impervious surfaces and changes in peak streamflow characteristics. The mapped impervious surface datasets will be further utilized as ground truth in Phases 2 and 3 that follow.
Phase 2:
The implementation of a regional-scale strategy that will provide impervious surface maps at the sub-watershed scale, 14-unit Hydrologic Unit Code (HUC), over the wider mid-Atlantic region. This effort will rely on circa 1990 high-resolution truth sets to develop and test the accuracy of impervious surface coefficients derived from the NCLD92.Phase 3:
An accuracy assessment of sub-pixel impervious surface maps that cover large portions of the mid-Atlantic region. Algorithms and truth data used in the previous phases will be utilized to assess the accuracy of the sub-pixel algorithm derived maps at the pixel scale.
Combined, these phases offer a multi-scale integrated approach that will use the local scale impervious surface maps (derived from high spatial-resolution imagery) to alternately model and determine the statistical accuracy of regional scale impervious surface estimates drawn from National Land Cover Data (NLCD) 92 and 00.
Phase 1:
The pilot phase shows that manual, interpretive impervious area mapping is highly accurate. However, the area calculations have been achieved through the use of time consuming and costly cartographic methods (high spatial resolution imagery and interpretive mapping techniques). It is inefficient to extend and implement this interpretive methodology over broad areas such as 8-unit HUCs. As such, ecological research that requires an impervious surface parameter over large areas is dependent upon the development of an automated analytical approach. It is this remote sensing/geo-spatial objective that drives Phases 2 and 3 of this proposal. Additionally, results from phase 1 indicate that the historical increase in streamflow discharge rates has a relationship with historical changes in impervious surface area, irrespective of precipitation. Of ecological interest is the possibility that streamflow thresholds may exist between the various levels of impervious cover.
Phase 2:
Results from phase 2 suggest that a coefficient process, utilizing the NLCD92 as a base, can reasonably estimate watershed impervious surface area compared to truth impervious surface maps. However, this method maps impervious surface area at the subwatershed scale and lacks the explicit placement of impervious cover on the landscape.
Phase 3:
It is anticipated that this phase will provide a statistically rigorous accuracy assessment at the pixel and watershed scales for sub-pixel data. This is important because sub-pixel mapping can provide an explicit placement of impervious cover on the landscape. This provides a mechanism to drive explicit, cell-to-cell, hydrologic models. To date, there has been minimal research emphasis on sub-pixel validation.
Future Remote Sensing Research:
It also should be noted that innovative remote sensing systems and image processing algorithms provide promise for mapping impervious surfaces. Future research needs to examine the alternative approaches to determine their accuracy and efficiency in mapping impervious surfaces. These include the analysis of non-traditional remote sensing datasets such as Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) and EO-1 Hyperion hyperspectral data and LIght Detection And Ranging (LIDAR) which has begun to show promise as a vehicle for extraction of building surface and road features. Also of note are sensor fusion investigations focusing on the ability of merged remote sensing datasets to increase the detection and classification accuracy of anthropogenic land-use features (Hodgson et al., 2003).
Additionally, the investigators involved with this research plan have completed an effort to model the impervious surface parameter from Census population data as part of a USEPA National Exposure Research Laboratory (NERL) internal Annual Performance Measure (Jennings and Jarnagin, 2004). As with models derived from the NLCD92, a successful impervious surface model using Census population data would provide a possible means for mapping impervious surfaces at the national scale, a goal of national level indicator assessments.
Future ecological research will extend the long-term streamflow methods derived in the pilot study and investigate the long-term relationship between urban development and streamflow in subwatersheds throughout the mid-Atlantic region. The initial regional focus will be on a sub-set of subwatersheds used in phase 2 and phase 3, but will be extended to a set of approximately one hundred and fifty (150) subwatersheds where USGS gauge stations have been active over the past 20-25 years.
Phases 1 and 2 are complete and have resulted in numerous publications and presentations (Jennings and Jarnagin, 2002; Jarnagin et al., 2004; Jennings et al., 2004). Phase 3 research, a collaborative effort with the United States Geological Survey (Jones et al., 2003) the Chesapeake Bay Program and the USEPA, is ongoing (Jennings et al., 2004).
References:
Carter, R.W. 1961. Magnitude and frequency of Floods in Suburban Areas. U.S. Geological Professional Paper, 424, B9 - B11.
CWP (Center for Watershed Protection). 2003. Impacts of Impervious Cover on Aquatic Systems, Center for Watershed Protection, Ellicott City, MD., p. 142.
Federal Water Pollution Control Act. 1972. Public Law 92-500. Washington, D. C.
Hodgson, M.E., J.R. Jensen, J.A. Tullis, K.D. Riordan and C.M. Archer, 2003. Synergistic use of LIDAR and color aerial photography for mapping urban parcel imperviousness. Photogrammetric Engineering and Remote Sensing, 69(9); 973-980, URL: http://www.asprs.org/publications/pers/2003journal/september/abstracts.html#973 
Jennings, D., S.T. Jarnagin. 2004. The derivation of impervious surface area from population and remotely sensed data and the application to projections of population growth in the Baltimore-Washington region, EPA/600/X-04/111. United States Environmental Protection Agency, Office of Research and Development, Washington, D.C., September 2004, 42p.
Jarnagin, S. T., D. Jennings, and D. Ebert. 2004. A Technique for Assessing the accuracy of sub-pixel estimates extracted from landsat TM imagery. Chapter 19 (pp. 269-280) in: Remote Sensing and GIS Accuracy Assessment, Lunetta, R. S., and J. G. Lyon (Editors). CRC Press, Boca Raton, FL (320 p.).
Jennings, D., S.T. Jarnagin and D. Ebert. 2004. A modeling approach for determining watershed impervious surface area from National Land Cover Data 92. Photogrammetric Engineering and Remote Sensing, 70(11):1295-1307, URL: http://www.asprs.org/publications/pers/2004journal/november/abstracts.html#1295
Jennings, D., S. T. Jarnagin, J. W. Jones and P. Claggett. 2004. An Accuracy Assessment of Multiple Mid-Atlantic Sub-Pixel Impervious Surface Maps. Poster presented at the USEPA Science Forum, 2004, Washington D.C. URL: http://www.epa.gov/nerl/news/forum2004/jennings_web.pdf
Jennings D., and S.T. Jarnagin. 2002. Changes in anthropogenic impervious surfaces, precipitation and daily streamflow discharge: a historical perspective in a mid-Atlantic sub-watershed. Landscape Ecology, 17(5):471-489, URL: http://www.kluweronline.com/issn/0921-2973
Jones, J.W., P. Claggett, S. T. Jarnagin, and D.B. Jennings. 2003. Shared assessment of USGS and NGO impervious surface datasets for the Chesapeake Bay Watershed. USGS funded PROSPECTUS PROPOSAL: 32p.
Klein, R.D. 1979. Urbanization and Stream Water Quality Impairment. Water Resources Bulletin. 15(4): 948-63.
Schueler T.R. 1994. The Importance of Imperviousness. Watershed Protection Techniques. 1 (3):100-111.
U.S. Environmental Protection Agency (USEPA). 1994. The quality of our Nation’s water:1992. United States Environmental Protection Agency # EPA-841-S-94-002. Washington D.C. USEPA, Office of Water.