Assessing Spatial Growth of the Washington Metropolitan Area Using Thematic Mapper Data

Andrew K. Johnston and Thomas R. Watters
Center for Earth and Planetary Studies, National Air and Space Museum, Smithsonian Institution


This work was presented at the 1996 ASPRS meeting in Baltimore.
The text below has been slightly modified from the paper that appears in the proceedings volume for that conference.
Click here for more information on this subject.


ABSTRACT

The Washington metropolitan area has experienced explosive growth in recent years. From 1982 the region has seen vast amounts of construction as the economy and population have grown. This has led to the creation and expansion of many urbanized corridors. Growth continues into the 1990s. As these and other developed regions extend further from the central urban core, the landscape is altered as thousands of hectares of forests and agricultural land are developed and paved. These urban growth patterns, common in many metropolitan areas, are visible throughout the region in satellite images.

Landsat TM images are used in this analysis to asses the extent of urbanization. Three images, from 1982, 1989, and 1993, are rectified and co-registered. A self-iterative, unsupervised clustering algorithm is used to identify developed land within each image. Alternative methods of identifying developed land using multispectral images are explored.

INTRODUCTION

Washington Metropolitan Area

The Washington area has experienced enormous growth in the years 1982-1993. As urbanized areas extend farther from central Washington, land consumed by development is altered in significant ways. The spread of impervious land surfaces and their effect on the physical environment, both geologic and ecologic, should be of great concern to regional planners. The Center for Earth and Planetary Studies maintains a collection of Landsat TM images of the Washington area. These images can be used to follow the sprawl of impervious surfaces as the metropolitan area expands. This paper is part of an ongoing effort to identify methods of achieving these goals. The study area covers 1,860 square kilometers. It is centered on the District of Columbia and extends between Dulles Airport to the west and the Patuxent River to the east.

Remote Sensing of Urban Areas

There are many strategies for mapping the extent of urbanization with satellite images. Most of these focus on efforts to perform multispectral classification algorithms. These methods could be used to frequently generate accurate land cover maps useful to planners and other authorities (Forster, 1985). The heterogeneous nature of urban areas, however, creates problems for classification (Haack et al., 1987). It may be possible to improve classification accuracy by using difference images and texture information (Jensen and Toll, 1982), and by submitting the imagery to a low-pass filter (Toll, 1985). Mutitemporal images can also be used (e.g. Martin and Howarth, 1989) to detect changes in land cover classes. An alternative method of describing urban land cover is the V-I-S model proposed by Ridd (1995). This model describes biophysical composition by assigning values for three components of the urban landscape: impervious surface, soil, and vegetation.

DATA AND METHODOLOGY

Landsat Thematic Mapper images of the Washington area were used in this analysis. Three TM scenes (path/row: 15/33) were used; November 1982, October 1989 and October 1993. The scenes were geometrically corrected and registered to a UTM projection using a second order polynomial.

An ISODATA classification was performed on bands 2,3,4,5 and 7 of the TM data, producing 16 classes. These were manually interpreted and grouped into "paved" and "non-paved" ground cover types. EASI/PACE software by PCI was used for the classification.

For the purposes of comparison, another method was tested. A normalized difference vegetation index (NDVI) image was computed for the 1993 scene from bands 4 and 3. It should be possible in the future to use NDVI values to explain the vegetation component of the V-I-S model. For the current study, a threshold was applied to the NDVI image to indicate areas lacking vegetation.

Field surveys were conducted to assess the accuracy of the classifications. Stratified random points were generated. These were located on the ground using GPS and visually interpreted to be "paved" or "non-paved."

RESULTS

The extent of developed impervious surfaces has increased 1982-1993. Using the estimated "developed" area in 1982 as a baseline, the area covered by impervious surfaces with the study area increased by approximately 5% between 1982 and 1989. Between 1989 and 1993, the area covered by impervious surface increased by about 2%. An image graphically demonstrates the spatial distribution of the spread of developed surfaces.The fastest pace of this development took place along major suburban highways, especially in northern Virginia. These preliminary results suggest that the rapid urban growth experienced in the study area during the 1980s has continued into the 90s. An accuracy assessment shows that the multispectral classification provides a reasonable estimation of urbanization in 1993.

The NDVI method provided a high-resolution view of urban structure. However, it failed to discern impervious urban surfaces from bare soil, which is commonly found in agricultural areas. An NDVI may prove useful as a part of a V-I-S model, but it needs to be combined with a spectral classification to separate similar land cover types.

REFERENCES

Forster, B.C. (1985), An Examination of Some Problems and Solutions in Monitoring Urban Areas from Satellite Platforms, International Journal of Remote Sensing, 6:139-151
Haack, B., Bryant, N., Adams, S. (1987), An Assessment of Landsat MSS and TM Data for Urban Digital Classification, Remote Sensing of Environment, 21:201-213
Jensen, J.R., Toll, D.L. (1982), Detecting Residential Land-Use Development at the Urban Fringe, Photogrammetric Engineering and Remote Sensing, 48:629-643
Martin, L.R.G., and Howarth, P.J. (1989), Change-Detection Accuracy Assessment Using SPOT Multispectral Imagery of the Rural-Urban Fringe, Remote Sensing of Environment, 30:55-66
Ridd, M.K. (1995), Exploring a V-I-S (Vegetation-Impervious Surface-Soil) Model for Urban Ecosystem Analysis Through Remote Sensing: Comparative Anatomy for Cities, International Journal of Remote Sensing, 16:2165-2186
Toll, D.L. (1985), Landsat-4 Thematic Mapper Scene Characteristics of a Suburban and Rural Area, Photogrammetric Engineering and Remote Sensing, 51:1471-1482



CEPS Home
CEPS Research

Revised: April 2, 2001 (jh)
1994-2001 National Air and Space Museum