The Baltimore-Washington Spatial Dynamics and Human Impacts Study builds on earlier research efforts that mapped urban land use change for the San Francisco Bay area (Acevedo and Bell, 1994; Bell and others, 1995; Kirtland and others, 1994). In developing a temporal database (Acevedo and others, in press), the team participants hope to provide data that can be used to study patterns of urban growth; assess ecological, environmental, and climatic impacts of urban change; and model and predict future urbanization patterns and impacts (Clarke and others, 1996). Both the San Francisco and Baltimore-Washington regions were selected because of the rapid urban growth and resulting impacts on their ecosystems. The Chesapeake Bay region in particular has undergone extensive environmental agitation due to the hydrologic problems that have arisen from the increase in impermeable surfaces and structures, that is buildings and pavement that physically cover the soil. Because of the inability of water to percolate into the ground, little purification occurs by filtration. Water runs over paved surfaces and quickly washes high levels of toxins directly into the water system. Toxins like gasoline, oil, and fertilizer have dramatically affected the local streams, rivers, and the bay.

This paper describes the techniques used to map the extent of urban areas for Phase I and does not discuss Phase II in detail because the work is still in progress. In this study, urban development is defined as areas of intensive use, with much of the land covered by structures. The built-up areas are characterized by the existence of a systematic street pattern, and the relative concentration of buildings and associated intensive use areas, such as parking lots. Using this definition, urban development does not refer to political boundaries and may include incorporated or unincorporated areas as well as military reservations.

To build the urban component of the temporal database, a multidisciplinary team was assembled and a phased approach initiated. Expanding on procedures developed for the San Francisco Regional Study (Bell and others, 1995), the team developed data definitions, a classification scheme, compilation criteria, mapping specifications, guidelines for source materials, and metadata specifications to support development of a logically consistent dataset. Extensive documentation procedures were established to ensure consistency in data collection, and for subsequent application to other regions. Phase II was the implementation of the regional mapping effort.

The study area for Phase I consisted of an approximate area of 15- by 15-minute segment centered around the city of Baltimore (fig. 1). Phase I was used as a prototype for the technique development and integration that the multiagency collaborative effort would require. The regional study, Phase II, encompassed a 2-degree square centered on Washington, D.C. With more than 7 million people spread across 39 counties, the Baltimore-Washington region is one the Nation's fastest growing metropolitan areas. The two cities are rapidly merging into one.

Figure 1. Extent of Baltimore-Washington study, Phase I - inner box, Phase II - outer box. The Phase II area is composed of two 1:250,000-scale topographic maps.

The Mixed Urban or Built-up Land category was used when specific identification of features was limited by the available source materials.

Urban Classification Scheme

Table 1. The land use and land cover classification system used in the temporal mapping of urban development.

The mapping of urban development in our study differs from the USGS LULC mapping program (Loelkes, 1977) in a number of ways. First, this study was only mapping the Urban or Built-up Land category and omitting the rest of the features from the scheme, such as Agricultural Land, Rangeland, Forest Land, Barren Land, and Tundra. Second, this study used a much more stringent density criteria in defining the residential class. Housing and road density were used to delineate residential areas. Third, this study included a temporal database as opposed to a single date. Fourth, this study mapped the actual extent of built-up areas, structures that undeniably disturbed the natural terrain, as opposed to the broadly interpreted features, such as residential areas, where for every 10 acres only 4 houses were represented from the USGS LULC mapping program. Fifth, this study did not require a distinction between all the Level II Mixed Urban or Built-up Land categories. The detail in historical maps and the resolution of Landsat data precluded a rigorous effort to differentiate between the Residential, Commercial, Industrial, and Transportation categories. As a result, polygon delineations were assigned to the Mixed Urban or Built-up Land category when specific identification was not possible.

Table 2. Source materials used for the development of the Baltimore urban database Phase I Source Materials

Year	Data Source	Scale
1792	Baltimore County Plan	1:12,474
1801	Baltimore County Plan	1:12,276
1822	Baltimore County Plan	1:13,200
1851	Baltimore County Plan	1:7,200
1878	Baltimore County Plan	1:158,400 - 1:31,680
1890	USGS Topographic Maps	1:62,500
1900	USGS Topographic Maps	1:62,500
1925	Election District Map	1:62,500
1938	USGS Topographic Maps	1:62,500
1953	USGS Topographic Maps	1:24,000
1966	USGS Topographic Maps	1:24,000
1972	Landsat MSS USGS GIRAS Land Use and Land Cover Digital Data	57 by 79 meters per pixel 1:250,000
1982	Landsat TM USGS Digital Line Graphs (DLGs)	30 meters per pixel 1:100,000
1992	Landsat TM	30 meters per pixel

Historical Sources For 1792 through 1966, historical maps (fig. 2) were acquired from the USGS Cartographic Information Center, Maryland Historic Trust's archives, historical urban plans with proposed development from the Library of Congress, and the USGS National Mapping Division map archive.

Figure 2. One example of the many historical sources use for urban collection. Located within the limits of Baltimore City, dated 1923 from an Election District map.

An unusually large number of maps were acquired for Phase I due to the small spatial extent of the area and the rich cartographic history of Baltimore. Topographic maps covering the regional area, Phase II, were not as abundant. The maps were inventoried and examined to determine which would be the most readily available, and which would offer the best available data at the most desirable scale. The Election District maps that covered an entire county at a scale of 1:62,000 were deemed the most appropriate for regional coverage.

The time frames in the database were selected by the average map date. A range between plus or minus 3 years was the goal and what was achieved. The dates of the maps for the different time periods was reflected in the metadata documentation. The maps dated from 1890 to 1966 were exclusively USGS maps. They offered enough detail to allow collection for all Level II features and for Level III for residential areas. The older maps did not offer as much detail and thus only the Mixed Urban or Built-Up Land distinction was made.

Contemporary Sources For 1972 through 1992, satellite, DLG, and GIRAS LULC data were used to derive the urban area. Landsat satellite data provided the most continuous and uniform data available at the regional scale. The Landsat MSS, with a 57- by 79-square meter pixel spatial resolution, and the TM scanner, with a much better 28.5-square meter resolution, provided the data for monitoring development at 5 year intervals starting in 1972. The LULC data provided an additional resource for 1972. The USGS GIRAS LULC data provided a digital source of LULC maps and associated overlays at 1:250,000 scale (Fegeas and others, 1983). The 1:100,000-scale DLG transportation feature data were used to determine the extent of the urban areas in the 1980's.

General Guidelines Residential areas were characterized either by housing density, road density, spectral reflectance, or degree of land disturbance depending on the source materials. A residential density of three houses per 2.5 acres was approximately the level of development interpreted as urban. Dense residential areas were determined by the existence of dense systematic street patterns and the relative concentration of buildings.

The Transportation, Communications, and Utilities category contained only the actual extent of the intensive area. For example, airport facilities included the runways, terminals, parking lots, and other intensive use areas but did not include the surrounding open land. Major freeways and highways were not delineated unless they were surrounded by urban development. Major roads were instead captured as a separate principle transportation data layer (Clark, in press).

All areas identified as built-up, where individual uses could not be separated using the available source material, were assigned to the Mixed Urban or Built-up category. This category typically contained areas of residential, commercial, industrial, and transportation land uses.

Land having less intensive uses may have been located in the midst of urban development. Less intensive land that was surrounded by built-up land in an urban setting was delineated as Other Urban or Built-up Land. This category consisted of areas such as urban parks, golf courses, cemeteries, and undeveloped land within the urban setting.

Specific Guidelines The minimum mapping unit for all features was defined as 10 acres and the minimum polygon width was defined as 1/10 mile (528 feet). These criteria were selected with the anticipation of using 1-meter resolution image data in later phases, and were based on the current LULC data collection standards (USGS, 1992). Using a 10-acre minimum polygon size allowed us the flexibility of using a wide variety of source material resolutions.

The pink and gray tint depicted on the USGS topographic maps represents dense residential areas where there were too many houses and other building structures to be individually symbolized. Within tinted areas, all residential houses were replaced by the tint; only selected individual structures were shown, such as public use buildings. When pink or gray tint was present on the maps (USGS, 1961), it was automatically categorized as high density residential. Pink tint helped to expedite the collection process when it was available. Residential appendages that typically developed along transportation routes that radiated from the cities were captured as low density residential. Small towns and villages that were not linear were also collected in this category.

Linear residential developments that grew along highways were defined by the following:

• a minimum width of 1/10 of a mile.
• at least 1/2 mile in length.
• at least 12 houses total, represented on both sides of the road - taking care to note that one "symbolized" house may represent more than one house, depending upon the scale of the source.

• systematic road network.
• at least 3 city blocks, and at least 12 houses depicted within the 10-acre minimum polygon size - taking care to note that one "symbolized" house may represent more than one house,depending upon the scale of the source.
• a main highway or railroad, or both, travels through the city.
• town's name in weight and text size used as a general que-in to alert us of a potentially large village or city to be defined.

Figure 3 . A historical perspective of urban development for the Baltimore area from 1792 through 1992.

Collection from Historical Sources When using historical maps, urban extent was compiled by map interpretation and delineation on mylar overlays (fig. 3). Some cities that were not otherwise identified were assigned a seed point if they were listed as historically significant cities by the U.S. Bureau of the Census tabular list of cities and dates with at least 500 people. After the mylar overlays were annotated (fig. 3), they were digitized into a geographic information system (GIS), edited, and projected to a Universal Transverse Mercator (UTM) grid. For each time period there were from one to six maps used to compile the Baltimore area. A GIS file was as generated for each of these maps and when verified they were appended into one file that represented the time period. Much more detailed process information are available in the metadata database.

Collection from Contemporary Sources When mapping from contemporary sources (fig. 4) urban development was determined by various image processing procedures that combined the spectral information from Landsat data with the dense systematic street pattern derived from the DLG road network, and the GIRAS land use and land cover interpretations. The Landsat-derived boundaries were determined by the spectral signatures commonly interpreted as concrete, asphalt, buildings, roads, residential neighborhoods, and commercial buildings. The 30-meter per pixel size TM data were first resampled to a UTM projection, zone 18, North American Datum of 1927. Next, an unsupervised training and multi- spectral classification of the registered data were computed using a maximum likelihood classifier. Unsupervised classes were then interpreted and urban classes identified. The DLG road network was rasterized and processed to derive a built-up delineation established by the proximity of roads. The GIRAS data were converted to a 30-meter UTM grid and compared to the satellite data. Misclassification or spectral confusion in the TM data were minimized by combining the Landsat interpretations with ancillary data such as the GIRAS LULC or the DLG-derived interpretations.

Data Integration A rigorous data integration effort was necessary to improve data quality and consistency because the database was compiled from different sources at varying scales and by different participants of the team. The collection criteria helped to minimize inconsistencies between the different groups but inconsistencies due to the different source materials required a closer evaluation of the individual polygon coverages. The Internet was used to view and transport the data between team members. Problem areas were illustrated, discussed, and resolved in weekly conference calls. One of the key potential errors we looked for was evidence of built-up areas diminishing as we viewed the database from past to present. When the data passed quality verification, a time-series animation of the database was generated (Masuoka, in press). Metadata Documentation To conform to National Spatial Data Infrastructure requirements, the study complied with the Federal Geographic Data Committee (FGDC) Content Standards for Digital Geospatial Metadata (FGDC, 1994). All techniques, lineage of source materials, and procedures were documented. We evaluated, and are currently still in the evaluation stage, the use of ARC/INFO DOCUMENT, a program written by the USGS and the Environmental Protection Agency that aided in the management of metadata collection efforts.

Because of the collection of the metadata, users will be able to determine if the dataset meets their needs, what data were available, and how to obtain them by referring to:

• lineage of data
• standards, methodology, and criteria applied
• terminology and definitions
• intended use of these data
• data quality

RESULTS

Acevedo, W., Foresman, T., Buchanan, J. T., in press, Origins and Philosophy of Building a Temporal Database to Examine Human Transformation Processes, Proceedings, ASPRS/ACSM Annual Convention and Exhibition, Baltimore, MD.

Anderson, J.R., Hardy, E.E., Roach, J.T., and Witmer, R.E., 1976, A Land Use and Land Cover Classification System for Use with Remote Sensor Data: U.S. Geological Survey Professional Paper 964, 28 pp.

Bell, C., Acevedo, W., and Buchanan, J.T., 1995, Dynamic Mapping of Urban Regions: Growth of the San Francisco-Sacramento Region, Proceedings, Urban and Regional Information Systems Association, San Antonio, TX, pp. 723-734.

Clark S. C., in press, Development of the Temporal Transportation Database for Analysis of Urban Development in the Baltimore-Washington Region, Proceedings, ASPRS/ACSM Annual Convention and Exhibition, Baltimore, MD.

Clarke, K.C., Gaydos, L., and Hoppen, S., 1996, A Self-Modifying Cellular Automaton Model of Historical Urbanization in the San Francisco Bay Area, Environment and Planning B, Santa Fe, NM.

Kirtland, D., Gaydos, L., Clarke, K.C., De Cola, L., Acevedo, W., and Bell, C., 1994, An Analysis of Human-Induced Land Transformation in the San Francisco Bay/Sacramento Area, World Resource Review, Vol. 6, No. 2, pp. 206-217.

Federal Geographic Data Committee, 1994, Content Standards for Digital Geospatial Metadata (June 8) Federal Geographic Data Committee, Washington, D.C., pp. 54.

Federal Geographic Data Committee, 1995, Development of a National Digital Geospatial Data Framework, Federal Geographic Data Committee, Washington, D.C., pp. 23.

Fegeas, R.G., Claire, R.W., Guptill, S.C., Anderson, K.E., and Hallam, C.A., 1983, Land Use and Land Cover Digital Data, U.S. Geological Survey Circular 895-E, 21 pp.

Loelkes, G.L., 1977, Specifications for Land Use and Land Cover and Associated Maps, U.S. Geological Survey Open File Report 77-555, 103 pp.

Masuoka, P., Acevedo, W., Fifer, S., Foresman, T.W., and Tuttle, M.J., in press, Techniques for Visualizing Baltimore Regional GIS Data, Proceedings, ASPRS/ACSM Annual Convention and Exhibition, Baltimore, MD.

U.S. Geological Survey, 1992, The National Mapping Program Technical Instructions, draft Standards for Digital Line Graphs for Land Use and Land Cover, U.S. Geological Survey, Reston, VA., 52 pp.

U.S. Geological Survey, Department of the Interior, 1961, Topographic Instructions of the U.S. Geological Survey.

FOOTNOTES********************************

{1.} Any use of trade, product, or firm names is for descriptive purposes only and does not imply endorsement by the U.S. Government.