Jim Meldrum Programmer Analyst Connie Blackmon Director of Data Services Atlanta Regional Commission 100 Edgewood Avenue NE, Suite 1801 Atlanta, Georgia 30335 LAND USE INFORMATION SYSTEM: ADAPTATION OF THE USGS SYSTEM FOR REGIONAL PLANNING ABSTRACT. Under cooperative agreements with the U.S. Geological Survey (USGS), the Atlanta Regional Commission (ARC) successfully adapted the national land use and data analysis system to meet the needs of its regional planning program. Components of the implemented system include a high quality map and computerized file of 1975 land use and cover, 1980 updates of urban land use changes, digital files of geographic boundaries, and an operational software system for statistical analysis and computer mapping. As part of the 1985 update in progress, ARC will implement additional techniques to increase efficiency and improve the accuracy of urban land use data. Included are plans to link the file to ARC's data base on apartment complexes using a geographic point reference file. The implemented system has given ARC a flexible and powerful planning tool with acceptable data accuracy and at considerably lower cost than expected. ARC has used the land use data and maps to develop small area forecasts for ARC's Regional Development Plan and to prepare transportation corridor studies and a number of special studies for business and government, including a pending project of the U.S. Forest Service to project future canopy coverage in the Atlanta Region. This paper discusses these applications as well as the process involved in modifying the USGS system, updating urban land use data and improving the accuracy and efficiency of the system. INTRODUCTION Comprehensive data on land use and cover are an integral part of ARC's planning process. The data are used to assess regional growth trends, to prepare small area forecasts, and to monitor implementation of the Commission's regional plans and policies. Land use data and maps also support a variety of special studies for transportation corridors, watersheds, and other environmentally sensitive areas such as the Chattahoochee River Corridor. In addition, ARC produces summary statistics and maps to meet the needs of local governments and developers for current information on land use patterns and trends.. 29 -
ARC'3 land use Information needs are Intensified by the rapid pace of development in the Atlanta Region. The seven-county region of 2,000 square miles has a current population total of 2.1 million. Between 1970 and 1985 the region added almost 500,000 jobs and more than 300,000 households to reach 1985 totals of 1,118,600 jobs and 747,000 households. In one fiveyear period from 1975 to 1980, almost 102 square miles of agricultural and forested land were converted to urban uses. ARC/USGS COOPERATIVE AGREEMENTS When this growth first exploded in the early 1970's, ARC began investigations to Identify a computerized system to handle both land use data and maps that could be easily updated at frequent (five-year) intervals and at acceptable costs. At that time USGS was launching its Land Use and Data Analysis (LUDA) Program to provide systematic and comprehensive collection and analysis of land use and land cover data on a nationwide basis* As part of this program, USGS had agreed to prepare for the State of Georgia a 1975 land use/cover overlay for a special 1:100,000 scale Greater Atlanta Region topographic map. Since the scope and timing of the LUDA Program came close to meeting ARC's need, in 1975, the Commission entered into the first of three cooperative efforts with USGS to adapt their system to the specific needs of the regional planning process. The national program provides data at levels 1 and 2 of the USGS land use/cover classification system shown in Table 1. Under the first agreement, USGS prepared a separate overlay of certain third-level urban land uses including apartment complexes, mobile home parks, golf courses and cemeteries needed by ARC. They also compiled the ownership boundaries of major institutions and parks as supplied by ARC. This ownership information is needed to support ARC's plans for public facilities and forecasts of land available for development. (The final land use types used by ARC are also shown in Table 1.) In exchange for this compilation, ARC participated in the review and quality control of the 1975 land use overlay. In 1979, due to program delays at USGS, ARC agreed to edit the 1975 digital land use/cover files with USGS technical assistance. The 1975 land use file provided by USGS was a laser-digitized version of a hard copy map scribed as shown in Figure 1. ARC's task was to review this file and correct errors that originated in the digitizing process. ARC also edited a series of geographic overlays covering political boundaries, census tracts, and river basins. - 30 -
TABLE 1 USCS CLASSIFICATION SYSTEM WITH ARC URBAN LAND USES 1. Urban or Built-up Land *11 Residential *112 High Density 115 Mobile Home Parks *12 Commercial and Services *12l Institutional *125 Extensive *13 Industrial *14 Transportation, Communications, Utilities *145 Limited Access Highways *15 Industrial and Commercial Complexes 16 Mixed Urban or Built-up Land *17 Other Urban or Built-up Land *171 Coif Courses *172 Cemetarles *173 Parks 2. Agricultural Land *21 Cropland and Pasture *22 Orchards, Vineyards, and Nurseries *23 Confined Feeding Operations *24 Other Agricultural Land 3. Rangeland 31 Herbaceous Rangeland 32 Shrub and Brush Rangeland 33 Mixed Rangeland 4. Forest Land *41 Deciduous Forest *42 Evergreen Forest *43 Mixed Forest 5. Water *51 Streams and Canals *52 Lakes *53 Reservoirs 54 Bays and Estuaries 6. Wetland *61 Forested Wetland *62 Nonforested Wetland 7. Barren Land 71 Dry Salt Flats 72 Beaches 73 Other Sandy Areas *74 Bare Exposed Rocks *75 Strip Mines and Quarries *76 Transitional Areas 77 Mixed Barren Land 8. Tundra 81-84 9. Perennial Snow or Ice 91--92 * Land use/cover types in the Atlanta Region. Source: USCS Professional Paper 964. - 31 -
Figure 1 Sample From Original Land Use Mylar Then in 1980, under a formal memorandum of understanding, USGS and ARC began a joint technical effort to update the Atlanta land use/ cover map and digital data file. USGS provided 1980 aerial photography, the LUDA software, and technical assistance. ARC undertook the photointerpretation of urban land use change and compiled digital files of 1975-1980 land use change and 1980 land use and land cover. During this process ARC made a number of refinements to the USGS software that have increased both data accuracy and system efficiency. SYSTEM OVERVIEW The Greater Atlanta Region map, which Includes the seven counties in the Atlanta Region, covers the equivalent of 96 7.5 minute quadrangles. In digital form, the amount of land use detail requires subdividing the map into 48 independent sections for processing. Figure 2 shows the 35 sections which form a rectangular plotting window for ARC's seven counties. All or part of 31 sections fall into the region. A single section file may contain as many as 32,000 coordinates. Table 2 below presents summary statistics for three of the digital overlays in ARC's current land use data base. - 32 -
Table 2 LUDA Land Use Statistics for the Atlanta Region Overlay Number of Arcs Number of Coordinates Number of Polygons Number of Sections 1975 Land Use 1975-80 Change Traffic Zones 47,675 3,198 2,418 1,009,428 66,672 105,408 17,680 1,850 875 48 29 4 The USGS Land Use and Data Analysis (LUDA) system consists of three distinct phases, each of which involves several steps. 1. Compilation, Editing and Mapping - Land use interpretation from aerial photography, - Arc digitizing and editing, - Polygon ID digitizing and editing, - Quality control, - Polygon and land use summary statistics, - Shaded land use plots, - Arc plots with polygon ID's. 2. Polygon-to-Grid Conversion (Includes all data restructuring and coordinate conversion) - Merge all sections of a single overlay, - Merge arc and polygon information, - Rotate, translate and scale data, - Convert polygons to gridded data, 3. Statistical Phase - Merge land use with other overlays - Produce master file of gridded data, - Tabulate land use by census tract or other area ARC's adapted land use system follows essentially the same steps outlined above for LUDA with the exception of the statistical phase, which ARC has streamlined into one step. - 33 -
Figure 2 INDEX TO MAP SECTIONS FOR ARC'S LAND USE DATABASE Update Compilation The photographic resources provided by USGS for the 1980 update include NASA high-altitude, color-infrared aerials (original scale 1:130,000) plus black-and-white (B&W) positive transparencies enlarged to 1:100,000 - the scale of the Atlanta base map. In addition, they provided a mylar and scribecoat of the original 1975 land use/cover overlay with polygons and ID's as shown in Figure 1. From the edited 1975 digital file, ARC produced a shaded land use plot with simplified patterns to overlay the mylar and give visual emphasis to its polygon outlines and ID's. To compile 1975 to 1980 changes in urban land use, ARC overlayed the last three resources in the following order: 1. Mylar with 1975 polygons and IDs. 2. Shaded 1975 land use plot. 3. B&U positive transparencies. - 34 -
Then acetate overlays were prepared for each of the 31 map sections covering the ARC region. These acetate sections were registered to the mylar and placed over the photos that were aligned to the underlying land use. Change polygons were outlined on the acetate in permanent felt tip pen and labeled with old (1975) and new (1980) land uses as a single ID value to be placed in the digital file. The color-infrared photography was used as a resource when the resolution of the B&W positives was inadequate. The fine resolution of the color infrared allowed use of ten-power magnification to scrutinize areas in question. The next step involved transfer of change polygons to section plots to improve digitizing accuracy. This step allowed digitizing along original arcs where they coincided with change polygon boundaries thus improving accuracy and eliminating potential sliver problems. Digitizing was performed on ARC's 42" by 60" backlit Altek digitizer using ARC developed software to build polygon arc and ID files which were then input to standard U.S.G.S. editing programs. As a last step, plots of 1975-1980 land use change were compared to the original 1975 land use maps, the photography and the acetate overlays as quality control checks. The update compilation turned out to be the smoothest part of the adaptation process. The interpretation of land use change took only two man months to complete. Digitizing the change polygons was also a relatively fast process. However, there are some data quality problems associated with the scale of the high-altitude photography and the resolution of land use compilation. For the 1985 update, ARC has acquired low-altitude B&W contacts at a scale of 1:31,200 (1"=2,600'). The 1980 land use will be plotted in sections at this scale on transparent mylar with relatively open shade patterns. The plots will be matched to the photography over a light table and changes will be drawn carefully on the mylar in erasible ball point. Enlarged photos at a scale of 1:10,500 (1" = 880*) will be used as a resource where needed. Digitizing will be done directly from the section mylars, thus eliminating the transfer step taken in 1980. Mapping Process At the end of the 1980 compilation, ARC had two digital polygon files of land use data - the original 1975 file and a new file of 1975-1980 changes in urban land use. Each file was ready for mapping. However, there was no automatic procedure available to produce a consolidated file for mapping 1980 land use/cover. To produce this map file, ARC developed a method that involved full editing of the 1975 file to reflect the 1975-1980 changes. Section arc plots with polygon IDs were made at a scale of 1:24,000 (1" = 2000') for both the 1975 land use and the 1975-1980 change. The 1975 land use plot was superimposed over the change plot on ARC's backlit digitizer. Then the 1975 polygon - 35 -
arcs and IDs were edited to incorporate the 1980 land use updates. This process proved to be very complicated and time consuming. As a result, it was completed for only the densest areas and some others of particular interest. Plotting of 1980 land use is now done by superimposing change plots over 1975 plots where consolidated 1980 data do not exist. This technique presents little problem for plots that assign no pattern to agriculture and forestry because most change occurs in these areas. For the 1985 update, ARC plans to obtain software that will convert gridded data back to a polygon file structure. This software will streamline production of a 1985 land use file for shade plotting. Also, the capability of converting grids to polygons will preclude the need for full editing or dual file mapping. Polygon-to-Grld Conversion As mentioned in the System Overview, this phase begins with a set of programs that merge and restructure the section files into one large file in the USGS GIRAS* format with UTM coordinates. Following these steps, the polygon data are converted to grid cell data to facilitate statistical analysis of multiple overlays. The polygon-to-grid (PTG) program is run separately for each land use overlay and each of the related geographic files. The choice of grid cell size is specified at this point. The choice is critical under the LUDA system due to computer time and file storage considerations, which must be weighed against acceptable accuracy. (See Table 3 and Figure 3). In 1980 ARC chose a grid cell size of 100 meters square (2.5 acres) because projected CPU estimates for smaller cell size made such choices impractical. Since 1980 ARC has developed an alternative approach that greatly reduces both the CPU and storage requirements of the LUDA system. This approach uses the condensed raster files output from the USGS PTG program for tabulation rather than LUDA's master grid file, which has one 52 byte record with up to 11 overlay ID fields for each grid cell. ARC's condensed grid files maintain only one record for each row of data. The record length varies with the complexity of polygons intersected by the row. On average, there are 200 fields per row for the most complex land use overlay. * Georgraphic Information Retrieval and Analysis System. - 36 -
TABLE 3 COMPARISON OF ARC AND USGS SYSTEM REQUIREMENTS Grid Cel1 Number of Units (000) File Storage (Megabytes) Tabulation (CPU Minutes) Size* USGS ARC USGS ARC USGS ARC 200 400 125 20 1.7 36 0.75 100 1,700 250 85 3.4 143 1.25 40 10,500 625 525 6.5 600 3.00 20 40,000 1.250 2,000 17.0 2,200 6.00 * Meters square. NOTE: Table 3 data were calculated on an IBM 3031 using ARC's most complex land use overlay. ARC's revised approach also involves a new tabulation program. This program uses an indexed table look up to accumulate data for two overlays, e.g., land use by census tract. On the other hand, the USGS program tabulates data sequentially. The net effect of ARC's condensed file structure and revised tabulation technique is the drastic reduction in CPU time shown in Table 3. As a result, It is now feasible to use more accurate, smaller grid cells. To determine accuracy levels by grid cell size, ARC's traffic zone overlay was gridded at 200, 100, 40 and 20 meters square. Statistics were then tabulated for the 135 smallest zones (300 acres or less). These zones were chosen for the analysis because they are uniquely identified areas close to the size of land use polygons. Total zone acreages were calculated for each of the four grid files and then compared to acreages calculated at the polygon level. Figure 3 reveals serious accuracy problems with both the 200 and the 100 meter size, previously used by ARC. This year ARC will convert its land use data base from UTM to State Plane coordinates so that other ARC data bases can be interfaced more readily to the land use file. After the conversion, both 1980 and 1985 land use data will be gridded at 50 or 100 feet square to achieve the higher accuracy levels associated with the 20 and 40 meter sizes shown in Figure 3. - 37 -
RCCURRCY LEVELS BY CELL SIZE Figure 3 JoonSOUBBE too*sbupae»onsaus«;onjoum GHIO CELL SIZE IN METERS ED 40 Z OR HIGHER ERROR 3 20.0 TO 39.9 Z ERROR ED 10. 0 TO 19. 9 Z ERROR QD s. 0 TO 9. 9 Z ERROR LESS THAN 5.0 2 ERROR RESOLUTION/DATA QUALITY The standard minimum mapping unit specified by USGS is four hectares (10 acres) for all urban and built-up land, water, strip mines, quarries and certain agricultural land. All other land cover areas have a minimum size of 16 hectares (40 acres). In practice, polygons with smaller dimensions are not unusual on the Atlanta map. Analysis of the 1975 file shows that urban land use areas of two hectares (5 acres) and rural areas of four hectares (10 acres) are the norm, which ARC has used for its edit and update. USGS also specifies minimum width criteria that originally precluded mapping central portions of Atlanta's interstate system. They have now been added and the entire interstate system is classified as a separate third-level land use. Even with these finer levels of resolution, some imprecision in land use classification is unavoidable. The understatement of residential. land in rural areas is a common problem because strip developments often fail minimum width requirements. ARC added rural residential land to both the 1975 and 1980 files in many areas that fall below USGS standards. Even so, comparison with 1980 Census counts suggests that residential land is understated in the region's more sparsely settled census tracts. - 38 -
During che small area forecasting process, more serious problems were encountered with high-density residential, commercial and industrial land. In particular, high-density residential land is understated in areas where apartment complexes are mid-to-high rise buildings. ARC plans to offset this problem by linking the land use file to its data base on apartments to develop the precise density measures needed in central Atlanta. The problem with the commercial and industrial land is more difficult to solve. In the first place, interpretation sometimes fail to separate the classes properly. Also, the types of activities defined for each class do not correspond well with the Standard Industrial Classification (SIC). Since ARC's small area forecasts are based, in part, on correlations of land use by type to jobs by SIC, this problem is an area of continuing concern. APPLICATIONS For regional planning purposes, the primary application of the land use system has been the production of summary statistics used in the development of ARC's long-range forecasts of jobs and households for census tracts and traffic analysis zones. In addition, the LUDA shade plotting program, as modified by ARC, has been used extensively to map a wide variety of demographic and economic data. These maps were an important aid to the evaluation and analysis of the forecasts and policies of ARC's Regional Development Plan adopted by the Commission in 1984. In addition to ARC's planning applications, the land use system has been used to produce customized statistics and maps for business and government. ARC digitized Fulton County's planning districts and developed a summary file of 1975 and 1980 land use by district for their planning data base. The project included a number of maps showing land use change within districts, also, if funding permits, the U.S. Forestry Service plans to use 1975-1985 land use/cover data to develop a model for projecting future canopy coverage. Business applications of the system have included AT&T's use of data and maps for staff training in household forecasting techniques as well as real estate market development studies. CONCLUSION The land use information system now implemented by ARC gives the agency a powerful and flexible planning tool with comprehensive facilities for data manipulation, analysis, update and mapping. Using the system, ARC can produce statistical summaries of land use data for counties, census tracts, traffic zones or any userdefined area. ARC also produces a wide variety of computerdrawn, color-shaded maps of land use at variable scales. The maps can be produced for the entire region or selected areas. In addition, ARC can map and tabulate land use changes, such as new residential areas. - 39 -
The system as modified by ARC has potential for other regional planning agencies as well as state agencies, particularly those involved in transportation planning. ARC does not recommend that others begin, as they did, with edits of raw digital data. However, this step is no longer necessary. Despite budget cuts, the LUDA Program is progressing reasonably well. As of May 1986, USGS had completed land use overlays for 90 percent of the nation's 1:250,000 scale quadrangle. Digitizing of the overlays in more than 50 percent complete and fully edited files of land use and geography are available for 22 states. Moreover, both LUDA and a similar, special program for Alaska are funded for completion. To meet regional planning needs, LUDA requires a substantial initial investment of staff time and system resources. For ARC this investment has paid off by providing a flexible, easily replicated system to maintain a consistent data base on land use trends. - 40 -