Data used as input to the models and for spatial analysis were stored in ArcGIS 9.1 in an Oblique Mercator NAD83 projection with units of meters. These data were converted to a raster format for modeling and analysis and then registered to a 26.5-m National Elevation Dataset derived for the Muskegon River Watershed (MRW). Our backcast model was developed using a unique configuration of the Land Transformation Model or LTM (cf. Pijanowski et al. 2002a, Pijanowski et al. 2005, and below). The groundwater travel-time model was developed using MODFLOW-2000 (Harbaugh et al. 2000) and ArcGIS 9.1.
Two dates of aerial photography were used to develop land-use/cover maps from 1978 and 1998 for use in land-use/cover change modeling. For 1978, 1:24,000 color infrared aerial photographs were used to develop the land-use/cover maps as part of a larger statewide management program for mapping natural resources organized by the Michigan Department of Natural Resources (MDNR); this mapping database is called the Michigan Resource Information System, which was acquired from the MDNR in digital form. The land-use/cover database contains land-use/cover features organized in a level-II and -III modified Anderson classification scheme with 83 categories. The minimum polygon size was 0.3 ha, and the minimum mapping distance was 100 ft (30.5 m). To develop our 1998 land-use/cover map, 1:40,000 nominal-scale color infrared photography was rectified to the 1978 land-use/cover database, then land-use/cover polygons were updated to create the second land-use map that could be used to calculate changes in land use/cover. Updating the 1978 map with 1998 information helped to ensure accuracy for change detection using the two land-use/cover datasets. The 1998 land-use/cover database was developed by staff at Michigan State University for all local government units falling wholly or partially in the MRW. The 83 original categories of land use/cover were summarized, for modeling and analysis, into seven Anderson level-I land-use/cover categories: urban, agriculture, shrub/grass, forest, open water, wetland, and barren. Data were rasterized to the 26.5-m cell size creating a grid with dimensions of 7562 columns x 7831 rows and an effective modeling area of 15.2 million cells.