Feature For any time period, the geophysical model can predict deformation at hundreds of discrete points over the Great Lakes region. These discrete predictions can be interpolated using the spline interpolation methods, available in the ArcGIS Spatial Analyst extension, to yield a deformation raster at each time of interest. interest Just as the locations of modern rivers and lakes are a result of topography, so too did topography in past ages control surface hydrology. To determine past topography, the deformation raster map was subtracted from a modern digital elevation model (DEM). The Shuttle Radar Topography Mission (SRTM) DEM was used over the land areas and combined with Great Lakes bathymetry data from the National Geophysical Data Center of the U.S. National Oceanic and Atmospheric Administration (NOAA). Because the bathymetry dataset was not complete, it was supplemented with Lake Superior bathymetry provided by the Large Lakes Observatory of the University of Minnesota. To speed calculations, the raster resolution was changed from 90 meters to 200 meters. The result was a series of PaleoDEMs that spanned the last 20,000 years. Mapping Great Lakes History Through extensive fieldwork, geologists have determined the extent of the ice sheet as it retreated from the Midwest. Maps of the ice sheet extent were digitized. The ice sheet north of the margin was assigned an arbitrary thickness of 1,000 meters. By adding ice sheet raster maps to the Paleo-DEMs, the actual topography during late-glacial and post-glacial times was re-created. With this topography, the hydrology tools in ArcGIS Spatial Analyst were used to determine paleohydrology. A first step in modeling modern hydrology, given a DEM, is usually to fill any closed depressions in the topography to correct for any small imperfections in the DEM that would incorrectly serve as sinks for water. In the present application, these filled depressions are critical to the solution because the lakes that formed on the ancient landscape are exactly those that result from filling the Paleo-DEM. Subtraction of the Paleo-DEM from the filled Paleo-DEM yields a raster of Paleo-lakes and their bathymetries. Not only can the ancient distribution of lakes and shorelines be predicted but so can the www.esri.com volumes of those lakes. Using the Hydrology Flow Accumulation tool, which determines the number of raster cells above a given cell that would contribute overland water to that cell, the locations of ancient stream channels could be predicted on the filled Paleo-DEM. In addition, the outlet of each lake was located. It was identified as the single raster cell on the margin of each lake with the highest flow accumulation value. While the locations of hundreds of lakes are predicted for each time period, most of these lakes are very small. To limit the analyses to the largest lakes, the paleolakes raster was searched for all lakes with contiguous areas larger than a specific minimum area. These steps were automated using a Python script. A record of the 20,000-year history of the Great Lakes was produced after only about three days of computational time on a PC. Modeling Results When ice covered much of the Great Lakes basin, water flowed out through Chicago, down the Illinois River, and into the Mississippi River. When ice had retreated north of the Great Lakes, an outlet in Canada that had been depressed under the weight of the ice became the lowest outlet. It captured the drainage as soon as the outlet became ice free. This caused the lakes to drain rapidly. Lake Michigan, which experienced a 125-meter drop in lake level, became separated from Lake Huron. Subsequently the northern controlling outlet rose as the earth readjusted to the reduced load of the melting ice until the outlet reached the same elevation as the more southern outlets. When that happened, drainage of the Great Lakes shifted to the south and the modern outlets. Lake Erie, which almost drained completely because its northern outlet was low, gradually filled as its outlet slowly rose to its present position. Conclusion These lake-level predictions were suggested by a century of fieldwork by geologists mapping old shorelines of the Great Lakes. (This work is summarized in Jack L. Hough's classic work, Geology of the Great Lakes.) These geologists observed that ancient shorelines were tilted relative to modern shorelines and attributed the tilting to earth deformation forced by ice sheet loading. The results of this GIS methodology: The present topography (A) is warped vertically by the predicted deformation to yield the Paleo-DEM (B). The ice sheet is added (C) to form the actual surface landscape at given times in the past. The Paleo-DEM is filled to eliminate closed depressions (D) and the difference between the filled Paleo-DEM and the Paleo-DEM (C) is the predicted lake. Outlet location is on the lake margin where the greatest flow accumulation value occurs. analysis demonstrated that when the physics of viscous mantle flow is combined with ArcGIS methods, these ancient shorelines can be reproduced. Using this GIS methodology makes it possible to verify shoreline positions by uploading shapefiles of the predicted shoreline to ArcPad installed on GPS units and locating shorelines in the field. Differences between the observed and predicted locations suggest how the geophysical model should be improved ,either by altering the viscosity structure of the earth model or adjusting the ice sheet thickness chronology. Continued on page 26 ArcUser