Transportation is a high priority issue in Minnesota. The Minnesota Department of Transportation (Mn/DOT) maintains and upgrades the fifth largest highway system in the United States. Many highway projects receive federal funds, requiring review of each project for potential impacts on cultural resources under the National Historic Preservation Act (NHPA). In the past, this meant surveying large areas that often had little potential to contain archaeological sites. Because these surveys followed planning and design phases, it sometimes meant costly redesigns when significant archaeological resources were discovered. There was no way to anticipate and avoid previously undiscovered sites early in the project development process.

Before the Model

   In 1994, Mn/DOT’s Chief Archaeologist, G. Joseph Hudak, proposed developing an archaeological predictive model for the major Minnesota river valleys to support anticipated bridge replacements. By mapping areas of high, medium, and low archaeological site potential, this model would permit advance planning to avoid potential archaeological resources and minimize project delays. Mn/DOT and the Federal Highway Administration (FHWA) thought it was such a good idea that it should not be limited to the river valleys. They authorized funding to model the entire state.

   Predictive modeling had been successful on a smaller scale elsewhere. Many of these models were subjective, based on expert opinions of the archaeologists who constructed them. The earliest models were based on measurements taken from paper maps. But GIS and statistical analysis were becoming more common for this purpose. Notably, archaeologist Kenneth Kvamme published several studies using multiple logistic regression to model the environmental parameters of known archaeological sites. Minnesota adapted Kvamme’s methods and became the first state to develop a statewide GIS-based statistical archaeological predictive model (Figure 1).

   At the same time, Mn/DOT recognized the need for information to help predict locations of sites buried more than a meter below the surface. Because only a handful of such sites had been discovered in Minnesota, statistical methods could not be used to evaluate their distribution. Moreover, there was no subsurface environmental data for developing such a model. To adequately address this issue and provide a subsurface model needed to support future bridge replacements, the Mn/Model project included geomorphic mapping and modeling of several major river valleys. 

Building the Model

Conceptual Challenges

   In 1995, Mn/DOT assembled a team of consultants and advisors to develop the model, which by then was being called “Mn/Model.” The first challenge was to get everyone on the interdisciplinary team pulling in the same direction. Although Mn/DOT knew that GIS was the best technology for building the model, not all team members were convinced. In the early stages of the project, one archaeologist insisted on developing alternative models based on measurements from paper maps and fuzzy logic. The GIS team eschewed the idea of the time-consuming measurements, and anything “fuzzy” was anathema to Mn/DOT’s engineers. Fortunately, this effort was not completed, and an alternative to GIS was never needed.

   Geographers and archaeologists worked together to develop the conceptual model. This team made decisions on such matters as raster cell size, archaeological sites to include in the database, environmental variables to use as predictors, how to symbolize the model (e.g., how many probability classes and what colors they should be when mapped), and how to judge whether the model was successful. The goal of the model was to predict the locations of at least 80 percent of known archaeological sites, while classifying 33 percent or less of Minnesota’s land area as having high or medium site potential.

   At the same time, geomorphologists developed a conceptual model for classifying Minnesota’s landscape sediment assemblages and began mapping the Minnesota River Valley. When their first maps became available, geographers on the GIS team helped refine their system to work better for GIS analysis. The result is a high-resolution hierarchical GIS mapping system that provides geomorphic information necessary to model both surface and subsurface potential for preservation of archaeological sites.

Technical Challenges

   Timing of the work could not have been better. High resolution (1:24,000) GIS data were just becoming available for the state in 1995, when work began. Through the cooperation of the state’s Land Management Information Center (LMIC) and Department of Natural Resources (DNR), these data were funneled to the Mn/Model team as soon as they were available. These agencies even prioritized their data conversion schedules to provide the data in the order needed by Mn/Model.

   Data conversion tasks were daunting. First, all incoming data required extensive quality control and some correction of missing or miscoded data. Because of the high resolution required, a 30-meter raster cell size was selected for modeling. All data were converted to grids, and spatial analysis techniques were used to develop data layers that could be used in analysis. For example, elevation data were used to derive slope, aspect, and surface roughness grids that were used as independent variables in the regression model.

   Archaeological data posed another set of challenges. Minnesota suffers from the sparse site dilemma. Compared to other, warmer, parts of the country, the state has a very low density of known archaeological sites. Moreover, many of these sites contain very little information, sometimes only a few flakes or a single artifact, or were improperly recorded. Consequently, it is difficult to assign time periods and cultural affinities to them. To obtain a large enough sample size for modeling, we needed to combine sites from different time periods and cultures into a single database and to relax our definition of what could be considered an acceptable survey strategy.

   To efficiently manage the data conversion activities, data were organized by counties and regions. Regions with the best archaeological data were given highest priority for modeling, so their data were converted first. This allowed modeling to begin before data conversion was completed for the entire state. It also allowed the team to remove data for one region from the hard drives while converting another region. A one-gigabyte drive was enormous and expensive in those days, and processing the high resolution grids created many gigabytes of data. Another advantage of modeling by regions is that it compensates for the varied landscapes and ecology of the state, which includes prairies, oak savanna, deciduous forest, northern coniferous forest, and extensive wetlands. In all, Minnesota was divided into 20 regions, each with its own model. The statewide model is simply a mosaic of the regional models.

   The initial GIS models were admittedly crude, as they were based on small data samples and a limited number of available environmental data layers. As the project progressed through three modeling phases, the team included more archaeological sites in the analysis, added more and better environmental layers as they became available, and refined the modeling techniques.

The Model in Action

   Through the three modeling phases, each taking approximately one year, the models improved dramatically (Figure 2). The 1998 Phase 3 models exceeded project goals. Statewide, they predict 86 percent of all known archaeological sites while classifying only 23 percent of the land area as high and medium site potential. More important, the model predicted 77 percent of 977 sites discovered after the model was developed.

   Mn/DOT uses the archaeological predictive model extensively for project planning and reviews. It enables planners and designers to develop alternatives that avoid archaeologically sensitive areas when possible. When this is not possible, it alerts them to the need for archaeological surveys. Mn/DOT’s archaeologists use the model to determine which areas will be surveyed. Because we must both find any archaeological resources that might be affected and provide a test of the model, surveys cover areas with high, medium, and low site potential. However, 100 percent surveys of Mn/DOT project areas (the most detailed survey) are very rare these days, saving Mn/DOT both time and money.

   Mn/DOT staff consult the Mn/Model’s geomorphic model when projects might disturb deeply buried sites. To support this model, Mn/DOT is developing a field protocol to test for these sites. In the research to develop this protocol, previously undiscovered buried sites have been found in two of three places surveyed where they were predicted by the model (Figure 3).

Lessons Learned

   Developing and using the model highlighted several ideas worth sharing:

• Team building may take time and education. Some members of the research team are likely to know more about GIS than others. For Mn/Model, team communication improved after the archaeologists and geomorphologists attended a day-long seminar on GIS concepts.

• Data are everything—and quality can trump quantity. Time spent on data quality control is well worth the results. Mn/Model activities since 1998 have focused on improving both the archaeological and environmental data for developing the next round of models.

• Incorporate information about data quality into the model. Evaluating the Phase 2 models demonstrated to the Mn/Model team the extent of bias exhibited by archaeologists in deciding where to survey. That experience was turned into a model of survey bias, which helps define probability classes in the final working model. In particular identifying bias helped define areas where archaeological site potential is unknown because of inadequate information.

• Return on Investment. Mn/DOT’s investment in Mn/Model was paid back in time-savings during the first three years of its use. Mn/DOT continues to invest in data improvements to assure that future models will perform even better.

About the Author

   Elizabeth Hobbs is a Geographer in the Cultural Resources Unit at the Minnesota Department of Transportation. She served as Principal Investigator for GIS on all phases of Mn/Model and is currently Mn/Model’s Research Director. She can be contacted at mnmodel@ dot.state.mn.us.

©Copyright 2005-2021 by GITC America, Inc. Articles cannot be reproduced,
in whole or in part, without prior authorization from GITC America, Inc.

PRIVACY POLICY