The U.S. Department of Defense (DoD) is legally required to inventory and evaluate archaeological sites, Native American resources, and other cultural assets on lands it administers. To date, the agency has inventoried less than 40% of its holdings and has another 13.4 million acres to inventory. More than 110,000 sites are recorded, of which more than 20,000 are listed in or eligible for listing in the National Register of Historic Places (NRHP). Many other sites lack determinations of eligibility and must be treated as if they are eligible until their status is confirmed.

DoD must take into account the effects of military actions on thousands of potential historic properties on lands that have not been inventoried and resources that have not been evaluated, so refined technologies are needed to streamline the evaluation process. One particularly effective technology that can be adapted to reduce cost and effort associated with cultural resource management (CRM) requirements is archaeological predictive modeling. To be effective tools for cultural resource management, archaeological predictive models must be operationalized in a database using Geographic Information System (GIS) technology, refined as new data become available, statistically validated to demonstrate their accuracy, and incorporated into programmatic agreements (PAs) that will streamline compliance with the National Historic Preservation Act (NHPA, Section 106) and National Environmental Policy Act (NEPA).

Over the last 30 years, a number of DoD installations, especially those with large land holdings, have developed and used predictive models as planning tools. Few installations, however, have operationalized, refined, and validated their predictive models, and none have incorporated their models into PAs.

The overarching objective of this project was to demonstrate that predictive models of prehistoric archaeological site locations can be sufficiently accurate to serve as the foundation for programmatic approaches to compliance that, when implemented, can achieve greater efficiency and lower costs for administering CRM programs. The specific performance objectives—improving surface, subsurface, and “red flag” predictive models; developing Section 106 PAs; and demonstrating that models integrated into compliance protocols can significantly reduce the level of effort, cost, and number of evaluated sites—were met. Existing models at Fort Drum and Eglin Air Force Base (AFB) were used successfully to demonstrate the technology and their potential. Additional modeling work was conducted at Saylor Creek Range and Utah Test and Training Range; however, these were not formal demonstration sites.

Technology Description

The project team designed a multiphase process to demonstrate that highly effective archaeological predictive models can be developed to inform management decisions and streamline compliance through the creation of installation-specific PAs. The process began with (1) the collection and evaluation of relevant archeological and environment data, (2) the development of a formal model that can be operationalized with GIS technology, and (3) validation procedures that test the model’s accuracy and determine whether it meets predefined performance criteria. Once these steps have been taken, modelers may refine the model with new or better data to improve its performance and then repeat the validation process. With the development of one or more accurate, validated models, the process continues with a fourth phase; the creation of a zonal management model that synthesizes the results of each underlying model. It is this zonal model that DoD managers and stakeholders use to make decisions about inventory and site evaluation protocols in different probability or “sensitivity” zones for finding sites. Through consultation, the final phase is the preparation of a PA that stipulates how Section 106 requirements will be met.

Demonstration Results

Using the aforementioned phases, the project team demonstrated that three types of predictive location models could be developed or refined and subsequently integrated into a zonal management model that has been incorporated into draft PAs. The ESTCP project ended before the draft PAs could be finalized and executed, but the project team developed alternate methods to demonstrate the efficacy of using predictive models to manage cultural resources. These alternative methods, which use historic data from each installation on the level of effort and cost of past archaeological inventories, demonstrate considerable time and cost savings when effective models are used.

For Eglin AFB, the project team formalized and tested an existing surface sites model, refined and tested this model, created a model for information-rich habitation sites that would be expensive to mitigate (“red flags”), and created a model for deeply buried or subsurface archaeological sites. The first two models met and exceeded the specified performance criteria for a successful model. The subsurface model could not be tested due to a lack of appropriate data. Team members used the refined surface model, the red flag model, and the subsurface model to create a zonal management model that has been included in a draft PA for managing archaeological resources on Eglin AFB.

For Fort Drum, the project team formalized and tested an existing lowland surface sites model and an existing upland surface sites model, refined and tested both models, and created a model for deeply buried or subsurface archaeological sites. The refined lowland surface model met and exceeded the performance criteria; however, the upland surface model did not. Insufficient data were available to test the subsurface sites model, but a preliminary test using available data suggests that the model is close to meeting the criterion. Team members used the refined surface model and the subsurface model to create a zonal management model that has been included in a draft PA for managing archaeological resources on Fort Drum.

Implementation Issues

Future efforts to create or improve predictive models of archaeological site location now have a tested process for their development, refinement, validation, and integration into the compliance process. Web site guidance on how validated and accurate predictive models can be created will serve as the medium of technology transfer.

Future efforts should consider four implementation issues. First, the weakest link in developing and refining formal, inductive predictive models is the quality of the archaeological and environmental data. To build models efficiently, relevant archaeological data should be maintained in computerized databases usable by GIS. Similarly, environmental data should be of sufficient accuracy and resolution to facilitate the measurement and correlation of site locations with natural features. Second, to efficiently create and test predictive models, modelers and installation staff need to work together early and often to ensure that key variables are included in both the underlying model and the resulting management model. Third, for predictive models to be incorporated into PAs, installation CRM staff must involve their consulting parties (State Historic Preservation Office [SHPO] staff, Native American groups, and other interested parties) from the beginning of the modeling process and maintain regular contact. Consulting parties will need assurance to maintain confidence in the value of modeling for finding and protecting sites as well as enhancing knowledge of past cultural systems. Finally, it is critical to view modeling as a process and not an event; models improve with more data, allowing DoD to meet its stewardship and mission goals more efficiently and with better results.