It is apparent that model-building is a complex and time-consuming process. There is no set of standards to follow because of the variety and variability in modelling approaches and management objectives, as well as regional physiographic and cultural differences (Judge/Martin 1988:573).
The first issue is that of the complexity of the process, modelling past human behaviour/activities, is not an easy task. Humans do not behave mechanistically, and thus generalizations about their behaviour are difficult to derive and can never be completely accurate. The relationships among humans, their activities, and past landscapes are very complex to begin with, and this complexity is compounded by subsequent changes in these landscapes, and by the difficulty of the quantitative methods that one must employ to model these relationships: methods that are frequently beyond the expertise of those who wish to use them. Modelling is a tool, but it is by no means (Judge/Martin 1988:574) a simple tool and not a panacea. As a complex tool, its uses are limited, and it requires expertise to implement correctly. As with any tool, modelling can be abused, and the value of the results diminishes accordingly. Used properly, however, modelling can be of inestimable value to both the manager and the research archaeologists.
Predictive modelling of archaeological site locations can never be a complete substitute for actual field inventory (intensive survey). Human behaviour is too complex to permit this kind of modelling accuracy, and too many variables have intervened between the time the behaviour took place and the present to achieve, through modelling, the accuracy available with field inventory. For this reason, it is unlikely that predictive modelling could, in the foreseeable future, be sufficiently accurate to satisfy the identification requirements (Judge/Martin 1988:575) in an area. By the same token predictive modelling is unlikely to satisfy the needs of a research archaeologist whose research design requires accuracy at a similar level.
Modelling can, however, provide research archaeologists with estimates of probable site densities in unsurveyed areas, and this same capability is of great potential benefit to the manager. In the short term, for example, the ability of models to project areas of low site density or to indicate probable locations of sites for data recovery can be extremely helpful to the manager, not as a substitute for inventory but as an aid in designing cost-effective inventory.
Modelling's greatest strengths, however, lie in its contributions to the long-term planning process. It is here that models developed with resource planning, interpretation, and evaluation in mind can be of tremendous value to the establishment of management priorities and to the integration of cultural resource management with other resource management responsibilities.
Further, such model-based management can facilitate research, quite apart from the preservation and protective responsibilities of the manager. Since a fundamental purpose of cultural resource preservation is to maintain the scientific potential of the resource, that is, to preserve its information content, modelling as a component of long-range planning is of particular value to managers and researchers alike (Judge/Martin 1988:576).
As a planning tool, predictive models should permit the projection of resource potential, scaling of inventory effort to meet expectations, estimating resource values, projecting project costs, and checking on the validity and reliability of surveys conducted. Models will allow managers to project expectations onto project area maps, or to display expectations graphically. Job costing, time estimations, and other project planning activities will be facilitated. Survey efforts can be concentrated in areas of highest expectations and reduced in areas of low expectation. Sampling strategies demonstrated by predictive modelling to be most effective and efficient could be designed to maximize the amount of information collected for the effort expended (Kohler 1985:15).