Remote Sensing for Detection of Prehistoric Landscape Use in NW Arizona, USA

Optimal maize field locations possibly used by prehistoric agriculturalists in the Mt. Trumbull portion of the Colorado Plateau in NW AZ were modeled using remotely sensed data and ground based observations. Over 400 prehistoric archaeological sites have been recorded in the study area; in some areas site density is ~120 sites/mi2, including many 1-2 room structures traditionally referred to as "field houses" that archaeologists have long assumed were located on or immediately adjacent to maize fields. Other site types are larger C-shaped pueblos with up to 20 rooms and somewhat smaller multi room structures. We collected and used ground-based field measurements and satellite image data from ASTER (Advanced Spaceborne Thermal Emission and Reflection Radiometer) to produce GIS layers to predict ancient maize fields and compare these with known "field house" sites. Input data layers for the model included early spring maximum solar illumination, surface gradient, surface radiant temperature, water surface flow collection, water infiltration, and soil type. We constructed 2 types of optimality models: "restrictive" (or classification) models and "fuzzy logic" (or grouping) models. Highest values were assigned to pixels with more surface water, warmer temperature, better soils, etc. and then assigned a color for display. Analyses of patterns for the "green" restrictive model shows a disproportionate number of large sites found within 200 m of the green optimal zone; for the yellow optimal zone there is a statistically significant relationship between larger sites and the yellow zones at 100 m or less. For the blue fuzzy logic model, again there is a strong relationship between the number of large sites and a blue zone both at 100 m and 200 m distances. So-called "field houses" are not located preferentially close to our optimal areas. Rather, there is a clear preference for larger sites to be found closer to optimal areas. Using the proportion of site types from the training area, we performed a chi square test using those proportions against the actual values found in a previously unknown area (area B). It was found that the proportions of large sites close to the fuzzy logic blue optimally zone is indistinguishable from the test area, meaning essentially the same pattern is found in area B; viz., there are disproportionally more large sites found closer to blue optimal areas in the fuzzy logic model than would be expected by chance alone. These smaller structural sites are not located closer to the most optimal places as might be expected if they are in fact "field houses". Smaller sites may have been established only after ~ AD 800 when the larger C- and L-shaped pueblos were settled near the most optimal field locations. These smaller structural sites did in fact act as field houses-- but in more marginal locations and later in time. As this portion of the Mt Trumbull area got increasingly "packed" during the later periods, it may be that kin groups from the larger residential sites established field houses to monitor their more marginal fields. This process might have intensified in the 12th and 13th centuries as environmental conditions deteriorated, or at any time when summer monsoonal rains needed for successful agriculture became reduced for long periods.