A Predictive Model for Prehistoric Archaeological Sites within the Hill Air Force Range in Utah's West Desert Using GIS and Remote Sensing Data

Tilton, T. K., and B. L. Johnston

Introduction

Objectives: The purpose of this study is to identify the environmental factors that most influenced the location of prehistoric sites within the study area. To identify these factors, we made use of Remote Sensing data and available GIS coverages. Various factors were chosen and tested to see what influence, if any, they have on site location. After suitable factors are discovered, we will create coverages delineating the optimal conditions for each factor and then intersect them. This will reveal the areas in which all factors chosen are favorable and should have been attractive for prehistoric settlement. Data concerning the location of sites in the study area as well as the amount of survey coverage will be overlayed in order to realize any correlation that may exist and the strength of the model to predict site location.

Study Area: The first step in implementing the study was to choose a viable study area. The west desert of Utah was chosen because of the amount of data already obtained for the area and because it is an area where it can be assumed that environmental factors had a strong influence on human activity. We decided to limit the study to the Hill Air Force Range, which is the 371,000 acre bombing range that borders the Great Salt Lake. This area should be small enough to manage easily and large and diverse enough to have a variety of environments. We suspect that elevation will have an effect on site location, making the North range attractive, with elevations from about 4200ft. to about 5900ft.

The Archaeology of the West Desert

The Great Basin area, in which our study area falls, is believed to have been inhabited by people as far back as 11,000 years ago. For the most part. these people survived as hunters and gatherers, although agriculture was practiced for a short period by the Fremont people. Preservation of artifacts is extremely good in the dry desert air of the Great Basin, especially in the many cave sites and rock shelters that are found there. In fact, it is not uncommon to find such perishable items as basketry and even human feces that has been preserved for hundreds or thousands of years.

Well stratified sites with highly preserved artifacts has led to a fairly clear picture of the cultural chronology of the area. Two sites in particular, Danger Cave and Hogup Cave, are the basis of the accepted cultural chronology. Danger Cave was inhabited from approximately 8300BC until after AD1400, Hogup Cave is believed to have been inhabited intermittently from 6350BC to AD1470. When combined, the cultural record obtained from the two sites are the basis of the following cultural chronology for the Northern Great Basin:

The Bonneville Period (9000BC to 7500BC)- This period is believed to have been characterized by small, highly mobile groups of people whose subsistence was based on exploiting marsh and lake ecosystems. They were seasonally nomadic which means that they traveled to take advantage of various plant and animal resources according to their seasonal availability in various areas. The artifacts associated with this period are flaked stone tools consisting of many thrusting type spear artifacts. In the eastern Great Basin, there are only three sites that have been excavated that have revealed a Bonneville period component, among them is Danger Cave.

The Wendover Period (7500BC to 4000BC)- This period is characterized by the increased habitation of dry caves and rockshelters. The thrusting spear was replaced by the atl-atl (or spear thrower) and composite dart. Other artifacts that first came into use during this period were coiled and twinned baskets and ground stone tools for processing seeds. The most common resources exploited during this period were pickleweed, bulrush, and cattail seeds and both large and small game animals.

The Black Rock Period (4000BC to AD500)- This period is characterized by an appearent increase in population in the area and a more diverse settlement pattern. Groups started to occupy upland areas on a seasonal basis. It is believed that this was a period of climatic change, becoming somewhat hotter and drier. The bow and arrow became the weapon of choice late in the Black Rock Period, replacing the dart and the atl-atl.

The Fremont Period (AD500 to AD1300)- The Fremont people practiced agriculture supplemented by hunting and gathering. They grew a variety of maize called 'Fremont Dent' that had a short growing season and was fairly drought resistent, making it ideal for the Great Basin climate. It is during the Fremont Period that ceramic pot production is first practiced in the area.

The Late Prehistoric (AD1300 to AD1850)- This final period is characterized by the replacement of the semi-sedentary Fremont by Numic speaking people who subsisted as highly mobile hunter and gatherers. Artifacts typical to this period are small side notched triangular gis2213s (known as Desert Series) and unpainted brown and gray ware ceramics.

Environmental Setting

The Hill Air Force Range study area ranges in elevation from about 4200ft. to about 5900ft. AMSL. There are three major physiographic units associated with the study area: the Great Salt Lake shoreline, the Lakeside Mountains, and the Great Salt Lake Desert. Vegetation is characterized as belonging to the Great Basin Desertscrub plant community, known also as the Northern Desert Shrub Biome. Members of this community include: pickleweed, greasewood, shadscale, winterfat, cheatgrass, wheatgrass, amaranth, bulrush, sunflower, horsebrush, and juniper. Animals found in this area include: black-tailed jackrabbit, desert cottontail, Townsend ground squirrel, coyote, kit fox, bobcat, mule deer, ducks, hawks, gulls, and owls.

The climate of the region is characterized by cold winters and hot and dry summers. Summer temperatures often exceed 110F, and rainfall averages less than 12 inches per year. Most of the rain that falls in the region falls during the winter months, with occasional summer thunderstorms.

GIS and Remote Sensing Database

Building the Database:

Once the study's goals were more firmly defined along with the study area and time period, we began to populate our GIS database. The first step was to digitize a coverage that represents the Hill Air Force Range. Once this coverage was digitized, it was put in 'real world' coordinates and topology was built. This enabled us to use this coverage as a 'cookie cutter' to cut out relevant coverages from the Utah database for our area. By clipping out just our study area, we make the files smaller and the data more manageable. The clip command in ARC was used to clip out vector coverages, latticeclip (in ARC) was used for clipping DEM data. In order to clip out images such as TM scenes we used ERDAS Imagine. Due to problems with the vector module of our software, we could not simply use the vector coverage of our study boundary as an AOI, which would usually be the preferred method. To sidestep this problem, we converted our study area coverage to a grid in ARC with the 'polygrid' command. Once this was accomplished, we imported the grid into imagine as an image file. We then tried to subset images of our area but were unsuccessful due to unequal projections between the boundary grid and the desired images. Eventually, these images were cut out.

ARC Coverages:

North_bnd: The study area boundary in real world (UTM) coordinates, used to clip other relevant coverages for the study area.

Geol_nth: Geology for the Hill Air Force Range, clipped from the geology coverage for Utah (u104).

Soil_nth: Generalized soil types for the Hill Range, clipped from the soil coverage for Utah(utsoil-utm).

Eco_nth: Ecological zones clipped from the state coverage, generalized to the fact that there is only one zone in our study area - of limited use.

Prom_wat: 100k hydrology coverage - Promontory Point quad.

Bnvle_wat: 100k hydrology coverage - Bonneville Salt Flats quad.

Newf_wat: 100k hydrology coverage - Newfoundland Mts. quad.

Surface hydrology of study area.

All_wat: Above coverages appended into one.

100k_nth: Index of all 100k quad maps that cover the Hill Range.

24k_nth: Index of all 24k quad maps that cover the Hill Range.

North_slp: a polygon coverage delineating 8 slope classes within the study area, using the default classes defined by the software. This coverage was created with the 'latticepoly' command in ARC from the raw dem data using the 'slope' designation.

North_asp: a polygon coverage delineating 8 aspect classes within the study area,using the default classes defined by the software. This coverage was created with the 'latticepoly' command in ARC from the raw dem data using the 'aspect' designation.

North_con20: a polygon coverage delineating 20 meter contours within the study area, giving the user defined contour interval at 20 meters. This coverage was created with the 'latticepoly' command in ARC from the raw dem data using the 'range' designation. The 20m interval may be a bit course for our purposes so a more detailed coverage may need to be produced using a smaller contour interval.

ARC Grids

North_grd: Digital Elevation Model clipped to the study area boundary. This data will be used to create datasets concerning elevation, slope, aspect, and relief - factors that we suspect will be among the best predictors of site location.

Nth_shade: Shaded relief image of the study area created in the Grid module of ARC with the 'hillshade' command. This provides a good visual representation of the area's topography.

N_bnd_grd: A grid created with the 'polygrid' command in ARC, was imported into Imagine as an image file for the purpose of subsetting images for the study area.

Imagery

Unsupervised Classification of TM image, twelve classes.

Gap vegetation data for the HAFR

TM image clipped for the HAFR

Archaeological Site Location Data:

We acquired archaeological site data from LaDawn Neilson at the Space Dynamics Lab. The data was in the form of archaeological field reports and location of sites had to be discerned from somewhat rough copies of 7.5' quad maps, as SDL was reluctant to release location data to us.

In choosing sites suitable to our study, we read individual site descriptions and retained for the model only those sites of late Prehistoric, Fremont, or Numic affiliation (apx. 500 years AD to present). These sites are ones that we assume were inhabited under environmental conditions similar to the present ones. Unfortunately, in narrowing down the sites in this fashion, we reduced the number in our model to ten. While this is an extremely low sample size, it is made up for by the fact that we also have survey blocks outlined in the study area in which no sites whatsoever were discovered. With this information, we can choose random 'non-site' locations that can be analyzed as to thier relation to surrounding environmental conditions.

In order to put the site location and survey block data into digital form, it was first necessary to locate the 1:24000 quads on which they occur and transfer them over. Once this was accomplished, we were able to digitize them in using the quad's UTM coordinates to properly locate them within the study area. While there is certainly a loss of accuracy involved with this method, it was the best we could do under the circumstances. Once the survey blocks were digitized, chose random non-site locations within them and treated them the same as sites as far as buffering and analyzing them.

The following archaeological sites were chosen to be included in the study ...

42BO669 -"a pristine cave site that appears to contain a reletively deep cultural deposit, perhaps one to two meters in depth. The upper stratum of the deposit seems to be completely dry, indicating that the preservation of organic artifacts ... would be quite good." The site is classified as being Late Prehistoric in age due to the presence of one jasper Rose Spring Corner-notched arrow gis2213.

42BO681 -"a pristine, open-air prehistoric site located on a low-lying sand ridge at the southwestern base of the Newfoundland Mountains on Hill Air Force Range. This site has a Late Prehistoric cultural affiliation, and probably functioned as a seasonal camp for one or more groups of Gosuite Shoshone. Surface artifacts include Paiute-Shoshone Brown Ware sherds, a Cottonwood Triangular arrow gis2213, and one jasper Desert Side-notched gis2213.

42BO684 -"a pristine open-air site located along an ephemeral drainage ... on Hill Air Force Range. The site is Late Prehistoric in age, consists of five distinct loci, and appears to contain a shallow subsurface cultural deposit."

42BO685 -an open-air prehistoric site measuring 60m X 104m. It is classified as Late Prehistoric based on the find of one obsidian Desert Side-notched projectile gis2213 base.

42BO686 -a site classified as being of either Fremont or Numic cultural affiliation. "One Rose Spring Corner-notched arrow gis2213 was collected."

42BO687 -"a prehistoric open site situated on a linear dune ridge ... The site appears to have been occupied on a seasonal basis by one or more Fremont groups."

42BO688 -"a prehistoric open site of Fremont cultural affiliation..." Among the artifacts found were manos and various types of pottery.

42BO689 -"a prehistoric open site of Fremont cultural affiliation..." The site contains three loci.

42BO690 -"a prehistoric open site of Numic affiliation located at the northeast base of Gunsight Ridge." Artifacts include various sherds of Paiute-Shoshone Brown Ware and flaked stone.

42BO691 -"an open prehistoric site of Numic cultural affiliation situated just west of the northern base of Gunsight Ridge ... consisting of 40 - 50 Paiute-Shoshone Brown Ware sherds and several pieces of fire fractured rock."

Sites (blue) and non-sites/survey lots (red)
"Non-sites"

Since the strength and accuracy of our predictive model depends on a large sample size, and there was only limited archaeological site data provided for the Late Prehistoric Period, we chose to include non-sites. These were obtained by randomly chosing sites within survey blocks that were reported to have no archaeological sites within them. These non-sites can then be compared to sites as to the environment surrounding them. We chose 21 non-site locations to bring the total sample size of the model to 30, still perhaps too small.

Methodology

Once we had all relevant coverages and images in our database, we began our analysis. First, we took the GAP vegetation image and clipped it to our study area. We then buffered our site and nonsite images 15 pixels and overlayed them with vegetation to get a summary report. The summary report told us the type of vegetation classes that occur around sites and which occured around the random nonsites, this gave us an idea of whether certain vegetation classes could influence site location. To gain information regarding site and nonsite elevation, we used the 'identity' command in ARC. To do this however, we first had to build a polygon coverage from our DEM lattice. This was accomplished with the 'gridpoly' command. Likewise, polygon coverages were built from our DEM for slope, elevation, and aspect classes. Sites and nonsites were attributed as to thier class identity using the 'identity' command. The information concerning the attributes of sites and nonsites were then compared to try to see which attributes had the greatest difference, and therefore (we assume) play a greater role in site location. Once the attributes that were most condusive to site location were identified, coverages were built using the 'reselect' command for each one. The coverages we built to identify areas of high likelihood for containing sites were: 1) elevation polygons representing a range of 1275 meters to 1302 meters, this represented one standard deviation from the mean elevation of all sites in the study. 2) a buffer of 1km from any potable water (ephemeral or otherwise), and 3) vegetation polygons representing the two classes most in proximity to sites in the study, these classes were Greasewood (a plant known to be archaeologically significant to the people of the Great Basin), and Desert Scrub. Once these three coverages were built, they were intersected so as to delineate the areas where all factors would be suitable (according to our major assumptions) for Late Prehistoric site location.

Results

Concerning the vegetation classes most associated with (in proximity to) late prehistoric archaeological sites, the following information was obtained from a 'summary' in Imagine. Zone 0 represents the vegetation at the site itself, the following zones correspond to the number of pixels out from the site. As one moves farther from the sites, the number of pixels increases, as too should the diversity of vegetation classes. Still, the results are suprisingly homogenous, especially when compared to the nonsite data obtained in the same manner.

Pinyon/Juniper Sites

Archaeologically Significant Veg. Classes:
  class 36 - Greasewood
  class 37 - Pickleweed
  class 16 - Perennial Grass
  class 8  - Pinon/Juniper

Sites:
Zone number 0
Class Class Name                                 Count    %          Hectares
  ----- ---------------------------- ------ --------- ------ ---------------
     32 Class 32 (Salt Desert Scrub)                   5  55.56           0.450
     36 Class 36 (Greasewood)                          4  44.44           0.360
     
  Total                                                9 100.00          0.810

Zone number 1
  Class Class Name                                 Count    %          Hectares
  ----- -------------------------------------- --------- ------ ---------------
     32 Class 32                                      31  50.82           2.790
     36 Class 36                                      25  40.98           2.250

  Total                                               61 100.00          5.490

Zone number 2
Class Class Name                                 Count    %          Hectares
  ----- -------------------------------------- --------- ------ ---------------
     16 Class 16  (Perennial Grass)                   13  14.44           1.170
     32 Class 32                                      45  50.00           4.050
     36 Class 36                                      32  35.56           2.880
     
  Total                                               90 100.00          8.100
Zone number 3
 Class Class Name                                 Count    %          Hectares
  ----- -------------------------------------- --------- ------ ---------------
     16 Class 16                                      17  14.29           1.530
     20 Class 20                                       1   0.84           0.090
     32 Class 32                                      66  55.46           5.940
     36 Class 36                                      35  29.41           3.150
     
  Total                                              119 100.00         10.710

Note that for nonsites there is a considerable difference in the diversity of vegetation classes and in the percentage of Greasewood and Desert Scrub found in proximity. This is due in part to the fact that the nonsites are more numerous and less clustered, but is perhaps still significant.


Nonsites:
Zone number: Zone 0
 Class Class Name                                 Count    %          Hectares
  ----- -------------------------------------- --------- ------ ---------------
   
      1 Class 1                                        6  28.57           0.540
     16 Class 16                                       1   4.76           0.090
     20 Class 20                                       3  14.29           0.270
     32 Class 32                                       6  28.57           0.540
     37 Class 37                                       5  23.81           0.450 

  Total                                               21 100.00          1.890

Zone number 1
  Class Class Name                                 Count    %          Hectares
  ----- -------------------------------------- --------- ------ ---------------
      
      1 Class 1                                       39  25.00           3.510
     16 Class 16                                      12   7.69           1.080
     20 Class 20                                      17  10.90           1.530
     32 Class 32                                      46  29.49           4.140
     36 Class 36                                       1   0.64           0.090
     37 Class 37                                      41  26.28           3.690
   
  Total                                              156 100.00         14.040

Zone number 2
 Class Class Name                                 Count    %          Hectares
  ----- -------------------------------------- --------- ------ ---------------
     
      1 Class 1                                       57  24.57           5.130
     16 Class 16                                      23   9.91           2.070
     20 Class 20                                      22   9.48           1.980
     32 Class 32                                      65  28.02           5.850
     37 Class 37                                      62  26.72           5.580
     38 Class 38                                       3   1.29           0.270

  Total                                              232 100.00         20.880

Zone number 3

  Class Class Name                                 Count    %          Hectares
  ----- -------------------------------------- --------- ------ ---------------
      1 Class 1                                       75  24.83           6.750
     16 Class 16                                      35  11.59           3.150
     20 Class 20                                      34  11.26           3.060
     32 Class 32                                      72  23.84           6.480
     36 Class 36                                       6   1.99           0.540
     37 Class 37                                      77  25.50           6.930
     38 Class 38                                       3   0.99           0.270

  Total                                              302 100.00         27.180

Elevation was determined in two ways for both sites and nonsites. At first, we determined elevation based on the range that the sites or nonsites fell into. It became appearent that this was a little too general, so we decided to find the exact elevation (according to the DEM information) as well. Below is a summary of that information.

Site Elevation in Meters:
GRID-CODE                  =     1279
GRID-CODE                  =     1280      Mean Elevation = 1288.11m
GRID-CODE                  =     1281      Std.Dev. = 13.80m
GRID-CODE                  =     1281      1275m - 1302m
GRID-CODE                  =     1281
GRID-CODE                  =     1289
GRID-CODE                  =     1289
GRID-CODE                  =     1290
GRID-CODE                  =     1323

Non-site Elevation in Meters: 
GRID-CODE                  =     1269
GRID-CODE                  =     1268
GRID-CODE                  =     1347
GRID-CODE                  =     1381
GRID-CODE                  =     1429
GRID-CODE                  =     1344
GRID-CODE                  =     1368
GRID-CODE                  =     1300
GRID-CODE                  =     1266       Mean Elevation = 1301.29m
GRID-CODE                  =     1282       Std. Dev. = 45.09m
GRID-CODE                  =     1283       1256m - 1346m    
GRID-CODE                  =     1289
GRID-CODE                  =     1266
GRID-CODE                  =     1266
GRID-CODE                  =     1286
GRID-CODE                  =     1286
GRID-CODE                  =     1274
GRID-CODE                  =     1280
GRID-CODE                  =     1281
GRID-CODE                  =     1282
GRID-CODE                  =     1280

Elevation Ranges:
0  1276:1
1276 1286:2
1286 1292:3
1292 1402:4
1402 1584:5
1584 3000:6



Non-site Elevation (from nonsite2_elv.pat):
GRID-CODE                  =        1
GRID-CODE                  =        1
GRID-CODE                  =        1
GRID-CODE                  =        1
GRID-CODE                  =        1
GRID-CODE                  =        1
GRID-CODE                  =        2
GRID-CODE                  =        2         Mean elevation class = 2.38
GRID-CODE                  =        2             
GRID-CODE                  =        2
GRID-CODE                  =        2
GRID-CODE                  =        2
GRID-CODE                  =        2
GRID-CODE                  =        2
GRID-CODE                  =        3
GRID-CODE                  =        4
GRID-CODE                  =        4
GRID-CODE                  =        4
GRID-CODE                  =        4
GRID-CODE                  =        4
GRID-CODE                  =        5
Elevation of Negative Survey Area (stats generated from non_elv_area.pat):
  RANGE  GRID-CODE FREQUENCY           SUM-AREA     PERCENT    
    
  0-1276      1      4           15526057.125000     .3268   Mean elevation
  1276-1286   2     22           15961158.750000     .3360   class = 1.83
  1286-1292   4     11            9238283.664062     .1944 
  1292-1402   3     13            4486310.484375     .0944 
  1402-1584   5      6            2288158.445312     .0482 

Site Elevation (from site_elv.pat):
GRID-CODE                  =        2
GRID-CODE                  =        2
GRID-CODE                  =        2
GRID-CODE                  =        2
GRID-CODE                  =        2          Mean elevation class = 2.56
GRID-CODE                  =        3
GRID-CODE                  =        3
GRID-CODE                  =        3
GRID-CODE                  =        4

As far as distance to water, we arbitrarily decide that 1km was a good round figure to use. We buffered all streams occuring in our study area, overlayed it with site data, and got the following results...


Streams occuring within 1km of non-sites (statistics generated 
from nonsite_hydro.aat):

# of nonsites   # within 1km of water      SUM-LENGTH       Length/nonsites buff
     21                 6  (29%)           8588.728210         408.99

Streams occuring within 1km of sites (statistics generated from 
sites_hydro.aat):

# of sites     # within 1km of water       SUM-LENGTH       Length/sites buff
     9                  4  (44%)            6351.463623          705.72

Since aspect and slope appear to play a minor role in site location (at least in our study, maybe not in reality), they are only briefly summarized below...


Aspect Classes:

0    40 = 1
40   80 = 2
80  120 = 3
120 160 = 4
160 200 = 5
200 240 = 6
240 280 = 7
280 320 = 8

nonsite:  Mean aspect class = 5.21   site: Mean aspect class = 5.00


Slope Classes (in degrees):
0 5   = 1
5 10  = 2
10 15 = 3
15 20 = 4
20 25 = 5
25 30 = 6
30 35 = 7
35 40 = 8
40 45 = 9
45 50 = 10 
50 90 = 11

nonsite: Mean slope class = 1.14     site: Mean slope class = 1.44

1km buffer of all water within Hill Air Force Range

Areas between 1275m and 1301m elevation and within 1km of water

Here for enlarged view.

Areas of high probability for containing late prehistoric sites within Hill Air Force Range. Between 1275m and 1301m, within 1km of water, and in Greasewood and Desert Scrub vegetation classes.

References

Arkush, Brooke S., and William B. Fawcett. 1991. A 1991 Archaeological Assessment of U.S. Air Force Designated Lands in Utah. MS on file, Antiquities Section, Utah Division of State History, Salt Lake City, Utah.

Arkush, Brooke S. 1993. An Archaeological Assessment of the U.S. Air Force Utah Test and Training Range: The 1992 Field Season. MS on file, Antiquities Section, Utah Division of State History, Salt Lake City, Utah.

Arkush, Brooke S. 1993. An Archaeological Assessment of the U.S. Air Force Utah Test and Training Range: The 1993 Field Season. MS on file, Antiquities Section, Utah Division of State History, Salt Lake City, Utah.

Ramsey, Douglas 1995. All the extensive knowledge that has been spewed forth, and utilized.