A dissertation submitted in partial fulfillment
of the requirements for the degree of
Doctor of Philosophy


Gregory Vogel, B.A., M.A., RPA

University of Iowa, 1995
University of Arkansas, 2000

December 2005
University of Arkansas


Quick links:
Aerial photograph tables
Reconstructed DEMs of pre-reservoir topography
Full dissertation in PDF format

This dissertation is a study of late prehistoric mounds constructed in the Northern Caddo Area of northwestern Arkansas, northeastern Oklahoma, and southwestern Missouri. The study concentrates on the relationship of the mounds to the landscape - how they are situated in relation to arable bottomland, and from how large of an area on the landscape the mounds were visible. Both of these factors have implications for the environmental adaptations and social structures of the people who built the mounds.

You can read the full abstract of the dissertation below (click here).

The entire dissertation is available below in PDF format, in 11 separate sections. Be aware that some of the files are fairly large. To download them directly to your computer, right-click on the highlighted link and select "save as".


DEM of the study area. Click image for larger version.


A Few Points of Interest:

The Sites:

A great deal of the work for this dissertation consisted of compiling information about the sites themselves. Many of them have never been summarized in the published literature, and through archival research, early aerial photographs, and field visits, I was able to gain new insights into several of them. Site descriptions in Appendix A include significant and previously unpublished material concerning Point Remove (3CN4), Page (3LO15), Logan Eddy (3LO208), Scotia (3PP23), Guy Brittain (3SB1), Cavanaugh (3SB3), Bluffton (3YE15), and novel material relating to Reed (34DL1) and Lillie Creek (34DL41). New radiocarbon dates acquired for Norman (34WG2) and Elkins (3WA1) are reported here for the first time as well.


GIS Procedures:

Several of the mound sites are currently under the water of artificial reservoirs, and available digital elevation model (DEM) datasets show the elevation of the modern water level. To reconstruct a closer approximation of the prehistoric landscape, I digitized numerous early paper maps of river topography, and incorporated these elevations into the DEMs. I reconstructed the pre-reservoir topography of Eufala Lake, Robert S. Kerr Lake, Fort Gibson Lake, and Grand Lake O' the Cherokees in this way. While far from perfect, these reconstructed topographic models served very well for the purposes of this dissertation, and may be of use to other researchers as well. Information about the generation of these DEMs can be found in Chapter 5. You can download the DEMs of reconstructed reservoirs for you own use here.

In order to operationally define 'bottomlands' within a large and topographically diverse study area, I developed a novel procedure for de-trending an entire landscape for local stream gradient. Details concerning this procedure are given in Chapter 7 and Appendix C.

This dissertation involved a large-scale viewshed analysis employing Monte-Carlo modeling to determine the statistical background for viewsheds within different topographic regions of the study area. Details concerning these procedures is found in Chapter 6.

I also conducted elaborate proximity analysis to determine how much arable bottomland was available to the mound sites. This type of study is often referred to as 'catchment analysis', and may be conducted several different ways. I conducted the study using four different theoretical assumptions about travel across a landscape:

1) Topography is irrelevant to travel (equivalent to straight-line distance or circular catchments).
2) Topography is relevant to travel (it is harder to travel over steep, rough terrain) but rivers are not.
3) Topography is relevant to travel and rivers were boundaries (difficult to cross).
4) Topography is relevant to travel and rivers were corridors of easy travel (by boat).

Bottomland proximity (catchment) analysis with GIS returned different results depending on which theoretical assumption about travel across the landscape was used. Details and results of the proximity analysis are found in Chapter 7.


Aerial Photographs:

I made extensive use of early aerial photographs (from the 1930s to 1950s) in this study, to great effect. These aerial photographs were generally taken when much of the bottomland (where the mound sites are) was cleared of vegetation for farming, but before the reservoirs were constructed and before much large-scale landscape modification or urban development. These photographs helped me reconstruct some of the early landscape, locate some of the mounds that are gone now (and whose precise location was unknown), and to interpret prehistoric landscape modifications around some of the mound centers (I use the aerial photographs extensively in the site descriptions: Appendix A).

Aerial photograph flight plans were flown over most of the central U.S. in the 1930s and 1950s, and I believe they are a vastly under-used resource in North American archaeology. Unfortunately, scans of the photographs (if they are detailed enough to be useful) are so large that it is impractical to put them on-line at this time. If you are interested in any of the aerials I use in this dissertation, I can send them to you in either raw TIFF or georeferenced GEOTIFF format, just contact me: ggvogel@gmail.com. You can view a list of the aerials I have available here.



The overall conclusions I drew from this study are:

1) Many of the mound sites are situated preferentially on the landscape with regard to viewsheds. Mound sites in the Arkansas and Neosho River valleys, particularly the larger sites, are situated in locations with viewsheds far larger than would be expected by chance alone. Mound sites in the Ozark uplands, on the other hand, appear to be situated in locations with preferentially small viewsheds.

2) The size and elaboration of sites in the Arkansas and Neosho River valleys corresponds fairly well with the amount of locally available arable bottomland. In other words, larger sites are situated near larger tracts of good farmland. This is fairly intuitive, and holds true no matter which of the theoretical assumptions about travel across a landscape (explained above) is used. Sites in the Ozark uplands, however, do not express this pattern. In this region, the mounds are all situated along streams, but the amount of locally available bottomland does not seem to be an influence on the size or elaboration of the sites.

3) Both the viewshed and bottomland proximity analyses suggest that there are quite different meanings or uses of the mound sites between the major river valleys and the uplands. Prehistoric people in these areas were certainly in communication with one another and part of an overall integrated regional system, but there appear to be significantly different cultural practices and processes between the regions.

4) While a large amount of bottomland in close proximity appears to be necessary for sustaining large sites in the large river valleys, the bottomland alone was not sufficient for founding or sustaining a large center. Large tracts of bottomland exist along the Arkansas and Neosho Rivers within which there are no large mound centers. In other words, large amounts of bottomland resources may have created the potential for the development of large mound centers, but this potential was not realized in some locations.

Please note that these are somewhat simplified versions of the overall conclusions from the dissertation. You can read the rest of the details in Chapter 8.





This dissertation is a study of late prehistoric mound centers in the northern Caddo area of northwest Arkansas, eastern Oklahoma, and southwest Missouri, concentrating on the relationship of the mounds to the natural landscape through viewshed and bottomland proximity analyses. The mounds are interpreted from published accounts, archival sources, early aerial photographs, and fieldwork. The regional landscape is modeled in Geographic Information Systems (GIS) as a digital elevation model (DEM), modified to more accurately reflect topography prior to river impoundment and other historic landscape alterations.

Monte Carlo statistical methods show that the mounds are located preferentially with respect to viewsheds. Mounds in the Arkansas and Neosho River valleys and Ouachita Mountains are situated in locations with statistically large viewsheds, and at least some in the Ozark Plateaus are situated in locations with statistically small viewsheds.

The mounds are also located preferentially with respect to alluvial bottomland. Alluvial bottomland within the study area was delineated with a novel GIS procedure, and the amount of bottomland near each mound site was determined through four models of proximity: straight-line distances, a simple friction-based cost surface model, a cost surface model considering rivers to be barriers to travel, and a cost surface model considering rivers to be corridors of easy travel. In general, the size and elaboration of mound centers corresponds to proximate bottomland. The correspondence is regional and follows a trend of increasing mound-center size downstream, correlating with increasing access to alluvial bottomlands.

Together, the viewshed and bottomland proximity results suggest different cultural adaptations in the uplands from those in the larger river valleys. The amount of proximate bottomland has a greater influence on mound centers in the river valleys than in the uplands, and while large tracts of bottomland may have been necessary for large mound centers, this alone was not a sufficient criterion for their development.


Full dissertation in PDF format by chapter:

Chapter 00
(139 kb)

Title pages
Acknowledgments, dedication
Table of contents and list of figures

Chapter 1
(1,069 kb)

Introduction and Statement of Research Questions

  The study area
Research questions
Chapter 2
(728 kb)
The Natural Background

Physical characteristics
Climate history

Chapter 3
(420 kb)
Archaeological Background
The "Northern Caddo Area"
A typology of mounds in the Arkansas Basin
Civic ceremonial centers
Social organization and mounds
Chapter 4
(1,435 kb)
Site Summaries
Mound Comparisons
Chapter 5
(795 kb)
Building the GIS Model
The landscape
The sites
Chapter 6
(12,422 kb)

Viewshed studies
Quantitative aspects of viewsheds 1: mound comparisons
Quantitative aspects of viewsheds 2: the statistical background
Results of randomization tests
Qualitative aspects of viewsheds

Chapter 7
(4,341 kb)
Bottomland Proximity Analysis
  Delineating bottomlands
Four models of proximity
Determining bottomland areas
Results of bottomland proximity analysis
Chapter 8
(1,428 kb)
Summary and Conclusions
Further considerations
Further questions
Appendix A
(54,161 kb)
Site Descriptions
  Aerial photograph interpretations
Prairie mounds
Site descriptions: Arkansas sites
Site descriptions: Missouri sites
Site descriptions: Oklahoma sites
Other possible mounds
Non-mound site descriptions
Appendix B-F
(654 kb)
Appendix B: Aerial Photograph Tables
Appendix C: GIS Procedures
Appendix D: Bottomland Proximity Tables
Appendix E: Nearest-Neighbor Analysis and Proximity Models
Appendix F: Contents of the DVD
Back to Gregory Vogel's home page
Back to top

Much of the archival work for this project was made possible by a grant from the National Science Foundation (DDI Grant 0341068: Location and Patterning of Mississippian Mounds in the Central Arkansas River Valley), and a Dissertation Research Travel Grant from the University of Arkansas.

This web site is hosted by ProjectPast.org.