Chaco Canyon Cross-Cutting: Identifying Community Signatures with NASA Earth Observations

EarthzineDEVELOP Spring 2017 Article Session

This article is a part of the NASA DEVELOP’s Spring 2017 Article Session. For more articles like these, click here

This project identifies areas within the San Juan basin that have a high probability of containing an environment suitable to Chacoan sites and generates a risk map of sites at risk of disturbance from expanding infrastructure.

Authors:
Dashiell Cruz
Sydney Neeley
Kelsey Herndon
Ryan Schick

The project partners for this study expressed a need for a product that could remotely identify Chacoan community signatures in the San Juan Basin in New Mexico to possibly aid in the expansion of the park boundaries as well as cut the cost of time-consuming field studies. This project partnered with Tom Lincoln, the Intermountain Region Assistant Regional Director for Cultural Resources at the National Park Service (NPS), Dr. Ruth Van Dyke of Binghamton University, Dr. Carrie Heitman of the University of Nebraska-Lincoln, and Dr. Steve Lekson of the University of Colorado Boulder. Currently, these partners rely on Google Earth and Landsat series imagery to identify locations of ancient Chaco roads and houses. The project partners also locate Chacoan roads and communities through extensive field surveys and excavations, which are both time-consuming and expensive.
Our science advisor Dr. Thomas Sever, retired NASA archaeologist and research scientist and professor at the University of Alabama in Huntsville, provided the team with research he conducted for his doctoral dissertation as a starting point for the methodology. This provided the team with a historical understanding of Chacoan rituals and community practices. Dr. Sever also suggested the use of vegetation indices to train our site distribution model. Our research received further advising from Dr. Jeffrey Luvall, whom suggested that the team utilize thermal data to tease out spectral characteristics of ancient Chacoan communities. He suggested that using thermal radiance could delineate paths traveled by the Chacoan people by picking up temperature fluctuations from shadows cast by depressions in the ground caused by frequent treading of the landscape.
The first step in creating a Chacoan site suitability map was to determine which indices would be used to train our machine learning program, Princeton University’s Maximum Entropy modeling tool (MaxEnt). In ArcGIS 10.3, the following environmental layers were resampled to a cell size of 30 meters: Normalized Difference Vegetation Index (NDVI), Terra Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) Global Emissivity Dataset (GED), Shuttle Radar Topography Mission version 2 (SRTM-v2) Digitial Elevation Model (DEM), slope, aspect and LANDFIRE vegetation type.

Chacoan Site Suitability Map. This map was created using Princeton’s MaxEnt Modeling software. (From left to right): Final Chacoan Site Suitability Map created using MaxEnt, aspect map, Normalized Difference Vegetation Index (NDVI) map, slope map, soil map, and Digital Elevation Model (DEM). Image Credit: NASA DEVELOP

NDVI is an index calculated by using Red and Near-Infrared wavelengths. This is useful in identifying vegetation such as ragweed, which is commonly found growing in areas that have been disturbed by the ancient Chacoan people. The geophysical regions that the Chacoan people were accustomed to erecting great houses became apparent in the MaxEnt model with indices obtained from SRTM-v2. These layers were then converted to ASCII format for analysis in MaxEnt. Since site locations were not identified throughout the entire extent of the study area, we created a bias file that limits the background samples in the calculation of the model to only counties containing known sites within the San Juan Basin. We used 80 percent of the known Chaco site locations to train the model in MaxEnt and the remaining 20percent to test the performance of the model.
Using MaxEnt we created a site suitability map and also conducted an accuracy assessment on the input data. The model returned a relatively high area under the curve (AUC) for the test data of 0.814. Results also showed a moderate-to-high suitable environment for Chaco sites in both the southern and western parts of the San Juan Basin. Areas highlighted by the model also have a high correlation with the elevation and position of other known Chacoan sites. The model was run multiple times, however higher resolution data is needed to ensure the accuracy of these conclusions.
Our next investigation was to determine which of the 123 known Chacoan roads and sites were under threat from encroaching infrastructure. In order to create a site risk map, the team determined that risk would be defined as proximity of a Chacoan feature to areas with an expected population increase from 2015 to 2020, existing roads, existing or planned oil and gas drills and perennial hydrological features.

Chacoan Risk Map. This map was created using the Fuzzy Logic tool in ArcMap 10.3. (From left to right): Chacoan Sites Risk Map, OpenStreetMap roads map, perennial hydrological features map, NASA SEDAC projected population increase from 2015-2020 map, and existing/planned oil and gas drills map. Image Credit: NASA DEVELOP

Fuzzy memberships were assigned to each variable using the Fuzzy Membership tool in ArcMap 10.3. By combining the fuzzy membership with the “or” function in the Fuzzy Overlay tool, the function can identify suitability based on the input variables. This function is beneficial because it highlights which Chacoan features have a close proximity to any one of the harmful input variables. Results from the Fuzzy Logic model indicated that 44 of the 123 (35 percent) known Chacoan sites were under threat from developing infrastructure. Fortunately, 19 (43 percent) of these sites are already protected by our project partner, the National Park Service.
In addition to creating a site suitability map and site risk map, we focused a portion of our time identifying known Chacoan roads and sites with various image processing techniques. We were successful in identifying Chacoan roads using Thermal Infrared Multispectral Scanner (TIMS) data (Image C). TIMS data is thermal radiance data with a 4-meter resolution. This data is not georeferenced, so the team relied on finding referenced points in Google Earth to identify areas where Chaco features are located. The most useful TIMS band combination was 1(450-515 nm), 3(630-680 nm), and 6(2080-2350 nm) with a decorrelation stretch to exaggerate contrast in the imagery to tease out known Chacoan features.

Aerial Emissivity Imagery.
Top Left: “Visual Analysis of Chacoan Roads using Google Earth and Thermal Infrared Multispectral Scanner (TIMS) Imagery”: Identified Chacoan road (77) is referenced in this image on the left using Google Earth Imagery from 06/24/2014. Identified Chacoan road (77) is visible in the Thermal Infrared Multispectral Scanner (TIMS) 1989 aerial image with a band combination of (1, 3, 6) and a decorrelation stretch applied on the right.
Bottom Left: “Visual Analysis of Chacoan Roads Using Google Earth and ASTER L1T At-Sensor Radiance Imagery”: Identified Chacoan Road (77) is referenced in the image on the left, using Google Earth Imagery from 06/24/2014. On the right, identified Chacoan Road (77) is not clearly visible in the ASTER L1T image from 05/12/2016 using a band combination of (3, 5, 1) with a decorrelation stretch applied.
Right: Visual Analysis of The Holmes group (top) and the Morris 39 site (bottom) using HyTES land surface temperature (LST) and emissivity data. Individual structures were identifiable in these data. Image Credits: NASA DEVELOP

We were also successful at identifying two known sites, Morris 39 site and Holmes group, and a known road using Hyperspectral Thermal Emission Spectrometer (HyTES) data. HyTES provides emissivity and land surface temperature data at 5-meter resolution. Some features appear naturally with this aerial imagery; however, when we experimented with band combinations and a decorrelation stretch, additional features became apparent. We performed similar experiments using 90-meter Terra ASTER surface radiance data; no features were identifiable due to the coarse nature of this imagery.
This project revealed both the potential and challenges of using remote sensing to identify and describe ancient Chaco ruins in the San Juan Basin. The Chacoan Site Suitability Map will aid the project partners in narrowing the locations of field surveys to identify new archaeological sites. The Chacoan Sites Risk Map will aid the project partners in allocating resources for site preservation by highlighting known Chaco sites that are at the highest risk of being destroyed by developing infrastructure in the San Juan Basin. Additionally, the results of the project indicate that high resolution thermal data is useful for delineating known Chacoan sites and roads. Furthermore, very high resolution satellite imagery would be useful in further delineating known ancient Chaco sites and roads and to identify new sites and roads.
References
[1] R. Van Dyke et al., “Chaco Landscape: Data, Theory, and Management.” CESU, Boulder, CO, MA P14AC00979, Project Number: UCOB-109, 2016.
[2] T. Sever, “Remote Sensing Applications in Archaeological Research: Tracing Prehistoric Human Impact upon the Environment,” Ph.D. dissertation, Anthropology/Archaeology Dept., Univ. Colorado at Boulder, Order number 911780, 1990.
Author Biographies
Dashiell Cruz graduated from the University of Alabama in Huntsville with a bachelor’s in political science and works with DEVELOP as the impact analysis fellow and assistant center lead at Marshall Space Flight Center.
Sydney Neeley is a graduate from Duke University with a bachelor’s in international relations, a master’s in global health, and a certificate in geospatial analysis. Neeley consults for the Universidad del Valle de Guatemala as a GIS and remote sensing analyst.
Kelsey Herndon is an Auburn University graduate with a bachelor’s in anthropology. She earned a master’s in anthropology at the University of Alabama in Tuscaloosa and is pursuing her master’s in Earth system science at the University of Alabama in Huntsville. Upon leaving DEVELOP, she accepted a graduate research assistantship with NASA SERVIR.
Ryan Schick graduated from the University of South Alabama with a bachelor’s in meteorology and works for ADS Environmental Services as a data analyst. During his tenure at DEVELOP he served as the communications fellow for the National Program Office and independent research consultant for the six projects from the summer of 2015 to summer of 2016.