- Home
- >
- Preservation Archaeology Blog
- >
- UAV Photogrammetry in Southwest New Mexico
Kent Mead, University of Arizona
(March 14, 2018)—In 2017, Archaeology Southwest sponsored an internship for a photogrammetry and 3D-modeling project in southwest New Mexico, and I was fortunate to be that intern. I worked with Preservation Archaeologists Doug Gann and Karen Schollmeyer to coordinate a series of quadcopter drone flights and conduct low-altitude aerial photogrammetry on a select group of archaeological sites in the Upper Gila River watershed near Cliff.
The internship was initiated as an opportunity to learn unmanned aerial vehicle (UAV) flight skills, photogrammetry techniques, and 3D-modeling procedures. Dr. Gann taught me innovative techniques aimed at creating photorealistic 3D models of each site. Dr. Schollmeyer helped guide the research and facilitated access to the sites.
Unmanned aerial vehicle photogrammetry is an increasingly popular method for archaeological mapping and data collection. Drones offer a low-cost and relatively easy to use platform to capture low-altitude aerial imagery. Three-dimensional models created with specialized image-analysis software offer a new way to analyze and interpret spatial data. For this internship, we created 3D models with a mosaic of hundreds of high-resolution, digital aerial photographs compiled and processed using AgiSoft Photoscan software. We used a DJI Phantom 4 Pro quadcopter drone with built-in 20 megapixel to collect the imagery.
The project included drone flights and low-altitude aerial photography of five sites on Nature Conservancy lands (TNC-1, TNC-4, TNC-5, TNC-6, and the Gila River Farm site) and one site on Museum of New Mexico lands (Woodrow Ruin). We chose these six sites because collectively they contain visible surface features such as rock alignments, rubble mounds, and pit depressions, and in some cases show evidence of disturbance, such as looters pits or bulldozer scrapes. We chose the UAV photogrammetry and 3D modeling to document these subtle expressions in the surface terrain of sites because these techniques offer a new perspective on the areal limits of a site and the larger spatial extent of features as they are situated in the landscape.
The resulting models offer an innovative way to study archaeological sites with a dynamic and interactive perspective. For example, sites with obvious surface features documented in this manner offer the potential for mapping with precision comparable to traditional tape-and-compass maps or sites mapped with GPS (Global Positioning System). In addition to horizontal spatial documentation, 3D models of sites can display accurate topographic changes of sites and the surrounding landscape, including height and depth of feature in relation to the surrounding ground surface. This method of remotely sensed site mapping offers yet another method to document archaeological sites with a bird’s-eye-view perspective, one that is typically unavailable to researchers stuck on terra firma.
Each of the six sites documented in this manner offers a distinct set of archaeological and environmental conditions that test the usability and effectiveness of the technology in the field and for in terms of the resulting 3D model’s clarity for interpretation.
The Sites
TNC-1 is a small site located on a mesquite-covered hilltop within the Gila River floodplain. Three possible pit features and a small sample of Mogollon Red-on-brown and Three Circle Red-on-white pottery indicate that people were there in the Late Pithouse period. A single masonry rubble mound feature dates to the Early Pueblo period, and a single Mimbres Style III Black-on-white sherd suggests that people were indeed here in the Classic Mimbres phase. The rubble mound measures 1.3 m x 2.7 m. There is also a variety of historic artifacts, including datable metal objects and glass food-storage containers from the early twentieth century. Much of the site is overgrown with mesquite thickets that obscure large portions of the site surface.
TNC-4 is a Late Classic period Mimbres site with some evidence for use in subsequent decades. It is located on an escarpment edge of the first floodplain terrace of the Gila River near the community of Riverside. Multiple habitation features are arranged in two contiguous room groups. There is also an artifact scatter of flaked stone, ground stone, and pottery sherds. A large J-shaped pit disturbance nearly bisects the site, causing significant impacts to subsurface deposits. This is probably a backhoe trench related to pot-hunting.
The southern room group exhibits a cimiento-style foundation of parallel upright facing stones on interior and exterior sides that once supported adobe walls. Two rooms are visible on the surface with a single interior dividing wall of coursed unshaped cobbles. This structure was probably looted, because the sediments of the room interior are lower than the exterior ground surface and interior construction is visible.
The northern room group has evidence for cimiento and coursed cobble-masonry construction defining at least three complete rooms and part of a fourth. Additional surface cobbles suggest that this room group may have been much larger, but surface erosion probably displaced much more material off the edge of the escarpment. The presence of Mimbres Style II and III pottery, non-local flaked stone of Zuni-spotted chert and obsidian, and adobe architecture suggest this site was continuously inhabited through the end of the Mimbres Classic period into the Reorganization phase in the 1100s. The site surface is mostly open with no canopy, and surface vegetation consists of mostly medium grasses and sparse mesquite shrubs.
Doug Gann’s model of TNC-4 is available on Sketchfab.
This model resolution has been reduced from 5 million polygon faces down to 300,000 faces. The model shows a small ridge and escarpment adjacent to the Gila River. TNC-4 is confined to the top of the widest portion of the ridge landform. A zoomed-in view of the site shows the J-shaped depression which is also distinguished by a dense cluster of small mesquite bushes. This model offers a good example of the potential for 3D models to show aspects of an archaeological site that cannot be fully captured with 2D maps or photographs. The TNC-4 model shows a photorealistic landscape that is geographically accurate in location and physically accurate in dimension.
TNC-5 is located approximately 80 meters from TNC-4. It sits on portions of the same first terrace and escarpment overlooking the Gila River to the west. The two sites are separated by a small wash that may have formed after people left the settlement. TNC-5 is also a Late Classic period Mimbres habitation site that may have been used into the 1100s. The close proximity of the two sites and contemporaneous use indicate they were perhaps a single continuous site that has since been divided by a wash. The site suffers from extensive disturbance from mechanical grading of surface deposits. The western portion of the site may contain undisturbed deposits. It shows evidence of poorly defined rectilinear rock alignments made of unshaped cobbles; however, these were not identified a masonry room features. The site is open with surface vegetation of sparse grasses and mesquite scrub.
TNC-6 was used during the Mimbres Classic period and during the early 1900s. It is located near the mouth of Box Canyon where the Gila River transitions from the mountains to open valley. It occupies a first floodplain terrace overlooking the river to the west. This site was not fully documented by the field school because of limited time. They did document a variety of artifacts, including a high density of Mimbres Style III Black-on-white ceramics. Despite a profusion of cobbles, no definitive architectural alignments were defined. The second component contains possible historic masonry features and a scatter of artifacts including diagnostic glass. A Google Earth satellite image of the site shows east–west trending striations across the site, which suggest previous bulldozer disturbance. Mechanized looting may account for the lack of visible Classic period surface architecture despite the high density of surface artifacts.
Doug Gann’s model of TNC-6 is available on Sketchfab.
The resolution of this model has been reduced to 450,000 polygon faces in order to decrease the file size and be able to share it on Sketchfab. The TNC-6 model shows a grassy covered floodplain terrace adjacent to the Gila River. A small arroyo flanked by juniper trees partially bounds the east side of the landform. TNC-6 covers the entirety of the landform. Because a majority of the site was disturbed by what is thought to have been a bulldozer, however, there are no longer any obvious features in relief on the surface of the site. Additionally, the thickness of the grasses obscured any other subtle remains of site features.
The Woodrow Ruin site (LA 2454) is a large Classic period Mimbres village with an underlying Late Pithouse period village. It is located on an escarpment and first floodplain terrace of the Gila River, near the northern terminus of Grant County Road 293. One of the largest sites in the upper Gila region, Woodrow Ruin has been investigated on several occasions, beginning with its initial recording in 1929. A 1957 survey of the site estimated that there were more than 100 cobble masonry rooms with evidence for adobe wall architecture in places. Additionally, underlying pit house depressions in association with high trash mounds were thought to be reminiscent Hohokam sites. A second survey of Woodrow Ruin in 1969 identified discrete roomblock concentrations with possible open plaza areas. This surveyor estimated that there were 33 pit house depressions and surmised the site was inhabited from A.D. 700–1200.
In recent fieldwork at Woodrow Ruin, Jakob Sedig analyzed surface artifacts and conducted test excavations. He performed the first comprehensive surface survey of the site, including GPS mapping and magnetometry to delineate surface and subsurface features. The GPS survey produced a map of surface topography with 10-cm contours. Based on his results, Sedig estimates that the Late Pithouse period occupation at Woodrow “included approximately 75 contemporaneously occupied pithouses” with a population of between 300–450 individuals…the Classic period occupation contained at least 100 surface cobble rooms distributed amongst 10–13 roomblocks,” and “[I]n some instances, these rooms and roomblocks were built directly on top of the earlier Transitional and Late Pithouse period structures.”
The site model for Woodrow Ruin was created by Doug Gann. It is available on the Sketchfab website.
The resolution of this model has been reduced to 500,000 polygon faces. Similar to TNC-4 and TNC-6, the model for Woodrow Ruin shows a prominent floodplain terrace defined by an escarpment. The boundary of Woodrow is easily visible in the model because the site is bounded by a fence. The vegetation inside of the site fence appears as a mottled gray and lime-green color, easily contrasting with the tan-colored surface outside of the area. Zooming in on the Woodrow model, you can see rectilinear shapes of masonry roomblocks and the slightly square-shaped depression of the great kiva near the center of the site.
The Gila River Farm site (LA 39315) is a large Cliff phase (1360–1450) adobe village site of the Salado phenomenon. It is located on the first floodplain terrace of the Gila River adjacent to the New Mexico Nature Conservancy Lichty Ecological Research Station off Grant County Road 293. The full extent of the site is unknown because it is buried by sediments; however, buried architecture is suspected at two locations based on the presence of conspicuous mounded sediments and differences in surface vegetation growth showing linear patterns. In 2016 and 2017, the Preservation Archaeology Field School, led by Karen Schollmeyer, conducted limited-scale excavations at three locations within the site. Work revealed intact adobe architecture at all three locations, including preserved wall alignments and plastered floors with artifacts on them in situ.
There are two site models for the Gila River Farm (GRF) site, both created by Doug Gann. They are available on Sketchfab, here:
The first GRF site model is depicted in 292,500 polygon faces. The first link directs to a model of GRF just after the Preservation Archaeology Field School completed their 2017 excavations. It shows a number of open excavation blocks and trenches. Through the middle of the site is a long excavation trench in which the field school exposed the long adobe wall of a roomblock. At the end of this trench is an excavated room with a number of floor features.
The second model consists of 407,800 polygon faces. The second model shows a different perspective of the GRF site, taken later in the year, after the excavations were closed and backfilled. This model shows the site area in greater scale. Although no site features are visible and the excavated earth from months prior is now overgrown, you can see a number of modern surface features, including a dirt road that crosses through the site area.
***
The four Nature Conservancy site descriptions (designated with “TNC”) are summarized from Evan Giomi’s 2016 Survey of Nature Conservancy Properties along the Upper Gila River Grant County, New Mexico, which details survey results for the Upper Gila Preservation Archaeology Field School 2016 field season. The description of Woodrow Ruin (LA 2454) is summarized from Jakob Sedig’s 2015 dissertation, The Mimbres Transitional Phase: Examining Social, Demographic, and Environmental Resilience and Vulnerability from A.D. 900–1000 in Southwest New Mexico, which describes recent and prior investigations of Woodrow Ruin. The summary description for the Gila River Farm site (LA 39315) is based on Karen Schollmeyer’s report, Preservation Archaeology at the Gila River Farm Site: Research Update 2016.
One thought on “UAV Photogrammetry in Southwest New Mexico”
Comments are closed.
Good job, Kent! I’ve developed an interest in UAV phogrammetry as well so I really enjoyed seeing this. One quick and simple question … Which version of AgiSoft Photoscan was used? I’m assuming it was the professional edition. I’ve been working with the standard edition which has given me good results with artifacts and unit excavations but doesn’t seem to be powerful enough to handle large-scale projects like what you did here.