This presentation is part of the International Cemetery Preservation Summit, April 8-10, 2014 Niagara Falls, NY.
Finding Avondale: Remote Sensing for an Unmarked Cemetery in Difficult Subsurface Conditions by Hugh B. Matternes and Valerie Davis
Hugh: Good morning, I’m hoping you’ll learn not only from the results we have here but some of the mistakes we made and get some practical out of all of that. Cemetery researchers frequently turn to remote sensing technics when there are little to no trace of a burial ground visible on the surface. The effectiveness of these methods has been evaluated by numerous case studies however, these studies tend to be conducted under optimal and under more controlled conditions then we tend to find in the field. In this study we used real world situation where the adverse settings encountered at the Avondale burial place also known as 9BI164, an unmarked cemetery in southern Bibb County Georgia.
In short, records were nonexistent, informative data was sparse and we only had a rough estimate of where the cemetery was located. The grounds were over a century old. There were no surface features and it was situated in Georgia red clay, a notoriously difficult substrate for successful remote sensing. This is the location where it is in space. For those of you who know anything about Georgia, we’re about 10 miles south of Macon right almost in the center of the state. The cemetery had been abandoned, reforested and blurred by subsurface plants and animal activities. Though these conditions made survey difficult it provided us with a means of evaluating multiple remote sensing methods in conditions that were not uncommon in central Georgia. In 2009, New South Associates was asked by the Georgia department of transportation to determine the size and placement of the Avondale burial place.
Unfortunately the cemetery was discovered during the final planning phases of a future construction corridor and had to be relocated. The alternatives to avoiding it would have entailed moving another and as we determined later a much larger cemetery or impacted one of several very large wet land areas. There were no maps or historic records of the cemetery and previous archeological field surveys had failed to detect it. The site came to light when a local informant recalled that there was an old cemetery in an overgrown area along a neglected fence line. His memory of it was over 50 years old and admittedly a little sketchy. The Georgia DOT opened exploratory test units and found several unmarked graves thus verifying that the story was actually true. These units provided us with a general idea of where the cemetery was located.
We eventually determined that the Avondale burial place likely began as a pre-emancipation burial ground and was used throughout the late 19 and early 20th centuries by a dispersed community of African American domestic and farm laborers. The cemetery held the remains of 101 individuals consisting of 40 adults and 61 children. These decedents were placed in east west oriented grave pits arranged in loose family clusters. There were no grave markers present and the ephemeral scatter of brick, glass and shiny pottery emphasized that the Avondale burial place followed many African American cemetery folk traditions. These materials however, were too widely distributed to be useful in defining the cemetery or individual graves.
Our first attempt at defining the cemetery was to excavate a series of stratographic pits to determine the soil quality and composition. Avondale was situated squarely on the fall line. This is a geologic transition between the northern Piedmont Hills and the southern coastal plain. Along the fall line hills of undisturbed, deeply weathered saprolite, more affectionately known as Georgia red clay rise from the surrounding low lands. The cemetery was positioned on one of these low weathered hills. These bulbs of red clay are not very fertile and the farmers tend to avoid them whenever possible. The fact that the cemetery was situated in a corner of four land lots and the soil was relatively infertile probably contributed to why this corner had never been plowed. Soil stratigraphy in the strat pits indicated that the cemetery was surrounded by sandy loams.
We observed a very clear plow zone in units right along the peripheral of the cemetery while those closer to an inside the cemetery exhibited undisturbed natural horizon development comprised largely of sandy saprolitic clays. The cemetery was apparently left forested by the local farmers and as a result tree roots and animal activities blurred the top 20 to 30 centimeters of the sub soil. This bioturbation destroyed the feature definition of the upper soils and as we later discovered had adverse consequences on some of the remote sensing methods. You can see the edge of a grave pit coming up here, comes into the wall goes up to the base of the root line and at this point and everything above it we completely loose any vision at all of where this particular grave is located. This is pretty uniform through most of the graves we found across the site.
We lost practically everything above the root lens up there. We had to get below that before we could actually physically see where this cemetery was located. The project area was surveyed in two phases. An initial survey of the southwestern third was conducted in spring of 2009 and the remaining area was examined a year later in spring 2010. In the intervening year, 38 graves were removed from the 2009 block. Three methods of remote sensing soil compaction, ground penetration radar and canine scent detection were employed to help to find the cemeteries size and contents. The location of each potential grave defined by these detection methods was mapped. Ultimately the true distribution of the cemetery was determined by removing all surface soils and exposing the underlying grave shafts.
We placed a 30 foot wide grave free perimeter around all graves to insure that the true cemetery margins had been identified. During the 2009 survey a 30 by 40 meter area was examined using soil compaction. In clay based soils individual particles align themselves parallel to the ground surface, when left undisturbed they remain compact. When disturbed however, their orientations shift and the soils become less compacted. Over time displaced particles will shift, re-compact and realign with the ground surface. Since grave shafts are soil disturbance the fill is generally less compact then the surrounding undisturbed soils.
Penetrometers were used to measure the soil compaction in pounds per square inch, also known as PSI. Individual probe locations were spaced 50 centimeters apart and when possible readings were taken at about 25 centimeters below the ground surface. In this map the green indicated extremely compacted soils, those rating at about 350 PSI which grades into this red that we see in locations like this where the soil was much looser and tended to average as little as 100 PSI. Low compaction areas were found around mature trees, which is what we see here, here and here, along fence and property lines. This is an old fence line right here and in parts of the cemetery where the Georgia original test units were located which is this location right here. The results showed a highly irregular soil structure but it did not reveal any potential grave shafts. We felt that bioturbation and soil re-compaction had rendered this method ineffective and we abandoned it after our initial testing.
Prior to probing scent detection using dogs trained by the Alpha Team Search and Rescue were employed. These animals were able to detect molecular compounds associated with decomposition represented by only a few parts per million. The Alpha Team brought their dogs out twice. In 2009 the dogs were limited to about a half acre area around the probe test track where we felt that the core of the cemetery was originally located. They kept wandering off to the north east out of the original project recovery area. This is the general location where we originally started our work. In hindsight, we now realize that the animals had actually detected the highest concentration of graves. Between the first and second visits by Alpha Team New South excavated the first third of the cemetery and then back filled the entire area.
This was completed seven months prior to the 2010 visit. In their return in 2010 the dogs were allowed to wander more freely over about a one and a half acre square area. The previous excavated area was ignored during the 2010 survey. During both surveys teams of professional handlers allowed their dogs to systematically investigate the grounds. The same six animals were employed in both surveys. When a dog alerted on a potential grave site the location was flagged, recorded on a project map and then the flags were removed to prevent biasing the other dogs and their handlers. All dogs successfully identified the location of the cemetery. Of this 60 alerts were generated. We defined positive alerts as those were the center point which the dogs alerted us at was within 50 centimeters of a known grave. All told, 13 graves representing 12.87% of the true cemetery assemblage was identified.
As we learned later the upper partitions of three alerts graves had been impacted by previous testing. Potentially bringing decayed compounds to the surface and likely biasing the number of graves identified from an undisturbed setting. Among the undisturbed sample a more conservative estimate of 10 or 9.9% of the true cemetery assemblage was detected. While these figures may seem low they compare very favorably with the 15% achieved by Lassiter’s test using modern human remains which had been buried for only a few months. Now recall that our sample, we’re dealing with remains that have been buried for over a century. We found that no relationship existed between the grave form, it’s size, the deposition date, decedent age or sex and whether or not the grave have been alerted.
Success at the detection method could also be based on it’s ability to be reliable. From the 60 alerts generated by the six animals 25 were positive alerts. About half of the 13 graves identified were detected by more then one animal. As noted however, the upper aspects of three graves had been impacted. To examine success among undisturbed graves the nine associated alerts were removed from the sample and among the remaining 51 alerts 16 were positive. The distribution of scent dog alerts demonstrated that most false positives defined as alerts where no graves were found tended to be within the cemetery limits. As noted by Duprey in 2003, decayed compounds in the cemetery may have been too concentrated for these animals to discriminate specific locations.
During the 2010 survey New South was responsible for false alerts along the eastern and north western portions of the projects area. Alerts corresponded with equipment storage areas, parking, lunch and staging areas. You’ll note these are here, here and up here. Soils from previously excavated graves were undoubtedly scattered across the area from the teams equipment and clothing salting the ground surface with decay compounds. We also have to note with an element of humility that the mortuary archeologists had to leave during the canine survey because out clothing had been contaminated and despite the fact that it was about a year since we had last been out here on our site and our clothing had been washed many times it was still contaminated enough that the dogs kept alerting on us and we were kind of a … We haven’t figured out whether or not to take that as a compliment or not yet.
When you eliminate these 14 biased false alerts canine success rates jump from 42 to 54%. Another method we used was ground penetrating radar. This transmits pulses of radar energy into the ground and measures the velocity of the reflected signals. Signal velocity shift as they encounter buried surfaces. Unfortunately Georgia red clay readily absorbs radar energy significantly degrading the signals that are returned to the transmitter. Since graves in Avondale had been deposited for upwards of a century, their contents were decomposed and the surrounding soil partially re-compacted producing decidedly weaker signals then were typically encountered in more recently deposited intermits. GPR equipment requires ground contact.
Accordingly the GPR surveys were conducted last so that the project area could be completely cleared of vegetation. Results from earlier surveys were not disclosed to the GPR operators until their findings were completed. The 2009 radar survey addressed the same area covered in the compaction survey. The 2010 survey addressed an area just under a square acre in size. A GSS SIR3000 control unit with a 400 megahertz antenna which is capable of producing data for good resolution at depths of up to two meters was mounted on a drag sled for the survey. Radar data was collected along parallel transects oriented roughly north south and spaced about a quarter of a meter apart. In the lab this data was translated into horizontal slices representing common depths.
One of which we see here. The areas where significant shifts and signal amplitude were identified are reflected in green and orange colors. You can see a number of the green ones down here and these orange ones up in this area right here. GPR results from the two surveys were pooled and a total of 71 anomalies exhibited amplitude signals consistent with potential graves. Using the same criteria as above anomaly center points within 50 centimeters of a known grave were considered positive identifications. A total of 36 graves were detected however, four of these overlapped between the two surveys reducing the true score to 32 graves. When contrasted against the true number of graves GPR was able to identify 31.8% of the graves in the cemetery. We found no relationship between GPR detection, the grave form, it’s size, deposition date, the decedents age or sex.
Since two radar surveys were independent, our anomalies were pooled to reveal how well GPR was able to detect Avondale’s graves. True positive occurred among 37 of the 71 anomalies. Mechanical stripping included all areas where GPR anomalies were identified. Many of the false positives were in areas where ultimately determined to be outside of the cemetery. While the source of some of these anomalies was never identified most turned out to be soil interfaces and large roots. Recognizing that probing was ineffective, how did scent detection and GPR compare? Since the true population of the graves was the same for both methods a simple comparison of the number of graves identified emphasized that GPR found 19 more unique graves then canine detection.
Effectiveness can also be measured by the ratio of true to false positives since sample sizes for potentially identified graves were not equal we employed the odds ratio statistic with about a 95% confidence interval to assess whether differences varied significantly. For those unfamiliar with these statistics odds ratio values above 1.0 indicated a ratio in favor of one method while values below favor another. Scoring exactly 1.0 indicates that the methods were equally effective. An odds ratio confidence interval that does not include 1.0 as a possible value is considered significantly variable enough to infer that one methods effectiveness is truly different from the other.
An odds ratio value of 0.694 was calculated favoring GPRs ability to detect more graves. However, since the confidence interval range included 1.0 and values on both sides of the ratio the methods were not significantly different enough to reveal a clear superiority. A two tailed fishers exact value of 0.38 emphasized that detection values did not vary strongly. Another way of looking at effectiveness is in the terms of how well the methods were able to pinpoint the location of the cemetery. When we examined where false positives were located our GPR data placed a number of these on the southern and eastern sides. By these estimates we initially defined the grave size as somewhere around 3,150 square meters. In contrast, scent dogs were predominantly either within the cemetery or along it’s eastern or northwestern sides.
As noted earlier, these hits corresponded with areas that were contaminated by us, the archeologist. If the spurious alerts were eliminated, the dog alerts were confined to a much smaller area. The resulting cemetery size was close to 1000 square meters focusing largely on the cemetery’s core. We estimated the true size of the cemetery to be 1,350 square meters while canine success rates for identifying individual graves were lower then the GPR rates canine alerts were more concentrated within the actual core of the burial ground.
We know recognize that Avondale was an unmarked cemetery with no recorded history placed in a very difficult soil environment and in a place were bioturbation had erased the tops off of the grave shafts. Three remote sensing methods were used to figure out it’s structure. In some environments soil compaction can be a very effective means for detection. In one particular case we’ve been able to identify up to 80% of the graves in a ground truth cemetery that we worked on over in Alabama. Unfortunately Avondale was not in one of these environments. Clay re-compaction and bioturbation meant that the cemetery simply could not be detected using this method.
We concluded that soil compaction was ineffective for this particular environment. GPR and canine detection, scent detection were much more effective. In terms of defining the cemetery we found that the scent dogs were more accurate in terms of being able to identify the cemetery’s location and it’s general size. Once inside the burial grounds however, the dogs had trouble distinguishing individual grave sites. As a result there were a lot of false alerts. GPR was more effective in identifying the location of individual grave sites but it had spuriously identified a number of non grave anomalies which were clearly outside of the true burial area.
In retrospect we’re surprised that anything worked at all out at Avondale. It was a tough place to figure out where the cemetery was. The results of our remote sensing emphasized that the application, the multiple methods can result in different yet complimentary information. Some methods will turn out to be ineffective particularly when faced with adverse subsurface environments. In Georgia’s red clay environments however the use of scent detection dogs to define where the burial ground was located which would then be followed up by an intensive GPR sweep seems to be a more effective means of identifying where unmarked burials may be located. The use of at least two and optimally more remote sensing methods is advocated to provide estimates of the location and size of unmarked graves and burials in this particular part of Georgia.
Again, I want to go ahead and plug, this is a small part of this project. We had a huge amount of interaction with the decedent community. It was a joy working with these people and the results of all this are up on our website which has been dedicated to the Avondale burial place. The other thing which I’ll point out is that at the end of it all in working with the descended community we were able to place the cemetery in another local church which turns out to be an important part of the … This cemetery and this church as it turns out had a common history that was unknown before all this and it came out through the work with the descendants that this was a very appropriate place for it.
We put these guys back in the ground in 2012 and there has been because of this, there has been a lot of collaboration between the church congregation who did not know about this descendant community and the descended communities who are just ecstatic to find other members in the community that they didn’t even know were still out there. There we go.
One of the most challenging tasks for cemetery researchers is to positively identify the location of unmarked cemeteries. In central Georgia the problem is compounded by the presence of undisturbed beds of Saprolite, the deeply weathered remains of metamorphic rock, more affectionately known as Georgia red clay. These clays are extremely dense, are prone to re-compact relatively soon after disturbance, and they contain significant amounts of iron, making the application of non- and minimally-invasive remote sensing techniques a challenge. When left undisturbed for many decades, bioturbation will erase surface characteristics that can normally be used to help identify unmarked grave details. Our need to identify the location of the Avondale Burial Place (9BI164), an unmarked cemetery in southern Bibb County, Georgia, provided an opportunity to compare how different techniques fared in a worst-case scenario.
The Avondale Burial Place was a nineteenth- through early twentieth-century rural folk cemetery used by a dispersed community of African-American sharecroppers and farm laborers. The cemetery occupied the corner of a land lot on a slight rise of red clay that in all likelihood was never plowed. We ultimately determined that the grounds held the remains of 101 individuals, placed in an east-west orientation, and arranged in what appeared to be loosely organized family clusters. There were no grave markers present, but the ephemeral scatter of glass, potsherds, toys, and tile fragments emphasized that the community probably decorated grave surfaces following African-American folk cemetery traditions. These materials only approximated the general area of the cemetery and provided no information on individual grave locations.
While limited testing confirmed the presence of the cemetery, its exact size and distribution was unknown. Three methods of remote sensing, including soil compaction (or probing), ground-penetrating radar, and canine detection (cadaver dogs) were employed to help define the cemetery’s boundaries. The location of each potential grave was mapped during each survey. Based on the distribution suggested by these methods, we used heavy equipment to remove all surface deposits to expose the underlying grave shafts. While this provided us with the true location of each interment, it also generated a means of examining the effectiveness of each remote sensing method.
The results of the soil compaction survey revealed a highly irregular pattern of soil compaction immediately beneath the surface that likely reflected more recent natural intrusions (burrows and roots). Potential grave shafts were not detected and the method was determined to be ineffective for the given environment. Using ground-penetrating radar, a total of 71 anomalies were interpreted as potential graves. Only 32 (45%) of these ultimately corresponded with true graves. Many of the false positives were in areas that were determined to be outside the cemetery. Scent-detection dogs specially trained in grave identification by the Alpha Team Search and Rescue group, were employed to define the cemetery. Of the 60 positive alerts generated by these animals, 16 (26%) positively corresponded with true graves. While canine success rates for identifying individual graves were lower than GPR rates, canine alerts were more concentrated to within the burial ground. Most of the false positives outside the cemetery were associated with areas where equipment used in previous burial recoveries by New South had been stored.
While ground-penetrating radar was able to more accurately define the location of individual graves, its results included extra-cemetery anomalies that inflated the size of the burial ground. In contrast, canines were less able to define exact locations for individual graves; however, their alerts more accurately pinpointed the cemetery’s true location. The results from our triad of remote sensing techniques emphasized that the application of multiple methods can result in different, yet complimentary information. Given that some methods may be ineffective, particularly when faced with adverse subsurface environments, the use of multiple remote sensing methods is advocated to provide more accurate estimates of the location and size of unmarked graves and cemeteries.
Dr. Hugh B. (Matt) Matternes of New South Associates is the Director of the Cemetery Studies Program, where he is responsible for the identification, preservation, and recovery of cemetery sites, and the professional and ethical treatment of human remains. He received his B.A. in Anthropology from the University of North Carolina, Greensboro, his M.A. and Ph.D. in Anthropology with a focus in physical anthropology and mortuary archaeology from the University of Tennessee, Knoxville. He is a Registered Professional Archaeologist (RPA) with over 28 years devoted to studying cemeteries. He is the author or co-author of over 80 technical reports and professional publications. His research interests currently focus on community dynamics as reflected in post-Reconstruction era African American cemeteries.
Valerie S. Davis
Valerie Davis is a professional Mortuary Archaeologist and Physical Anthropologist at New South Associates, Inc. where she supervises field and laboratory cemetery investigations. She is a Registered Professional Archaeologist (RPA) with over 13 years of experience, and is proficient in identifying folk and ephemerally marked cemeteries, documenting surface and subsurface mortuary data, as well as conducting in-field and laboratory analysis of human remains. She received her BA in Anthropology and her MA in Anthropology with a focus on Bioarchaeology from Mississippi State University.