The application of ground-penetrating radar (GPR) to the search for unmarked graves is well established in formal cemeteries. The use of GPR in other contexts, such as landscapes of Indian Residential School (IRS) institutions, shows considerable utility, though refinements are ongoing.
In this section, we provide a short and concise guide to our view of best practices and their justifications. We are currently working on a published version of this guidance, but we want to provide information as soon as we develop it.
The content of this page is reproduced within the course files, but this is a more succinct version. If you wish to reference this guidance, please cite the authors as Andrew Martindale, Steven Daniel, Aeli Black, Kevin Wilson, and Elizabeth Campbell.
Methodological Challenges
The identification of unmarked graves in cemeteries using GPR is facilitated by 1) the preference for locating formal cemeteries in geological zones of homogenous, small-sized sediments, which tend to create a non-reflective background against which unmarked graves are highly visible; 2) the pattern of rectilinear grave shafts, the shape of which often creates characteristic GPR signal patterns; 3) the absence of non-grave features in the subsurface that can mimic unmarked burials in GPR signals; 4) the ordered arrangement of grave alignments, which allows for the anticipation of grave locations within rows and permits GPR data collection parallel to these alignments, and creates angles of incidence which enhance the appearance of unmarked graves in GPR data; and 5) the prior knowledge that graves will be present in cemeteries, which creates an additional non-GPR line of evidence, an important corroboration of any GPR results.
In informal landscapes, such as those associated with IRS institutions, some or all of the above may be absent, creating additional challenges for GPR. We argue that as work progresses, it would be useful for communities collecting GPR results on IRS landscapes to 1) conduct methods tests for best practices and parameters of data collection in their context, and 2) share the results of GPR visualization patterns (i.e., the kinds of traits associated with unmarked graves) between communities to better refine best practices.
Methods Test
We recommend that GPR operators conduct a methods test in an area of known ancestors (i.e. marked or known graves). This provides operators who are new to the location with useful experience in local conditions and generates valuable data for further work and for general refinement of the method. We recommend choosing a location of 4-6 ancestors (about 5 x 6 m grid). Operators should conduct data collection across this grid, using our suggested standard configuration, and then repeat while varying one GPR parameter setting each time (see below).
Standard GPR Parameter Settings. We have several decades of experience in the use of GPR in the location of unmarked graves (mostly within formal cemeteries). We have conducted informal and formal (published results pending) tests of GPR parameters (see below), and we recommend the following data collection settings:
- Bidirectional grids (i.e., in perpendicular directions)
- Forward-only grid lines (i.e., collect lines from the same baseline, avoid Z pattern collection)
- Grid aligned with grave orientation (i.e., grid lines should cross the grave shafts perpendicular to their axes)
- 25 cm spacing between grid lines
- 2 cm step size (i.e., the distance between trace initiations (GPR signals) along a line)
- Dynamic stacking (i.e., the number of redundant signals in a singe trace – multiple signals can be averaged for clearer results; most manufacturers use a dynamic algorithm that monitors the signal noise and increases of decreases stacking accordingly. If this is not available, then stacking should be set at 4 as a minimum and higher in sedimentologically busy contexts)
- 200-800 Mhz frequencies.
Testing Parameters. We encourage users to collect data of a known grid while varying parameters (one at a time) to permit the evaluation of best practices. Here are our preliminary results of varying parameters:
- Unidirectional grids: While this practice saves time, we argue that it prevents a key feature of data interpretation – the ability to inspect graves at different angles of incidence. This reduces grave identification confidence. If it cannot be avoided, we argue that lines perpendicular to the short axis of a grave produce the best radargrams for interpretation.
- Conduct a test either by collecting unidirectional data or by removing bidirectional data from a standard grid. The impact is most pronounced on amplitude map visualization and on the ability to confirm grave identification using alternate radargrams.
- Z Pattern data collection: This is also a time-saving measure. However, if your odometer calibration is in error or if ground conditions generate odometer errors, Z pattern data will create a noisy signal that is more difficult to interpret. If these conditions can be avoided, Z pattern data collection works well. We recommend checking your odometer calibration each day.
- Conduct a test of Z pattern data with an incorrectly calibrated odometer, coherent pattern in GPR visualizations will take on a zigzag quality.
- Grid alignment: Most GPR data of unmarked graves has been conducted in cemeteries and thus is usually collected at parallel/perpendicular angles to the grave orientation. If the angle of incidence between the GPR and the grave varies beyond 90, the quality of the trait data in radargrams deteriorates. At 45 degrees, some graves are not visible in either radargrams or amplitude maps.
- Conduct a test by resetting the test grid to 45 degrees from the standard. Data quality will deteriorate in both radargrams and amplitude maps, especially for marginal identifications.
- Grid line spacing: We have conducted tests of grids at 10 cm, 25 cm and 50 cm. 10 cm spacing at 500 Mhz has minimal impact of data quality (but is considerably more work). 50 cm intervals generate considerable data loss, especially in amplitude maps and in opportunities for multiple radargrams.
- Conduct a test by collecting data at 50 cm intervals or by removing every second line in a standard grid using software. Marginal patterns in amplitude maps will become considerably less obvious and options for reviewing radargrams are reduced. In a standard grid, a standard-size burial (about 1 x 2 m) will be intersected by between 10 and 14 radargrams. At 50 cm lines spacing this is reduced to 4-6. We argue that traits of burials do not appear in each radargram of a burial – though we are not yet sure why. We conclude that more radargrams per burial is necessary for the identification of unmarked graves. This is likely more important when grave contexts and forms vary from patterns found in cemeteries.
- Step size: 2 cm is standard on most equipment and, with modern processing and file storage capabilities, permits the collecting of detailed radargrams. Radargrams are the foundation of burial identification in GPR, especially in non-cemetery contexts. If data are collected at larger intervals, the quality of the radargrams deteriorates. This is especially true of contexts that are noisy.
- Conduct a field test with step sizes larger than 2 cm. The resulting radargrams will have less data that the standard.
- Stacking: Stacking creates a less noisy signal in a single trace (a point of signal emanation). The utility of this parameter is influenced by the nature of the target signal and the noisiness (i.e., presence of non-target features generating reflection) of the context. Dynamic stacking accommodates this. There is likely a threshold for each context below which signal noise will increase and cause deterioration of radargram quality against the standard.
- Conduct field tests of the standard grid with manual stacking at 1 to observe the degree of data loss.
Field Practices
We recommend standard GPR parameters, as defined above, when conducting grids. Our method of field collection follows these steps:
- Conduct a series of test grids in a known burial context, as discussed above if possible, to confirm utility of standard parameters or to create a modified standard for your context
- Establish project reference points for spatial (i.e., mapping) data
- Conduct a roaming survey to:
- Locate high potential areas
- Define grave orientation, if possible
- Establish project spatial datum and locate against project reference points.
- Map targets located with roaming survey
- Establish project datum and baseline to facilitate grids covering areas of interest/potential. In complex landscapes, multiple baselines may be needed. Further roaming surveys may be useful at this stage.
- Collect grid data using standard or modified standard parameters
- Collect spatial data of all grid reference points
- Ensure all sketch maps and gird forms are completed
- Copy and back up digital data files
- Conduct drone photomapping as necessary to create a high resolution base map for locating GPR results