Drone aerial imagery for the simulation of a neonate burial based on the geoforensic search strategy (GSS)

A woman reporting the homicide and burial of an infant in 2004 prompted the creation of an experimental simulated neonate grave shortly before the real search commenced. The real case, documented here, did not use aerial imagery, but used ground‐penetrating radar (calibrated to the test site described here) to identify two locations that were probed for gas release and the deployment of victim recovery dogs. We suggest technological advances in remotely sensed aerial imagery that have developed since 2004 will demonstrate their use in focusing such searches by informing a Geoforensic Search Strategy (GSS) and suggesting locations accessible by a perpetrator to identify a burial location using the still‐existent analogue site. To test this, in the spring of 2020 a DJI Mavic Pro drone was flown over the control site containing the simulated 2004 burial. Aerial image processing included the creation of orthomosaics, Normalised Difference Vegetation Index (NDVI), Visual Atmospheric Resistance Index (VARI), and photogrammetry. Conventional ground‐based geophysical surveys using ground‐penetrating radar, guided by this new type of information integrated into the GSS, confirmed that anomalies seen in drone data were the 16‐year‐old burial. We test this strategy using both the original simulated burial in Northern Ireland and further evaluate it in two recent simulated graves in the United States in more complex scenarios, but with successful results.

production of a separate Conceptual Geologic Model (CGM), whose purpose is to provide a preliminary geological assessment of the subsurface around the target, which informs which resources are needed to execute the search [5]. The GSS incorporates conventional techniques such as geological mapping and associated geomorphological observations, analysis of traditional aerial photography, and the use of engineering geophysics (e.g., ground-penetrating radar, and electromagnetic, magnetic, conductivity, and electrical resistivity surveys). Subsequently, the GSS has developed and evolved to combine this geological approach with other search and law enforcement techniques and now comprises a rigorous yet adaptable workflow, outlined below in Figure 1.
We highlight aspects of the GSS relevant to the experimental part of this work, as the historic case this is based on occurred prior to the development of the GSS. Furthermore, we evaluate what developments in aerial imagery may bring to this tried and tested, if always evolving, search model. The GSS was not developed as a prescriptive strategy to be followed for each and every search. Instead, the GSS provides a framework to work within, which can be adapted to suit the individual circumstances that are unique to each crime and search area, and enables measurable and proportionate resources to become applied to each task in a timely and cost-effective manner. The GSS has 30 stages divided into three distinct phases known as "Pre-search," "Search," and "Post-search." The work in this paper aligns predominantly with the "pre-search" phase of the GSS framework. The stages included in this work are listed below, and their integration into the full GSS is shown in Figure 1.

| Aerial imagery
Data from satellites and aircraft have long been used in the search for mass burials [6]. However, the issue of low image resolution at the scale of an individual adult burial, let alone that of a child or neonate (as in this study) remains a problem, especially with the passage of time due to body decomposition, land-use changes, and vegetation growth. For some time, satellites, helicopters, and fixed-wing aircrafts have been capable of performing regional assessments of the ground using remote sensing data, which has been complemented by high-resolution ground-based tools such as geophysical surveys, but an intermediate solution has been lacking, which can now be filled by drone technology.
For this reason of scale, drones (also known as unmanned aerial vehicles or "UAV's") have found increasing use in supporting ground searches, some of which are summarized, below. An overview of a range of applications can be found in Murray et al. [7], Horsman [8], and Mishra [9] provide preliminary introductions to the technology of the time, largely based on crime scene recording as opposed to search-the topic of this work. Mendis et al. [10] and Urbanová, et al. [11]  and Parrott et al. [13] work were primarily focused on the efficacy of the drone-based survey method, as opposed to our approach which is to integrate primary and processed aerial imagery with an understanding of the overall solid/drift geology and topography (part of the GSS), with attendant controls on vegetational variations and digital elevation modeling using photogrammetry.
In a 2020 study by Butters et.al., an infrared sensor was mounted to a DJI drone and flown over a simulated burial using two large cuts of pork. The infrared sensor imaged a thermal anomaly as decomposition from insect activity progressed [14]. Remote sensing such as this and drone-mounted LiDAR [15] are relatively new to crime scene detection and are not within the scope of this study.
From the summary above we can conclude that a significant surveying method may complement the Geoforensic Search Strategy by integrating drone data such as orthophotography/orthomosaics, digital elevation modeling, and vegetation indices (such as NDVI) in order to focus a search.

| Background
In March 2004 a Caucasian female informed police in England that she wished to report a murder (more correctly, a homicide). As a teenager some 20 years earlier, in a rural area of NW Ireland, she said she became pregnant during an affair with a foreign national of different race, which at the time would have carried deep social stigma. Upon home delivery of the infant, the woman (then teenager) believed her mother drowned the baby, and went alone with the corpse in an old cloth handbag with wooden looped handles (no metal parts) to bury the body. Such traumatic events would remain memorable and the witness recalled her mother's boots and shovel had abundant fragments of what she described as "bog rushes" on them (possibly Juncus sp.), upon return to their house. The witness did not know the location of the burial. As her mother had exited their house via the backdoor, which led to a sloping field with a waterlogged, marshy ground (bog) at its far extreme, the burial was likely somewhere in this general location. Upon interview, the English police and their psychologists thought the witness testimony to be credible and contacted the local police in Ireland. The daughter's motive was to provide the child with a proper burial, and although she risked being charged, she had waited until her mother died so  the remains were returned to the mother via a funeral service and F I G U R E 1 Full 30 stages of the GSS divided into "Pre-Search," "Search," and "Post-Search" phases, with highlights in red showing which stages this work intersects as listed above. (Source: Laurance Donnelly, modified after Donnelly [24]; Harrison and Donnelly [25] [Colour figure can be viewed at wileyonlinelibrary.com] interred. No charges were brought forward by either police force to the authorities.

| Control site summary
The simulated burial within the control site was created in 2004 after the homicide was reported but before the official search had begun. It is located on the author's own property (AR) in Northern Ireland, shown in Figure 3, and thus no external permissions were required. A full description of the site is provided below. In summary, the location was analogous to the child burial site, comprising gently undulating fields with bog-filled soil at its edges, overgrown with the common Irish Rush, Juncus (probably Juncus effusus). However, the underlying geology of both the original 2004 search and the 2020 control sites are different, the former being Carboniferous Sandstone and the latter Palaeozoic greywackes. This could cause a potential problem, but as both are covered by glacial till (unknown thickness, but at least some meters thick), and the peat (described below), the efficacy of the test should be unaffected as it was suspected the 2004 child grave would be shallow given the witness testimony.
To test the resolution and depth limitations of the GPR in a 2004 experiment, a simulated, rectangular grave, 1 m deep, 0.75 m wide, and 1.7 m long was cut into the peat until the underlying glacial till was reached. A cloth handbag containing some woolen clothes was placed in the ground and covered. It was noted at the time that having dug and buried the bag, workers' boots had peaty soil and cut Juncus fragments adhering. This kind of control burial is common practice in geophysics and especially forensic geophysics [16].

| Desk study
In the Geoforensic Search Strategy (GSS) [3,4], Stage 4 recommends a desk study to be completed, including the collation of pre-existing data, information, and intelligence, which is considered good practice [17]. Typically, this includes a review of topographic, land-use, soils, and geological maps, local history records, and anecdotal accounts. This should be followed by a reconnaissance walk-over survey to observe the general ground conditions including the geology, geomorphology, and vegetation changes, and perform an evaluation of the diggability of the ground if a burial is being considered. The geomorphology will be addressed later, as will the case intelligence.  Mavic camera has a consumer-grade filter above its sensor, which cuts off most of the light below UV and above IR wavelengths, but it does let in enough near-infrared (NIR) light to create a meaningful NDVI analysis, and recent studies have shown genetic algorithms using standard RGB cameras can produce similar results to NDVI taken with a modified or multispectral/hyperspectral camera [21]. NDVI uses the following equation:

| Focused survey
The orthomosaic imagery and possible burial locations were integrated with the topographic map, wherein it was noted that Location 1 is elevated (Figure 4C

| The burial at location 3
Negating two locations with Juncus vegetation on the basis of visibility from neighboring dwellings or lack of thick, soft, easily diggable soil, our trial focused on Location 3 (see Figure 6A), which contained the simulated burial from 2004. The increased resolution of the topography around this area of bog and Juncus vegetation also produces an elevation map that focuses the search to the north of the red-yellow elevated ground in the low-lying green and blue areas ( Figure 6B). NDVI of the location overlain on topography ( Figure 6C) shows a discrete area with similar response to normal bog and Juncus vegetation, but distinctly separate from the main bog area.
VARI overlain on topography ( Figure 6D) did not show this anomaly.
When the original, high-resolution orthomosaic of this area is examined ( Figure 7A), during the period of dry weather, when the imagery was obtained, this anomaly can be discerned some 16 years after the burial took place. In less-dry conditions, this may not be visible, as is often the case in archeological analysis of aerial photography during droughts [22]. A further test of NDVI and other indices when the soil has an elevated moisture content is recommended as future test work.
To complete this process of focusing the survey in order to (theoretically) advise whether Location 3 and its vegetation anomaly should be archeologically excavated, GPR was deployed (Figure 7).
Reconnaissance 2D 100 MHz profiles using a Mala rough terrain antenna were gathered to assess peat bog thickness (Figure 7B)

| DISCUSS ION
Our contribution to an actual search using aerial imagery is different from those who precede us [12,13] in that we placed the use of drone technology within the context of the Geoforensic Search Strategy to provide a basis for understanding ground conditions in and beyond the immediate area of interest (Control Site Survey Area). Subsequently, there was a more detailed focus on the survey locations (Focussed Survey Area and Location 3), using desk study combined with drone image analysis. Our burial is also considerably older (16 years) than that created by Parrott et al. [13]. A limitation of this experimental work is that the place of the burial was already known, as this was a simulation. However, the process was realistic and based on an actual case. The use of aerial imagery in this context, as opposed to assessing the response on specific sites is novel and