Search, Detect, Recover: A Systematic Review of UAV-Based Remote Sensing Approaches for the Location of Human Remains and Clandestine Graves

Highlights

What are the main findings?

• This systematic review found that UAV-based remote sensing is a low-cost and low-altitude alternative for clandestine grave location. However, the knowledge base is fragmented due to inconsistent reporting on seminal aspects of experimental designs and aerial survey parameters.
• Several key trends were identified through a synthesis of the main themes across all studies reviewed, specifically related to operational challenges and findings from experimental designs.

What is the implication of the main finding?

• The lack of standardisation in burial and aerial survey conditions impedes the development of effective strategies for locating human remains and clandestine graves.
• More robust experimental designs will contribute to forensic realism of studies, while the integration of artificial intelligence in image processing could lead to the development of more reliable automated detection models, increasing the accuracy of human remains and grave location for application in forensic investigations.

Abstract

Several approaches are currently being used by law enforcement to locate the remains of victims. Yet, traditional methods are invasive and time-consuming. Unmanned Aerial Vehicle (UAV)-based remote sensing has emerged as a potential tool to support the location of human remains and clandestine graves. While offering a non-invasive and low-cost alternative, UAV-based remote sensing needs to be tested and validated for forensic case work. To assess current knowledge, a systematic review of 19 peer-reviewed articles from four databases was conducted, focusing specifically on UAV-based remote sensing for human remains and clandestine grave location. The findings indicate that different sensors (colour, thermal, and multispectral cameras), were tested across a range of burial conditions and models (human and mammalian). While UAVs with imaging sensors can locate graves and decomposition-related anomalies, experimental designs from the reviewed studies lacked robustness in terms of replication and consistency across models. Trends also highlight the potential of automated detection of anomalies over manual inspection, potentially leading to improved predictive modelling. Overall, UAV-based remote sensing shows considerable promise for enhancing the efficiency of human remains and clandestine grave location, but methodological limitations must be addressed to ensure findings are relevant to real-world forensic cases.

1. Introduction

The first 48–72 h after an individual has gone missing is a crucial window for gathering evidence, interviewing witnesses and identifying suspects in order to locate the victim [1,2]. After this window, the chances of finding the victim alive decrease significantly. To hide the crime and the evidence, perpetrators employ a range of methods to dispose of the victims’ remains [3,4]. Locating victims’ bodies, especially in cases where it is suspected that they have suffered serious harm and where it is presumed that this has led to their death, becomes necessary to provide the family with answers related to the circumstances of a victim’s death and for justice to prevail. Several approaches, including geophysics [5,6,7,8], forensic archaeology [9,10,11,12,13], and forensic botany [14,15,16], have attempted to develop frameworks and approaches for the location of human remains and clandestine graves. Similarly, remote sensing has been employed to detect mass graves, cemeteries and tombs through satellite imagery [17,18,19]. Various aerial imaging systems have been mounted on aircrafts (a hyperspectral image sensor CASI/SASI (Itres Research Ltd., Calgary, AB, Canada) mounted onboard a Twin Otter aircraft [20,21], and a HyMap II sensor (Hyvista, North Ryde, NSW, Australia) mounted on a WB57 aircraft [22]) and a helicopter (an infrared imaging system (L-3 Wescam Sonoma Model 12DS200 (L-3 Communications Sonoma EO, Santa Rosa, CA, USA)) [23]) to locate graves. Portable imaging sensors (ASD FieldspecFR spectrometer spectrometer (Analytical Spectral Devices, Boulder, CO, USA) [24], Scott’s Eagle Thermal Imager (Scott Safety, Monroe, NC, USA) [25], and a field spectroradiometer (Field Spec 4 Std Res 350–2500 nm, ASD Inc., Longmont, CO, USA) [26]) have been employed previously to detect human remains and clandestine graves. While these early approaches demonstrate the potential and effectiveness of remote sensing in forensic contexts, they are often expensive, labour-intensive, and logistically complex (considering they have a large footprint and are heavy) [27,28]. Additionally, they require skilled pilots and technicians to operate the sensors and aircraft [29,30]. Portable imaging sensors while smaller, still require additional resource input (additional field team members, site installations and complex field sampling strategies) and are time-consuming to operate [31]. These studies underscore the need for more efficient, low-cost and scalable approaches to aid in forensic investigations. The advancement and miniaturisation of imaging sensors over the last decade have enabled their integration with Unmanned Aerial Vehicles (UAV) platforms, allowing forensic teams to survey large or hazardous areas more efficiently and with fewer personnel [32]. UAVs or unmanned aircraft systems (UAS), commonly referred to as drones, are autonomous or remotely piloted vehicles. UAVs can be pre-programmed to fly a specific flight plan to map and capture images of a specific area as part of an aerial survey. The increasing availability of low-cost and commercially available UAVs has facilitated their integration into a wide range of applications, including agricultural monitoring [33,34], search-and-rescue operations [35,36], archaeological surveys and mapping [37,38] and transportation of medical supplies [39]. As a non-invasive [40,41] approach, UAVs equipped with colour camera, thermal, hyper- and multispectral sensors have garnered attention in forensic and law enforcement applications [42,43]. UAVs equipped with a colour camera capture images in three colour planes, Red, Green and Blue (RGB), ranging from 400–700 nm (Figure 1). The images can be either incorporated into a structure from motion workflow to generate a 3D model to reveal surface variations derived digital surface and terrain modelling [44], or into an orthomosaic workflow to create a large area map at high resolution [45]. Multispectral and hyperspectral imaging, on the other hand, are remote sensing techniques used widely in both environmental monitoring and crop management to detect changes in vegetation health and soil conditions, through the calculation of vegetation indices (Normalised Difference Red Edge Index (NDRE), Normalised Difference Vegetation Index (NDVI), and Visible Atmospherically Resistant Index (VARI) [46,47,48]. Spectral imaging relies on measuring the reflectance of light across multiple wavelengths, including visible and near-infrared bands, to assess plant physiological status and stress, particularly through chlorophyll content [20,46,47]. The difference between these two methods is that multispectral imaging uses fewer and broader wavelength bands (such as Red (620–700 nm), Green (495–570 nm), Blue (450–495 nm), Red-Edge (700–750 nm) and Near-Infrared (NIR, around 750–1100 nm)) [31,49], while hyperspectral imaging uses more, but much narrower bands across the light spectrum (RGB, Red-Edge NIR, and short-wave infrared (SWIR, 1400–3000 nm)) [21,26,50]. UAVs equipped with thermal imaging sensors are also used in remote sensing to detect heat (infrared radiation) emitted from objects on the Earth’s surface, which is translated into visual images that are a representation of temperature differences [51]. Two main types of sensors exist within thermal imaging cameras: radiometric and non-radiometric. Radiometric sensors produce a temperature value at each pixel location within the image, enabling direct temperature readings [52]. Non-radiometric sensors produce a heatmap displaying relative temperature contrasts without any actual measurable temperature data [52]. UAV mounted radiometric sensors are preferred within agriculture [53,54], archaeology [55,56] and forensics [31,50,57,58,59,60] applications and can be uncooled or actively cooled. Radiometric uncooled cameras are often mounted on low-cost UAVs as they are cheaper and easier to integrate than actively cooled cameras. However, they are susceptible to internal and external temperature fluctuations therefore relying on radiometric corrections and calibration [61]. Actively cooled cameras are more expensive but provide better image quality as thermal noise is reduced within the system [62].

The cost-effectiveness of UAVs with imaging cameras compared to more expensive alternatives, such as Light Detection and Ranging (LiDAR) systems, makes UAV with imaging cameras a preferred alternative [64,65]. LiDAR uses repeated intense beams of light (laser) to map topography by measuring the time it takes the laser pulse to reflect from the surface [66]. This produces a dataset made up of densely spaced georeferenced points called a point cloud, which can be used for high-accuracy scaling and 3D representations [66]. UAVs with LiDAR cost between approximately GBP 11,000–GBP 45,000 [67,68], with this increased cost being a key factor limiting its accessibility to police forces. On the other hand, UAV models used in previous studies cost approximately GBP 1000–GBP 6000, depending on sensor payload [63,69,70,71]. Due to their availability and affordability UAVs have been incorporated in crime scene investigations [41,42,72,73], for their utility in locating mass graves [74] and the ability to locate missing persons and clandestine graves [40,75]. Despite the growing interest in the use and application of UAVs for forensic science [76], the current body of literature on UAV-assisted clandestine grave location is fragmented due to inconsistent reporting of key aspects such as sensor types, operational contexts, and methodologies. The lack of synthesis impedes the development of effective strategies for locating human remains and clandestine graves that could aid in providing answers to families and the closure of missing person cases. This paper presents a systematic review of peer-reviewed studies from 2014 to 2024, examining how UAVs have been used in forensic science to answer the question: How have UAVs been used to locate human remains and clandestine graves? By synthesising experimental designs, operational parameters, and sensor technologies, this review identifies prevailing trends and methodological gaps. Lastly, this paper will conclude by offering insights to inform future research and approaches in locating clandestine graves for forensic investigations.

2. Materials and Methods

2.1. Search Strategy

A systematic search was conducted to identify relevant articles, based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [77]. In the PRISMA workflow (Figure 2) journal articles are sourced from pre-identified databases based on the selected keywords. Articles are filtered for relevance through a three-step process including: identification, screening and eligibility. This systematic review protocol was not registered on a public registry. For this systematic review, articles published between January 2014 and December 2024 were sourced. This date range was chosen to explore the use of UAVs in forensic science research over the last decade, which coincides with an expansion of UAV use across private and commercial sectors. During the last decade, the rapid expansion of UAVs has led to their increased inclusion within academic and forensic-related research. Drones have become increasingly useful in archaeological surveys [78], ecological monitoring [79,80], mass casualty events [81] and for police and forensic investigations [42,82]. The use of drones for research and commercial purposes has led to the establishment of several legislative frameworks governing their use. For example, from 2013, the Civil Aviation Authority (CAA) mandated drone operations in the United Kingdom (UK) through the Permission for Commercial Operation, which was replaced by the CAA Operational Authorisation for Specific Category in 2020 [83]. Globally, several other countries have adopted legislation and regulations related to remotely operated vehicles between 2014 and 2024. South Africa gazetted their first regulations in 2015 as the Eighth Amendment of the Civil Aviation Regulations, Part 101: Remotely Piloted Aircraft Systems, under the Civil Aviation Act 13 of 2009 (Civil Aviation Regulations) [84]. The first comprehensive federal regulations permitting routine commercial drone operations (Part 107) took effect in 2016 in the United states of America (USA) [85].

The databases inspected during this systematic review included: Medline (19 May 2025), Copendex (19 May 2025), Scopus (13 May 2025) and Web of Science (12 May 2025). These databases were chosen as they form the main repositories where journal articles related to forensics and remote sensing are published. The search strategy included words and Boolean operators related to the main keywords: “drone”, “grave” and “forensics” (

Table 1. Keywords used in search fields.

Main Keyword	Synonyms and Boolean Operators
Drone	drone* OR “unmanned aerial vehicle*” OR ”UAV” OR ”unmanned aerial system” OR ”UAS” OR thermal OR infrared OR “near-infrared” OR multispectral OR “hyperspectral” OR “LIDAR” OR “Light Detection and Ranging” OR “low altitude" OR “low-cost” OR “non-invasive” OR “remote sensing” OR ”Unmanned Aerial Devices”
Grave	grave* OR burial* OR “forensic grave*” OR “clandestine grave*” OR “clandestine burial*” OR “mass grave*” OR “mass burial*” OR cemeter* OR “human remain*” OR “buried remain*” OR cadaver* OR “unmarked grave*” OR "hidden grave*” OR "hidden burial*”
Forensics	“forensic investigation*” OR “crime scene*” OR ”detect*” OR ”locati*” OR ”search*” OR ”survey”

* Truncation or wildcard operator allowing for the inclusion of multiple word variations in the keyword search.

). These keywords were derived from the research question, with additional terms identified through relevant synonyms associated with the primary keywords found in the literature. In the database searches, the ‘All fields’ category was selected, which searches for relevant articles by locating the specific keywords in the title, abstract, author keywords and main text of articles within each database.

2.2. Screening Process (PRISMA Flow Diagram)

The PRISMA flow diagram (Figure 2) illustrates the screening process [77]. The initial search yielded 15,008 records. After removing duplicates (n = 1242) and articles not available in English (n = 120), 13,646 records were screened for relevance, of which the majority (n = 13,594) were excluded. Full-text articles for the remaining 52 reports were sought for retrieval and assessed for eligibility, resulting in the exclusion of 35 articles considered out of scope for this review. During the screening phase, reference lists were also scanned for additional citations that could be relevant to the main research question but that were missed in the initial search. As such, two additional publications were identified, screened for relevance, retrieved and included in the review. In total, 19 papers met the eligibility criteria and were included in the final systematic review.

2.3. Inclusion and Exclusion Criteria

The search for this review was limited to peer-reviewed studies published in English from 2014 to 2024. Only articles available in full text were included, while dissertations, theses, news articles, reports, grey literature, book chapters and conference proceedings were excluded. Studies were excluded if they (1) did not focus on the use of UAVs for grave identification; (2) used other types of sensors; (3) used other grave detection methods (like geophysical approaches, airborne or helicopter-based sensor identification, or cadaver dogs); (4) were other types of publications; (5) were papers in which results were discussed in previous publications; or (6) were not available in English. Additionally, due to the nature of the search terms, articles dealing with Graves’ disease, gravel or archaeological burial mounds were also excluded. For this systematic review, studies were thus included if they (1) investigated the use of UAVs equipped with imaging sensors for identifying graves and human remains; and (2) were published in peer-reviewed journals between 2014 and 2024.

2.4. Qualitative Data Extraction and Thematic Synthesis

Following full-text screening and inclusion, data from the 19 eligible studies were extracted into a Microsoft Excel sheet (Microsoft 2025), which included study aims, geographic location, sensor type, burial characteristics, and key findings. A thematic analysis was then conducted. This is a qualitative research method used to systematically identify, analyse, and report common patterns or themes within qualitative data, such as interviews or systematic reviews [87]. Its main purpose is to gain deeper insight into a specific research topic by organising information (through the creation of keywords and codes) and describing the data in detail according to the identified themes, revealing potential trends within the data [88]. For this systematic review, initial coding was performed manually in Microsoft Excel and refined iteratively. Keywords were extracted from highlighted statements in the reviewed papers. The keywords were grouped into wider codes that summarised the general idea. Codes were then grouped into broader themes that reflected common approaches, challenges, and findings in the identification of clandestine graves and human remains using UAV-based remote sensing [87]. Any inconsistencies or further clarity on the themes and data extracted were addressed by all authors through discussion during the writing and editing of this paper. Five principal themes were identified: (1) UAV platforms and sensor technologies; (2) operational and practical considerations; (3) identification in relation to burial conditions and environmental constraints; (4) image processing and artificial intelligence (AI) integration; and (5) multidisciplinary approaches. It is noted, however, that the themes identified in this systematic review are based on the limited number of studies currently published focusing on UAV-based remote sensing for clandestine grave location, as identified through the above-mentioned search strategy.

3. Results

3.1. Overview and Characteristics of Studies

Nineteen published articles met the specified inclusion criteria for this systematic review. All of the studies were published from 2018 onward in various journal publications (Figure 3), most likely coinciding with the growing interest in the use of UAVs in forensic science and investigations.

To gain an understanding of each study’s characteristics and identify trends from the experimental designs, information regarding UAV and sensor type, aerial survey method, burial context and conditions, image processing software used, and application of any other detection methods were extracted into Table 2. This information provided the context for each of the studies, highlighting further the strengths and limitations of the various experiments in terms of their robustness and forensic realism.

The studies included were conducted in various countries, including Australia [58,89], Colombia [90,91], England [69,92,93,94], Italy [95], Kuwait [49,57], Malaysia [60], Mexico [31], Northern Ireland [96,97] and the USA [50,59,64]. Due to the sensitive nature of one study reporting on a forensic case, the European country the research was conducted in was not specified [98]. The studies used different types of UAVs, such as multirotor [49,89], quadcopter [31,50,57,58,59,60,64,90,91,92,93,94,96,97,98], hexacopter [69], and octocopter [50], which were equipped with either standard or custom imaging sensors. Two studies used a custom UAV [50,69].

Table 2. Details of the studies included in this systematic review.

Study ID	Study Details	Drone and Sensor	Aerial Survey	Burial Context	Image Processing Software	Other Detection Methods Used
ST1	Blau et al. (2018) [89] Australian Facility for Taphonomic Experimental Research facility Yarramundi, New South Wales, Australia	Lightweight drones, and multirotor and fixed-wing aircraft Multispectral	Time of the flight: - Aerial Survey: - Altitude: 40 m GSD: -	Model: Human donor cadavers Burial conditions: A combination of single and mass graves with human cadavers. Cadavers were also buried with objects like clothes, cell phones, and wallets. Burial depth: 0.3 m, 1 m and 1.4 m	Specialised photogrammetric software	Airborne LIDAR
ST2	Evers and Masters, (2018) [92] Westmill Woodland Burial Ground, Oxfordshire/Wiltshire, England	DJI Phantom 1 model with the Naza-M flight control transmitter GoPro Hero 3 and a Zomei NIR filter	Time of the flight: Middle of the day Aerial Survey: Sequential grid pattern Altitude: 10–40 m GSD: -	Model: None Burial conditions: Single unmarked graves in a natural burial ground Burial depth: -	MATLAB (MathWorks R2017a) Agisoft Photoscan Professional	Foot/ground-based survey
ST3	Murray et al. (2018) [50] Forensic Anthropology Research Facility, Texas State University, San Marcos, Texas, USA	A custom octocopter UAV and tow Phantom 4 Pros (quadcopters) Hyperspectral (Headwall Nano-Hyperspec), thermal (FLIR Systems Vue Pro), RGB (standard camera from Phantom Pro)	Time of the flight: Solar noon (11 a.m.–3 p.m.) Aerial Survey: - Altitude: 200 ft GSD: RGB—2 cm/pixel at 200 ft; Hyperspectral—6 cm/pixel at 200 ft	Model: Human donor cadavers Burial conditions: Four sections: (1) 3 shallow graves; (2) visible cadaver decomposition islands (CDIs) from bodies that had been previously removed; (3) uncaged and scattered skeletal remains; (4) approximately twenty caged remains in various states of decomposition. Burial depth: Surface depositions	– 1	–
ST4	Bodnar et al. (2019) [59] Bowling Green, Ohio, USA	DJI Inspire1 Model T600 UAV Zenmuse XT Longwave Infrared Thermal Camera (FLIR)	Time of the flight: 11:00 a.m. DST Aerial Survey: Not specified Altitude: 10′ (3.04 m), 25′ (7.62 m), 50′ (15.24 m), and 100′ (30.48 m). GSD: -	Model: Pigs Burial conditions: Four burials containing pig carrion and one soil only burial as control. Burial depth: 6″, 12″ and 24″ beneath the surface	FLIR Tools desktop application	–
ST5	Parrot et al. (2019) [69] Land located to the south of Chester, England	Custom DJI F550 Flame-wheel hexacopter A GoPro Hero 4 (Black Edition) Unmodified RGB camera	Time of the flight: Late Afternoon Aerial Survey: Raster pattern Altitude: 2 m, 5 m, 10 m and 20 m GSD: -	Model: Disturbed soil only (empty) burial Burial conditions: Roughly dug burial 1 m2 grave size Burial depth: Shallow	MATLAB (MathWorks R2017b)	–
ST6	Butters et al. (2021) [58] Queensland Police Service Driver Training Facility, Brisbane, Australia	DJI Inspire Zenmuse XT FLIR	Time of the flight: Morning and midday for the first 30 days, then once daily for the following 8 days, and sporadically until the end of the project. Aerial Survey: Images captured from stationary positions directly above the site. Altitude: 4, 8, 16, and 30 m GSD: -	Model: Pigs Burial conditions: Surface and buried; unwrapped, dismembered body parts and whole bodies. Burial depth: Surface, 50 cm and 60 cm	–	–
ST7	Rocke et al. (2021) [97] Northern Ireland *	Mavic Pro drone RGB	Time of the flight: - Aerial Survey: Grid pattern flown autonomously at 60 m spacing. Altitude: 100 m and 150 m GSD: 4.9 cm/pixel at 150 m altitude	Model: None Burial conditions: Simulated grave. Rectangular pit (0.75 m wide and 1.7 m long) containing a buried handbag with woollen clothes inside. Burial depth: 1 m	DroneDeploy [99]	GPR (100 MHz Rough Terrain Antenna, Mala Geoscience, Mala, Sweden)
ST8	Silván-Cárdenas et al. (2021) [31] Site Y: Yautepec, Morelos, Mexico Site M: Milpa Alta, Mexico City, Mexico	DJI Phantom 4 Advance RGB camera DJI Inspire V1 drone Xemus XT thermal camera	Time of the flight: Thermal images captured around dawn and noontime Aerial Survey: Double-scan flights were flown; the camera was oriented forward with a tilt of 70 degrees. Altitude: 50 m GSD: -	Model: Pigs Burial conditions: Y-site: In total, 7 pits of 2 × 2 m were excavated. In total, 10 complete carcasses (83–90 kg each) were deposited as follows: Y2—three pigs, Y4—two pigs, Y6—one pig and Y7— four pigs. Graves Y1, Y3 and Y5 were empty controls. M-Site: In total, 7 pits 2–3 m by 1–1.5 m were excavated with a loader machine. In total, 6 pigs were distributed as follows: M3—two pigs, M5—one pig, M6—one pig, M7—two pigs. Graves M1 and M4 were empty controls, while M2 contained a metal rod and clothes. Burial depth: Y-site: 1.5 m; 1.2 m; 1.1 m; 1 m; M-Site 0.9 m; 1.2 m; 1.3 m; 1.4 m	Pix4dmapper (Pix4D) MATLAB (MathWorks)	Portable field spectroradiometer (Field Spec 4 Std Res. with 350–2500 nm@ 1 nm by ASD Inc.)
ST9	Molina et al. (2022) [90] Antonio Nariño University, USME Campus, south of Bogota, Colombia	Parrot Bluegrass Field (PF726300) quadcopter Parrot Sequoia multispectral sensor and RGB camera	Time of the flight: - Aerial Survey: Automatic flight plan to monitor the 5 m × 10 m site. Altitude: - GSD: -	Model: Pigs Burial conditions: In total, 5 simulated clandestine graves (0.6 × 0.6 m). Clothed and unclothed and dismembered body parts were placed into four graves, one grave remained empty as a control. Burial depth: 0.5 m	Pix4Dfields (Pix4D)	GPR (ProEx model, Mala Geoscience, Mala, Sweden) and electrical resistivity surveys (GeoAmp 303 system, Subsuelo3D S.A.S, Bogotá, Colombia)
ST10	Rocke and Ruffell, (2022) [94] England *	DJI Inspire 2 Sentera 6× multispectral sensor and RGB camera	Time of the flight: - Aerial Survey: Flown at 90° facing directly down. The site was flown in North/South transects at 60 m altitude with 80% front and side overlap. Altitude: Not specified GSD: -	Model: Human burials Burial conditions: Three natural human burial grounds in the UK with interments ranging from 2005 to 2021. Burial depth: 1.8 and 1.4 m	Pix4Dfields (Pix4D)	–
ST11	Spera et al. (2022) [64] East End Cemetery, Richmond, Virginia, USA	DJI Mavic 2 Pro RGB	Time of the flight: - Aerial Survey: Image frontlap of 85%, and sidelap of 75% Altitude: 53.3 m GSD: 1.18 cm/pixel	Model: Human burials Burial conditions: Unmarked human burials in known cemetery (burial period 1980–2002) Burial depth: -	Pix4Dmapper (Pix4D) ArcGIS Pro (ESRI)	-
ST12	Alawadhi et al. (2023) [57] Jahra Pools Nature Reserve, Kuwait	Parrot Anafi thermal (quadcopter) Thermal (FLIR) Lepton 3.5 microbolometer sensor	Time of the flight: - Aerial Survey: - Altitude: 10, 30, and 50 m GSD: -	Model: Sheep carcasses Burial conditions: Two graves were simulated, 5 m apart 5 m × 2 m in size. G1—Empty control burial; G2—contained 8 sheep carcasses, clothing and 9 mm handgun shell casing. Burial depth: 1.5 m	Pix4Dmapper (Pix4D version 4.6.4) FLIR (FLIR Systems Inc., Wilsonville, OR, USA, Version 6.4.18039.1003) ArcMap (ESRI v.10.8.1.14362) ArcGIS Desktop (ESRI v.10.8.1)	–
ST13	Pringle et al. (2023) [93] East Midlands, England	DJI Mavic Pro RGB	Time of the flight: - Aerial Survey: Images had 75% overlap Altitude: 65 m GSD: 10 cm/pixel	Model: None Burial conditions: Forensic case looking for a missing child burial. Burial depth: -	DroneDeploy ArcMap (ESRI ArcGIS v.10.7)	Desktop survey; Metal detector (Compact Metal Detector, CEIA systems, Arezzo, Italy), Bulk ground conductivity survey (CMD Mini-explorer conductivity meter, GF Instruments, Brno, Czech Republic) and GPR (PulseEKKOTM 1000, Sensors & Software, Mississauga, ON, Canada)
ST14	Ruffell et al. (2023) [98] Europe *	DJI Mavic Pro 2 RGB	Time of the flight: - Aerial Survey: - Altitude: - GSD: -	Model: None Burial conditions: Unmarked burial in a park as part of a forensic case Burial depth: -	DroneDeploy	GPR (450 MHz, GuideLineGeo (Mala Geosciences) Solna, Sweden), ground probing, cadaver dogs
ST15	Alawadhi et al. (2024) [49] Jahra Pools Nature Reserve, Kuwait	Parrot Anafi RGB (21 MP Sony IMX230 1/2.422) (multirotor) Parrot Bluegrass Parrot Sequoia multispectral (multirotor)	Time of the flight: - Aerial Survey: - Altitude: 30 m GSD: 10 cm/pixel	Model: Sheep cadavers Burial conditions: G1—single control grave (50–60 cm); G2—shallow grave with a single sheep; G3—single deep (100–150 cm) grave; G4—deep grave with single sheep; G5—deep (150 cm) mass grave; G6—mass grave occupied with eight sheep. Burial depth: 30 cm, 40 cm, 60 cm, 80 cm, 150 cm	Pix4Dcapture (Pix4D version 4.6.4) ENVI (version 5.6, Exelis. Visual Information Solutions, L3 Harris Geospatial, Boulder, CO, USA)	–
ST16	Gaudio and Betto, (2024) [95] Costa d’Agra, Italy	Drone not specified	Time of the flight: - Aerial Survey: - Altitude: - GSD: -	Model: None Burial conditions: Missing WWI soldier unmarked burials Burial depth: -	–	–
ST17	Molina et al. (2024) [91] Site 1: Marengo Agricultural Center, Universidad Nacional de Colombia, Colombia Site 2: Barcelona Experimental Farm, the Universidad de Los Llanos, Colombia	DJI Matrice 300 Micasense Altum-PT multispectral sensor	Time of the flight: - Aerial Survey: Automatic 15 min flight plan, flightpath was at a NW–SE direction. Altitude: 70 m GSD: -	Model: Pig and human skeletal remains Burial conditions: Site 1: Donated human cadavers, pig bodies and forensic objects were buried in eight simulated graves with dimensions of 2 m × 2 m. Four burials were 0.8 m deep, and the other four burials were 1.2 m deep. Site 2: Donated human skeletons, pig bodies and forensic objects were buried in four simulated graves with dimensions of 0.7 m × 1.7 m, 0.5 m deep. Burial depth: 0.5 m, 0.8 m and 1.2 m	Pix4Dmapper (Pix4D version 4.6.4)	GPR (ProEx model 200 MHz and 600 MHz, Mala Geoscience, Mala, Sweden) and electrical resistivity surveys (GeoAmp 303 system, Subsuelo3D S.A.S, Bogotá, Colombia), bulk ground conductivity (CMD Mini-explorer conductivity meter, GF Instruments, Brno, Czech Republic)
ST18	Ruffell and Rocke, (2024) [96] Omagh Town, County Tyrone, Northern Ireland (Case study 2)	DJI Mavic Mini Pro 2 RGB	Time of the flight: - Aerial Survey: Autonomous flight Altitude: Low GSD: -	Model: Human burials Burial conditions: Unmarked human burials in the cemetery Burial depth: -	DroneDeploy	GPR (450-160 MHz, GuideLineGeo (Mala Geosciences) Solna, Sweden)
ST19	Syed Mohd Daud et al. (2024) [60] Universiti Teknologi MARA, Sungai Buloh Campus, Selangor, Malaysia	DJI Matrice 300 RTK Zenmuse H20T (Thermal camera)	Time of the flight: Between 9 a.m. to 11 a.m. Aerial Survey: Images were captured directly at 90 degrees above the rabbit carcasses Altitude: 15 m, 30 m, 60 m, 70 m, 80 m, 90 m and 100 m GSD: 1.333 cm/pixel at 15 m	Model: Rabbit carcasses Burial conditions: 24 rabbit cadavers, clothed and unclothed, were placed on the soil surface in cages; 3 live rabbits in cages were used as controls. Burial depth: Surface deposition	DJI Thermal Analysis Tool 3 software (DJI, Shenzen, China)	–

¹ The hyphen (-) indicates that the information was not specified or included in the study. * Specific town or county not specified due to the sensitive nature of the forensic case (Author personal comment).

Table 2. Details of the studies included in this systematic review.

Study ID	Study Details	Drone and Sensor	Aerial Survey	Burial Context	Image Processing Software	Other Detection Methods Used
ST1	Blau et al. (2018) [89] Australian Facility for Taphonomic Experimental Research facility Yarramundi, New South Wales, Australia	Lightweight drones, and multirotor and fixed-wing aircraft Multispectral	Time of the flight: - Aerial Survey: - Altitude: 40 m GSD: -	Model: Human donor cadavers Burial conditions: A combination of single and mass graves with human cadavers. Cadavers were also buried with objects like clothes, cell phones, and wallets. Burial depth: 0.3 m, 1 m and 1.4 m	Specialised photogrammetric software	Airborne LIDAR
ST2	Evers and Masters, (2018) [92] Westmill Woodland Burial Ground, Oxfordshire/Wiltshire, England	DJI Phantom 1 model with the Naza-M flight control transmitter GoPro Hero 3 and a Zomei NIR filter	Time of the flight: Middle of the day Aerial Survey: Sequential grid pattern Altitude: 10–40 m GSD: -	Model: None Burial conditions: Single unmarked graves in a natural burial ground Burial depth: -	MATLAB (MathWorks R2017a) Agisoft Photoscan Professional	Foot/ground-based survey
ST3	Murray et al. (2018) [50] Forensic Anthropology Research Facility, Texas State University, San Marcos, Texas, USA	A custom octocopter UAV and tow Phantom 4 Pros (quadcopters) Hyperspectral (Headwall Nano-Hyperspec), thermal (FLIR Systems Vue Pro), RGB (standard camera from Phantom Pro)	Time of the flight: Solar noon (11 a.m.–3 p.m.) Aerial Survey: - Altitude: 200 ft GSD: RGB—2 cm/pixel at 200 ft; Hyperspectral—6 cm/pixel at 200 ft	Model: Human donor cadavers Burial conditions: Four sections: (1) 3 shallow graves; (2) visible cadaver decomposition islands (CDIs) from bodies that had been previously removed; (3) uncaged and scattered skeletal remains; (4) approximately twenty caged remains in various states of decomposition. Burial depth: Surface depositions	– 1	–
ST4	Bodnar et al. (2019) [59] Bowling Green, Ohio, USA	DJI Inspire1 Model T600 UAV Zenmuse XT Longwave Infrared Thermal Camera (FLIR)	Time of the flight: 11:00 a.m. DST Aerial Survey: Not specified Altitude: 10′ (3.04 m), 25′ (7.62 m), 50′ (15.24 m), and 100′ (30.48 m). GSD: -	Model: Pigs Burial conditions: Four burials containing pig carrion and one soil only burial as control. Burial depth: 6″, 12″ and 24″ beneath the surface	FLIR Tools desktop application	–
ST5	Parrot et al. (2019) [69] Land located to the south of Chester, England	Custom DJI F550 Flame-wheel hexacopter A GoPro Hero 4 (Black Edition) Unmodified RGB camera	Time of the flight: Late Afternoon Aerial Survey: Raster pattern Altitude: 2 m, 5 m, 10 m and 20 m GSD: -	Model: Disturbed soil only (empty) burial Burial conditions: Roughly dug burial 1 m2 grave size Burial depth: Shallow	MATLAB (MathWorks R2017b)	–
ST6	Butters et al. (2021) [58] Queensland Police Service Driver Training Facility, Brisbane, Australia	DJI Inspire Zenmuse XT FLIR	Time of the flight: Morning and midday for the first 30 days, then once daily for the following 8 days, and sporadically until the end of the project. Aerial Survey: Images captured from stationary positions directly above the site. Altitude: 4, 8, 16, and 30 m GSD: -	Model: Pigs Burial conditions: Surface and buried; unwrapped, dismembered body parts and whole bodies. Burial depth: Surface, 50 cm and 60 cm	–	–
ST7	Rocke et al. (2021) [97] Northern Ireland *	Mavic Pro drone RGB	Time of the flight: - Aerial Survey: Grid pattern flown autonomously at 60 m spacing. Altitude: 100 m and 150 m GSD: 4.9 cm/pixel at 150 m altitude	Model: None Burial conditions: Simulated grave. Rectangular pit (0.75 m wide and 1.7 m long) containing a buried handbag with woollen clothes inside. Burial depth: 1 m	DroneDeploy [99]	GPR (100 MHz Rough Terrain Antenna, Mala Geoscience, Mala, Sweden)
ST8	Silván-Cárdenas et al. (2021) [31] Site Y: Yautepec, Morelos, Mexico Site M: Milpa Alta, Mexico City, Mexico	DJI Phantom 4 Advance RGB camera DJI Inspire V1 drone Xemus XT thermal camera	Time of the flight: Thermal images captured around dawn and noontime Aerial Survey: Double-scan flights were flown; the camera was oriented forward with a tilt of 70 degrees. Altitude: 50 m GSD: -	Model: Pigs Burial conditions: Y-site: In total, 7 pits of 2 × 2 m were excavated. In total, 10 complete carcasses (83–90 kg each) were deposited as follows: Y2—three pigs, Y4—two pigs, Y6—one pig and Y7— four pigs. Graves Y1, Y3 and Y5 were empty controls. M-Site: In total, 7 pits 2–3 m by 1–1.5 m were excavated with a loader machine. In total, 6 pigs were distributed as follows: M3—two pigs, M5—one pig, M6—one pig, M7—two pigs. Graves M1 and M4 were empty controls, while M2 contained a metal rod and clothes. Burial depth: Y-site: 1.5 m; 1.2 m; 1.1 m; 1 m; M-Site 0.9 m; 1.2 m; 1.3 m; 1.4 m	Pix4dmapper (Pix4D) MATLAB (MathWorks)	Portable field spectroradiometer (Field Spec 4 Std Res. with 350–2500 nm@ 1 nm by ASD Inc.)
ST9	Molina et al. (2022) [90] Antonio Nariño University, USME Campus, south of Bogota, Colombia	Parrot Bluegrass Field (PF726300) quadcopter Parrot Sequoia multispectral sensor and RGB camera	Time of the flight: - Aerial Survey: Automatic flight plan to monitor the 5 m × 10 m site. Altitude: - GSD: -	Model: Pigs Burial conditions: In total, 5 simulated clandestine graves (0.6 × 0.6 m). Clothed and unclothed and dismembered body parts were placed into four graves, one grave remained empty as a control. Burial depth: 0.5 m	Pix4Dfields (Pix4D)	GPR (ProEx model, Mala Geoscience, Mala, Sweden) and electrical resistivity surveys (GeoAmp 303 system, Subsuelo3D S.A.S, Bogotá, Colombia)
ST10	Rocke and Ruffell, (2022) [94] England *	DJI Inspire 2 Sentera 6× multispectral sensor and RGB camera	Time of the flight: - Aerial Survey: Flown at 90° facing directly down. The site was flown in North/South transects at 60 m altitude with 80% front and side overlap. Altitude: Not specified GSD: -	Model: Human burials Burial conditions: Three natural human burial grounds in the UK with interments ranging from 2005 to 2021. Burial depth: 1.8 and 1.4 m	Pix4Dfields (Pix4D)	–
ST11	Spera et al. (2022) [64] East End Cemetery, Richmond, Virginia, USA	DJI Mavic 2 Pro RGB	Time of the flight: - Aerial Survey: Image frontlap of 85%, and sidelap of 75% Altitude: 53.3 m GSD: 1.18 cm/pixel	Model: Human burials Burial conditions: Unmarked human burials in known cemetery (burial period 1980–2002) Burial depth: -	Pix4Dmapper (Pix4D) ArcGIS Pro (ESRI)	-
ST12	Alawadhi et al. (2023) [57] Jahra Pools Nature Reserve, Kuwait	Parrot Anafi thermal (quadcopter) Thermal (FLIR) Lepton 3.5 microbolometer sensor	Time of the flight: - Aerial Survey: - Altitude: 10, 30, and 50 m GSD: -	Model: Sheep carcasses Burial conditions: Two graves were simulated, 5 m apart 5 m × 2 m in size. G1—Empty control burial; G2—contained 8 sheep carcasses, clothing and 9 mm handgun shell casing. Burial depth: 1.5 m	Pix4Dmapper (Pix4D version 4.6.4) FLIR (FLIR Systems Inc., Wilsonville, OR, USA, Version 6.4.18039.1003) ArcMap (ESRI v.10.8.1.14362) ArcGIS Desktop (ESRI v.10.8.1)	–
ST13	Pringle et al. (2023) [93] East Midlands, England	DJI Mavic Pro RGB	Time of the flight: - Aerial Survey: Images had 75% overlap Altitude: 65 m GSD: 10 cm/pixel	Model: None Burial conditions: Forensic case looking for a missing child burial. Burial depth: -	DroneDeploy ArcMap (ESRI ArcGIS v.10.7)	Desktop survey; Metal detector (Compact Metal Detector, CEIA systems, Arezzo, Italy), Bulk ground conductivity survey (CMD Mini-explorer conductivity meter, GF Instruments, Brno, Czech Republic) and GPR (PulseEKKOTM 1000, Sensors & Software, Mississauga, ON, Canada)
ST14	Ruffell et al. (2023) [98] Europe *	DJI Mavic Pro 2 RGB	Time of the flight: - Aerial Survey: - Altitude: - GSD: -	Model: None Burial conditions: Unmarked burial in a park as part of a forensic case Burial depth: -	DroneDeploy	GPR (450 MHz, GuideLineGeo (Mala Geosciences) Solna, Sweden), ground probing, cadaver dogs
ST15	Alawadhi et al. (2024) [49] Jahra Pools Nature Reserve, Kuwait	Parrot Anafi RGB (21 MP Sony IMX230 1/2.422) (multirotor) Parrot Bluegrass Parrot Sequoia multispectral (multirotor)	Time of the flight: - Aerial Survey: - Altitude: 30 m GSD: 10 cm/pixel	Model: Sheep cadavers Burial conditions: G1—single control grave (50–60 cm); G2—shallow grave with a single sheep; G3—single deep (100–150 cm) grave; G4—deep grave with single sheep; G5—deep (150 cm) mass grave; G6—mass grave occupied with eight sheep. Burial depth: 30 cm, 40 cm, 60 cm, 80 cm, 150 cm	Pix4Dcapture (Pix4D version 4.6.4) ENVI (version 5.6, Exelis. Visual Information Solutions, L3 Harris Geospatial, Boulder, CO, USA)	–
ST16	Gaudio and Betto, (2024) [95] Costa d’Agra, Italy	Drone not specified	Time of the flight: - Aerial Survey: - Altitude: - GSD: -	Model: None Burial conditions: Missing WWI soldier unmarked burials Burial depth: -	–	–
ST17	Molina et al. (2024) [91] Site 1: Marengo Agricultural Center, Universidad Nacional de Colombia, Colombia Site 2: Barcelona Experimental Farm, the Universidad de Los Llanos, Colombia	DJI Matrice 300 Micasense Altum-PT multispectral sensor	Time of the flight: - Aerial Survey: Automatic 15 min flight plan, flightpath was at a NW–SE direction. Altitude: 70 m GSD: -	Model: Pig and human skeletal remains Burial conditions: Site 1: Donated human cadavers, pig bodies and forensic objects were buried in eight simulated graves with dimensions of 2 m × 2 m. Four burials were 0.8 m deep, and the other four burials were 1.2 m deep. Site 2: Donated human skeletons, pig bodies and forensic objects were buried in four simulated graves with dimensions of 0.7 m × 1.7 m, 0.5 m deep. Burial depth: 0.5 m, 0.8 m and 1.2 m	Pix4Dmapper (Pix4D version 4.6.4)	GPR (ProEx model 200 MHz and 600 MHz, Mala Geoscience, Mala, Sweden) and electrical resistivity surveys (GeoAmp 303 system, Subsuelo3D S.A.S, Bogotá, Colombia), bulk ground conductivity (CMD Mini-explorer conductivity meter, GF Instruments, Brno, Czech Republic)
ST18	Ruffell and Rocke, (2024) [96] Omagh Town, County Tyrone, Northern Ireland (Case study 2)	DJI Mavic Mini Pro 2 RGB	Time of the flight: - Aerial Survey: Autonomous flight Altitude: Low GSD: -	Model: Human burials Burial conditions: Unmarked human burials in the cemetery Burial depth: -	DroneDeploy	GPR (450-160 MHz, GuideLineGeo (Mala Geosciences) Solna, Sweden)
ST19	Syed Mohd Daud et al. (2024) [60] Universiti Teknologi MARA, Sungai Buloh Campus, Selangor, Malaysia	DJI Matrice 300 RTK Zenmuse H20T (Thermal camera)	Time of the flight: Between 9 a.m. to 11 a.m. Aerial Survey: Images were captured directly at 90 degrees above the rabbit carcasses Altitude: 15 m, 30 m, 60 m, 70 m, 80 m, 90 m and 100 m GSD: 1.333 cm/pixel at 15 m	Model: Rabbit carcasses Burial conditions: 24 rabbit cadavers, clothed and unclothed, were placed on the soil surface in cages; 3 live rabbits in cages were used as controls. Burial depth: Surface deposition	DJI Thermal Analysis Tool 3 software (DJI, Shenzen, China)	–

¹ The hyphen (-) indicates that the information was not specified or included in the study. * Specific town or county not specified due to the sensitive nature of the forensic case (Author personal comment).

Different imaging sensors (such as a colour camera (RGB), hyperspectral, multispectral, thermal, and Near-Infrared (NIR)) were used across the studies (Table 2), with the majority capturing RGB images. Depending on the specific study aim, visible (RGB) [31,49,50,64,89,90,92,97,98], thermal [31,50,57,58,59,60], NIR [92], hyperspectral [50] or multispectral [49,64,89,90,91,94] imaging was used to identify changes related to burial activity or the decomposition process.

In these papers, the primary applications of UAVs were: (1) mapping changes in the topography of the landscape (through characterisation of burial-related surface variations) due to burial activity or the decomposition process [64,89,96]; (2) mapping of boundaries or edges of features to detect grave-shaped anomalies [91,92]; (3) mapping of visual soil colour and vegetation changes related to disturbance from burial activity or decomposition process [69]; (4) mapping of spectral signatures of soil and vegetation, indication of disturbance from burial activity or decomposition process [49,90,91,94,97]; (5) mapping of heat signatures related to microbial thermogenesis and/or insect activity during the decomposition process [57,58,59,60]; (6) mapping disturbed ground and subsurface voids through differential thermography [31]; (7) mapping variations in topsoil moisture content [57]; and (8) mapping of the survey area for the creation of an orthomosaic map used in other analyses [93,95,96,98].

The experimental designs across the papers varied in burial conditions, burial environment and the models which were used as proxies for human remains. Within the burial conditions, burial depth varied across the studies. Eleven studies reported burial depths between 0.15 m to 1.8 m below the surface (Figure 4) [31,49,57,58,59,60,89,90,91,94,97]. One study reported the depth as shallow [69], while 6 studies did not specify burial depth at all [64,92,93,95,96,98]. Although not constituting burial depth per se, surface depositions were nonetheless recorded and included for 4 studies using pig [58,59], rabbit [60] and human [50] remains. In two cases, the surface deposition was studied in conjunction with buried remains [58,59]. The remaining two studies [50,60] focused solely on surface depositions. While the deposition of bodies on the surface does have a different rate and progression of decomposition compared to buried bodies [100], surface-deposited remains do alter the surrounding environment. This leaves a distinctive physical (soil colour changes and decaying vegetation), microbial and biochemical signature that is present in the subsurface below the remains, which can be detected by geophysical approaches [101] as well as thermally (heat signatures) and spectrally (vegetation indices) with UAVs [50]. Understanding these above-ground signatures and anomalies provides useful information for forensic cases where perpetrators haphazardly conceal remains in the subsurface as fast as possible [102,103,104,105], where remains are partially covered [106] or in cases where remains have been translocated [107]. The diverse range of burial depths is similar to that reported previously by Mannheim [108] in a retrospective study of 87 forensic cases from the USA [31,49,50,57,58,59,60,64,69,89,90,91,92,93,94,95,96,97,109,110,111].

Diverse models were used (Figure 5) in the studies reviewed, including human donor cadavers [50,89] or mammalian remains (pigs [31,58,59,90,91], sheep [49,57] and rabbits [60]) as proxies for human remains. One study also incorporated human skeletal remains instead of whole bodies, as well as beheaded and burnt human skeletal remains in shallow graves to simulate forensic scenarios commonly encountered in Colombia [91]. In the absence of human donor cadavers or mammalian models, human burials in cemeteries [64,92,94,96] and forensic cases [93,95,96], soil-only burials (pits containing disturbed earth with no addition of remains) [69], and a simulated grave with buried objects [97] also formed part of the experimental designs. Mammalian models are used as proxies for human remains within forensic taphonomy research, where the focus is on studying the natural and cultural processes that affect the body from death until discovery [109,110]. The need for the use of mammalian proxies is especially true in cases where the use of human donor cadavers for forensic taphonomy-based research is prohibited due to cultural sensitivities or legal and ethical guidelines [110,111]. Within these experimental designs, pigs are often the preferred model due to the similarity in hair-to-skin ratio and bacteria found in the intestinal tract [111]. Burial conditions in the studies incorporated several unique elements as part of the experimental designs, such as clothed proxies [31,89,91], dismemberment of the bodies [58], covering of pig cadavers with lime, a concealment method commonly used in Mexico [31], and the burial of other objects, such as a 9 mm handgun shell casing [57], a metal rod [31], wallets [89], cell phones [89], and a handbag [97] to simulate forensic scenarios.

3.2. Thematic Synthesis Findings

A thematic synthesis was undertaken to identify recurring themes across all papers, providing a more nuanced understanding of UAVs as a tool for human remains and grave location. Five principal themes were identified: (1) UAV platforms and sensor technologies; (2) operational and practical considerations; (3) identification in relation to burial conditions and environmental constraints; (4) image processing and AI integration; and (5) multidisciplinary approaches.

3.2.1. UAV Platforms and Sensor Technologies

The reviewed studies demonstrated a growing reliance on consumer-grade UAVs equipped with imaging sensors for non-invasive grave location (

Table 3. The various UAV types and image sensor payloads (grey highlights) which are reported in the studies reviewed.

Study	UAV	RGB	MS *	HS *	Thermal	NIR	Specific Sensor Type
ST1	Lightweight drone [89]						-
ST2	DJI Phantom 1 [92]						GoPro Hero 3 and a Zomei NIR filter
ST3	Custom octocopter UAV [50]						Hyperspectral (Headwall Nano-Hyperspec), Thermal (FLIR Systems Vue Pro), and RGB (standard camera from Phantom Pro)
	Phantom 4 Pros [50]
ST4	DJI Inspire1 Model T600 [59]						Zenmuse XT Longwave Infrared Thermal camera (FLIR)
ST5	DJI F550 Flame-wheel [69]						A GoPro Hero 4 (Black Edition) and Unmodified RGB camera
ST6	DJI Inspire [58]						Zenmuse XT FLIR
ST7	Mavic Pro drone [97]						-
ST8	DJI Phantom 4 Advance [31]
	DJI Inspire V1						Xemus XT Thermal camera
ST9	Parrot Bluegrass Field (PF726300) [90]						Parrot Sequoia Multispectral sensor and RGB camera
ST10	DJI Inspire 2 [94]						Sentera 6× Multispectral sensor and RGB camera
ST11	DJI Mavic 2 Pro [64]						-
ST12	Parrot Anafi [57]						Thermal (FLIR) Lepton 3.5 microbolometer sensor
ST13	DJI Mavic Pro [93]						-
ST14	DJI Mavic Pro 2 [98]						-
ST15	Parrot Anafi [49]						RGB (21 MP Sony IMX230 1/2.422)
	Parrot Bluegrass						Parrot Sequoia Multispectral sensor
ST16	Not specified [95]						-
ST17	DJI Matrice 300 [91]						Micasense Altum-PT Multispectral sensor
ST18	DJI Mavic Mini Pro 2 [96]						-
ST19	DJI Matrice 300 RTK [60]						Zenmuse H20T (Thermal camera)

The hyphen (-) indicates that the information was not specified or included in the study. * MS stands for multispectral, and HS stands for hyperspectral.

). Within the consumer-grade UAV space, both DJI (De-Jiang Innovations, China) and Parrot (France) are market leaders, with their drones popular not only with hobbyists but also with researchers and law enforcement agencies [112,113,114,115,116,117,118]. This is mostly because these UAVs are widely available and accessible, and incorporate a suite of different imaging sensors. In the studies reviewed for this paper, DJI UAV platforms were the most commonly used (n = 10) [31,50,58,59,60,64,69,91,92,93,94,96,97,98], followed by Parrot UAVs (n = 3) [49,57,90]. Some studies did not specify the type of UAV platform used [89,95], which limits reproducibility and comparative analysis for grave location.

In general, four different imaging sensors were commonly used in human remains and clandestine grave location. RGB imaging was used primarily to detect changes in vegetation patterns, soil colour and surface disturbances, the edges or boundaries of graves, and soil mounds around areas where graves were dug [69,95,96,97,98]. Surface variations form as the body decomposes, causing soil to fall in and compact around the abdominal cavity as it caves in, forming a shallow depression on the surface of the grave [119]. These changes can be visually recorded using RGB. RGB images formed part of photogrammetric workflows to survey sites and generate orthomosaic maps [31,49,64,90,91,93,94,95,96,97,98], which in some instances were used for further geospatial analysis, including digital elevation models [89], greyscale 3D models [92], digital surface [49,64,93] and terrain models [31,49]. In one study, and although not statistically validated, it was found that while RGB imaging provides clearer visual indicators of grave-related anomalies than greyscale NIR images, neither of the imaging methods was consistent in locating graves on the site [92]. Vegetation cover had a big influence on identifying the locations of the graves, as vegetation over older graves (From Area 1 and Area 3) concealed them in both RGB and NIR imaging. In some cases, RGB outperformed NIR in locating graves, especially in instances where RGB was able to capture differences in soil colour and highlight grave shape in areas of dense vegetation. NIR was more successful in locating graves in areas with less dense vegetation because it could detect the reflectance differences between disturbed and undisturbed areas from the vegetation and soil profiles [92]. The variation in vegetation and age of graves highlights that there is no one universal method that worked across the site. Based on this, the successful location of graves is influenced by several variables, which necessitate that methods be tested in a range of different burial scenarios, allowing practitioners to draw on multiple methods.

Environmental changes associated with clandestine graves, such as disturbances in soil and changes in vegetation, have been detected using multispectral [49,64,89,90,91,94] and hyperspectral [50] imaging. When a grave is dug, the natural soil layers are disturbed. The subsequent burial of a body and the decomposition of organic material alter the soil chemistry [120]. This, in turn, affects the overlying and surrounding vegetation, often leading to changes in chlorophyll content observed through plant health or stress [47,121]. This is also due to the concentration of volatile organic compounds and body fluids that are leaching into the soil [14,122]. Based on this, multispectral and hyperspectral imaging sensors can be used to identify burial-related anomalies from vegetation growth and soil reflectance patterns through the calculation of vegetation indices, which can be used to quantify vegetation health, aiding in identifying areas of plant stress most likely related to subsurface disturbances [49,94].

UAV-based thermal imaging was noted for its ability to identify anomalous heat signatures within a landscape, in which the thermal signatures deviated from the expected ambient temperature. These thermal anomalies were associated with heat generated from insect activity [58,60], the decay process [50,57,58,60], heat trapped by wrapped remains [58] and from the disturbed burial soil [59]. In one study, thermal anomalies for unrelated graves were the result of the regeneration of vegetation that was stimulated by the burial activity, the burial of pig carcasses, and a plastic suit that trapped water, which increased the pit’s heat capacity and thermal conductivity [123], resulting in a more stable soil temperature [31]. The persistence of the thermal signature from dismembered and wrapped bodies could also be used to locate bodies deposited on the surface and buried in shallow graves (50 cm) after the first few days post-burial [58]. In the same study [58], the authors also tested the reliability of thermal imaging to detect heat signatures from buried remains (a 40 kg pig) at 50–60 cm and noted that no heat signature was detected during the study period. Excavation of the grave revealed that the wet conditions of the burial had led to adipocere formation, effectively stopping the decay [58]. The formation of adipocere or “grave wax” [124] on the buried body effectively halted the decay process. This, by extension, also stopped the generation of decomposition-related heat signatures that could potentially be detected with thermal imaging, especially in the early post-burial period. This indicates that the persistence and detectability of the thermal signature can be correlated with the burial environment and soil conditions. Thermal imagery was also able to distinguish between experimental graves (containing sheep carrion) and soil-only burials (no-sheep controls with disturbed earth only) from an arid environment in Kuwait [57]. In the same study [57], the temperature of the experimental graves was higher than that of the natural background environment and the control graves. The higher temperatures were attributed to the decomposition processes, which led to a thermal signature that was distinct from the rest of the site [57]. The temporal sensitivity and ephemeral nature of thermal signatures remain, however, a limitation, particularly in varying environmental conditions and diminishing maggot masses as the remains are consumed and decay.

3.2.2. Operational and Practical Considerations

Operational feasibility and cost-effectiveness are central considerations for the adoption of UAV-based grave location in forensic and law enforcement contexts. Across the reviewed studies, factors such as UAV specifications, sensor payloads, flight altitude, and environmental conditions were shown to significantly influence data quality and detection outcomes. Low-cost UAV platforms were evaluated frequently for their accessibility, but often at the expense of image quality. One study [92] noted that a GoPro Hero 3 camera fixed to a consumer-grade drone produced images with noise and blur. Several images were excluded from the final dataset due to motion blur distortion that resulted from camera movement during the aerial survey [92]. This highlights a trade-off between affordability (using commercially available cameras and consumer UAVs) and the quality and reliability of the data that can be captured. On the other hand, another study from England demonstrated that an unmodified RGB camera mounted on a UAV capturing video footage is a reliable tool for grave location [69]. The study consisted of an aerial survey in a raster pattern over the identified search area with a single site of disturbed soil, simulating a clandestine grave. During the survey, the gimbal was positioned parallel to the ground, capturing video footage at 25 frames per second [69]. Leveraging a computer vision algorithm, the video footage was processed in MATLAB (MathWorks) where single frames were extracted and individually assessed. Each frame was compared against a still reference image of the disturbed site to automatically detect the disturbed soil before moving to the next frame. The algorithm was successful in detecting disturbed soil, in the various frames at different altitudes (2, 5, 10 and 20 m) and in randomly selected perspectives from the dataset [69]. While there is a trade-off between image quality and data reliability, there is potential for low-cost UAVs and cameras paired with appropriate processing algorithms. They could offer an alternative solution for grave location in light of operational time and resource limitations faced by law enforcement. However, the study was preliminary [69]. Further validation is needed under varied environmental and burial conditions. Sensor payloads were also a key consideration.

The type of analysis and post-processing pipeline followed depended on the sensor payload used during the aerial survey, as such studies either analysed thermal data through thermograms [31,50,57,58,59,60] or multispectral data through various vegetation indices [49,89,90,91,94]. None of the reviewed studies provided any justification for the type of image sensors used in the various experimental designs, suggesting a gap in methodology that may stem from practical constraints instead of scientific rationale for experimental designs. The use of specific RGB, multispectral and thermal sensors for human remains and clandestine grave location likely reflects factors such as (1) previous forensic taphonomic studies have validated the use of thermal and multispectral sensors mounted on aircrafts [21,125], encouraging their validation on UAVs as a low-altitude, low-cost alternative; (2) due to budget constraints, academic teams rely on equipment that is pre-existing within the institution or to that they can afford to purchase additionally [126]; and, (3) the type of analysis is determined by the drone platform used and its compatible sensor [127]. Within this systematic review, teams in three cases used multiple UAVs to capture different image datasets within a single project, namely thermal, hyperspectral and RGB [50], RGB and thermal data [31], and multispectral and thermal data [49,57]. As such, a current limitation is that the heterogeneity in UAV platforms and sensor payloads employed across the studies and the lack of data integration between the payloads, which hinders data comparability and the development of standardised predictive models for clandestine grave location.

The required spatial resolution (Ground Sample Distance, GSD) and the area that needs to be covered (Field of View, FOV) are influenced directly by the choice of the altitude that will be flown during an aerial survey (

Table 4. The reported GSD along with the specific drone, image sensor and survey altitude.

Drone and Image Sensor	cm/per Pixel	Altitude (m)
DJI Phantom 4 Pro with RGB [50]	2	60.06 *
Hyperspectral (Headwall Nano-Hyperspec) ** [50]	6	60.06 *
Mavic Pro with RGB [97]	4.9	150
DJI Mavic 2 Prowith RGB [64]	1.18	53.3
DJI Mavic Pro with RGB [93]	10	65
Parrot Anafi RGB (21 MP Sony IMX230 1/2.422) [49]	10	30
Parrot Bluegrass Parrot Sequoia multispectral [49]	10	30
DJI Matrice 300 RTK Zenmuse H20T (Thermal camera) [60]	1.33	15

* Original paper reported altitude as 200feet. ** Not specified on what type of drone this sensor was mounted on.

). From the reviewed studies, it appears that depending on the altitude, UAV type and payload, the optimal GSD for human remains and clandestine grave location is a GSD below 10 cm/pixel [49,50,60,64,93,97], as this allows for good resolution within the images. Within forensic cases, images with a high resolution that indicates burial-related anomalies clearly and in detail are important to distinguish the potential location of clandestine graves or human remains from the natural background environment [119]. Imaging sensors need to show and distinguish potential burial areas clearly from the natural background environment for police to locate the remains of victims and clandestine graves. However, while a lower altitude would provide the best spatial resolution, the decision for the optimal altitude is often motivated by the survey time available, UAV battery life, sensor capabilities, size of the search area, the expenses associated with conducting the aerial survey, and available post-processing time [61,128]. Five of the nineteen studies considered within this review explicitly reported GSD values, limiting the ability to compare results across studies [50,57,60,93]. Notably, five studies did not report either flight altitude or GSD, highlighting a lack of standardisation in reporting key operational and aerial survey parameters [90,94,95,96,98]. Without this information, it becomes increasingly difficult to validate findings across diverse biogeographic regions, test sensor accuracy, and develop frameworks for human remains and clandestine grave location, limiting the impact and usefulness of UAVs with imaging sensors in forensic investigations.

No universally optimal altitude was identified across the reviewed studies (Figure 6). Due to the diverse range of low-cost UAVs currently available on the market, the different methods of postmortem body treatment and body disposal strategies employed by perpetrators, as well as the wide range of biogeographic conditions, an optimal flight altitude will most likely not be determined. This is not to say that bigger payloads and more expensive UAVs might be able to identify graves at a similar optimal altitude. However, within the current financial resources of law enforcement, low-cost UAVs will need to take ground conditions and the context of the burials being investigated into account when determining the altitude. In this review, lower altitudes were generally preferred for thermal surveys due to improved spatial resolution and thermal contrast between burial and natural, undisturbed areas. Thermal imaging was employed in 6 studies [31,50,57,58,59,60], which consisted of eight different flights, with six of these surveys operating at altitudes between 3 m and 40 m (Figure 6). For example, it was found that thermal imaging at 4 m above ground provided the highest contrast between body fragment heat signatures and the natural environmental background [58]. Although detection remained adequate at 8 m, higher altitudes introduced image distortion such as fisheye effects and reduced resolution [58]. Testing thermal data quality at different altitudes (10 m, 30 m, and 50 m), it was reported that an altitude of 30 m maintained the clearest GSD [57]. RGB imaging was tested across a broader range of altitudes (0–150 m), reflecting its flexibility in the visual spectrum for grave identification [31,49,50,64,89,90,92,97,98]. In contrast, multispectral imaging was evaluated at fewer altitudes (30 m, 40 m, and 70 m) [49,89,91]. This potentially indicates that the use of UAVs equipped with multispectral imaging for grave location still requires further field testing at different altitudes, for different burial depths and burial conditions. Only one study used hyperspectral imaging at 60 m, and while another study used NIR at 20–40 m [50]. RGB video footage was captured in one study at various altitudes (2 m, 5 m, 10 m, and 20 m) [69].

Based on the studies discussed, the optimal altitude is case-specific, and dependent on several variables such as the UAV type and sensor payload, environmental conditions, burial context, search area and the pilot’s discretion. The available UAV and sensor payload are variables that will determine the survey strategy, which speaks to the overall purpose of the aerial survey and the specific methods and image sensors that will be used in the mission, and it will also include any legal and safety considerations that need to be taken into account [129,130]. The survey strategy will dictate the required survey parameters implemented to achieve the specified survey strategy’s objectives. This will include deciding on the optimal altitude or required GSD, the camera settings and the flight speed [128,131]. The survey parameters, in turn, are also dependent on the size of the search area and the specifics of the forensic case being investigated, which is why the survey parameters are more site-specific. Both the survey strategy and survey parameters form part of the flight preparation and planning stage [129]. Before a mission, these parameters need to be considered carefully to ensure that the image data captured during the mission meets the objectives of the survey strategy (Figure 7). The specific camera settings will also be dependent on the imaging sensor type. The survey strategy, on the other hand, is more universal in application to different sites, as similar overall aims and methods can be implemented in relation to the type of UAV and sensor payload available. For instance, to survey large areas of interest with the purpose of narrowing down priority areas (objective), the pilot might opt to fly at a higher altitude to cover a wider FOV, with a quicker shutter speed and adjusted side and frontal overlap (survey parameter). The need to cover a larger expanse of land with a wider FOV (survey strategy) is also informed by the type of drone and the resolution of the sensor payload. In cases where smaller, more manageable areas are surveyed (objective), the pilot might opt for a lower altitude and slower speed (survey parameter), increasing the resolution of the images captured with the UAV (survey strategy) of the area of interest. It should be noted that the final flight parameters and, by extension, the flight mission will, as such, be decided at the discretion of the pilot based on their knowledge and experience, which will include an assessment of the current terrain and climatic conditions [130].

3.2.3. Location of Graves in Relation to Burial Conditions and Environmental Constraints

Environmental variables, such as vegetation cover were identified as critical factors influencing detection efficacy and affecting the ability of UAVs to capture images [64,92]. Vegetation affects the visibility of surface features and the spectral signatures captured by UAV imaging sensors. This is due to vegetation such as tall grass and tree cover, which can obscure burial features [64,119]. In this review, all the studies were conducted in terrestrial environments. Some of the studies were reports on case work [93,95,97,98], or studies conducted in controlled environments like dedicated facilities for taphonomy research, demarcated research sites or university grounds [49,50,57,58,59,60,69,75,89,90,91], while others used unmarked natural burial grounds [92,94,96] or cemeteries [63]. In some of the controlled environments, vegetation is monitored and managed, which limits the variability and forensic realism of detection scenarios in experimental designs [110]. Forensic taphonomy facilities are often located in rural areas outside of busy cities. Due to the sensitivity of the research, maintaining the research integrity, and to prevent unwelcome intruders or scavengers the centres are surrounded by walls or fences [110]. While adopted for practical reasons, this setup could in some cases hinder the natural patterns of vegetation as sites are prepared for experiments, for example, where grass is cut or vegetation is cleared [91,132]. Similarly, natural burial sites are maintained by site management to ensure grave accessibility to family members, even when surface features have faded, which further limits the development of natural vegetation [133]. For remains that are buried in natural burial grounds, the surface features of the graves transition from bare soil to full vegetative growth over time, blending into the surrounding landscape and obscuring any visual surface markers. One study [92] identified 55 of the 138 individual burials in a natural burial ground in Oxfordshire/Wiltshire, UK, between 2000 and 2017 through visual identification based on surface features. However, the work struggled to locate graves in the same areas using NIR aerial images [92]. Of the six burial areas surveyed in the study [92], graves could only be identified in four areas using NIR (n = 9) and a Sobel edge detection filter (n = 11). It was found that regenerative vegetation and grave soil in extended post-burial periods obscured the burial pits, making identification using NIR of older graves difficult [92]. The NIR was, however, able to identify recent graves where vegetation had not yet regenerated [92]. In East End Cemetery in Richmond, Virginia, USA, unmanaged vegetation growth had overtaken the site, leading to the obscuring of grave markers and the loss of known grave positions [64]. Dense vegetation significantly hindered visual and remote sensing-based grave location for this site. Deeper graves (of 1.5 m) and certain soil types, such as moist clay or desert sand, can slow decomposition and nutrient release [102], reducing the visibility of vegetation stress indicators [49]. In such anaerobic conditions, adipocere formation may occur, halting microbial activity and thereby decreasing the heat generation detectable by thermal sensors [58,124,134]. Apart from environmental conditions, which had the biggest impact on the successful location of graves, other burial-specific conditions also played a significant role in grave location outcomes. These included the size of the body [93], the presence of wrappings [58], and the time elapsed since burial [49,57,91]. In the case of a neonatal burial [93], grave size is correlated with body size. Apart from the fact that the neonatal remains will decompose more quickly than adult remains, the smaller size of the grave also makes it more difficult to detect. Concealing remains, by wrapping the remains in carpets, cloth, and clothes, slows down the decomposition. The wrapping acts as a barrier which prohibits insect activity [135]. Additionally, it reduces access to oxygen, creating a more anaerobic environment, which slows down decomposition [136,137,138]. Finally, the longer a body is buried, the more difficult it becomes to locate it, as vegetation can regrow [64] and redevelopment of the site [98,139] can obscure surface features.

Given the difficulty in locating burials in these environmental and burial conditions, a common solution appears to draw from geospatial analysis to aid in grave detection. Hydrological and elevation analyses were employed to detect surface variation (such as positive or negative topography) [31,49,64,89,96]. These methods were, however, sensitive to environmental noise and natural variation in topography and required high-resolution data. This highlights the importance of integrating environmental and forensic taphonomic knowledge when interpreting UAV-derived data for grave location. These studies underscore the importance of testing UAV identification methods in diverse, real-world settings and in different environmental conditions.

Tools developed in other disciplines can be implemented for grave location. For example, vegetation indices, which are widely used in ecological monitoring, are a tool that maps vegetation health and plant stress in response to various environmental conditions [140]. While visible vegetation changes alone may not reliably indicate the presence of graves, multispectral and hyperspectral imaging can enhance detection through vegetation indices. In forensic applications, these indices can reveal anomalies in vegetation caused by buried objects or disturbed soil, which affect root systems and moisture availability, contributing to plant stress or plant growth. Several indices were applied in the reviewed studies, including the Green Normalised Difference Vegetation Index (GNDVI), NDRE, NDVI, and VARI [49,90,94]. These tools were particularly effective during dry periods, when vegetation stress due to environmental conditions allowed for burial-related anomalies to become more pronounced and easier to identify visually and spectrally [97]. Comparing graves from three natural burial ground in the UK: Site 1 (burials dating 2011–2021); Site 2 (burials dating 2014–2021), and Site 3 (burials dating 2005–2021), Rocke and Ruffell, [94] found that using multispectral imaging and specifically a vegetation index or digital terrain model, recent graves (2021–2019) are easily detected. At all three sites, the burial depth of graves was constant at 1.4 m, with the only exception being earlier graves (early 2006) at Site 1 that were 1.8 m deep. The authors did, however, find that older graves are difficult to detect using the same post-processing tools (burials post-2016/2017) [94].

Climatic and environmental conditions can affect a UAV’s ability to locate graves or human remains. One study [59] found that rainfall, for instance, had an impact on the thermal camera’s ability to detect graves. Puddles formed as a result of rain obscured surface anomalies, complicating the visual identification of grave locations [59]. It was suggested that the puddles above the burial pits affected the soil’s thermal inertia. Specifically, the moisture most likely led to a reduced temperature contrast between the graves and the background environment, making them harder to detect with a thermal camera [59]. While climatic conditions can affect the ability of imaging sensors to detect graves and human remains, adverse weather conditions like high winds, heavy rain or extreme heat also mean some UAVs cannot be operated [57,69].

3.2.4. Image Processing and AI Integration

The image post-processing pipeline varied across studies, likely attributed to differences in the resources, expertise, and software applications available to each research team at the time of study. Additionally, the choice of post-processing tools was determined by the specific analytical objectives and research questions of each study, such as thermal anomaly detection [57,58,59,60], or spectral classification of burial-related anomalies [49,90,91,94,97].

In UAVs equipped with image sensors, the images captured by low-cost drones were processed successfully using a range of software tools, including DroneDeploy [96,97], Pix4D software [31,64,90,94] (version 4.6.4 [49,57,91]), Agisoft Photoscan Professional [92], ArcMap (ESRI ArcGIS v.10.8.1.14362 [57], ESRI ArcGIS v.10.7 [93]), ArcGIS (ESRI) [64], ArcGIS Desktop (ESRI, v.10.8.1) [58], ENVI [49], FLIR Desktop application (version 6.4.18039.1003 [57,59]), DJI Thermal Analysis Tool 3 (DJI, Shenzen, China) [60] and Mathworks MATLAB ([31], R2017b [68], R2017a [92]). DroneDeploy is specialised software designed specifically for UAVs. It is an internet-based platform that is used to plan aerial missions, manage flight routes and set flight parameters [99,141]. Pix4D and Agisoft Photoscan Professional are photogrammetry software. Within the reviewed studies, the majority incorporated Pix4D (including Pix4Dfields and Pix4Dmapper) into the photogrammetry workflows, forming the first step in creating accurate orthomosaic maps by stitching aerial photographs together [44,142]. While similar processing can be undertaken with Agisoft Photoscan Professional, it was mainly used in one study [92] to render a 3D model of a survey area. ArcMap (ESRI ArcGIS v.10.7 [93], ArcGIS (ESRI) [58,64,93]), ENVI (version 5.6) [49] are geospatial software that can be used to enhance and process images through a variety of tools. The ArcPro’s Hydrology Toolset within ArcGIS Pro (ESRI) was used to distinguish grave depressions from the natural undisturbed environment [64], while the Spatial Analyst Tools in ArcMap was used to explore moisture and temperature variations [57]. Orthomosaics can also be exported to ENVI (Exelis. Visual Information Solutions, L3 Harris Geospatial, Boulder, Colorado, USA) for image processing and analysis [143]. ENVI (Exelis. Visual Information Solutions, L3 Harris Geospatial, Boulder, CO, USA) provides access to a range of manual and automatic tools for advanced visualisation, classification, and anomaly detection like the RX Anomaly Detection Tool [49,144]. Visualisation and variation in temperature was conducted using the FLIR Desktop application [57,59] and DJI Thermal Analysis Tool 3 (DJI, Shenzen, China) [60]. Image processing and analysis were performed with MATLAB (MathWorks), a mathematical computing software [145], through tools such as the edge function that highlights areas of rapid intensity change [92], the creation of terrain models [31] and the classification, labelling and detection of objects or features through the computer vision toolboxes [69]. Apart from the DJI Thermal analysis tool 3, all of the used software and programmes are subscription-based. While the DJI Thermal Analysis Tool 3 is free, its drawback is that it only supports DJI thermal camera payloads. Although the strength of these platforms is their reliable and robust image processing and geospatial toolsets, they are limited by their inaccessibility to non-specialists, requiring complex skill sets and knowledge of some of their processing tools.

In many of the studies (n = 14), grave location relied on manual inspection of images [31,57,58,59,60,64,89,90,91,93,94,95,96,97,98]. For these studies, graves were located through visual observation and interpretation of images by the team based on the context of the study, expert knowledge and manual comparison of the images to datasets from the same site (Figure 8). One study [50] reported on the use of Structure-from-Motion (SfM) photogrammetry, which was employed to create 3D models of the site, enabling the visual identification of elevation changes and unusual surface anomalies. This approach offers a cost-effective method for generating 3D models from 2D imagery but relies on the user to interpret 3D images to identify any anomalies that could be burial-related [50]. However, the effectiveness of SfM and the resolution of the images are dependent on altitude, image overlap, sensor specifications, ground control accuracy, and environmental conditions [50].

Automated detection algorithms (Figure 8), where advanced models and algorithms are employed to aid in the identification of anomalies that could be graves, based on shape, size, elevation, colour and vegetation cover, were used in five studies [49,50,64,69,92]. Evers and Masters [92] employed a Sobel edge-detection algorithm, an image processing method, on converted greyscale images taken of burials in the Westmill Woodland Burial Ground, Oxford, UK. Through the edge detection function, the intensity of pixel values is examined. Abrupt changes in the pixel values are emphasised by the algorithm, highlighting edges of objects and features. Boundaries and edges are outlined but require visual identification by the user to interpret edges as graves [92]. Murray et al. [50] used an untrained deep learning algorithm YOLO [146], to detect postmortem bodies in thermal images at the Texas State University Forensic Anthropology Research Facility. A noted limitation of the YOLO algorithm was that while it could successfully detect bodies in the early postmortem stages, it struggled to detect bodies in advanced stages of decomposition, often misclassifying anomalies [50]. Spera et al. [64] leveraged an automated hydrological flow model to identify grave-shaped depressions, which acted as hydrological sinks (closed depressions where water can accumulate but it does not have an outlet) across the East End Cemetery in Richmond, USA. While the model successfully identified several depressions, manual inspection of each site was still needed to ensure depressions fell within the parameters for burials. The authors noted that tree cover and leaf canopies across the cemetery prevented effective mapping of burial-related depressions overall [64]. Alawadhi et al. [49] buried sheep in an arid environment in Kuwait and compared the Reed–Xiaoli Detector (RXD) [147] and the Uniform Target Detection (UTD) [148] algorithms using ENVI software (version 5.6) for grave identification. RXD and UTD are unsupervised anomaly detection algorithms used in remote sensing, designed to detect anomalies or targets in multispectral images. The differences in these algorithms are the methods used to identify and flag areas/anomalies of an image, and whether human input is required for modelling. RXD identifies anomalies (pixels) that are spectrally distinct from the calculated background statistics determined by user choices and thresholds [149]. UTD is fully automated in background modelling and anomaly detection by building a model from uniformly distributed background signatures and then flagging areas (pixels) that deviate from it [148]. This study [49] found that while RXD was effective for identifying all six graves during the early (1 day–2 months) and late (15 months) post-burial period, UTD demonstrated consistency and temporal robustness (from 3 months to 12 months post-burial), especially as the spectral signatures of the graves changed with time. During image analyses, the RXD algorithm outperformed UTD in 56% (9 of 16 times) of cases, while UTD was best in 44% (7 of 16 times) of the cases [49]. The authors further noted that over the entire study period, a hybrid of the RXD and UTD algorithms (RXD-UTD) had a better outcome in detecting graves (56.25%), than the individual UTD (31.25%) and RXD (12.5%) algorithms [49]. While useful and demonstrating promise for grave location, the reliability of automated detection algorithms remains limited without manual validation from the user and environmental considerations. Parrott et al. [69] deployed a UAV equipped with an unmodified RGB camera to detect disturbed soil from video footage captured near Chester, England [69]. The captured video footage was processed in MATLAB (MathWorks) using computer vision toolboxes, while a custom software was developed for detection. The authors noted that the simulated grave site was located successfully across different altitudes and perspectives when the site was in frame across the video footage [69]. This study, which used a UAV similar to those employed by UK police forces [149], illustrates the potential for operational deployment of AI-enhanced UAV systems to aid in reliable human remains and clandestine grave location in forensic contexts [69].

The variation in automated detection tools used in these studies underscores the early and experimental nature of these approaches in forensics, and the need for further development, validation, and standardisation of approaches and experimental designs. Also, a current limitation in the implementation of AI and automated algorithms is that while drone operators can process data through manual inspection methods, they might not have developed this specific skill set to execute and interpret data generated by AI and automated algorithms. However, as the discipline and technology advance, teams will become more multidisciplinary, with these skills becoming more transferable between operator and programmer.

3.2.5. Multidisciplinary Approaches

Several studies (n = 7) emphasised the value of integrating UAV data with ground-based forensic methods to enhance detection accuracy and validation. These included desktop surveys [93], foot surveys [92], metal detecting [93], a portable field spectroradiometer [31], cadaver dogs [98], ground probes [98], bulk ground conductivity [91,93], airborne LiDAR [89], ground-penetrating radar (GPR) [90,91,93,96,97,98], and electrical resistivity tomography (ERT) [90,91]. Culture-based microbial analysis has also been used to investigate whether the thermal signatures observed during the aerial surveys could be correlated with microbial and insect activity [59]. Multidisciplinary approaches were not only complementary but often necessary to overcome the limitations of individual methods. For example, while Molinia et al. [91] were able to identify seven of eight burials using multispectral imagery; the deeper (0.8 m) eight burial was difficult to detect. Similarly, the shape and boundaries of five graves could be identified using NIR imagery captured over the same site, while the remaining three graves were more difficult to identify due to low contrast at the edges when compared to the surrounding area [91]. Geophysical surveys of the site were useful to complement the findings of the aerial surveys. ERT showed electrical resistivity anomalies for all graves, most likely due to the difference in porosity and permeability of the grave soil compared to natural undisturbed soil, while GPR detected isolated hyperbolic reflection events over the graves. The anomalies detected by both these geophysical approaches confirmed subsurface disturbances [91]. Interpretation of the subsurface disturbances as graves is however almost always based on the context of this study. Within forensic cases, subsurface anomalies identified by geophysical approaches will need further investigation and ground truthing (such as excavation) to confirm the presence or absence of human remains.

4. Discussion

4.1. Key Findings

UAVs with imaging sensors are proven to be effective in various applications, including agriculture, archaeology, military scenarios and environmental monitoring, while their potential in forensic contexts is increasingly recognised [40,71,72,117]. In this systematic review, the thematic synthesis highlights the practical use and limitations of UAVs for human remains and clandestine grave location. The results in the analysed studies revealed that UAVs have considerable potential to locate clandestine graves and to aid forensic investigations, but that environmental factors such as vegetation and climate, as well as burial conditions, significantly influence detection accuracy [150,151]. Evidence from multiple studies demonstrated that UAVs equipped with RGB, thermal and multispectral imaging sensors are effective in locating burial-related anomalies. This has been tested across several environments consisting of arid, tropical, temperate, woodland, and open grassland [49,60,90,93,98]. However, effectiveness remains context-dependent, influenced by factors such as sensor payload, aerial survey design, vegetation cover, soil composition, and burial conditions. Approaches adapted from ecology and hydrology [64] further enhance detection capabilities. For instance, the use of vegetation indices (e.g., NDVI, GNDVI) has become significantly useful for identifying vegetation stress that is potentially indicative of burial-related activities [49,90,97]. This suggests that UAVs can locate graves indirectly by mapping ecological responses, such as vegetation changes, soil colour changes, and topographic changes resulting from the decomposition process and the burial activity. Entomological indicators such as heat signatures from insect activity and larval masses have become promising markers to be used in grave identification, particularly for shallow graves and surface depositions, but are limited to early phase of the decomposition timeline [58,60].

4.2. Methodological Gaps

Within forensic taphonomy and forensic archaeology, experimental design is critical, not only for studies to reflect forensic realism but also for inclusion within the legal case work [10]. The integration of advanced technologies such as geophysical approaches and UAVs within these fields becomes increasingly useful in the movement towards multidisciplinary approaches. The use of UAVs for forensic science is a growing field involving the collection, preservation, and analysis of data and evidence collected from crime scenes for use in investigations [41,71,152]. However, for research and data generated by these technologies and multidisciplinary approaches to have forensically relevant real-world impact, particularly in country specific legal contexts, experimental designs must adhere to standards, similar to, for example, the Daubert criteria, a standard and legal benchmark that determines the admissibility of scientific evidence in the USA-based court system [153]. Experimental designs using UAVs need to demonstrate that the methodology and experimental design are based on sufficient facts and relevant published data, and the approach uses reliable principles and methods that have been tested and evaluated [110,154,155,156]. This is especially the case if research groups want to ensure their research remains relevant and significant and can be included as a validated approach within current medico-legal investigations. From the studies reviewed in this paper, this does not yet appear to be the case. Current experimental designs using UAVs to locate clandestine graves sometimes lack robustness in data collection, while often failing to report motivations for experimental designs thoroughly. It is noted that while experimental designs often lack standardisation, they do attempt to mimic real-world cases as closely as possible, given budget constraints and the availability of resources. The inclusion of dismembered body parts or skeletal remains, on the other hand, can provide relevance that might apply to specific forensic cases [50,58,90,91]. As an example, in order to hide remains or avoid detection, perpetrators might dismember their victim and conceal the body parts at different locations [157,158]. Another example is the secondary burial of remains after concealing them at different primary locations [159], which can form part of the modus operandi of perpetrators in homicides [84,85,86]. Notwithstanding this, to provide robust and reliable results, research teams will have to weigh the benefits of testing novel equipment and tools to locate human remains and clandestine burials, against the need for forensic realism as well as reliability and usability of these tools and methods by police and practitioners in real cases.

In the studies reviewed, a few papers (n = 2) presented experimental designs that included soil-only burials (disturbed earth pits without buried mammalian remains) and the burying of objects like bullet casings, wallets and clothes, separately or on top of mammalian models [49,69], with the aim of replicating forensic scenarios that often contain these objects. The burial of a body has a significant chemical and microbial impact on the environment, which can affect the spectral and chemical signature of the soil and vegetation [160,161,162]. Additionally, clothes have been demonstrated to have an impact on the decomposition rate and process [136,163]. Unless the intention is to develop approaches to locate concealed evidence such as money, drugs or weapons, omitting a body from a grave cannot produce accurate burial-related changes, as the changes are the combined result of the soil disturbance, the complex chemical changes and microbial activity [161,164]. Incorporating these elements in experimental designs is useful for the forensic validity of the studies and improves the applicability of findings to casework. However, it is recommended that objects should be included in a context-appropriate manner.

Considering the ethical and legal restrictions on using human donor cadavers, as well as the cost implications for setting up experimental studies using mammalian remains, the use of natural burial grounds and established cemeteries is a resourceful means to test and develop grave location processes [92,94]. However, a limitation of these facilities is that as established burial facilities, graves are organised in a specific burial pattern and layout, making blind identification almost impossible. Additionally, the use of wooden coffins and burial depths within these facilities might not necessarily reflect shallow clandestine conditions. Clandestine burials are often haphazard and less than a metre deep due to perpetrators’ need to dispose of and conceal the body rapidly, with minimal energy expenditure, while spending as little time as possible at the deposition site to avoid detection [102,103,104,105]. This leads to irregularly dug burials in remote locations that are different from interments in formal burial grounds, with their concealed nature making them challenging to locate [165]. While natural burial sites offer an excellent resource for proof-of-concept validation, studies need to be tested in forensically relevant scenarios to evaluate the results from proof-of-concept studies.

It is crucial for researchers to report on the reasons for specific experimental designs, as exemplified by explaining the burial conditions that reflect current cases reported in Kuwait [57] or burial depths that are based on clandestine graves discovered in Colombia [90]. Additionally, both these studies also reported on the constraints the 2019 COVID pandemic had on data collection, which explained some gaps in the reported results. Incorporating similar elements in experimental designs, or motivations for specific experimental parameters, leads not only to the development of better identification models but also creates transparent and reliable methods and results that can be tested and validated across geographic regions and burial conditions, which can be used in forensic investigations and court proceedings for justice to prevail.

Within the reviewed studies, the lack of robustness of data collected, especially the absence of replicates within experimental designs, affects the reliability of the reported results. This is especially significant since the decomposition process is highly variable and dependent on several biotic (living components such as insects, fungi, microbial communities and scavengers) [166,167,168,169] and abiotic (non-living components such as soil pH, temperature, soil moisture, burial conditions, treatment of the body, and individual characteristics of the victim) [163,170,171,172,173,174] factors that influence the rate and progression of decay [111,175,176,177]. The lack of standardisation and of replicates in experimental designs, like using multiple burial depths or multiple models/proxies in a pit without replication [31,50,58], make it difficult to establish if decomposition-related signatures observed are consistent patterns or per-chance isolated anomalies. While the purpose of research is not duplication, validation across different environments, geographic regions and burial conditions is crucial to develop reliable predictive models for time since burial estimations and clandestine grave location.

5. Conclusions and Future Directions

The widespread availability of low-cost consumer-grade UAV platforms with imaging sensors has become a useful tool for forensic applications. This systematic review aimed to analyse papers discussing human remains and clandestine grave location using UAVs published between 2014 and 2024 to determine the prevailing trends and identify gaps in current research. The studies reviewed demonstrated that UAVs equipped with multispectral, RGB and thermal sensors can be used to detect surface variations as well as locate human remains and clandestine graves within the first few days to years after burial [49,91,94]. In this review, many of the studies used commercial, subscription-based software (DroneDeploy, Pix4D, Agisoft Photoscan Professional, ArcMap (ESRI), ArcGIS (ESRI), ENVI, FLIR Desktop application and MATLAB (MathWorks)), which may not be readily accessible to all forensic practitioners or police forces, particularly in resource-constrained settings. The limiting factor of these software packages is their dependence on user knowledge and technical expertise in image processing, photogrammetry, or geospatial analysis, which may not be available within local forensic teams or police forces. This reliance on specialised software and skills presents a barrier to operational adoption, especially in local police force contexts, and highlights the need for more accessible, user-friendly, and possibly open-source alternatives for grave location and detection.

Some limitations have been identified for this systematic review. Although chosen intently by following PRISMA guidelines to develop a targeted focus, the specific inclusion and exclusion criteria led to the analysis of 19 peer-reviewed papers, hence the exclusion of any data presented in book chapters and reports. To avoid misinterpretation and misrepresentation due to ineffective translations, the authors adopted the accepted and professional practice of excluding publications that were not written in English as their academic language. Nevertheless, this option means that potentially relevant data on UAV-based remote sensing to locate human remains and clandestine graves, from studies published in other languages, have been omitted. Therefore, future primary research articles and critical reviews should test and build on the themes identified in the current systematic review, incorporating relevant pre-2014 outputs, including books, book chapters and grey-literature reports, and non-English publications for enhanced inclusivity in knowledge development.

Drawing on Table 2, it is recommended that future studies report on the survey strategy and survey parameters implemented for the different experimental designs. For future studies to be replicable, standardised and reliable, it is recommended that researchers also include key information about the motivation regarding the experimental design as well as the aerial survey parameters (including the flight pattern, flight time, flight altitude, and GSD). As part of research and knowledge development, future studies could focus on “sensitivity analyses” for each UAV type/detection techniques across different geographical regions, for similar experimental designs [178]. Discussions about the strengths and limitations of the survey parameters to meet the objective of the survey strategy should be included. Specifically, authors can reflect on UAV and sensor type, aerial survey method, burial context and conditions, and the image processing software used. Future research could also explore the use of UAVs in different burial contexts, such as extended postmortem periods and translocation under different biotic and abiotic conditions.

The recent advances in AI (machine learning and computer vision) can enhance the capability of UAVs deployed with imaging systems in locating graves [49,69]. The inclusion of automated algorithms can lead to the development of predictive models, improving the accuracy and efficiency of locating clandestine graves. While AI offers a novel method to differentiate between single and mass graves, a current limiting factor is the generalisability of these models because they are built and tested on simulated burial scenarios and small training datasets. Therefore, larger datasets are needed across diverse biogeographic regions to account for overfitting. Datasets should be expanded by surveying larger areas and by incorporating different angles/perspectives and multiple altitudes. This could aid in the development of more robust predictive models for clandestine grave location and the validation of these models in real forensic case work. Lastly, future studies should also focus on validating the location of human remains and clandestine graves using UAVs for all climatic regions and burial conditions. Importantly these empirically designed studies should incorporate scenarios that reflect real-world forensic cases.