Remote Sensing: Satellite, Aerial, and UAV Applications

Stressed, disturbed, or anomalously nourished vegetation reflects light differently from healthy undisturbed surroundings. Plants over a compacted backfill tend to show reduced near-infrared reflectance (less photosynthetic activity) in the early months post-burial. Later, as nutrient release from decomposition peaks, they may over-perform relative to background, producing a paradoxically high NDVI patch.

A single image is difficult to interpret because many factors cause NDVI variation. A time series comparing pre-event baselines to post-event imagery is far more diagnostic. A patch showing a sudden NDVI drop that persists for 3-6 months and then recovers is consistent with vegetation disturbance and recovery. An NDVI spike above background in a patch that previously tracked its surroundings is consistent with nutrient enrichment from decomposition.

Microbial decomposition releases heat as a byproduct of aerobic and anaerobic metabolism. Shallow burials, particularly in warm seasons, generate a measurable thermal anomaly above the grave. Studies in the UK, Australia, and North America have confirmed temperature differentials of 1-5 degrees Celsius above background over active decomposition, depending on burial depth, soil type, and ambient temperature.

Satellite TIR bands (Landsat 8-9 Band 10, ECOSTRESS) have a ground resolution of 30-100 m, which is far too coarse for a single grave. Airborne TIR surveys at 0.5-2 m resolution are operationally useful. UAV-mounted uncooled microbolometer cameras (FLIR-type) have become the standard tool for targeted surveys because they cost a fraction of a manned aircraft and can be positioned over a suspected zone within hours.

• Optimal timing: pre-dawn surveys minimise solar heating confounders. The soil thermal signature is most detectable when surface solar radiation has been absent for several hours.
• False positives: pipes, cables, animal burrows, and differential moisture from recent rain all generate thermal anomalies. Correlation with NDVI and GPR is necessary before a thermal anomaly becomes a search priority.
• Temporal window: the thermal signal is strongest during active decomposition (weeks to months post-burial) and diminishes once tissue is fully skeletonised. Old graves may not produce a detectable thermal anomaly.

Synthetic Aperture Radar is operationally invaluable in regions with persistent cloud cover. The Sentinel-1 SAR constellation (C-band, 10 m GSD, 6-12 day revisit) provides free archive data for most of the globe and is the standard starting point for SAR-based forensic investigations. Sentinel-1 cross-polarisation (VV-VH) combinations are particularly sensitive to volumetric moisture content and surface roughness, both of which differ between disturbed and undisturbed soil.

A freshly dug and backfilled grave pit has higher surface roughness than the surrounding compacted soil and, for weeks after burial, higher moisture content from the loosened, more permeable fill. These two properties combine to elevate SAR backscatter over the grave relative to background. Change detection between a pre-disturbance and post-disturbance Sentinel-1 acquisition can flag the anomaly even when the site is under cloud or in a vegetated but not fully canopy-closed environment.

UAVs close the resolution gap between satellite reconnaissance and ground survey. A standard forensic UAV mission uses a fixed-wing or multirotor platform carrying RGB cameras for photogrammetric surface models, multispectral sensors for vegetation indices, and thermal cameras for heat mapping. At 50-100 m altitude over a 1 ha target zone, a 20-minute mission can produce a sub-centimetre RGB orthomosaic and a 5-10 cm thermal map.

Regulatory constraints vary by jurisdiction. In most countries, UAV operations over a crime scene require coordination with aviation authorities and police command, but this has become routine for major investigations. The main operational limitations are battery endurance (typically 20-40 minutes per sortie for multirotors), wind sensitivity, and the need for a skilled pilot-in-command who understands the geophysical rationale for the mission as well as the flight safety requirements.

Change detection works by subtracting a pre-event baseline from the post-event image. The baseline captures normal soil variation, shadows, crop stages, and seasonal patterns; the post-event image contains the same background plus any disturbance. Their difference isolates the change.

1. Baseline selection
   Select the most recent cloud-free image before the estimated date of burial. Sentinel-2 archives extend to 2015, providing baseline data for cases as old as a decade. Match seasonal conditions (same month, ideally same phenological stage) to minimise false changes from seasonal vegetation cycles.
2. Image co-registration
   Align the baseline and post-event images to sub-pixel accuracy using tie-point matching or a common reference grid. Misregistration of even 0.5 pixels introduces edge artefacts that can be mistaken for change.
3. Normalisation
   Apply radiometric normalisation to correct for differences in atmospheric conditions, sensor gain, and solar angle between acquisition dates. Pseudo-invariant features (stable asphalt, bare rock outcrops) serve as reference targets.
4. Differencing and thresholding
   Subtract the baseline band values from the post-event values. Apply a statistical threshold (typically mean plus 2 standard deviations of the difference distribution) to identify pixels with significant change. These become the candidate anomaly mask.
5. Contextual filtering
   Remove anomalies that are inconsistent with the forensic context: large agricultural fields (ploughing), known infrastructure works, and water bodies. What remains is a short list of unexplained change polygons for ground verification.

Optical and near-infrared sensors, including thermal, cannot penetrate a closed forest canopy. The signal reaching the sensor is from the canopy surface, not the soil. A grave under 15 m of mixed-species temperate forest is essentially invisible to Sentinel-2. SAR C-band (Sentinel-1) penetrates thin vegetation and can detect surface-level disturbance under light canopy, but X-band and Ku-band systems needed for sub-canopy penetration are generally not available at forensic-relevant resolution from free platforms.

Built environments present a different problem: spectral and thermal complexity. Tarmac, metal roofing, and concrete all have strong thermal signatures that swamp a grave-scale anomaly. In these environments, remote sensing plays a reduced role. Ground-penetrating radar, cadaver dogs, and witness-led search are the primary tools. Remote sensing can still contribute by identifying the best access routes, mapping building footprints for exclusion, and providing a spatial reference frame for recording negative outcomes.