Environmental DNA and Emerging Detection Technologies

Environmental DNA is DNA that organisms shed into their surroundings. In aquatic systems, fish and amphibians constantly release cells through their skin, gills, and faeces. A water sample from a live-animal market tank, a customs holding facility, or a suspected transport vehicle can contain eDNA from every species that was recently present, even after the animals are gone.

Metabarcoding amplifies a short, taxonomically informative region of the genome using universal primers that work across many species simultaneously. The amplified products are sequenced on a high-throughput platform, and each sequence is compared against a reference library to assign a species identity. A single water sample can yield a species list covering fish, amphibians, reptiles, and invertebrates. Studies have demonstrated that eDNA metabarcoding can detect CITES Appendix I and II species in market water samples with sensitivity comparable to traditional inspections, while covering species the inspector might not have noticed or recognised.

The forensic challenges are real. eDNA degrades rapidly in warm, sunlit, or turbulent conditions. Contamination during collection or laboratory processing can introduce false positives. The reference libraries for many traded taxa are incomplete, so an eDNA detection may match a protected species or a closely related non-protected one depending on database quality. None of these problems is fatal, but they mean eDNA results in casework require careful validation, quantitative controls, and conservative interpretation.

Satellite imagery became a conservation tool in the 2000s with the availability of medium-resolution sensors like Landsat and MODIS. By 2020, commercial operators including Planet Labs, Maxar, and Airbus were offering daily revisit cycles at resolutions of 30 to 50 centimetres per pixel over specific areas. This resolution is sufficient to resolve vehicle tracks, tent structures, camp fires, and cleared areas within protected zones.

Change-detection analysis compares a time series of images and flags areas where forest cover has decreased, bare ground has appeared, or spectral signatures have shifted in ways consistent with fire, clearing, or vehicle traffic. Tools like Global Forest Watch, which is maintained by the World Resources Institute using Landsat and Sentinel data, allow anyone to set an alert for a specific protected area and receive notification when cover change exceeds a threshold. Enforcement agencies use the same data to generate patrol priorities.

• Deforestation mapping: detects illegal logging at the perimeter of protected areas, which is often the first stage of habitat fragmentation that enables poaching. Time-stamped imagery provides evidence that clearing occurred within a protected zone.
• Camp and road detection: high-resolution images can show the establishment of illegal hunting camps, access roads cut into forest, and temporary structures used by poaching teams on multi-day operations.
• Night-light analysis: infrared sensors can detect campfires and vehicle headlights in otherwise unlit protected areas at night, a signal consistent with poaching activity.
• Habitat connectivity modelling: overlaying land-cover data with species range data identifies corridors where wildlife is likely to cross agricultural or logging land, and where poaching pressure is therefore highest.

Satellite data as direct evidence in criminal proceedings requires establishing an authenticated chain of custody for the imagery: who acquired it, what processing was applied, and whether the processing could have introduced artefacts. Courts have accepted satellite imagery as evidence in international law contexts, and its use in domestic wildlife crime prosecutions is increasing as awareness of the technology grows among prosecutors.

Large protected areas present a persistent enforcement problem: patrol coverage cannot be continuous across the full area, and a poaching team can kill an animal and leave before rangers arrive. Acoustic detection systems address this by deploying listening devices throughout the park and using automated classifiers to distinguish gunshots from background noise, triggering real-time alerts.

The Rainforest Connection system, developed in partnership with conservation organisations and deployed in parks across Africa, Southeast Asia, and South America, uses repurposed Android devices with external microphones mounted in tree canopies. Audio is streamed via cellular or satellite to a server running a convolutional neural network trained on labelled recordings of gunshots, chainsaws, vehicle engines, and ambient forest sound. Alerts reach rangers' smartphones within 20 to 30 seconds of a trigger event, enabling a response while poachers are still at the kill site.

Acoustic data has also been used in a different investigation mode: passive recording over months to map temporal and spatial patterns of gunshot events. This produces intelligence about which areas are targeted most heavily, what times of day or night poaching activity peaks, and whether enforcement actions change the spatial distribution of activity.

Camera traps trigger on motion or heat, taking photographs of whatever passes in front of them. A network of cameras deployed across a protected area can generate tens of thousands of images per week. Manual review is slow, expensive, and inconsistent. Machine-learning classifiers trained on labelled camera-trap images can sort this volume in hours, categorising each image by species, flagging images with humans, and in some cases identifying individual animals from natural markings.

Wildlife Insights, a platform developed by Google and conservation partners, provides cloud-based species classification for camera-trap images from a model trained on millions of labelled photographs. Users upload raw images and receive automated species labels with confidence scores, which are then reviewed and corrected by a researcher. The platform contributes to a global dataset that improves the model iteratively. Accuracy varies by species and image quality: well-represented taxa in the training set are classified at high accuracy, while rare or visually similar species may perform poorly.

Individual identification extends the forensic value further. Spot patterns in leopards, stripe patterns in tigers, and ear notches in elephants are individually distinctive. Pattern-matching algorithms originally developed for astronomical star-field matching and adapted for individual whale shark identification from spot patterns can match individual animals across images taken months or years apart, supporting population size estimates and detecting the removal of known individuals.

SMART (Spatial Monitoring and Reporting Tool) is open-source software developed by a consortium of conservation organisations, including the Wildlife Conservation Society, WWF, and the Zoological Society of London, for recording and analysing ranger patrol data. Rangers enter data on standardised digital forms during or after patrols, recording their route, duration, observations of wildlife, and signs of illegal activity such as snares, carcasses, poaching camps, or human tracks.

The data is uploaded to a central server and visualised on maps showing patrol coverage, threat density, and trend lines over time. The key analytical output is the ability to compare where rangers are patrolling against where threats are occurring, and to identify gaps where coverage is thin relative to threat level. Park management can shift patrol assignments week by week based on what the data shows, rather than following fixed patrol schedules that poachers quickly learn to exploit.

• Threat mapping: snare density, carcass locations, and human sign plotted on a map reveal where poaching pressure is highest and whether it is shifting in response to enforcement.
• Patrol effort accounting: patrol hours and kilometres by sector allow managers to demonstrate effort to funders and to detect sectors where coverage is falling below minimum standards.
• Trend analysis: comparing illegal-activity rates across years reveals whether interventions, such as increased patrols, community outreach, or fence construction, are reducing threat.
• Evidence documentation: SMART records can be exported as time-stamped, GPS-located reports that can support a criminal case by showing when and where a poaching sign was found by a named ranger.

No single technology in this toolkit is ready to substitute for traditional forensic casework as evidence in court. eDNA can show a species was present in a location; it cannot prove a specific individual was transported there. Satellite imagery can show deforestation; it cannot identify the individual who authorised the clearing. An acoustic gunshot alert pinpoints a location and a time; it does not name the shooter. Camera-trap AI identifies a species; it may misclassify visually similar taxa. SMART records document patrol activity; they are not forensically validated digital evidence by default.

Integration is where these tools become most powerful. An eDNA detection of a protected species in a trader's water supply, combined with satellite imagery showing vehicle tracks from that trader's location to a known poaching area, combined with camera-trap records of the species at the source location before its disappearance, and supported by SMART patrol records showing the area was lightly patrolled during the critical period, builds a circumstantial case that no single technology could construct alone.

Validation for court use is the common gap. Each of these methods needs published error rates, documented protocols, chain-of-custody procedures for digital data, and case law establishing acceptance. eDNA is the furthest along this path, with a growing body of peer-reviewed forensic validation studies. Remote sensing and acoustic monitoring are primarily intelligence tools at present. As wildlife crime prosecutions increasingly rely on digital and remote evidence, the validation work must keep pace with the investigative use.