The future of surveying, mapping and global sustainability requires pairing satellite technology with human insight.

By Priscillia Moulin

I’d like to paint a picture — likely a familiar one — for anyone working in spatial data.

A geospatial analyst monitors a compliance dashboard when a cluster of pixels over a remote province shifts from deep green to red.

A machine learning algorithm fires off an automated alert, flagging a high-probability deforestation event. Within days, a smallholder farmer thousands of kilometres away is flagged by a corporate buyer and suspended from a global supply chain.

The catch? That farmer hadn’t cleared a single hectare of primary forest — instead, the alert could have been caused by technical and data limitations, environmental and biological factors, human activity, land-use misclassification, or analytical errors. As impressive as orbital observations undoubtedly are in 2026, they cannot accurately determine intent or causality, and they do not account for complex local dynamics.

This crucial flaw will become increasingly important with the landmark European Union Deforestation Regulation (EUDR) set to come into force at the end of 2026.

(Editor’s note: The EUDR applies to Australian companies that export certain goods such as beef and timber to the EU, and even derived products such as leather. Geolocation information about the products’ source must be supplied upon importation.)

Major commodity buyers are scrambling for compliance, with many looking to use the golden era of AI and satellite imagery to help them determine deforestation. But as I’ve highlighted above, relying exclusively on remote sensing to police agricultural supply chains is creating a highly damaging “ground truth gap”.

For the surveyors, geospatial professionals and data scientists building these systems, it is time to critically examine the limits of our pixels — and acknowledge that automated models built without human geography have fundamental flaws that need addressing.

The limits of the algorithm

In the rush to scale supply chain mapping, compliance platforms are relying heavily on automated land-cover classification. While modern satellite constellations offer staggering temporal frequency and spatial resolution, the machine learning models interpreting that imagery are not infallible.

Algorithms are exceptional at detecting changes in land cover, but they are not designed to account for context, intent or legality. For those of us working at the intersection of spatial data and tropical agriculture, there are several potential reasons for a deforestation alert we need to be aware of:

• Intentional clearing: Intentional, illegal deforestation to open up new land for cash crops.
• Natural events: Accidental or natural landscape disturbances, such as localised fires, floods, or severe storm damage.
• Permitted agriculture: Legal, planned rotational harvests that are fully aligned with recognised community land tenure rights.
• Civic development: Unrelated rural infrastructure projects — such as a new community road or power line — that are entirely detached from the agricultural supply chain.
• Land tenure invisibility: Vast swathes of global agricultural production occur on completely unregistered land without formal legal titles. Satellites cannot detect cadastral boundaries or the legal status of the land, making it impossible to comply with regulations requiring verifiable land ownership without ground-level administrative data.
• Protected area overlaps: While algorithms can detect trees, they cannot automatically cross-reference agricultural expansion against legally designated conservation zones or permanent forest reserves without integrated spatial mapping.

Alerts on water bodies

The image below is a case in point. It shows a sequence of red polygons — deforestation alerts — appearing across the surface of a lake and along its margins. Geomorphologically and ecologically, forest cover loss cannot occur on open water, because no canopied vegetation grows above an open water body. Alerts in locations like this are pure false positives.

Several technical factors commonly trigger errors like these:

1. Seasonal water-level fluctuation that exposes edge vegetation or substrate, so changes in spectral reflectance between acquisition periods are read as forest loss;
2. Residual thin cloud cover, cloud shadow, and atmospheric haze that are not fully corrected during pre-processing;
3. Sun glint on calm water surfaces, which produces NIR/SWIR signatures resembling vegetation loss; and
4. Water masks that haven’t been updated to reflect lake morphology dynamics (edge expansion or contraction).

The risk of leakage

What happens when spatial data is used to drive automated enforcement without human context? It oversimplifies complex geographies and creates a path of least resistance.

Faced with a red polygon on a compliance dashboard and the threat of severe EUDR financial penalties or compromising its own ‘No Deforestation, No Peat, No Exploitation’ (NDPE) commitments, a corporate buyer will almost always choose self-preservation. They drop the flagged supplier and source their commodity elsewhere.

This creates a dangerous illusion of compliance. The corporate dashboard looks clean, and the spreadsheet shows zero deforestation. But this is a failure of both data integrity and sustainable development.

When vulnerable smallholder farmers are abruptly cut out of premium, regulated markets due to an algorithmic false positive, they do not simply pack up their tools and stop farming. Economic necessity dictates that they find another buyer. This dynamic actively drives farmers into ‘leakage markets’ — regions or buyers with lower environmental standards, fewer NDPE policies, zero EUDR oversight, and lower prices.

In these grey markets, deforestation continues unmonitored. One unintended consequence of legislation like EUDR is that it’s essentially using advanced spatial technology to clean up European supply chains while quietly pushing the actual root causes of deforestation into the shadows. We cannot achieve accurate environmental monitoring by sacrificing ground-level realities.

People need to work with technology

To build monitoring systems that actually work, the geospatial industry should instead work towards a more combined methodology. Satellites and AI are tools, but they need accurate baselines and verified, on-the-ground data to validate them.

Integrating human verification into spatial workflows might feel like a backward step to those embedded in the tech world. But when it comes to hugely important matters like deforestation, we need to take the best of orbital tech and human geography to ensure NDPE policies and legislation like EUDR have a genuine, positive impact. Effective due diligence needs to move beyond simple risk screening; it requires an end-to-end workflow combining landscape analytics with field implementation.

First, monitoring systems need to be anchored in verified baselines. Deforestation alerts cannot be generated from generic, outdated global datasets. They require a rigorous methodology that combines high-resolution spatial data of planting areas with digital cadastral records and official protected area maps.

By linking suppliers to specific plot polygons with stable IDs, validating those polygons, and attaching physical tenure documents, companies can build a legally defensible master dataset. Without this granular traceability, spatial alerts cannot translate into meaningful accountability.

Secondly, technology should trigger a conversation… not an automatic suspension. High-risk tree-cover loss alerts should be paired with response protocols that mandate on-the-ground verification. A local surveyor or agronomist should then establish the true context of the clearing to determine if it was intentional, legally permitted, or linked to complex land tenure disputes. This includes assigning risk classifications and defining specific mitigation actions for flagged plots.

The shift to Verification-as-a-Service

The broader market is already waking up to the liabilities of relying solely on algorithms. We are witnessing a distinct, rapid shift in AgTech investment trends: capital is flowing away from standalone SaaS (Software-as-a-Service) monitoring platforms and heavily toward Agriculture Technology-as-a-Service (ATaaS) and Verification-as-a-Service (VaaS) models.

Investors and major brands now realise that a satellite image is just raw material — the true value lies in the verified insight. VaaS models combine scalable remote sensing with robust networks of local agronomists, surveyors, and field agents.

This hybrid approach provides the high-fidelity, legally defensible data required by strict frameworks like the EUDR, while significantly reducing the reputational and supply chain risks associated with automated false positives.

Conclusion

The EUDR is a landmark piece of legislation, and the geospatial industry has provided indispensable tools for unprecedented visibility into global supply chains. But we must remember that they are exactly that — tools.

Laws and satellites do not save forests — people do. Extracting compliance data from space is an incredible first step, but it is incomplete without accurate baselines and verifiable on-the-ground action.

The future of surveying, mapping and global sustainability requires pairing our best satellite technology with human insight, acknowledging the messy complexities of the ground truth, and working toward a fairer, more accurate future for everyone.

Priscillia Moulin is co-founder and Director of Strategy at MosaiX, and Senior Advisor with Earthqualizer Foundation and Inovasi Digital.