Unlocking Landscape Intelligence for a Sustainable Future
Matt Lythe
Director, Lynker Analytics
The convergence of Artificial Intelligence (AI) and Earth observation has transformed how we understand and manage our landscapes. At Lynker Analytics, we’ve spent much of the last decade building intelligent systems that transform satellite imagery, aerial photography, terrestrial photos, and geospatial data into actionable insights for environmental planning, climate resilience, and sustainable agriculture.
Our work has evolved from foundational projects like city or region scale land cover mapping to ecosystem or landform detection targeting things like wetlands, beaver habitat, critical source areas or pest plants. We have also undertaken many infrastructure monitoring projects across the built environment including road and building mapping, flood monitoring, drain detection and rooftop condition assessment. Additionally, we have applied this technology in the marine realm to detect, speciate measure the length of commercially harvested species in New Zealand and the United States.
In 2024, we partnered with Silver Fern Farms to launch Prism Earth Ltd, with the goal to apply these methods in a focused way to support the primary sector in addressing climate change, meeting emissions targets, and optimising land use. Prism administers our advanced carbon stock mapping system which combines machine learning with expert review and farmer acknowledgement to assess sequestration within vegetation across grazing land using aerial imagery, satellite data, and LiDAR.
One of our flagship projects, in partnership with the Meat Industry Association, exemplifies the power of AI in Earth observation to address the urgent need for robust, science-based tools that help producers meet emerging environmental regulations such as the European Union Deforestation Regulation (EUDR). This system identifies land used to raise and finish beef that has converted from forest to agricultural use. It enables importers operating in Europe to prove that their products are not linked to deforestation.
In our work we prioritise efficiency through automation. To this end we often find that models trained on specific regions or datasets can at times struggle to perform well in new contexts due to floralistic variations, lighting or reflectance conditions, climate, or seasonal variability. For example, a model trained to detect native vegetation in one region may not accurately classify the same species in another region or in a different season. This limits the reusability of AI tools and increases the cost and complexity of deploying them at scale – even within a single country.
The example below is an example of this with exotic hardwoods on grazing land showing substantial difference in reflectance between the “leaves on” and “leaves off” state.
As a result, in many of our solutions, we combine AI for Earth observation with expert (domain specific) human review using independent high-resolution imagery often checked across the seasons. This allows us to merge the scale and speed of AI enabled automated analysis with the contextual accuracy of human judgement.
This approach delivers speed, accuracy, and credibility, making it ideal for compliance, biodiversity monitoring, and land-use decisions.
Looking ahead, the potential of AI in Earth observation continues to grow. Tools we currently have under development at Lynker Analytics and Prism will integrate ecological models, spatial data, and expert independent verification inputs to deliver tailored recommendations and visualisations to support planning and decision making at farm and catchment scales.
Earth observation is no longer just about observing or even classifying the land - it’s about understanding it. With AI alongside independent expert human validation, we’re turning that understanding into action.
