One field, three readings.
Artificial intelligence and digital technologies for pest control across agricultural products and farmland, with environmental assessment and life cycle analysis behind every intervention.
Where the population is pressing
The model reads catches from camera traps, symptoms in the canopy and microclimate, then works out how close each plot sits to the intervention threshold.
- Inputs
- Camera traps, multispectral imagery, temperature and humidity, agronomist field notes
- Decision
- Where and when an intervention is justified, and over what area
Hover over a plot to read its value.
The project
Crop protection became a data problem without ceasing to be a field problem.
Pest cycles are shifting
Mild winters and extreme weather events change when pests appear and how many generations follow. Last season's calendar has stopped being a reliable guide, and scheduled applications miss their target more and more often.
The available tools are shrinking
Active substances are being withdrawn, residue limits are tightening, and resistance builds whenever the same solution is repeated. Every application made without cause is paid for twice, once at the till and once in future effectiveness.
The market asks for evidence
Food chains and export markets now ask for measurable figures on the environmental footprint of a product. Good practice without numbers goes unrecognised, and a story without a method does not survive an audit.
Three pillars
Detect, assess, document.
The three pillars do not run one after another. The same field database feeds all of them, so a recommendation to intervene always arrives with its environmental cost attached.
First pillar
Detection and forecasting with artificial intelligence
The aim is not to spray more or to spray less, but to spray where it is needed, when it is needed.
- What is recorded
- Camera traps that count catches in real time, multispectral imagery from drones and satellites, weather stations at canopy height, and the agronomist's notes from scouting. These heterogeneous sources are merged onto a single spatial base layer for each plot.
- What the model learns
- Species and life stage recognition from images, phenology models that translate accumulated degree days into the expected emergence of a generation, anomaly detection in spectral indices before a symptom becomes visible, and spatiotemporal forecasting of spread from one focus to the next.
- What is delivered
- A risk map for each plot with its uncertainty range, an alert as the intervention threshold is approached, and a proposal for targeted application by zone instead of blanket coverage of the whole area.
- Who decides
- The agronomist and the grower. The system carries out no intervention and does not replace knowledge of the field. It documents a proposal, shows the data behind it, and records what was finally decided.
Second pillar
Environmental assessment of the intervention
Every intervention has a recipient beyond its target. Assessment makes that recipient visible before the decision is taken.
- What is assessed
- Emissions from operations and inputs, runoff and leaching towards water bodies, effects on organisms that are not the target, load on the soil, water and energy use per operation.
- How it is done
- Primary farm data is combined with spatial sensitivity data such as proximity to water bodies, slope and soil type, neighbouring habitats and the position of beehives. The result is not one average for the holding, but a distribution across the map.
- How it is used
- Management scenarios are compared on a common basis, for example established practice against targeted application supported by the model. The comparison shows what is gained and what is displaced, so that moving a burden from one category to another is never mistaken for an improvement.
Third pillar
Life cycle analysis of the product
The analysis follows ISO 14040 and ISO 14044, so the results hold up to review by third parties.
- Goal and boundaries
- Two things are stated explicitly: the functional unit, meaning the quantity of product every result refers to, and the system boundaries, meaning which stages are included and which are left out. Without those two, no figure is comparable.
- Inventory
- Input and output flows are recorded at every stage, from the production of inputs and the field operations through to packaging, transport and residue management. The field data from the first pillar feeds this inventory directly.
- Impact assessment
- Flows are translated into impact categories using a recognised characterisation method, so that quantities of a different nature become comparable with one another.
- Interpretation
- The stages that dominate the result are identified, the sensitivity of the conclusion to assumptions is tested, and recommendations are written that feed back into the management of the next growing cycle.
Method
A loop, not a line.
Assessment does not arrive at the end to confirm what was already done. It returns as an input to the next monitoring cycle.
-
Step 1
Field recording
Traps, sensors, imagery and scouting feed a continuous stream of data for each plot.
-
Step 2
Integration
Data is cleaned, aligned in time and projected onto a common spatial base layer.
-
Step 3
Model
A risk estimate is produced with its uncertainty range, never as a single absolute figure.
-
Step 4
Recommendation
Intervening or waiting is proposed, with the reasoning and the environmental cost next to the decision.
-
Step 5
Application
The action is carried out in the field and recorded as it actually happened, with timing, dose and area.
-
Step 6
Verification
The outcome is checked in the field and the error in the forecast becomes training data.
Inputs
Production and transport of fertilisers, plant protection products, planting material and fuel.
Cultivation
Soil operations, irrigation, crop protection, energy and emissions in the field.
Harvest
Picking, first grading and product losses before leaving the holding.
Processing
Cooling, standardisation, packaging and the materials that come with it.
Distribution
Storage and transport to the points of sale, with the conditions they require.
End of life
Management of packaging waste and residues, recycling and recovery.
What we measure
Indicators under monitoring.
Agronomic indicators answer what is happening in the field right now. Environmental indicators answer what each decision leaves behind.
| Indicator | Unit | What it shows |
|---|---|---|
| Population density | catches per trap per day | The pressure of the pest at that particular position and moment. |
| Distance to threshold | percentage of threshold | How close the plot is to the point where an intervention is justified. |
| Accumulated degree days | degree days | The expected development stage of the population based on temperature. |
| Treated area | percentage of area | How much of the crop actually received an intervention rather than the whole of it. |
| Forecast accuracy | error in days | The gap between predicted and observed emergence in the field. |
| Impact category | Unit | What it shows |
|---|---|---|
| Climate change | kg CO2 eq | Total greenhouse gas emissions per functional unit. |
| Acidification | mol H+ eq | The load on soils and waters from acid deposition. |
| Freshwater eutrophication | kg P eq | Nutrient enrichment of water bodies and the consequences that follow. |
| Freshwater ecotoxicity | CTUe | The potential effect of active substances on aquatic organisms. |
| Water use | m3 world eq deprived | Water consumption weighted by how scarce water is in the area. |
| Land use | soil quality index | Pressure on the productive capacity of the soil from management. |
| Energy demand | MJ | Total primary energy embedded across all stages. |
The principle running through the project
A recommendation without its environmental cost is not a complete recommendation.
Questions
What gets asked first.
Does the system replace the agronomist?
No. The system gathers and interprets data that nobody can watch continuously across a whole holding. Judging the proposal, knowing the particular field and taking the final decision stay with people.
Does the holding need new equipment?
The approach is designed to start from what is already there. Traps and weather stations are added gradually, and targeted application by zone can be carried out with conventional machinery as long as the zones are large enough.
Who owns the farm data?
Primary data belongs to the grower. What is collected, for which purpose, how long it is kept and how it is anonymised when used in aggregate to train the models are all agreed from the outset.
How accurate is the forecast?
It depends on the species, the density of sensors and the depth of historical data. For that reason every estimate is presented with an uncertainty range rather than as a single figure, and the deviation from observation is recorded systematically.
Do life cycle results transfer to another crop?
Not as they stand. Results hold for the functional unit, the system boundaries and the data of the specific study. The method transfers, the numbers are calculated again.
What does less intervention mean in practice?
It means treating the zones where pressure justifies the cost rather than the whole area, at the moment when the stage of the population makes treatment effective. The reduction comes from targeting, not from leaving things out.
Terms
The words that keep coming back.
Functional unit
The quantity of product or service that every result of the analysis refers to, so that two systems can be compared on a common basis.
System boundaries
The explicit definition of the stages and flows included in the study, together with what is left out and for what reason.
Life cycle inventory
The systematic record of every input and output of the system, from raw materials and energy through to emissions and waste.
Intervention threshold
The population or infestation level above which the expected damage exceeds the cost of intervening.
Degree days
Heat accumulated above a lower development threshold, which predicts how fast a population progresses.
Spectral index
A numerical combination of image bands that captures the state of the vegetation before it becomes visible to the naked eye.
Integrated pest management
An approach that puts preventive and non chemical measures first, and uses chemical intervention as a last resort with documentation.
Burden shifting
The case where an improvement in one impact category causes a deterioration in another, which is why assessment has to weigh several criteria at once.