Crop modelling before you choose your next crop

Your maize crop has struggled through another dry summer. Should you change the variety, shift sowing dates or try a different crop? For farmers in Central Europe, crop modelling can help compare those choices. Its value depends on local records and clear assumptions about the conditions ahead.

Crop modelling starts with a scenario

A crop model estimates plant growth and yield under defined weather, soil and management conditions. Paired with climate projections, running several scenarios shows how a shift in rainfall, temperature or sowing date could affect the harvest.

Two process-based tools take complementary approaches. AquaCrop, developed by the Food and Agriculture Organization of the United Nations (FAO), focuses on yield response to water. The Decision Support System for Agrotechnology Transfer (DSSAT) simulates crop growth and lets users compare varieties, fertiliser rates and management options. Machine learning models work differently, learning patterns from historical datasets and requiring separate testing on new seasons.

At a regional scale, the Common Agricultural Policy Regional Impact Analysis (CAPRI) model examines agricultural markets and policy effects across Europe. Its macro-level outputs support policy planning, whereas a farm planting decision needs field-level crop simulation.

Every simulation depends on the time horizon you choose. A model using recent local weather answers a different question from a mid-century emissions scenario, so keep the baseline period and management assumptions attached to each projected yield.

Bring the field records together

For crop modelling, gather weather, soil profile, cultivar, sowing and management records from the same field. FAO’s input guidance outlines the core data groups required for a simulation, and including observed yields lets an agronomist check how well the model reproduces past seasons before running future scenarios.

• Keep locations, dates and units attached to each record.
• Log irrigation, fertiliser and major heat or drought events.
• Separate measured observations from estimated or missing inputs.

Weather station distance and soil sampling depth strongly influence model behaviour. When local records are incomplete, testing how different plausible soil or rainfall inputs affect the result shows how sensitive the forecast is before you rely on it for an investment.

In-season observations also help explain unusual years. Pairing field logs with sensor checks, as covered in our guide to plant stress detection, shows when water shortage or heat stress began in the canopy.

Check crop modelling against local evidence

Broad regional lists of climate winners and losers rarely hold true for every farm. Soybean is often promoted as a warm-weather option, yet a 2025 European Journal of Agronomy study by Barrera and colleagues using AquaCrop in two Italian locations and Ljubljana, Slovenia found mainly negative climate impacts on rainfed soybean between 2041 and 2100. Irrigation offset those losses under moderate scenarios in the Italian sites, whereas in Ljubljana irrigation had a weaker mitigating effect even though overall yield potential remained higher.

Because those simulations cover periods from 2041 to 2100, your next planting decision still requires current farm constraints, local trial evidence and a buyer for the harvested crop.

Data-driven yield models show similar local sensitivity. A 2024 Agronomy Journal study in Saxony using random forests found lower prediction errors for rapeseed, barley and wheat than for maize, potatoes and sugar beet. Regionally trained models outperformed broader ones, and accuracy dropped on unfamiliar datasets, showing why you should check which crops and seasons validated a model.

Choose a trial before a full change

Begin with a practical crop selection decision you can test on your own land. Compare your current practice with an adjusted sowing date, a drought-tolerant variety or a candidate crop such as sorghum, while reviewing water availability, machinery, rotation and buyer requirements with your agronomist.

Isolate one variable at a time to see what drives the result. Holding soil and management constant across rainfall scenarios highlights weather risk, while fixing the weather series lets you compare sowing dates. Crop modelling can narrow the choices for a small field trial across several seasons, and recording the trial harvest lets you verify the comparison on your farm.

Listen to the crop forecasting podcast episode for a wider discussion on future crops.

Contact AGRI-FOOD.AI with your location, crops, available records and the decision you want to compare. We can discuss the data workflow and scope of a possible project.