Keeping crop models in step with the field
The fourth interview with Lior Doron covered crop models. He compared them to recipes: they describe how to grow a crop under given conditions. The weakness is that they are set at the start of the season and know nothing about the rain, the heatwave or the broken valve that followed. Run on their own, they drift further from reality as the weeks pass.
FIGARO's main technical challenge was to link the models to field data so they could be corrected during the season. The models were never designed to receive real-time inputs, which made this hard.
The intended result: the system takes in what actually happened in the field each week, adjusts the model and updates its recommendations. If irrigation was heavier or lighter than planned, or if it rained, the rest of the season is recalculated.