Abhatoo - Centre Nationale de Documentation

The multi‐model approach to crop monitoring and yield forecasting was here tested using the BioMA models CropSyst and WOFOST for soft and durum wheat in Morocco.
Parameterizations derive from the activities carried out within E‐AGRI WP3 tasks, reported in deliverables D31.1, D31.2, D34.1, D34.2, D34.3, D34.4, and D35.1.
Within this task, we tested the multi‐model approach (i) for different species (durum and soft wheat), (ii) for different regions, (iii) for performing yield forecasting in different moments during the crop cycle (maturity decade or one month before maturity).
This design allowed demonstrating that the different approaches implemented in different crop models to formalize biophysical processes involved with crop growth and development could make a model more suitable than others under certain conditions (regions, species).
Results, indeed, showed that WOFOST achieved the highest reliability for both durum and soft wheat forecasts when the forecasting event was triggered at maturity on the whole Morocco; on the contrary, for forecast events triggered one month before maturity, CropSyst resulted the most accurate in case of soft wheat. At national level, R2 between historical official statistics and forecasted yields ranged between 0.61 and 0.83.
For durum wheat, at regional level, WOFOST achieved the best performance metrics in four out of seven regions for all the combinations species × forecasting moment, although the regions changed across combinations. For soft wheat, CropSyst resulted the most suitable in five out of seven regions at maturity, and in four out of seven regions one month before.
These results confirm the usefulness and the potentialities of the multi‐model approach to in season, large area wheat monitoring and yield forecasting.