Embedding expert opinion in a Bayesian network model to predict wheat yield from spring-summer weather

Wheat yield is highly dependent on weather, Therefore, predicting its effect can improve crop management decisions. Various modelling approaches have been used to predict wheat including process-based modelling, statistical models, and machine learning. However, these models typically require a large data set for training or fitting. They often also limited ability in capturing the effects of small-scale variability, time, duration extreme weather events. Here, we develop Bayesian Network (BN) model by interviewing experts farmers, embedding their knowledge from years experience within quantitative model. These identified period beginning anthesis end grain filling stage as critical maximum temperature, mean temperature precipitation key variables inclusion BN. To keep time input manageable, conditional probability table BN was constructed based anticipated impact different conditions. The predicted same neighbouring class (very low, medium, high very high) reported with low error rate ranging 9.1 15.2% and, when estimate median yield, R2 41 52%. Interestingly, successfully 1998, 2007, 2012 2020 which had most Additionally, more recent data, 2022 accurately, especially season not sown yet eliciting information recently added testing data. Little difference observed between predictions made using parameters only opinion farm manager test originated, average group 9 experts. causal provided insight into experts’ rationale, allowing unexpected results be explored. This methodology provides means rapidly successful predictive (or no) expert understanding. could tuned updated it becomes available.