Chapter 8 - Assessing the France 2016 extreme wheat production loss—Evaluating our operational capacity to predict complex compound events

Abstract

The 2016 wheat harvest in France suffered from an unforeseen and unprecedented production loss (Ben-Ari et al., 2018). At 5.4 t/ha, the wheat yield was the lowest recorded since 1986 and 30% below the 5-year average. Crop yield forecasting can be considered as near-real-time impact modeling, but unfortunately, none of the forecasting systems in place anticipated the extent of the impact. We summarize key insights and assess the reasons for our failure to forecast French wheat accurately in 2016, with wider implications for quantifying the impacts of extremes on agriculture. The 2015/2016 growing season was characterized by compounding warm autumn temperatures and abnormally wet conditions the following spring. High rainfall and high temperatures leading to fungal diseases, soil water lodging and anoxia, low radiation affecting grain filling, and leaching of nitrogen have all been suggested as important factors ultimately leading to the yield loss. The use of binomial logistic regressions accounting for warm autumn and wet spring conditions together with autumn precipitation and June temperatures, suggests that the odds of an extreme yield loss (i.e., <10th percentile) in 2016 was 35 times higher than expected. The challenge now is further untangling the variety of biotic and abiotic processes interacting at different timescales that can explain observed patterns. Collecting relevant insights from field trial experiments and confronting these with statistical and biophysical crop modeling will be key to achieve this. Besides improving forecasting methodologies, improving the collection, combination, and aggregation of novel real-time data streams will be necessary. Interestingly, a map of yield estimates based on Tweets of farmers harvesting their wheat was the first indication that something was amiss. Such novel data streams can help to anticipate impacts. Improved impact relevant predictors will need to be integrated into operational crop yield forecasting systems in preparation for future compound events.