Causes and implications of the unforeseen 2016 extreme yield loss in the breadbasket of France

Tamara Ben‐Ari, Julien Boé, Philippe Ciais, Rémi Lecerf, Marijn van der Velde, David MakowskiNature Communications 2018 · DOI

TL;DR

The 2016 extreme wheat yield loss in France, which surprised forecasting systems, was caused by a novel "compound extreme" event: warm autumns followed by wet springs. This study develops a statistical model that accurately predicts such losses and projects that these conditions will become more frequent in the future, necessitating adaptation in agricultural practices and forecasting.

What Problem Does This Paper Address?

Yield forecasting systems are crucial for global food security and agricultural economics. However, these systems often struggle to predict extreme yield losses, particularly those resulting from complex, interacting climate conditions rather than single extreme events like heatwaves or droughts. The 2016 wheat harvest in France, a major global producer, experienced its most severe yield loss in over 50 years, a phenomenon that public forecasting systems failed to anticipate. This highlights a critical gap in our understanding and modeling of compound extreme weather events and their impact on agriculture. The paper aims to: (i) characterize the exceptional climate conditions that led to the 2016 event, (ii) improve yield forecasting by identifying key predictive climate variables, and (iii) assess the future likelihood of such events.

Methods

The researchers employed a multi-faceted approach:

  • Data Analysis: They analyzed historical wheat yield data from 1959 to 2016 at the departmental level in France's breadbasket region. Climate data (temperature, precipitation, etc.) from the SAFRAN reanalysis dataset for the same period were also used.
  • Statistical Modeling: A core methodological contribution is the development of binomial logistic regression models. These models were designed to predict the probability of severe and extreme yield losses based on climate variables.
  • Variable Selection: A Bayesian Information Criterion (BIC) was used to identify the most influential climate predictors, focusing on both monthly and seasonal variables.
  • Interaction Analysis: Crucially, the models incorporated interactions between climate variables, particularly between autumn temperatures and spring precipitation, to capture compound extreme effects.
  • Model Training and Validation: Models were trained on data from 1959-2015 to predict the 2016 event (out-of-sample validation), demonstrating their predictive power.
  • Climate Projections: Future climate scenarios (RCP2.6 and RCP8.5) from CMIP5 models were used to project the frequency of the identified key climate conditions (warm autumns, wet springs) towards the end of the 21st century.

Datasets

  • Wheat Yield Data: Historical winter wheat yield data for France at the departmental level (1959-2016).
  • Climate Data: SAFRAN reanalysis data (8x8 km grid, daily time step) for France, covering October 1958 to July 2016. This dataset includes variables such as daily maximum and minimum temperatures, precipitation, solar radiation, potential evapotranspiration, and number of days with specific temperature ranges.
  • Climate Projections: CMIP5 (Coupled Model Intercomparison Project phase 5) model outputs for historical simulations and two Representative Concentration Pathways (RCP2.6 and RCP8.5) for the period 1951-2100.

Key Insights and Ideas

  • The "Compound Extreme" Revelation: The 2016 yield loss wasn't due to a single extreme, but a specific combination: abnormally warm late autumn temperatures followed by abnormally wet spring conditions. This highlights the growing importance of understanding synergistic climate impacts.
  • Vernalization and Spring Wetness Interaction: The study found a critical interaction: the negative impact of wet springs on yield is amplified when late autumn temperatures are warm, leading to fewer vernalizing days. This suggests that the "calm before the storm" (mild autumn) can set the stage for disaster when combined with adverse spring weather.
  • Predictive Power of Specific Climate Variables: The models identified a parsimonious set of key predictors (December days between 0-10°C, November precipitation, spring precipitation, and June minimum temperature) that, when combined, can effectively predict severe and extreme yield losses.
  • Future Likelihood of Compound Extremes: Climate projections indicate that the conditions conducive to the 2016 event, particularly the warm autumns, are likely to become more frequent under future warming scenarios, posing a significant challenge to agricultural stability.

Key Results

  • The 2016 wheat yield loss in France's breadbasket region was unprecedented, with yields dropping 17-45% below expected values, a magnitude not seen in over five decades.
  • A binomial logistic regression model, incorporating the interaction between autumn temperatures and spring precipitation, successfully captured historical severe and extreme yield loss events (e.g., 1976, 2003, 2016).
  • The model estimated that the odds of an extreme yield loss in 2016 were 35 times higher than expected based on historical averages, and severe yield loss odds were 1.8 to 4.6 times higher.
  • Climate projections suggest that the frequency of warm autumn conditions (specifically, fewer days between 0-10°C in December) will increase significantly by the end of the century under the RCP2.6 scenario, while spring precipitation patterns show less clear trends.

Limitations and Future Work

  • Model Scope: The study focused on the breadbasket region of France and did not attempt to extrapolate the models to neighboring countries, although the authors suggest similar agro-ecological areas might benefit.
  • CO2 Effects: The study did not address the potential impact of increased atmospheric CO2 concentration on wheat yields, which could have mitigating effects, particularly under dry conditions.
  • Data Granularity: While the models are powerful, the authors note that incorporating more spatially explicit data on soil conditions, rooting depth, and soil water-holding capacity could further improve predictions.
  • Future Work: The authors emphasize the need for farming systems and yield forecasting systems to adapt to the increasing likelihood of compound extreme events. They suggest that integrating real-time observational data (e.g., plant health bulletins, hydrological data) and exploring ensemble modeling approaches could enhance forecasting capabilities. The potential benefits of aggressive climate mitigation policies (RCP2.6) are highlighted as crucial for limiting the future frequency of these detrimental conditions.

Summary updated

Abstract

In 2016, France, one of the leading wheat-producing and wheat-exporting regions in the world suffered its most extreme yield loss in over half a century. Yet, yield forecasting systems failed to anticipate this event. We show that this unprecedented event is a new type of compound extreme with a conjunction of abnormally warm temperatures in late autumn and abnormally wet conditions in the following spring. A binomial logistic regression accounting for fall and spring conditions is able to capture key yield loss events since 1959. Based on climate projections, we show that the conditions that led to the 2016 wheat yield loss are projected to become more frequent in the future. The increased likelihood of such compound extreme events poses a challenge: farming systems and yield forecasting systems, which often support them, must adapt.
BibTeX
@inproceedings{benari2018causes,
  author = {Tamara Ben‐Ari and Julien Boé and Philippe Ciais and Rémi Lecerf and Marijn van der Velde and David Makowski},
  title = {{Causes and implications of the unforeseen 2016 extreme yield loss in the breadbasket of France}},
  booktitle = {Nature Communications},
  year = {2018},
  doi = {10.1038/s41467-018-04087-x},
}
Figure