Uncovering the past and future climate drivers of wheat yield shocks in Europe with machine learning

Zhu Peng, Abramoff Rose, Makowski David, Ciais Philippe. 2021. Uncovering the past and future climate drivers of wheat yield shocks in Europe with machine learning. Earth's Future, 9 (5):e2020EF001815, 13 p.

https://doi.org/10.1029/2020EF001815

Article de revue ; Article de recherche ; Article de revue à facteur d'impact Revue en libre accès total

Prévisualisation

Version publiée - Anglais
Sous licence

.
614325.pdf
Télécharger (2MB) | Prévisualisation

Quartile : Outlier, Sujet : GEOSCIENCES, MULTIDISCIPLINARY / Quartile : Q1, Sujet : ENVIRONMENTAL SCIENCES / Quartile : Q1, Sujet : METEOROLOGY & ATMOSPHERIC SCIENCES

Résumé : Recently, yield shocks due to extreme weather events and their consequences for food security have become a major concern. Although long yield time series are available in Europe, few studies have been conducted to analyze them in order to investigate the impact of adverse climate events on yield shocks under current and future climate conditions. Here we designated the lowest 10th percentile of the relative yield anomaly as yield shock and analyzed subnational wheat yield shocks across Europe during the last four decades. We applied a data-driven attribution framework to quantify primary climate drivers of wheat yield shock probability based on machine learning and game theory, and used this framework to infer the most critical climate variables that will contribute to yield shocks in the future, under two climate change scenarios. During the period 1980–2018, our attribution analysis showed that 32% of the observed wheat yield shocks were primarily driven by water limitation, making it the leading climate driver. Projection to future climate scenarios RCP4.5 and RCP8.5 suggested an increased risk of yield shock and a paradigm shift from water limitation dominated yield shock to extreme warming induced shocks over 2070–2099: 46% and 54% of areas were primarily driven by extreme warming under RCP4.5 and RCP8.5, respectively. A similar analysis conducted on yields simulated by an ensemble of crop models showed that models can capture the negative impact of low water supply but missed the impact of excess water. These discrepancies between observed and simulated yield data call for improvement in crop models.

Mots-clés Agrovoc : rendement des cultures, modèle de simulation, changement climatique, modèle mathématique, apprentissage machine, modélisation des cultures, sécurité alimentaire, adaptation aux changements climatiques, apprentissage

Mots-clés géographiques Agrovoc : Europe

Auteurs et affiliations