Copper Content and Export in European Vineyard Soils Influenced by Climate and Soil Properties

Copper-based fungicides (Cuf) are used in European (EU) vineyards to prevent fungal diseases. Soil physicochemical properties locally govern the variation of the total copper content (Cut) in EU vineyards. However, variables controlling Cut distribution at a larger scale are poorly known. Here, machine learning techniques were used to identify governing variables and to predict the Cut distribution in EU vineyards. Precipitation, aridity and soil organic carbon are key variables explaining together 45% of Cut distribution across EU vineyards. This underlines the effect of both climate and soil properties on Cut distribution. The average net export of Cu at the EU scale is 0.29 kg Cu ha-1, which is 2 orders of magnitude less than the net accumulation of Cu (24.8 kg Cu ha-1). Four scenarios of Cuf application were compared. The current EU regulation with a maximum of 4 kg Cu ha-1 year-1 may increase by 2% of the EU vineyard area, exceeding the predicted no-effect concentration (PNEC) in soil in the next 100 years. Overall, our results highlight the vineyard areas requiring specific remediation measures and strategies of Cuf use to manage a trade-off between pest control and soil and water contamination.B.D. was supported by a fellowship of the Region Grand Est and the Rhine-Meuse Water Agency (AERM). P.B. was funded by the EcoSSSoil Project, Korea Environmental Industry & Technology Institute (KEITI), Korea (Grant No. 2019002820004). The authors acknowledge the copper data set owners for sharing their data: Holger Tülp (Mosel, Germany), Denis Rusjan (Slovenia), Maria Concepcion Ramos (Spain), Nikolai Dinev (Bulgaria), several Swiss state environmental agencies (Schaffhausen, Neuchâtel, Genève, Tecino, Aargau, and Valais), the Central Institute for Supervising and Testing in Agriculture (ÚKZÚZ, Czech Republic), Bundesanstalt für Geowissenschaften und Rohstoffe (Germany), le Réseau de Mesures de la Qualité des Sols (RMQS), and de l’Institut National de Recherche pour l’Agriculture, l’Alimentation et l’Environnement (INRAE, France). The authors acknowledge Koens Oorts for sharing the toxicological model. The LUCAS topsoil data set, CORINE Land Cover, and Copernic data were funded by the European Environment Agency (EEA). MODIS data were supported by the NASA EOSDIS Land Processes Distributed Active Archive Center (LP DAAC) and the USGS/Earth Resources Observation and Science (EROS). The authors acknowledge Dimitri Rambourg, Raphaël di Chiara, Marwan Fahs, Gerhard Schäfer, and Benoit Masson-Bedeau for enabling computing time. The authors gratefully acknowledge Margaret Johnson for enabling significant improvement of the manuscript. Finally, the authors thank the two anonymous reviewers and Guillaume Drouin for useful comments.Peer reviewe