Influence of climate variability on the potential forage production of a mown permanent grassland in the French Massif Central

Abstract

Climate Services (CS) provide support to decision makers across socio-economic sectors. In the agricultural sector, one of the most important CS applications is to provide timely and accurate yield forecasts based on climate prediction. In this study, the Pasture Simulation model (PaSim) was used to simulate, for the period 1959-2015, the forage production of a mown grassland system (Laqueuille, Massif Central of France) under different management conditions, with meteorological inputs extracted from the SAFRAN atmospheric database. The aim was to generate purely climate-dependent timeseries of optimal forage production, a variable that was maximized by brighter and warmer weather conditions at the grassland. A long-term increase was observed in simulated forage yield, with the 1995-2015 average being 29% higher than the 1959-1979 average. Such increase seems consistent with observed rising trends in temperature and CO2, and multi-decadal changes in incident solar radiation. At interannual timescales, sea surface temperature anomalies of the Mediterranean (MED), Tropical North Atlantic (TNA), equatorial Pacific (El Ni\~no Southern Oscillation) and the North Atlantic Oscillation (NAO) index were found robustly correlated with annual forage yield values. Relying only on climatic predictors, we developed a stepwise statistical multi-regression model with leave-one-out cross-validation. Under specific management conditions (e.g., three annual cuts) and from one to five months in advance, the generated model successfully provided a p-value<0.01 in correlation (t-test), a root mean square error percentage (%RMSE) of 14.6% and a 71.43% hit rate predicting above/below average years in terms of forage yield collection.