Utility of dynamical seasonal forecasts in predicting crop yield

Abstract

Advance predictions of crop yield using crop simulation models require daily weather input for the whole growing season. Seasonal forecasts, based on coupled ocean–atmosphere climate models, are now available up to 6 mo in advance from a number of operational meteorological centres around the world. Seasonal forecasts are not directly suitable for crop simulations, because of model biases and mismatch of spatial and temporal scales. However, it is possible to utilise seasonal forecasts for yield predictions by constructing site-specific daily weather using a stochastic weather generator linked to seasonal forecasts. In our study, we use the LARS-WG weather generator and a subset of predictions by DEMETER (Development of a European Multimodel Ensemble system for seasonal to inTERannual climate prediction), i.e. seasonal ensemble hindcasts from the general circulation model (GCM) of ECMWF (European Centre for Medium-range Weather Forecasting) for 1980–2001. To assess the value of seasonal forecasts, 2 sets of scenarios were created, one based on seasonal forecasts and the other on historical climatology. The Sirius wheat simulation model was used to compute distributions of wheat yield at 2 locations in Europe and New Zealand. The main conclusion is that the use of dynamical seasonal forecasts at selected sites has not improved yield predictions compared with the approach based on historical climatology. The likely reason is that for dynamic seasonal forecasts, the skill score for temperature and precipitation is generally low for latitudes higher than 30° for northern and southern hemispheres, and our test locations are at 47.6°N and 43.6°S.