Climate Trends in a Seasonal Forecasting System

Deterministic forecasts produced with the second Historical Forecasting Project (HFP2) multimodel two-tier seasonal forecasting system are analyzed in order to assess the presence and importance of long-term trends on seasonal forecast skill. The results show that trends in land surface temperature, 850 hPa temperature and 500 hPa geopotential height in the National Centers for Environmental Prediction (NCEP) reanalysis data, with which the forecasts are initialized and verified, are much stronger than those in the forecasts. Since biases are removed from the forecast variables, this lack of trend is unlikely to be a consequence of model climate drift. Although the atmospheric initial conditions and oceanic boundary conditions for the forecasts contain greenhouse gas and aerosol forcing information, these forcing mechanisms are not explicitly represented in the atmospheric general circulation models with which the forecasts are made. It is possible, therefore, that long-term externally forced trends are weak or absent in the forecasts, especially over land, for this reason. The largest trends are in the December to February (DJF) season over land; their near absence in the forecasts can be at least partially overcome, and cross-validated forecast skill increased in some regions, by correcting the forecast trends a posteriori.