Recognizing Sources of Forecast Uncertainty in Extreme Low-Temperature Events: The C-NFSVs Method

The combined nonlinear forcing singular vectors (C-NFSVs) method combines initial and model perturbations and accounts for the collective effect of initial and model uncertainties in ensemble forecasts through the nonlinear forcing singular vector (NFSV; also referred to as CNOP-F) approach. We apply C-NFSVs to the Weather Research and Forecasting (WRF) model and investigate the forecast uncertainty of the 2 m temperature over southern China during four major sequential periods of the 2008 extreme cold event. The results show that compared with scenarios considering only initial or model perturbations, C-NFSVs can provide more reliable ensemble forecasts. Furthermore, the C-NFSVs reveal that the 2 m temperature forecast uncertainties are predominantly sensitive to uncertainties in the upstream circulation system, whereas sensitivity to initial and model uncertainties varies across different periods of the cold event. The early period of the extreme cold event tends to propagate forecast uncertainty, as represented by the C-NFSVs ensemble spread from the upstream circulation to southern China following the background circulation. However, the forecasts of later periods present spread-characterized uncertainty that persists in the upstream circulation while remaining in contact with the background circulation and continuously propagating its effect downstream to southern China. This mechanism indicates that forecast uncertainties are dominated by initial uncertainties in the forecasts of the earlier period, whereas model uncertainties play a much more significant role in the forecasts of later periods. These findings highlight the potential of the C-NFSVs method in identifying the source of forecast uncertainty and delivering skillful forecasts for extreme cold events.