Research: Forecast skill of numerical weather prediction models
Forecast skill for Atlantic-European weather regimes in sub-seasonal models (Dominik Büeler)
Daily weather can be relatively well predicted up to two around 10 days ahead. Beyond this time, or more precisely on sub-seasonal time scales (10 – 60 days), the prediction skill for daily weather substantially drops as the large-scale atmospheric flow is highly non-linear. Instead, it is possible to predict quasi-stationary, persistent, and recurrent large-scale flow patterns, so-called weather regimes, which explain most of the regional surface weather variability on sub-seasonal time scales. Weather regimes are thus of primary interest in sub-seasonal regional forecasting – not just for weather services but also for socio-economic sectors such as the energy industry or agriculture.
In this project, we assess the skill of sub-seasonal numerical weather prediction models in predicting 7 year-round weather regimes in the Atlantic-European region. To this end, we analyze sub-seasonal reforecasts from 11 different global modeling centers, obtained from the Subseasonal-to-Seasonal (S2S) Prediction Project Database. Our primary focus is on how forecast skill differs for different flow situations and seasons. Moreover, we investigate how forecast skill is modified by lower-frequency climate modes such as the stratospheric polar vortex, the Madden-Julian Oscillation, or the El Niño-Southern Oscillation. In addition to these dynamical questions, we elucidate how different kinds of model calibration techniques affect forecast skill. This project is strongly linked to the project “Influence of WCBs on forecast skill for Atlantic-European weather regimes in sub-seasonal models”, in which we investigate to what degree warm conveyor belts and their representation in models affects flow-dependent sub-seasonal forecast skill.
Figure 1: Qualitative depiction of sources of weather forecast skill on different time scales. In this project, we investigate the forecast skill for Atlantic-European weather regimes and how it depends on both synoptic-scale activity as well as lower-frequency climate modes. Source: White et al. (2017), https://doi.org/10.1002/met.1654.
Rossby wave packets in sub-seasonal models (Julian Quinting)
Equatorward and poleward perturbations propagating eastward along the jet streams in the upper troposphere are commonly referred to as Rossby wave packets (RWPs). These packets have been linked to extreme weather events such as winter storms, extreme temperatures, and precipitation. Hence, an adequate representation of RWPs instate-of-the-art weather prediction models is desirable to better predict these events.
For a large set of sub-seasonal forecast data, we verify fundamental properties of RWPs such as their climatological frequency of occurrence, their life time, and their mean propagation distance. Sub-seasonal models—especially those with a rather coarse horizontal grid spacing—struggle to adequately represent the decay of these waves in the Atlantic/European sector. Instead of decaying over the eastern North Atlantic, RWPs propagate into far eastern Europe likely due to an underestimation of the occurrence frequency of long-lasting and stationary high-pressure systems.
Figure 2: An illustrative RWP over the North Pacific in November 2014. Source: Julian Quinting.
Influence of WCBs on forecast skill for Atlantic-European weather regimes in sub-seasonal models (Jan Wandel)
Warm Conveyor Belts (WCBs) occur along the cold front of extratropical cyclones and can be decribed as regions with ascending airstreams and strong diabatic forcing. Due to moist processes in the ascent region, WCBs can be associated with increased error growth in numerical weather forecast models. This is especially important since WCBs are an important componant of synoptic-scale activity and influence the onset, maintenance, and decay or transititions of so called weather regimes. Recent case studies show that the misforecast of the onset of blocked regimes over the European continent can be linked to a wrong representation of WCBs (and especailly the outflow phase of the WCB) in the ensemble forecast.
This PhD project investigates the overall predictability and forecast skill of WCBs in current numerical weather prediction models that are available through the Subseasonal-to-Seasonal (S2S) Prediction Project Database. In order to identify WCBs in the forecast model, we use a recently developed global multivariate logistic model that is trained on a combination of meteorological parameters from ERA-Interim and which is designed for the inflow, ascent and outflow phase of WCBs. Furthermore, we link the WCB forecast to weather regime forecast and assess how the representation of diabatic outflow affects flow dependent forecast skill of the regimes and how well WCBs can be predicted in different regimes. This will give hint, if, how and when diabatic outflow dilutes forecast skill for the midlatitude flow on subseasonal timescales.