Evaluating decadal predictions of northern hemispheric cyclone frequencies

Abstract

A B S T R A C T Mid-latitudinal cyclones are a key factor for understanding regional anomalies in primary meteorological parameters such as temperature or precipitation. Extreme cyclones can produce notable impacts on human society and economy, for example, by causing enormous economic losses through wind damage. Based on 41 annually initialised (1961Á2001) hindcast ensembles, this study evaluates the ability of a single-model decadal forecast system (MPI-ESM-LR) to provide skilful probabilistic three-category forecasts (enhanced, normal or decreased) of winter (ONDJFM) extra-tropical cyclone frequency over the Northern Hemisphere with lead times from 1 yr up to a decade. It is shown that these predictions exhibit some significant skill, mainly for lead times of 2Á5 yr, especially over the North Atlantic and Pacific. Skill for intense cyclones is generally higher than for all detected systems. A comparison of decadal hindcasts from two different initialisation techniques indicates that initialising from reanalysis fields yields slightly better results for the first forecast winter (month 10Á15), while initialisation based on an assimilation experiment provides better skill for lead times between 2 and 5 yr. The reasons and mechanisms behind this predictive skill are subject to future work. Preliminary analyses suggest a strong relationship of the model's skill over the North Atlantic with the ability to predict upper ocean temperatures modulating lower troposphere baroclinicity for the respective area and time scales.

Decadal predictability of regional scale wind speed and wind energy potentials over Central Europe

“…Kruschke et al, 2014Kruschke et al, , 2015. This systematic bias is both dependent on the model generation and forecast time.…”

“…Therefore, a positive MSESS suggests that the initialised hindcasts are more accurate in representing the observed decadal climate variability than the uninitialised historical runs (Goddard et al, 2013), and a negative value indicates the opposite. The probabilistic RPSS (Wilks, 2011;Kruschke et al, 2014) is defined as…”

“…The RPS is an extension of the Brier score (scalar accuracy measure for binary events) to multi-category forecasts (Wilks, 2011). Following Kruschke et al (2014), three categories are used here for the calculation of RPS: below normal, normal and above normal. The categories are defined using the 33.3 and 66.6 percentiles of Eout and wind speed anomaly time series.…”

“…A positive RPSS therefore indicates that the initialised hindcasts have a higher probability to predict an observed anomaly category than the uninitialised historical runs, and vice versa for a negative RPSS. Following Kruschke et al (2014), we corrected the RPS for biases due to finite ensemble sizes (see also Ferro, 2007 …”

“…So far, only few studies investigated decadal predictions on the regional scale, and to our knowledge, none is dealing with wind energy. Kruschke et al (2014), for example, analysed the decadal forecast skill for cyclone activity over the Northern Hemisphere in the MPI-ESM and found some regions over the North Atlantic with positive predictive skill for intense cyclones. Mieruch et al (2014) investigated the decadal forecast skill for seasonal temperature anomalies and precipitation sums in dynamically downscaled MPI-ESM hindcasts, focusing on Europe.…”