Prior to IFS cycle 47r1, the EFI (and SOT) for both CAPE and CAPE-shear were based on instantaneous CAPE (ShortName=cape; parameter ID=59) and CAPE-shear (ShortName=capes; parameter ID=228044) fields, in 6-hour steps, from which 24-hour maximum values were computed (more details can be found in ECMWF Newsletter 144). For example for the nominal T+24-48h EFI forecast, four CAPE values, at T+30h, T+36h, T+42h and T+48h, were used to compute the maximum for this forecast period. In IFS cycle 45r1 two new model output parameters, maximum CAPE in the last 6 hours (ShortName=mxcape6; parameter ID=228035) and maximum CAPE-shear in the last 6 hours (ShortName=mxcapes6; parameter ID=228036), were introduced (see details here). In IFS cycle 47r1, for the respective EFI and SOT computations, mxcape6 and mxcapes6 will be used instead of cape and capes respectively. This change is aiming for a better sampling in the computation of the 24-hour maxima needed for the EFI. In effect with the change we extract the maximum within 24 hourly values, instead of using 4 6-hourly values. The impacts of this change are described in detail below.
Firstly this changes has a neutral impact in terms of the EFI skill measured by the area under the Relative Operating Characteristic (ROC) curve.
The EFI fields using mxcape6 and mxcapes6 generally look smoother and in the case of a tornadic outbreak in the USA earlier in March 2020 shown in Fig. 1 they even fit better to the severe weather reports. Please note that the EFI shows a wide area where the environment favours convective hazards whilst the actual severe thunderstorms develop along relatively narrow bands where sufficient lift is present to initiate deep, moist convection, e.g. along cold fronts or dry lines.
Fig. 1. T+24-48h CAPE-shear EFI/SOT forecast during the tornadic outbreak on 3rd March 2020 over Tennessee, USA.
As a result of taking instantaneous cape and capes at discrete time steps the EFI can exhibit a stripy structure in a case of fast moving squall fronts as in the example in Fig. 2. Apparently this is an issue because the cold front is not jumping rather, it is moving continuously from the west to the east. Replacing cape and capes with mxcape6 and mxcapes6 removes the stripy behaviour and provides a smoother and more realistic forecast field due to better sampling.
Fig. 2. The EFI for CAPE-shear as well as CAPE-shear high-resolution forecast in a case of a fast moving cold front. Stripy fields of instantaneous CAPE -shear and the EFI is due to taking instantaneous CAPE-shear values at discrete time steps. The EFI forecast looks smoother when using the smoother fields of maximum CAPE-shear in 6-hour periods. The sequence of Air mass RGB imagery shows snapshots of the cold front approaching the south-western parts of the Iberian Peninsular.
The EFI is a climate-related product which provides information how different cumulative distribution functions (CDFs) of the ensemble forecast and the model climate are. Therefore, an anomalous forecast does not necessarily translate into a high-impact one. In case of convection this means that anomalous values of CAPE does not necessarily suggest that convective hazards are likely. For example climatological values of CAPE in the cold season over the Arctic are so low that severe convection is impossible even if the forecast shows up extreme CAPE. To filter out anomalous but insignificant signals in the EFI, currently CAPE values less than 10 J.kg-1 are set to 0 before computing both model climate and ensemble forecast CDFs. Now with replacing cape with mxcape6 and capes with mxcapes6, CAPE over 10 J.kg-1 will happen more often and therefore this may show up on the EFI (Fig. 3). To correctly interpret the EFI one should always account for the model climate as well.
Fig. 3. The EFI for CAPE and maximum CAPE in the Siberia showing that the latter may show up a bit more often.
Finally, as a result of this change, the model climate globally also slightly changes (Fig. 4a). In the example in Fig. 4b the EFI suggests a bit more extreme CAPE values over NW Spain. The CDFs for a representative location there show what this signal corresponds to in terms of model climate and ensemble forecast distributions.
a)
b)
Fig. 4 a) Model climate 99th percentile for cape and mxcape6 valid for the end of April and b) an example of the EFI for cape and mxcape and CDFs of corresponding model climate and real-time ensemble forecasts.