Changes in CIN to concur with forecasting practice

In IFS cycle 47r1 the following two changes are implemented for the Convective Inhibition (CIN):

• CIN is computed with a completely revised code using virtual temperature correction;
• CIN and Convective Available Potential Energy (CAPE) both refer to the same (most unstable) parcel curve. 

Details of these changes are provided below.

In  parcel theory, CAPE and CIN are computed as vertical integrals of Buoyancy expressed in terms of the virtual temperature (or virtual potential temperature) difference between the lifted air parcel and its environment (see Fig.1 on the right). In IFS cycles up to and including 46r1, for computational efficiency, CAPE and CIN are approximated using the difference between the equivalent potential temperature of the parcel, which is conserved during pseudoadiabatic ascent (the condensate is removed immediately from the parcel as it forms), and the environmental saturation equivalent potential temperature, which is a function of environmental temperature only (for details see this technical memorandum). This approximation provides a relatively good estimate for CAPE but it can massively overestimate CIN. In practical terms, this means that the model output can suggest no chance of thunderstorm initiation whilst according to parcel theory CIN is so small that thunderstorm initiation is very likely. To correct this deficiency in the CIN output from the model, its computation is completely changed in IFS cycle 47r1. In new computational code, CIN is estimated using the difference between the virtual potential temperatures of the parcel and the environment, exactly as it is in parcel theory. This provides an estimate for CIN which is much more in line with parcel theory and forecast practice. Please note that in the CAPE and CIN provided from the IFS the surface parcel is not considered. Instead, for all the model levels in the lowest 60 hPa mixed layer parameters are used. This is in line with the notion that the updraught in thunderstorms will probably involve a deeper layer near the surface rather than just the surface air parcel. Please note that the CAPE computation is not changed in cycle 47r1 because of the implications that such a change may have on users, e.g. for re-forecasts and for the EFI. Instead, a set of new CAPE and CIN parameters are under preparation including most-unstable and mixed-layer CAPE/CIN which all will use the virtual temperature correction. These will become available later in an IFS cycle after 47r1.

In addition, CAPE and CIN both represent the most unstable parcel in cycle 47r1. In earlier cycles, CAPE represents the maximum value encountered when considering parcel curves departing from each model level, from the surface up to 350 hPa, whilst CIN represents the minimum encountered amongst those curves. As a result CIN and CAPE can represent different air parcels, which makes interpretation more difficult. In cycle 47r1 both CAPE and CIN refer to the same (most unstable) parcel which improves usability for diagnosing deep moist convection.

The plot on the right shows a Skew T - log p diagram. The dashed black curve represents the temperature curve for a mixed-layer parcel lifted from the lowest 30 hPa layer, as considered in the model. The red area between the level of free convection (LFC) and the equilibrium level (EL) is proportional to CAPE whilst the blue area under the LFC is proportional to CIN. Apparently, this is a quite unstable profile with some CIN which has to be overcome for the parcel to reach LFC. In the table CAPE and CIN values are displayed using different options for computing buoyancy. Prior to cycle 47r1, while the environment is very unstable (CAPE of the order of few thousands of J.kg^-1 ) the CIN is so large that - conceptually - no thunderstorms could really be initiated. In fact the true CIN is much lower as the thermodynamic diagram shows (only a few tens of J.kg^-1). In IFS cycle 47r1 CIN changes significantly across the Globe, to now be in accordance with parcel theory.

The old and new CIN forecasts are compared on the charts in Fig.2. CIN values over 50 J.kg^-1 are shaded in semi-transparent grey, to mask CAPE. From these plots it is apparent that 47r1 CIN is significantly lower than 46r1, in many areas such as Eastern Europe, France and parts of the USA, giving better guidance that convection/thunderstorms are likely to initiate in those regions.

Fig.1. Skew T - log p diagram showing a lowest 30-hPa mixed layer parcel curve (dashed black curve), the environmental temperature curve (red curve) and environmental dew point curve (blue curve). CAPE is proportional to the area in red, and CIN is proportional to the area in blue under the LFC. CAPE and CIN computations are also shown according to parcel theory (note that the overbar denotes the environment) and approximations used in different IFS versions. The change in the CIN computation is also highlighted. In this case the CIN computation prior to cycle 47r1 massively overestimates the CIN while with 47r1 its value is much closer to the parcel theory. For reference see the table below.

Table 1. CAPE and CIN values for the parcel shown on the Skew T - log p diagram computed with different approximations. Please note that CAPE is lower and CIN higher in the case when virtual temperature corrections are not used. The last row shows the IFS CAPE and CIN prior to cycle 47r1 - apparently, CIN is massive (shown in the red cell) compared to parcel theory estimation. With cycle 47r1 forecasters will get the values in green - no change in CAPE for the time being whilst CIN will generally change significantly providing the best estimate according to parcel theory.

Fig. 2. CIN (values above 50 J.kg^-1 shaded in semi-transparent grey) and CAPE (colour shading) comparing 47r1 ("NEW") with 46r1 ("OLD"). Significantly more areas with an unstable air mass, e.g. CAPE>100 J.kg^-1 are outside the grey mask, giving much improved guidance to forecasters on where convection/thunderstorms are likely to initiate. Note that the subtle differences also apparent in the CAPE fields are due to differences in the actual forecasts, as the CAPE formulation is unchanged.