Evaluating the impact of prognostic ozone in IFS NWP forecasts

The forecast impact of a recently developed, novel prognostic ozone scheme, interactive with radiation, is here evaluated within the ECMWF Integrated Forecasting System (IFS) on both medium-range and sub-seasonal timescales. The importance of radiatively interactive ozone has been demonstrated in terms of accurate representation of stratospheric heating rates, which has direct implications for the predictability of temperature and wind. At ECMWF, the capacity to implement prognostic ozone interactive with radiation, on an operational forecast basis, has existed within IFS for several years. However, such usage has been inhibited by insufficient signal for an overall skill improvement. The new Hybrid Linear Ozone (HLO) scheme, evaluated here, simulates the impact of chemistry on the tendency of ozone, using a blend of inputs from analyses and chemistry model calculations, in a realistic and affordable manner. The scheme has been developed to address a number of shortcomings associated with earlier linear ozone schemes, such as the widely used Cariolle parametrization. With the aid of different established verification metrics, the forecast skill impact of implementing radiatively interactive HLO is critically evaluated, in regard to both changes to the variability aspect of skill (e.g. pattern correlations) and the mean bias. A mixture of individual and aggregate (e.g. multi-month) cases are examined, using a combination of deterministic and ensemble forecasts (comparing interactive and control experiments). Large changes in stratospheric ozone in association with midwinter Sudden Stratospheric Warmings (SSWs) additionally provide natural candidates for verification of the HLO scheme performance, particularly as such events are known to impact tropospheric weather, and are thus a key focus of this report. We find a clear overall improvement in the representation of forecast temperature, wind and geopotential height in 15-day forecasts in the stratosphere, compared to using an ozone climatology, supporting the inclusion of radiatively interactive HLO within the IFS medium-range forecasting system. In contrast, implementation of HLO, interactive with radiation, over longer sub-seasonal timescales does not appear to improve sub-seasonal forecasts following SSWs. A key limitation of this focus however is the already enhanced forecast skill following these events, as the stratospheric state is inherently more predictable, and the performance of radiatively interactive HLO during more typical wintertime conditions has not here been assessed. Nevertheless, radiatively interactive HLO is demonstrated to yield enhanced lower stratospheric forecast temperature skill following SSWs, in conjunction with the persistence of a long-lived positive anomaly in ozone. Detailed investigation of the evolution in radiative heating tendencies following the major January 2009 SSW provides insight into the mechanisms of the response to radiatively interactive HLO and the reasons why such implementation does not necessarily translate to forecast skill improvement. The findings are used to highlight a series of recommendations for further verification and refinement of the HLO scheme.