Atmospheric Composition priority developments for Numerical Weather Prediction

One of the ECMWF’s strategic goals for 2025 is to develop an integrated global model of the Earth system to produce forecasts with increasing fidelity up to a year ahead. This will be achieved by incorporating an increased level of complexity of physical and chemical processes and of the interactions between the different Earth’s components into the model. Atmospheric composition (AC) has the potential to be one of the sources of predictability at different time scales. Development activities at ECMWF have built a capacity to simulate and assimilate a variety of AC species in the Integrated Forecasting System (IFS). Today this capacity is at the core of the Copernicus Atmosphere Monitoring Service (CAMS), but the complexity of many AC modules in the IFS often makes them computationally unaffordable at the resolutions used in numerical weather prediction (NWP). Only stratospheric ozone is a prognostic variable in NWP applications, and it is included interactively in the radiative transfer model used for radiance assimilation. The radiation scheme still relies on climatologies of aerosols, ozone and trace gases, although considerable collaborative effort has been made in the last few years to use the CAMS products to improve the realism of these AC climatologies and of the related AC-weather feedbacks.

This document recommends AC priority developments for NWP that has the potential to further improve the forecasts from days to seasons ahead. These are considered for testing and possibly operational implementation by 2022. The proposal is to improve the representation of AC in NWP to the level of complexity and coupling beneficial for NWP by leveraging the CAMS developments wherever possible. Hence, the focus is on ozone, aerosols, and CO₂. In addition to modelling and data assimilation aspects, attention will also be paid to code structure and code efficiency, and, provided a higher level of coupling between AC and meteorology is introduced, to how the model performance evaluation process might need to be adapted in the future.