Computationally efficient upper atmosphere chemistry schemes for a weather and climate model

Background / Motivation

The role of the Mesosphere and Lower Thermosphere (MLT) region in coupling the lower and upper levels of the atmosphere has become increasing apparent in the past decade or so. Both the ionosphere and thermosphere are sensitive to forcing from the lower atmosphere. The seminal paper by Immel et al. (2006) indicated connections between tidal patterns in the lower thermosphere and the F region ionosphere and noted that the tidal structure was linked to patterns of convection in the equatorial troposphere. Furthermore, numerous papers (e.g., Goncharenko, Chau, et al., 2010, Goncharenko, Coster, et al., 2010; Liu & Roble, 2002; McDonald et al., 2018; Pedatella et al., 2012) have
shown how planetary wave forcing, specifically via stratospheric sudden warmings (SSWs), can affect lower thermospheric tides and thus the ionosphere.

The importance of the Mesosphere and Lower Thermosphere (MLT, 50-100 km altitude) region has become increasingly apparent in the recent decade. Both the ionosphere and thermosphere are sensitive to forcing from the lower atmosphere. MLT coupling can also affect stratospheric ozone, through chemical changes linked to energetic particle precipitation and associated transport, and the ozone changes can then affect the evolution of troposphere on seasonal to decadal timescales. Within the MLT domain itself, chemistry plays a very important role. Here, heating from exothermic reactions becomes important (especially during polar night) and must be accounted for to correctly simulate temperatures. Accurate modelling of the MLT, and of coupling above and below, thus requires an accurate modelling of the chemistry in that region.

Models that span the Earth’s surface to the MLT are used for both operational space weather forecasts and climate studies. The chemistry schemes developed for these models reflect the computational constraints of the application. Forecast models (e.g. the Whole Atmosphere Model (WAM)) use simplified chemistry schemes to enable the model to run fast enough to produce forecasts in close to near-real-time. Only relatively few major chemical reactions are sufficient to adequately represent the large rise in temperature in the MLT and so WAM focuses on a self-consistent treatment of a few major species like O, O2 and N2 while using simplified ozone chemistry and omitting minor species. Models used for climate studies (e.g. the Whole Atmosphere Community Climate Model, WACCM), on the other hand, adopt a comprehensive, more expensive chemistry scheme which includes ion chemistry and which enables, for example, studies of climate change from 1850 and of climate impacts associated with long‐term ozone change.

A limitation of this approach is that few, if any, sensitivity studies have been carried out to assess whether the WAM (WACCM) scheme would benefit from an increase (decrease) in complexity and to assess the costs and benefits of such changes in complexity to a range of different model applications. Such a study is a major recommendation from the whole atmospheric commentary paper of Jackson et al. (2019).
Simplified schemes, such as that used in WAM, is one way of reducing computational expense. Another way is to exploit recent large advances in machine learning (ML) methods to open up the possibility of emulating more comprehensive chemistry schemes. This would enable the retention of most of the accuracy and adaptability of the comprehensive schemes but at a fraction of the cost. This approach enables us to perform a whole range of experiments hitherto thought to be too expensive, such as ensemble climate simulations, or very high resolution coupled chemistry climate simulations aimed at better understanding chemistry-dynamics interactions.

The Met Office provides operational space weather alerts and forecasts. A strategic goal of the research programme that supports this service is the development of a coupled Sun-to-Earth modelling system for improved forecast capability. An important part of this coupled system is the development of an extended version of the Unified Model (UM) which extends from the Earth’s surface to the thermosphere and ionosphere. The Met Office have developed a new extended version of the UM with a top boundary near 150 km and recent work has added neutral chemical reactions relevant to the MLT (both the major reactions and sodium chemistry) to the UM scheme. Future work is planned to adapt these changes, and the results from this project, for use in the Next Generation Modelling System, which will replace the UM for Met Office weather and climate use by the mid 2020s.


• Examine current UM MLT chemistry (limited range of neutral species, parametrized ion heating) for initial assessment of accuracy compared to the comprehensive WACCM scheme.
• Extend UM scheme where necessary (e.g. adding ion chemistry) to improve its applicability and to provide a reference point for development and testing of new computationally efficient schemes.
• Use simplified versions of either or both of the UM and WACCM comprehensive chemistry schemes to complete cost / benefit analysis of different weather / climate applications.
• Use machine learning methods to develop a new chemistry scheme with much of the accuracy and complexity of the comprehensive schemes but at much reduced cost.

Additional Benefits

• The project will provide a gateway for further studies that can exploit efficient but accurate chemistry schemes (e.g. use of ensembles, impact of high resolution on transport or resolution of gravity waves).

• The project will contribute to the further development of the UM to model the thermosphere and thus to contribute to the strategic objectives of Met Office space weather research.

• The project will form part of the growing focus on Artificial Intelligence (AI)/ML within the Met Office Academic Partnership.

Potential for high-impact outcome

This is a rapidly developing area of atmospheric science, with significant interest in the potential of AI techniques for parametrisation and, as a result, there are excellent opportunities for high impact outcomes including publications in top international journals. The Met Office have a keen interest in exploring the potential for AI and ML techniques and there is potential for feeding the work through into future model development.


The project addresses important science questions at the interface of meteorology, atmospheric chemistry and space physics and will provide excellent interdisciplinary training for the student. Further interdisciplinary training comes from the comparison of data analytic (machine learning) and physics-based (pre-existing WACCM and UKCA schemes) approaches for modelling atmospheric chemistry.
The student will work under the supervision of Prof. Martyn Chipperfield, Prof. John Plane and Dr Wuhu Feng at the University of Leeds, and Prof. David Jackson and Dr James Manners from the Met Office. The supervisory team includes expertise in atmospheric chemistry, dynamics, numerical models and weather forecasting. In addition Dr Edmund Henley (MO) will provide expertise on machine learning. Co-supervision will involve regular meetings between all partners and regular visits to the Met Office in Exeter to work closely with the Met Office supervisors as part of the CASE Met Office Academic Partnership. The successful PhD student will have access to a broad spectrum of training in numerical modelling, through to managing your degree or to preparing for your viva (

This project will provide a high level of specialist scientific training in:
• Numerical modelling and use of cutting-edge supercomputers;
• Analysis of atmospheric observations and reanalysis data;
• State-of-the-science application and analysis of global atmospheric reanalysis data;
• Computer programming languages (e.g. Python) to perform complex analysis techniques;
• Effective written and oral communication skills.

The student will be part of the Composition Group, which is embedded in the Institute for Climate and Atmospheric Science (ICAS) within the School of Earth and Environment. This is a large and active group of people working on a range of problems from the surface to the upper atmosphere. Group meetings provide an excellent opportunity to discuss your work as well as to learn more about what others are working on.

Student Profile

Candidates should have an undergraduate degree (2.1 or better) in Atmospheric/Climate Science, Environmental Science, Mathematics, Chemistry, Physics, Computing, Earth Sciences or similar. A keen interest in weather/climate modelling is required, although previous experience is not required as our training will equip the student with the necessary skills.


Akmaev, R.A., Whole atmosphere modeling: Connecting terrestrial and space weather, Rev. Geophys., 49, RG4004, doi:10.1029/2011RG000364, 2011.

Eyring, V., et al., Long‐term ozone changes and associated climate impacts in CMIP5 simulations, J. Geophys. Res., 118, 5029– 5060, doi:10.1002/jgrd.50316., 2013.

Jackson, D. R., T.J. Fuller‐Rowell, D.J. Griffin, M.J. Griffith, C.W. Kelly, D.R. Marsh, and M.T. Walach, Future directions for whole atmosphere modeling: Developments in the context of space weather, Space Weather, 17, 1342-1350., 2019.

Kovacs, T., Plane, J.M.C., Feng, W., Nagy, T., Chipperfield, M.P., et al., D-region ion-neutral coupled chemistry (Sodankyla Ion Chemistry, SIC) within the Whole Atmosphere Community Climate Model (WACCM 4) – WACCM-SIC and WACCM-rSIC, Geosci. Model Dev., 9, 3123-3136, doi:10.5194/gmd-9-3123-2016, 2016.

Keller, C.A and Evans, M.J. Application of random forest regression to the calculation of gas-phase chemistry within the GEOS-Chem chemistry model v10, Geosci. Model Dev., 12, 1209–1225,, 2019.

Marsh, D.R., et al., Modeling the whole atmosphere response to solar cycle changes in radiative and geomagnetic forcing, J.Geophys. Res., 112, D23306, doi:10.1029/2006JD008306, 2007.

Marsh, D.R., Janches, D., Feng, W., and Plane, J.M.C., A global model of meteoric sodium, J. Geophys. Res., 118, 11,442–11,452, doi:10.1002/jgrd.50870, 2013.