Impact case study database

Implementing our model parameterization

The impact of our research has been to improve a number of leading weather and climate models that are used for operational weather forecasting and for climate projections, contributing to a significant improvement in the accuracy of global weather forecasts. Our most significant relationship has been with the Met Office, but we have also worked with the European Centre for Medium-range Weather Forecasts (ECMWF; a multi-national agency primarily funded by European nations); and have recently started working with Météo-France (France’s national weather and climate agency). These are three of the world’s leading forecasting agencies and climate model developers. The Met Office’s forecasts are the second most accurate globally – according to agreed World Meteorological Organization (WMO) statistics – while the ECMWF’s forecasts are the most accurate globally (see MOSAC 2019; S1). The WMO is a specialized Agency of the United Nations.

We have collaborated with scientists at the Met Office from 2007-present . Between 2007-2014, this focussed on testing different settings for the existing surface exchange scheme for sea ice used in the Met Office Unified Model. From 2015, we collaborated on implementing and then testing a new surface exchange scheme based on prior theory and our observations [R4]. This collaboration has involved numerous visits to the Met Office, direct correspondence with key staff and the employment of Andrew Elvidge (UEA senior research associate from 2013-2015; 2017-present) at the Met Office from 2015-2018. Dr Elvidge worked with researchers at the Met Office to implement our surface exchange scheme in version 10.5 of the Met Office Unified Model [S2].

Impact on operational weather forecast accuracy

Our surface exchange scheme became operational at the Met Office on 25^th September 2018 as part of the ‘Parallel Suite 41’ (PS41) upgrade. From this date the Met Office’s global operational weather forecasts make use of our new surface exchange parameterization. For several months prior to this date the PS41 trial forecasts were compared to the Operational Suite 40 forecasts to ascertain if the new configuration provided more accurate forecasts (e.g., Figure 1a). The PS41 configuration gave an overall improvement in the deterministic global Numerical Weather Prediction ( NWP) Index of 1.6% (Figure 1a; S3). The NWP Index is a gross metric for forecast skill that is specified by the WMO. Note this was

“one of the largest improvements [in forecast skill] of the last decade” [S3].

We cannot attribute all of this forecast improvement to the new surface exchange scheme, because other changes to the forecasting system were also made during this upgrade. However separate ‘paired trials’ were also carried out where the only significant change to the model was to the sea-ice surface exchange scheme (Figure 1b; S3). In these trials there were significant improvements in several benchmark atmospheric quantities: a 1-2% improvement in mean-sea-level pressure and upper-troposphere wind in both the Southern and Northern Hemisphere extra-tropics, as well as a 5% and 1-2% improvement in mid-tropospheric geopotential height for the Southern and Northern Hemisphere extra-tropics. In total these constitute an improvement of 0.7% of the NWP index (Figure 1b; **[S3]**), equivalent to almost half of the PS41 improvement.

In addition to weather forecasts, the Met Office Unified Model is also a component in all of their climate models. Our new scheme was incorporated into the ‘Global Land 8’ (GL8) climate configuration in 2018, so will have an impact on future climate projections and predictions, including those made for future IPCC assessments.

Figure 1 | Met Office model performance [S3] illustrating the change in RMSE (root mean square error) for several meteorological quantities at different forecast times (in hours). Here green shows an improvement and the area shaded is the magnitude of the improvement (the maximum shading is a 5% improvement). The quantities are mean-sea-level pressure (PMSL), geopotential height at the 500 hPa level (H500) and the wind at the 250 hPa or 850 hPa level for the Northern and Southern Hemispheres (NH and SH) or Tropics (Trop). A weighted combination of these quantities makes up the ‘NWP index’. Panel (a) shows the change in model performance against observations of the new forecasting set up (PS41 – Parallel Suite 41) compared to the previous operational suite (OS40); results here are for a four-month trial that had showed an overall NWP index improvement of 1.59%. This improvement included the surface exchange parameterization change, but also included other changes to the model set up. Panel (b) shows the change in model performance for a so-called paired trial experiment, where only the surface exchange parameterization is different; results here are for a two-month period and indicate an overall NWP index improvement of 0.72%. There are dramatic improvements in the winter hemisphere (SH) for H500 and also for PMSL.

The Met Office summarise in their evidence letter [S3]

“It is worth stressing that the implementation of a new surface exchange parameterization for over sea ice would not have occurred without the research led by Prof. Renfrew. Furthermore, we are pleased to confirm that this parameterization led directly to a significant improvement in our weather and climate prediction modelling systems locally and globally, making a contribution to our world-leading forecast system and climate modelling capabilities.”

Coincident with our research, staff at the ECMWF were developing a new surface exchange scheme for sea ice and our research informed this development. We presented our provisional observational results from our Arctic field campaign [R4] to staff at the ECMWF during a visit there in September 2014. These provided evidence that supported their proposed new sea-ice drag parameterization and consequently supported the changes they were planning to make [S4]. The ECMWF note

*“The knowledge exchange with Prof Renfrew was extremely useful in making decisions. In fact, we changed the surface exchange of momentum over sea-ice parameterization for Cycle 41R1 of the IFS [Integrated Forecasting System], which became operational on 12 May 2015.” * [S4]

Our research is now also having an impact at other leading forecasting agencies. In summer 2020 Météo-France requested our observations and, in a joint meeting, a leading Météo-France researcher, outlined their plans to us to develop a surface exchange parameterization for their suite of weather and climate models that will be guided by our research [R4, R5].

Economic worth of our model improvement

It is difficult to quantify the economic value of meteorological model upgrades. Improvements in meteorological forecasts allow safer and more efficient activities throughout society. Both the Met Office and ECMWF provide hundreds of global forecast products every day to nations, businesses and the general public. It has been calculated that the economic benefits of the Public Weather Service for the UK, which is delivered by the Met Office, are

“very likely to exceed GBP1,000,000,000 per annum and are likely to be close to GBP1,500,000,000 per annum” (Met Office 2016 ; S5).

Consequently, any improvement to these services contributes to this economic benefit. The forecast improvement we contributed to has been one of the largest in the last decade and so our contribution towards delivering these improved services has been significant.