Improving high impact weather forecasts via an international comparison of ObServation error Correlations in data Assimilation (OSCA)

Approximately 4 million properties in the UK are at risk from surface-water flooding which occurs when heavy rainfall overwhelms the drainage capacity of the local area. In the future, as a result of climate change, the frequency and intensity of severe weather events, such as storms and floods, is likely to increase. Accurate forecasts of severe weather provide significant benefit, allowing households and businesses to take mitigating action and emergency services to mobilize resources.

Numerical weather forecasts are obtained by evolving forward the current atmospheric state using computational techniques that solve equations describing atmospheric motions and other physical processes. The current atmospheric state is estimated by a sophisticated mathematical technique known as data assimilation. Data assimilation blends previous forecasts with new atmospheric observations, weighted by their respected uncertainties. The uncertainty in the observations is not well understood, and currently up to 80% of observations are not used in the assimilation because these uncertainties cannnot be properly quantified and accounted for. Working in partnership with the UK Met Office, we have recently demonstrated in the NERC FRANC: Forecasting Rainfall exploiting new data Assimilation techniques and Novel observations of Convection project (NE/K008900/1), that it is now feasible to estimate spatial statistics for observation uncertainty. Our previous work in idealized systems has shown that better accounting for these errors in the assimilation is expected to provide significant forecast improvement. There are still a number of fundamental questions to address before the benefits can be realized in operational forecasts.

This proposal to the NERC International Opportunities fund will add value to the work carried out in FRANC, by supporting access to international observation data, numerical weather prediction models and assimilation systems. We will build a new collaboration with the Deutscher Wetterdienst (German Weather Service), and compare observation error statistics for Doppler radar wind data from Deutscher Wetterdienst with those from the UK Met Office. By considering the similarities and differences between the operational forecasting systems, and attributing these to features in the observation error statistics, we will obtain a detailed knowledge of the error sources. By carrying out theoretical and idealized studies and comparing their results with the statistics from the operational systems, we will gain understanding of the impact of differences in the assimilation systems on the diagnostic used to estimate the observation error statistics. In turn, this should allow the observation errors to be reduced, and therefore more of the expensively acquired observations to be utilised, rather than discarded. Ultimately, understanding the observation uncertainty will result in improved forecasts of severe weather events.

NATIONAL WEATHER SERVICES
In this project we will be developing a new international collaboration and working in partnership with two national weather services (Met Office, Deutscher Wetterdienst). Our results will also be disseminated to further international weather research services (see Academic beneficiaries). Our results should lead to a step-change in the understanding and treatment of correlated observation errors. This is expected to result in significant improvements in weather forecast skill, particularly for high impact weather such as storms and intense rainfall.

With an understanding of the source of observation error, it can be possible to have an operaiotnal impact on a short timescale. For example, previous research in the FRANC project diagnosed a bias in the SEVIRI satellite data assimilated by the Met Office into their UKV model. A correction was implemented by Met Office staff and made operational within a few months. To reduce other types of error, corrections may require a longer development period, or may not currently be computationally feasible within the time constraints of operational forecast production.

With the appropriate software infra-structure in place, that takes proper account of correlated observation errors, it may be possible to utilize a larger proportion of expensively acquired observations, rather than discarding them. This is expected to improve forecasts as well as providing better value for money for the various government agencies funding these weather services. We note that the UK Met Office has already begun to develop the necessary software infrastructure and we anticipate that this will be in place before the end of this project.

FLOOD FORECASTING
In the UK the Met Office/EA Flood Forecasting Centre (FFC) and Met Office/SEPA Scottish Flood Forecasting Service (SFFS) use direct output from Met Office numerical weather prediction systems to drive river routing models, coastal surge models and surface flooding tools. The output from these tools is used to provide flood warnings to Category 1 responders (fire service, police etc), and to provide driving data for Environment Agency regional offices who carry out predictions and give warnings for individual river catchments (such as those available to the public on EA floodline). The accuracy of all of these warnings is dependent on the accuracy of the numerical weather prediction inputs. Hence the expected improvements in weather forecasting skill due to this project have the potential for significant improvements in flood warnings. Since the FFC and SFFS already have Met Office numerical weather prediction outputs embedded, improvements in flood warning could be felt as soon as the improvements in weather forecasts, although some warning tools may need to be recalibrated to take account of the improved weather forecast skill. Even short lead-time flood warnings can provide useful information for the public and allow time for mitigating actions, evacuation etc. This should help to reduce injury and personal losses.

OTHER GENERAL BUSINESS USERS
The University of Reading has recently received funding from HEFCE to create a unique, flagship centre, "The Institute for Environmental Analytics". Harnessing the expertise of world-leading academics and commercial organisations (such as Airbus, Sainsburys and the Lighthill Risk Network), the Institute is developing the technologies and skills that are urgently required to translate cutting-edge environmental research into commercially-relevant solutions. Data assimilation and the use of earth observation data are expected to be a large part of the core business for the centre. The results of this project will feed into market-focussed research and development when the centre begins operations.