26 March 2014
Collaboration with scientists at operational and academic centres around the world is crucial for the development of the Met Office Unified Model (MetUM) — our seamless numerical modelling system. Before 1990, the Met Office used several different modelling systems for predicting the weather, including global, regional and fine mesh numerical weather prediction (NWP) and climate models. The move to a Unified Model consolidated different scales and meant that all the models could use the same code and run on our high-performance supercomputer.
Since those early beginnings, the Met Office Unified Model or 'MetUM' has grown to contain over a million of lines of code. Met Office UM Partnerships Manager, George Pankiewicz makes it possible for the MetUM suite of software to be accessed by partners across the world, both in national meteorological services and academia, who contribute to scientific and technical development.
Collaboration is key to the MetUM's ongoing success. "We simply don't have the resources and expertise to maintain and develop a world class model on our own, given the increasing level of scientific and technical effort that will be needed." explains George. Consequently the Met Office has teamed up with a number of international partners - including organisations in Korea and Australia, with India coming onboard soon - to input their expertise and enhance the model.
New consortium, new opportunities
The partnerships and the MetUM will be overseen by a new consortium headed by core partners; the Met Office, the Centre for Australian Weather and Climate Research and the Korea Meteorological Administration. The consortium has agreed that every core partner involved must assign four people a year to the scientific and technical development of the project, as well as financial support to pay for direct support and collaboration services.
"The ultimate goal is a model that's fit for purpose in many applications, from convective scale NWP in many locations around the world, through to complex global earth system modelling," explains George. "It's about building a robust system that you can use for operations but is also a research tool open to everyone."
The consortium's next goal is to consolidate all MetUM code into one repository, which all Met Office partners can access and work from online, rather than shipping out individual code across the world. Much attention is focused on the technical infrastructure, as Korea and Australia use different computer, observation and post-processing systems from the Met Office. With one consistent framework in place, setting up and using the model will become more straightforward for partners.
Lifesaving work in Australia
By bringing in partners from across the globe, the Met Office is ensuring that MetUM can handle a huge diversity of scenarios. For instance, Australia is particularly strong in tropical meteorology and the Southern Ocean, while partners in India are experts in monsoon forecasting.
The partnership with Australia began around seven years ago, when the Bureau of Meteorology (BOM) and the Commonwealth Scientific and Industrial Research Organisation (CSIRO) joined forces to look at various modelling options. The decision was taken to draw a line under their own independent developments and partner with the Met Office, with MetUM now at the heart of the Australian model.
Peter May, acting Deputy Director at the Bureau, is pleased with the progress so far. "If you look at our numerical weather prediction performance, you can see a very distinct step change in the reliability of our warnings," explains Peter. For instance, when there was severe flooding in Victoria, warnings were issued five days in advance - a timeframe that would only have been dreamt of ten years before.
There are similar advances in forecasting tropical variability, such as the monsoon onset, which has always been extremely difficult. The work is literally lifesaving. For instance, work in bush fire weather highlights the importance of local terrain effects and complex physics, with the potential to save many lives.
Blue sky thinking
Partnerships with academia in Australia are also proving extremely beneficial for the MetUM. Christian Jakob, Professor in Atmospheric Science at Melbourne's Monash University, is researching and developing atmospheric physics components for use in weather and climate prediction. He is part of the Australian Research Council's Centre of Excellence for Climate System Science, a team from five universities collaborating with the Met Office and helping to build the infrastructure. Thanks to academia's focus on research, the partnership can stretch and test the model even further than the Met Office's already rigorous standards.
"The Met Office wants to improve the model, which requires a lot of testing and often involves small, incremental steps," explains Christian. "The academic world doesn't have that constraint - so we can take more risks and engage in blue sky thinking."
Another key aspect of collaborating with academia is that it is a much more diverse community, and requires even more flexibility from the modelling system. Consequently the Met Office's Australian partners have huge experience in building infrastructure. One of the most important principles of science is reproducibility, whereby an experiment can be carried out by scientists anywhere and lead to the same verifiable conclusions. "By helping to develop a joint infrastructure we can be sure we're comparing apples with apples," says Christian.
Model development is becoming an ever more difficult task. Christian draws the analogy with a medical breakthrough, where millions of hours in the laboratory eventually lead to a small advance and the next step forward. That's what makes a consortium so attractive, as with different backgrounds, experiences and ways of looking at challenges, everyone involved is increasing the chances of the next breakthrough. "The models have been developed for thirty years, and all the easy problems have been solved," says Christian. "What's left now are the hard ones."
Where next for the MetUM?
George Pankiewicz, Peter May and Christian Jakob discuss their visions for MetUM's future.
George sees a future where more and more scientists will get involved, providing inspiring leadership and vision. Once the infrastructure work is complete, the partners will focus on more global model evaluation, including work on tropical, Southern Ocean and monsoon predictions. From air masses to land surface, from fresh water run-off to ocean temperature, there are multiple parameters in predicting the Asian monsoon, and it will take a lot of teamwork to rise to the challenge.
Peter believes that the 'big science' of weather prediction will lead to even greater achievements - saving more people's lives in the process. "One of the scientific triumphs of the twentieth century is the weather forecast," he says, "with millions of observations, using super computers that were simply unimaginable two decades ago."
Christian is hoping for 'MetUM sans frontières'. In other words, a future with seamless collaboration between science teams across the globe. So if someone is interested in mapping the monsoon, they can be part of it and make a vital contribution.
The Korea Meteorological Administration (KMA) is now a key part of the Unified Model Partnership working closely together developing the model collaboratively. Since the Koreans implemented the MetUM in 2010 it has shown excellent performance in East Asia.
As Sangwon Joo, Director of the Numerical Data Application Division at KMA explains: "We are very happy with the UM. In a recent survey on MetUM performance, Korean forecasters answered that the MetUM is much better than equivalent models over East Asia."
KMA and the academic community in Korea bring a range of skills to the partnership as part of a global modelling community focussing on specific phenomena and contribute to evaluating and improving the overall performance of the MetUM.
"The MetUM is a global model, but the Met Office has focussed on its performance over the European countries, but as a partner we can focus on the performance over East Asia, specifically typhoon prediction and heavy rainfall over the east part of the Korean peninsula. In future we will develop parts of the model, especially in terms of typhoon forecasting combined with our new forecasting technique," says Sangwon.
Importantly, KMA involvement doesn't just improve typhoon prediction over Korea but in different areas, including the Southern Oceans.
KMA not only runs a global version of the model but high-resolution versions as well. For several years, we have been able to improve the performance of the model through external evaluation but now other organisations in other parts of the world are now looking at the high-resolution models. At this year's user workshop a focus will be on joint development and how we can improve those models together through this partnership. KMA also works with the Australian Bureau of Meteorology (BOM) collaborating on how to incorporate observations into the model and developing software to use model output in different formats.
"We are trying to introduce more observations, for example from radar radiosonde, or wind profiler, into our high-resolution model system to improve its performance," explains Sangwon.
KMA scientist, Yoonjae Kim is currently working at the Met Office on the software for the observation assimilation system, working out how best to combine these with the high-resolution model to implement improvements at the Met Office that have already been made in Korea.
Similarly, Met Office Data Assimilation Scientist, Adam Clayton, is working at KMA. The KMA shares its computer resources and infrastructure making it possible to run large experiments. One element of the partnership is running a joint seasonal production system. Sharing resources in this computationally intensive area is of real benefit.
By engaging with staff, Adam understands how things work and how working practices and cultural differences vary between the two countries; all the while this serves to improve communication between the partners. The technical nature of the work, language barriers and time differences all make working in partnership difficult so these relationships are absolutely crucial to a successful partnership.
Sangwon spent two years working at the Met Office evaluating the impact of satellite data on the MetUM using the Forecast Sensitivity to Observations system, a way of working out the value of each observation, which has been implemented at KMA. "We can evaluate each individual observation to determine how it improves or degrades the forecasting system. We can then focus on certain quality control or other processes of each observation and reorganise the observation network", says Sangwon.