cover art for AI Special Episode 3: ECMWF's strategy for using AI predictions

The WeatherPod

AI Special Episode 3: ECMWF's strategy for using AI predictions

The guest this episode is Florian Pappenberger, Deputy-Director General & Director of Forecasts at the European Centre for Medium-Range Weather Forecasts, the ECMWF.

Hosts David Rogers and Alan Thorpe delve into the ECMWF's use of AI and how this may impact it's future products and services, especially when it comes to their use in the developing countries by meteorological services and other customers.

Florian leads the ECMWF's Forecast Department which is responsible for the production of weather forecasts, forecast quality control, and the development of novel forecast products.

ECMWF has for a while been experimenting with using deep learning to produce a data-based AI weather prediction model and last October Florian played a key role in the launch of the Artificial Intelligence / Integrated Forecasting System. This is the ECMWF's first forecasting system incorporating a machine learning prediction module.

More episodes

View all episodes

  • 2. AI Special Episode 2: Trustworthy Artificial Intelligence

    In this special AI Episode, hosts Alan Thorpe and David Rogers invite Amy McGovern into the studio to discuss the meaning of "trustworthy AI".Amy is Director of the National Science Foundation AI Institute for Research on Trustworthy AI in Weather, Climate, and Coastal Oceanography - or AI2ES for short.She's also a Professor at Oklahoma University's School of Computer Science and School of Meteorology.Working under the University of Oklahoma's leadership, AI2ES brings together researchers in AI, atmospheric science, ocean science, and risk communication. The thinking is that accelerated AI research in the environmental sciences can improve understanding of the rapid changes taking place in weather patterns, oceans, sea level rise, and disaster risk. Amy's research focuses on developing and applying machine learning and data mining methods for real-world applications, with a specific interest in high-impact weather. Much of this work involves weather analytics or physical data science and she and her students are developing physics-based trustworthy AI methods as well as explainable AI. Their aim is to apply their work to high-impact weather phenomena, including tornadoes, hail, severe wind events, flooding, drought, and aircraft turbulence.A key aim is to help build a diverse and flexible science, technology, engineering, and mathematics workforce. Amy's thinking is that diversity will bring new ideas to the forefront, while flexibility is crucial to dealing with rapid changes in technology. To help this process, Amy and her team have developed outreach projects to encourage students to pursue STEM careers.This work aside, Amy also directs the Interaction, Discovery, Exploration and Adaptation - or IDEA - Lab at Oklahoma University. The Lab's focus is on developing and applying data science, AI and machine learning techniques for high-impact real-world applications.
  • 27. Episode 27: A Global Interactive Forecasting System

    In this episode of The WeatherPod, hosts Alan Thorpe and David Rogers invite David Richardson of the European Centre for Medium-Range Weather Forecasts - the ECMWF - into the studio.About David RichardsonDavid Richardson has over 35 years’ experience in weather forecasting research and operations at the UK Met Office and, until recently, as Head of Evaluation at the European Centre for Medium-Range Weather Forecasts or ECMWF. His work on ensemble prediction includes the configuration of ensembles to represent the uncertainties in the initial conditions and modelling systems, development of products and tools for forecast users, and evaluation of forecast performance.  Previously, he worked for the WMO where he played a key role in establishing the THORPEX Interactive Grand Global Ensemble or TIGGE project on global ensemble forecasting - a project which led to improved understanding of predictability and also to improvements in forecasting skill. David has chaired several WMO committees, and is currently chair of the WMO Standing Committee on Data Processing for Applied Earth System Modelling, Prediction & Projection (SC-ESMP). SC-ESMP leads and co-ordinates the WMO Integrated Processing and Prediction System (WIPPS) which facilitates the provision of numerical earth-system model forecasts, including ensembles, to all WMO members.
  • 1. AI Special Episode 1: Using AI for weather forecasting

    In this special AI episode of The WeatherPod, hosts Alan Thorpe and David Rogers open a new series of discussions which focus specifically on the use of Artificial Intelligence across the weather enterprise.In this first discussion, our hosts invite Professor Kirstine Dale of the UK Met Office into the studio.About Kirstine DaleKirstine Dale is the Met Office's Chief AI Officer (CAIO) and Principal Fellow for Data Science. As CAIO she is charged with embedding AI in the Met Office’s core business - initially focusing on the use of AI in weather forecasting through leadership of the ‘AI for Numerical Weather Prediction’ (AI4NWP) programme. As Principal Fellow, Kirstine plays a leading role in shaping the future of Data Science (including Artificial Intelligence and Machine Learning) in the Met Office. Kirstine is also Co-Director of the Natural Environment Theme of the Turing Research and Innovation Cluster on Digital Twins (TRIC-DT) and an Honorary Professor at the University of Exeter.
  • 26. Episode 26: Anticipatory Action

    In this special episode of the WeatherPod, hosts Alan and David have invited into the studio Irene Amuron of the Red Cross Red Crescent Climate Centre, Paul Davies of the UK Met Office, and Terrence Fernando the University of Salford, to debate the question: “How do we improve outcomes for everyone at risk from the impact of hydrometeorological hazards?” Societies are increasingly challenged by the need to cope with complex interconnected threats. Either a single hazard leads to a cascade of additional perils, often more significant effects than the primary one, or multiple independent hazards occur simultaneously. In each case, lack of planning and early preparation leads to the realisation of major adverse impacts on people and their livelihoods. Economic development may falter, and political instability becomes an added risk.Climate change compounds existing threats and poses new ones; many, outside the bounds of experience, require us to harness new knowledge and take a more holistic view of the immediate and future risks facing us. The problem affects all of society and a “whole of society” approach is needed. This is immensely challenging. Societal compacts and partnerships among more than a few groups of stakeholders are rarely sustained and we proceed with compartmentalised approaches where each sector attempts to solve its own problems with little or no insight into how sectoral interdependencies compound and exacerbate risks or how to reduce them.Two key concepts have emerged in the past decade as the means to improve people’s ability to handle different types of hazards – Anticipatory action and impact-based forecasting. Anticipatory action is a set of actions people can take to prevent harm to themselves and others before acute impacts are felt. Impact-based forecasts focus on what the hazard will do rather than what the hazard will be. In practice, anticipatory action and impact-based forecasts are complementary. Neither is effective without the other. “Good forecast but shame about the outcome” is a repeated mantra. So, for those impacts of hazards for which anticipatory action driven largely or in part by weather forecasts would make a difference, what isn’t being done right? Irene Amuron is Head of Anticipatory Action at the Red Cross Red Crescent Climate Centre.Paul Davies is Principal Fellow, Meteorology and Chief Meteorologist at the UK MetOffice.Terrence Fernando is Director of the THINKlab at the University of Salford.Theme music by John Cockram
  • Episode 25: The ECMWF Copernicus Climate Change Service

    In this episode if The WeatherPod, hosts Alan Thorpe and and David Rogers invite Dr. Carlo Buentempo into the studio.Carlo is the Director of the Copernicus Climate Change Service (C3S), which is part of the European Union's Copernicus earth observation programme.After doing a PhD in Physics at the university of L'Aquila in Italy, Carlo joined the the UK Met Office Hadley Centre where he led the climate adaptation and services teams before taking his current role at ECMWF.The main mission of C3S is to make climate data more readily available and usable by society. As Director, Carlo leads a team of scientists at the European Centre for Medium-range Weather Forecasts (ECMWF) as well as coordinating the activities of many external research groups and contractors working on this programme to deliver its mission. The Service offers free and open data about past, present and future climate to its hundreds of thousands of users across policy, business and scientific domains all over the world.Theme music by John Cockram
  • 24. Episode 24: What next for Weather Forecasting?

    In this episode of The WeatherPod, hosts Alan and David meet Paul Davies, Chief Meteorologist and Principal Research Fellow of the UK Met Office, to discuss the implications of 'impact-based forecasting' for the future of weather forecasting and of the weather forecaster.About the guestPaul Davies has a number of years experience in meteorology and is now recognized nationally and internationally as an authority in the field of operational meteorology.As Principal Fellow, Paul will ensure the Met Office’s professional and cultural approach to weather forecasting is globally recognized and to head a respected and valuable operational meteorological profession. He will spearhead the innovation and use of new scientific, technological, and operational techniques to deliver impact and benefit to a wide range of customers. Paul will also be the external figurehead for all things to do with the meteorological profession be this with stakeholders or key customers.Paul was a Chief Forecaster in 2003 before being appointed as the Chief Hydrometeorologist and Deputy Head of the Flood Forecasting Center in 2008. From 2010, he created the Met Office’s Hazards Center and was the founder and inaugural Chair of the Natural Hazards Partnership (NHP) comprising a number of partner agencies, government departments and the Cabinet Office.Paul was the lead author to WMO’s Multi-Hazard Impact Based Forecasting and Risk Based Warnings guidelines in his capacity as WMO’s IMPACT task team chair. Paul later became UK representative on the WMO Commission for Basic Systems (CBS) OPAG Management group for GDPFS ensuring the flow of data and advice is cascaded between global, regional, and national systems.
  • Episode 23: Predicting the Uncertain World

    In this episode of The WeatherPod hosts David and Alan talk to Tim Palmer to explore some of the thoughts developed in his new book 'The Primacy of Doubt' (Basic Books) and their potential implications for weather forecasting. About the guestTim is an eminent research scientist in weather and climate prediction and predictability. He has worked at the Met Office, at ECMWF and now is a Professor of Physics at the University of Oxford. He is well known as an architect of ensemble weather prediction, but he is a polymath with interests across fundamental physics and maths. In his book 'The Primacy of Doubt' Tim discusses how incorporating uncertainty appropriately in the form of (random) noise into simulations and models of a range of natural phenomena, including weather and climate, is fundamental to improving the predictions of future behavior, i.e. the weather forecast.Tim is also a keen musician. See: music by John Cockram
  • Episode 22: Perspectives from early career scientists on the evolution of the GWE

    In this episode of the WeatherPod, hosts David and Alan talk to Dr Shipra Jain and Dr Marisol Osman about their experience as early career scientists of working within the global weather enterprise. Both Shipra and Marisol are members of the Young Earth System Scientists (YESS) community.Please send your feedback on The WeatherPod to us at: About the guestsDr Shipra Jain (University of Edinburgh – climate science) is a physicist by training and particularly interested in how society can benefit from improved knowledge of extremes. She conducts research to advance the physical understanding and early prediction of weather and climate extremes. For the last five years, she is working on the development of physical-mathematical frameworks that can be used to quantify the chance of record-breaking rainfall for the next season to next decades.Shipra has worked on diverse climate science-related topics including seasonal predictions, monsoon dynamics, stratospheric tropospheric processes, climate change and variability, climate extremes, and air pollution. She has collaborated with > 60 co-authors across different countries, organizations, and career stages. Very recently, she has joined the Centre for Climate Research Singapore, where she works on the seasonal and subseasonal forecasting and development of forecast products for Singapore and South-East Asia.Dr Marisol Osman (research scientist at CIMA/University of Buenos Aires – Meteorology/climatology) is a postdoctoral researcher working at Karlsruhe Institute of Technology. Coming from a town of farmers, she was always interested in how climate variability affects agricultural activity. During her Diploma and PhD, which she completed in Buenos Aires, Argentina, she worked on climate predictability and prediction on seasonal timescales over the Southern Hemisphere and South America. During her postdoc, she developed a probabilistic seasonal forecasting system for South America that is now used by the Argentinian Weather Service. As a researcher, she worked on interdisciplinary projects with anthropologists, economists as well as farmers to co-design and co-produce forecasting tools for the agricultural sector on subseasonal and seasonal timescales. Now at KIT she studies the predictability of different variables relevant for the energy sector on weather and subseasonal timescales.Theme music by John Cockram