Newsletter No. 149 banner

Météo-France hosts OpenIFS workshop

Glenn Carver
Sándor Kertész
Filip Váňa (all ECMWF)
Frédéric Ferry
Etienne Chabot (both École Nationale de la Météorologie)


OpenIFS workshop 2016
OpenIFS workshop 2016. Twenty-two postgraduate students from Météo-France’s École Nationale de la Météorologie took part in the workshop. (Photo: Sébastien Laflorencie)

An OpenIFS workshop dedicated to interpreting ensemble forecasts took place at Météo-France’s École Nationale de la Météorologie (ENM) in the city of Toulouse from 7 to 9 June 2016. The OpenIFS programme at ECMWF provides a supported version of the operational Integrated Forecasting System (IFS) under licence to national meteorological and hydrological services, research institutes and universities. OpenIFS encourages and promotes research, teaching and training on numerical weather prediction (NWP) and NWP-related topics. An important part of this activity is scientific outreach represented by the annual workshop held in institutions of ECMWF Member States. In previous years, the workshop has been organised by the universities of Helsinki, Stockholm and Oxford and has been open to all users of OpenIFS. This year the workshop was held solely for ENM students at the invitation of ENM staff.

École Nationale de la Météorologie

L’École Nationale de la Météorologie is a department of the French national meteorological service, Météo-France, dedicated to providing higher education and professional training in meteorology, climate and related sciences. It delivers undergraduate, graduate and postgraduate courses. ENM is part of a larger campus in Toulouse, unique in Europe, which brings together the Météo-France operational branches of forecasting, observation, IT, business, climatology, research and training, together with the European Centre for Advanced Research and Training in Scientific Computing (CERFACS) and the French National Centre for Hydrometeorology and Support for Flood Forecasting (SCHAPI). There are 30 to 40 masters students per year and they follow a three-year course in atmospheric physics, dynamic meteorology, climate dynamics, hydrology and oceanography, numerical weather prediction, computer science, statistics and data processing.

The approach taken for this year’s OpenIFS workshop was to work exclusively with ENM staff and develop a case study for their students to analyse using the same pedagogical approach and tools successfully used in previous workshops.

Hurricane Nadine during HyMeX

The topic for this workshop was the use of operational ECMWF ensemble forecasts during the HyMeX (Hydrological cycle in Mediterranean experiment) observational campaign in September 2012. As discussed in an article by Pantillon et al. in Q.J.R. Meteorol. Soc. 142 (2016), the presence of Hurricane Nadine in the Atlantic at this time caused high uncertainty for ensemble forecasts at medium range for the north-western Mediterranean. This presented the opportunity to create an interesting case for the students to study by applying ensemble products to the real-world task of forecasting for flight planning during the campaign.

The workshop started with keynote presentations to introduce the topic: Véronique Ducrocq, as chair of the Executive Committee for Implementation and Science Coordination of HyMeX and head of the Groupe de Météorologie de Moyenne Echelle (GMME) at the CNRM (Météo-France’s National Centre for Meteorological Research), spoke about the HyMeX field campaign; Jean-Pierre Chaboureau (Laboratoire d'Aérologie, University of Toulouse/French National Centre for Scientific Research) described the impact of Hurricane Nadine; François Bouttier (CNRM/GMME) talked about the new pre-operational ensemble prediction system at convective scale based on the AROME-France model (2.5 km horizontal resolution, 12 members).

Potential vorticity plots
Potential vorticity plots. Charts of potential vorticity at 320 K showing a single forecast from the ensemble representing the observed weak interaction case between Hurricane Nadine and the Atlantic cut-off low (left), a single forecast representing the strong interaction case (right), and the verifying analysis (middle).

The ECMWF operational ensemble forecast of 20 September produced a bifurcation in the ensemble spread downstream of Nadine, causing difficulties in planning flights to intercept precipitation events for the first special observation period (SOP1) during HyMeX. Key to understanding the ensemble spread was the predicted separation between Nadine and an Atlantic cutoff at their closest point; distances above 1,000 km produced the observed case of weak interaction between the vortices, resulting in strong precipitation over the Cevennes, France; below 1,000 km produced strong interaction in the forecast and little or no precipitation over southern France.

The forecast of 20 September was just three days before IOP6 (Intensive Observation Period number 6) during HyMeX SOP1. A strong mesoscale convective system crossed south-eastern France during the night from 23 to 24 September, then northern Italy during 24 September 2012 in the morning and afternoon.

Case study exercises

Hands-on exercises were created using ECMWF analyses and forecasts from 20 September. To make the workshop more interesting, the event was reforecast using the 2016 operational ensemble system, a resolution improvement of 18 km from the 36 km of the operational ensemble in 2012 (but using the original analysis and members of the Ensemble of Data Assimilations). The students were given a series of tasks to first understand the synoptic situation by looking at the analyses, then the HRES deterministic forecast of the event from September 2012, followed by studying the ensembles from 2012 and then the 2016 reforecast. Specially written Metview macros allowed the students to explore the datasets by animating their choice of parameters, plotting ensemble products such as ensemble spread, stamp maps and so on. A gallery of the various plotting tasks can be found at

The students could also construct clusters of the ensemble members from the 2012 and 2016 ensemble forecasts. The aim of clustering is to identify the two (or more) main forecast scenarios, in this case intense precipitation or no intense precipitation over south-eastern France. With clustering, the bifurcation in the forecast becomes more evident by plotting, for example, ensemble spread separately for each cluster. The students were first asked to choose and plot their own clusters by manually selecting the most appropriate members from their choice of parameter. They then used Metview to compute a principal component analysis (PCA) on the 500 hPa ensemble output at 00 UTC on 24 September and compare their own derived clusters with those from the PCA. The exercises and Metview macros were designed so that the students could choose different parameters and dates to create the clusters, allowing them to see how the bifurcation in the forecast was related to the time of the closest approach between Nadine and the cut-off low.

The forecasting instructor Etienne Chabot (ENM), who was the forecaster on duty for HyMeX SOP1 at the time, challenged the students to realise the same forecast that he personally made for the IOP6. Using the forecast products (convection-parametrized models only) they could understand the difficulty of helping a scientific manager to take a binary decision on whether or not to give the go-ahead for an aircraft research flight. The students had to assess statistical ensemble products such as quantiles, probabilities, ‘spaghettis’, stamp maps, and using clustering, to try to estimate the uncertainty of their forecast in terms of intensity, localisation and chronology. The students were asked to explore and compare the 2012 and 2016 operational ensembles to look at the improvements the new ensemble made and how decisions for flight planning for the HyMeX SOP1 would have been impacted.

Cross-sectional view.
Cross-sectional view. Cross-section through the approximate centres of Hurricane Nadine and the cut-off low, showing potential temperature (in Kelvin – contours), potential vorticity (shading), and three-dimensional wind projected onto the plane of the cross-section (arrows). The chart illustrates the different nature of the two weather systems: the warm core of Hurricane Nadine (left) and the cold tilted core of the Atlantic cut-off low (right). The red line in the small map of mean sea level pressure shows the position of the cross-section in the analysis of 22 September 2012 00 UTC.

A novel aspect of the OpenIFS workshops is the use of the Metview/OpenIFS virtual machine (VM). It incorporates a fully functional Linux desktop that will run on any host operating system and includes all the required software and forecast data used in the workshop. The VM is ideal for workshops run remotely from ECMWF, as all the development and testing can be done in-house at ECMWF, knowing the students will be using the same system in their classroom.

Feedback and outlook

The workshop was a great success and both the students and the tutors enjoyed it. The students were completing a course on ensemble forecasting, and this workshop complemented and increased their understanding of the use of ensembles in real-world forecasting. Some found using OpenIFS ensemble forecasts for a ‘real-world’ case study more useful than the ‘traditional’ courses! The students liked the ease of exploring ideas with the datasets using Metview. Another enjoyable aspect was the mix of French and English teaching. ENM staff have expressed a desire to repeat this workshop in collaboration with ECMWF for their students in the future.

The success of the workshop was due to the contributions and support of many people, including in particular Linus Magnusson at ECMWF and Thierry Barthet and David Pollack at ENM.

We look forward to the next OpenIFS general user meeting in Trieste, Italy, in June 2017.