Seminars / Informal seminars / Lectures by ECMWF Staff and Invited Lecturers

Seminars contribute to our ongoing educational programme and are tailored to the interests of the ECMWF scientific community.

Informal seminars are held throughout the year on a range of topics. Seminars vary in their duration, depending on the area covered, and are given by subject specialists. As with the annual seminar, this may be an ECMWF staff member or an invited lecturer.

The following is a listing of seminars/lectures that have been given this year on topics of interest to the ECMWF scientific community.  See also our past informal seminars

2017

23 June at
10:30-12:00

Room: MR1

Bias correction of tropical cyclone size and structure in the ECMWF global ensemble prediction system

Speaker: Jeff Kepert (Bureau of Meteorology, Australia)

Abstract

Global EPS systems provide very valuable warning of severe weather risk, including tropical cyclones. One limitation arises because the spatial resolution of the EPS is necessarily too coarse to capture the fine details of the tropical cyclone's core circulation, resulting in the predicted storms having systematic biases in structure and intensity. In particular, the cyclones are too weak and the radius to maximum winds is too large, with these biases increasing with forecast length. Ocean waves are a major impact mechanism of tropical cyclones, but these atmospheric biases result in errors in the predicted ocean wave field also. While these biases to not affect users who respond purely to the expected risk of a tropical cyclone in their vicinity, they are important for users who wish to make more quantitative use of the wind or wave predictions.

We have developed a method for bias correcting tropical cyclone structure and intensity in EPS systems, and applied it to cyclones off north west Australia in the ECMWF global EPS. The method diagnoses the properties of forecast storms within the EPS, and replaces them in the surface wind and pressure field with parametric vortices with the corrected intensity and structure. The correction scheme is statistical, based on regression analysis of historical errors. The bias-corrected wind field ensemble is then used to force a wave model ensemble.

This work was supported by Woodside, Shell, Inpex and Chevron, who operate in the waters off north west Australia.

8 June at 13:00-18:00

Symposium on dynamic meteorology and numerical weather prediction

Speakers:

  • Raymond T. Pierrehumbert (Halley Professor of Physics at the University of Oxford)
  • Nedjeljka Žagar (Associate Professor of Meteorology at the University of Ljubljana)
  • Hans Huang (Head of Weather Modelling and Prediction at the Centre for Climate Research Singapore)
  • Stephen Belcher (Chief Scientist at the UK Met Office)
  • David Burridge (Director of ECMWF 1991–2004)
  • Erland Källén (Director of Research at ECMWF)
  • Florence Rabier (Director-General of ECMWF)

Abstract: Raymond Pierrehumbert

Raymond Pierrehumbert will speak on multiple equilibria, bifurcations and climate regimes. Professor Pierrehumbert’s research on climate has spanned timescales from billions of years in the past to billions of years in the future. His work on atmospheric dynamics has included contributions to theories of weather regimes and blocking, baroclinic instability, stratified flow (with applications to mountain meteorology), atmospheric mixing problems and dynamical control of atmospheric humidity.

Abstract: Nedjeljka Žagar

Nedjeljka Žagar will talk about data assimilation in the context of tropical dynamics. Professor Žagar’s research interests range from modelling mid-latitude mesoscale weather to large-scale tropical variability and data assimilation. Her research has highlighted the role of large-scale equatorial waves in tropical data assimilation. She applies normal modes for the scale-dependent evaluation of the global unbalanced circulation and the growth of forecast uncertainties in ensemble prediction.

Abstract: Hans Huang

Hans Huang’s lecture will be devoted to high-resolution forecast modelling in the context of tropical dynamics. Dr Huang has extensive experience of data assimilation research, observation system design and assessment, and NWP system development. He was the lead scientist for data assimilation in the forecast system developed by the European HIRLAM programme, and the manager of the US-based Weather Research and Forecasting (WRF) Model data assimilation system.

Abstract: Stephen Belcher

Stephen Belcher will talk about climate change. Since 1990, Professor Belcher has published over 100 peer-reviewed papers on the fluid dynamics of atmospheric and oceanic turbulence. Between 2007 and 2010, he was the Head of the School of Mathematical and Physical Sciences at the University of Reading. In 2012, he joined the Met Office as Director of the Met Office Hadley Centre for Climate Science and Services before becoming the Met Office Chief Scientist in November 2016.

Abstract: David Burridge

David Burridge will speak on the past evolution of numerical weather prediction at ECMWF. Dr Burridge was the Head of Research at ECMWF from 1982 to 1991 and the Centre’s Director from 1991 to 2004. Subsequently he shaped and steered the implementation of THORPEX, a ten-year international research and development programme set up by the World Meteorological Organization to improve one-day to two-week forecasts of high-impact weather.

Abstract: Erland Källén

Erland Källén’s lecture will be about the Centre’s research and operations today. Between 1983 and 2009, Professor Källén was first an associate professor and later professor in dynamic meteorology at Stockholm University. His research areas are large-scale dynamics of the atmosphere, NWP and climate modelling. He has worked as a project leader for the HIRLAM programme and was responsible for initiating and heading the first Swedish climate modelling programme in the 1990s.

Abstract: Florence Rabier

Florence Rabier will talk about what the future holds for numerical weather prediction and related services at ECMWF. Dr Rabier has greatly contributed to delivering major operational changes at ECMWF and Météo-France. She is especially well known within the meteorological community for her key role in implementing a new data assimilation method (4DVar) in 1997. She was ECMWF’s Director of Forecasts before she became the Centre’s Director-General in January 2016.

25 May
at 10.30

Room: LT

20th Century trends in potential predictability and the role of land surface

Speaker: Bart van den Hurk (KNMI, The Netherlands)

Abstract:

Climate change does give rise to shifts in patterns of risks of high-impact extreme events. An important means to adapt to extreme events is early warning, based on skilful forecasts. The ability to forecast any signal is inherently limited by the chaotic nature of the system. A change in the variability patterns in the climate system may or may not impose shifts in the inherent potential predictability of phenomena at monthly to seasonal time scales, which may or may not change our ability to prepare to these phenomena.

 

During a short sabbatical stay I've explored signatures of trends in potential predictability and the role of land surface processes in this. Use was made of earlier 20th century reforecasts by Antje Weisheimer, where at 4 startdates per year an ensemble of 51 4-month forecasts was produced initialized from ERA20C. In order to explore the role of the land surface, a duplicate experiment was performed in which the land surface in each of the ensemble members was initialized from different states, removing the source of predictability arising from this component in the climate system.

 

The seminar will review some related studies reported earlier in the literature, and will present early results of the experiments carried out at ECMWF. Suggestions for further analyses from the audience are more than welcome.

22 May
at 9:30

Room: CC

Statistical postprocessing of multi-model ensemble
precipitationforecasts in the US National Weather Service

Speaker: Tom Hamill (NOAA Earth System Research Lab, USA)

Abstract

Taking concepts developed in programs such as THORPEX/TIGGE, the US NWS is embracing the use of multi-model ensembles and their statistical postprocessing using short training data sets.  The intended goal is to produce dramatically improved forecasts of important weather elements such as temperature, winds, and precipitation, both deterministic and probabilistic.  Generating reliable and skillful forecasts of precipitation will be the focus of this talk.  Postprocessing of precipitation is especially challenging since forecast bias depends on precipitation amount, and a short training sample (here, the previous 60 days) is often insufficient to estimate this conditional, often location-dependent bias. 

This talk will discussed a sequential algorithm for precipitation postprocessing over the US for forecast leads of +1 to +10 days using US and Canadian global ensemble data, an algorithm that provides a workaround for limitations imposed by using a short training data set. The steps in the postprocessing include: (a) pre-specification of "supplemental locations" for each model grid point -- i.e., the specification of other grid points with similar precipitation climatologies and terrain characteristics that are likely to have similar forecast biases. (b) generating forecast and analyzed cumulative distribution functions (CDFs) for each grid point using the last 60 days of forecasts, including data from supplemental locations. (c) Quantile mapping of each member forecast at each grid point; (d) Dressing each quantile-mapped forecast with amount-dependent noise to account for the under-spread character of ensemble forecasts; (e) estimation of probabilities from the ensemble relative frequency; and (f) Savitzky-Golay smoothing of forecasts in regions with little variation in terrain height.  The resulting forecasts are shown to be highly reliable and skillful

19 May
at 14:00

Room: LT

Application of a CubeSat-Based Passive Microwave Constellation to Operational Meteorology

Speaker: Al Gasiewski (University of Colorado, USA)

Abstract

In their most recent decadal assessment (Earth Application from Space, 2007) of Earth science space missions the U.S. National Research Council identified the Precipitation and Allweather Temperature and Humidity (PATH) mission as one of ten recommended medium cost missions. Based on the NRC’s outlined goals, PATH would have the unique capability of providing allweather temperature and moisture soundings and cloud and raincell imagery at spatial scales comparable to AMSU-A/B or ATMS, but at sub-hourly temporal resolution. The essential need is to provide the atmospheric penetrability and spatial resolution of operational microwave sensors but with temporal resolution commensurate with the natural rate of evolution of convectively driven weather. This seminar will focus on the merits of a constellation of passive microwave sounding and imaging CubeSats for achieving PATH goals from the multiple viewpoints of calibration accuracy, data assimilation and global sampling, downlink capability and latency, and orbital lifetime and launch availability. Microwave spectral imagery at 50, 118, and 183 GHz with spatial resolution of ~10-30 km and temporal resolution of ~15-60 minutes from such a fleet could be expected to significantly enhance forecasting of mesoscale convective weather and hurricane rain band evolution, along with provide valuable temporal gap-filling data for synoptic weather forecasting. It is argued that from a joint technology, science, and operational standpoint that a cost-effective realization of the PATH goals, but with the additional features of global coverage and improved NWP sensitivity, can be achieved by a low-cost random-orbit constellation of CubeSats supporting the ATMS and 118 GHz bands. The CU PolarCube mission will be discussed as a basis for this fleet concept.

18 May
at 10:30

Room: MR1

"How do I know if I’ve improved my continental scale flood early warning system?"

Speaker: Hannah Cloke (University of Reading, UK)

Abstract

Flood early warning systems mitigate damages and loss of life and are an economically efficient way of enhancing disaster resilience. The use of continental scale flood early warning systems is rapidly growing. The European Flood Awareness System (EFAS) is a pan-European flood early warning system forced by a multi-model ensemble of numerical weather predictions. Responses to scientific and technical changes can be complex in these computationally expensive continental scale systems, and improvements need to be tested by evaluating runs of the whole system. It is demonstrated here that forecast skill is not correlated with the value of warnings. In order to tell if the system has been improved an evaluation strategy is required that considers both forecast skill and warning value. The combination of a multi-forcing ensemble of EFAS flood forecasts is evaluated with a new skill-value strategy. The full multi-forcing ensemble is recommended for operational forecasting, but, there are spatial variations in the optimal forecast combination. Results indicate that optimizing forecasts based on value rather than skill alters the optimal forcing combination and the forecast performance. Also indicated is that model diversity and ensemble size are both important in achieving best overall performance. The use of several evaluation measures that consider both skill and value is strongly recommended when considering improvements to early warning systems.

16 May
at 10:30

Room: LT

Mesoscale aggregation of shallow marine cumulus convection

Speaker: Christopher Bretherton (University of Washington, USA)

Abstract

Over the oceans, shallow cumulus convection, often mixed with patchy stratocumulus, is a common cloud type.  It is usually 'aggregated' into mesoscale patches or polygons of deeper cumuli, with possible consequences for the mean vertical structure of cloud cover and cloud-precipitation-aerosol interaction.  Large-eddy simulations (LES) covering domains 50 km or more across also exhibit mesoscale aggregation of shallow cumulus convection, but it is not fundamentally well understood.   To further that understanding, we analyze the development of convective aggregation in multiday LES of a 108x108 km doubly periodic domain simulating mean summertime conditions at a location east of Hawaii.  The simulated convection aggregates within 12 hours.  Vertically resolved heat and moisture budgets on mesoscale subdomains elucidate this process.  Shallow cumulus deepen preferentially in more humid regions of the boundary layer, stimulating net moisture convergence into those regions.  Sensitivity studies show that the aggregation does not require precipitation.  Aggregation is weakened but not prevented if radiative cooling and surface fluxes are horizontally homogenized.  A unifying conceptual model explains these findings.

15 May
at 13:30

Room: LT

Measuring Large Scale Divergence and Vorticity by Aircraft

Speaker: Bjorn Stevens (MPI, Germany)

Abstract

During NARVAL-2 an extensive array of dropsondes were launched over the tropical Atlantic, near the ITCZ, to test methods for measuring large scale vertical velocity.  These measurements have been evaluated and show that as few as 12 dropsondes launched by an aircraft are well suited to estimating divergence on the scale of a flight leg.  Analysis of high-resolution (1-2 km) simulations over the tropical Atlantic during the ICON period are used to evaluate the credibility of the measurements, and the richness of the vertical structure observed, as well as to characterize the spatial and temporal correlation of divergence as a function of scale. The analysis and measurements open a new chapter in observing diabatic processes in the tropical atmosphere and underpin a major new field initiative planned for 2020.

5 May
at 10:30

Room: LT

Improved Climate Forecasting Service of Australian Bureau of Meteorology

Speaker: William Wang (Bureau of Meteorology, Australia)

Abstract

Australia Bureau of Meteorology has started seasonal climate outlook service since 1989 based purely on statistical methods using tropical SST conditions as predictors. In May 2013 the Bureau of Meteorology Climate Information Services has started issuing climate outlooks based on the Predictive Climate Ocean Atmosphere Model for Australia, a dynamical climate model developed by the Bureau of Meteorology and CSIRO Marine and Atmospheric research division. Since 2016, we have been developing our next generation of climate forecasting system and products based on UKMO Glosea5 or ACCESS-S. This talk is to give an overall picture of the climate prediction services in Australia through the following aspects:

  1. A very brief introduction of our Bureau’s climate forecast service history; why we did what we did, experiences and lessons we obtained from the past;
  2. The current dynamic climate forecasting system
    • A very brief introduction of (POAMA)
    • The revolutionary transition from statistic to dynamic and how did we make such a CHANGE;
    • What we investigate when we develop a new climate forecasting system and how, including verification metrics, such as percent consistence, LEPS, ROC skill, Brier skill score etc. and some major results as the benchmarks of a system;
    • The current services and the online products and relevant issues.
  3. The development of the next generation of Bureau’s climate forecasting system
    • The new ACCESS-S/UK Met Office Glosea5 model and why it was chosen;
    • Development of new forecasting system and products– the concept of seamless forecasting from weather to annual climate prediction.
    • Primary results of this development
  4. A brief introduction of a new verification metric, the so called weighted percent consistence

26 April
at 10:30

Room: LT

Compatible finite element methods for numerical weather prediction

Speaker: Jemma Shipton (Imperial College, UK)

Abstract

We present recent work on developing a compatible finite element model for numerical weather prediction. This work has been motivated by the requirement for numerical discretisations that are stable and accurate on nonorthogonal grids (such as icosahedral or equiangular cubed sphere grids) without sacrificing properties of conservation, balance and wave propagation that are important for accurate atmosphere modelling on the scales relevant to weather and climate (Staniforth et. al. 2012).

Compatible finite element methods are a type of mixed finite element method (where different finite element spaces are used for different fields) where the divergence of the velocity space maps on to the pressure space. This necessitates the use of div-conforming finite element spaces for velocity, such as Raviart-Thomas and Brezzi-Douglas-Marini, and discontinuous finite element spaces for pressure. The main reason for choosing compatible finite element spaces is that they have a discrete Helmholtz decomposition of the velocity space; this means that there is a clean separation between divergence-free and rotational velocity fields. Cotter and Shipton (2012) used this decomposition to demonstrate that compatible finite element discretisations for the linear shallow water equations satisfy the basic conservation, balance and wave propagation properties listed in Staniforth and Thuburn (2012).

In the talk we will show the progress we have made towards extending this approach to the fully 3D equations via a vertical slice model. Many current atmospheric models use a staggered Charney Phillips grid in the vertical to ensure a good representation of hydrostatic balance. In the finite element context the equivalent staggering requires the temperature field to be discontinuous in the horizontal direction but continuous in the vertical. We present a stable and accurate advection scheme for this field. The success of this approach is illustrated by benchmarking results from our model, implemented in Firedrake (www.firedrake.org).

25 April
at 10:30

Room: LT

Developments in ensemble variational data assimilation at Météo-France

Speaker: Benjamin Ménétrier (Météo-France, France)

Abstract

A brief overview of the work that is being carried out at Meteo-France in data assimilation will be presented. Short term developments include a better resolution for the ensemble of global 4DVars, whose size might double. Significant steps are also made towards the operational implementation of an ensemble of 3DVars for the regional model. Long-term plans for both global and regional models are based on a transition from 3D/4DVar to 3D/4DEnVar algorithms. After 3 years of development within the OOPS framework, numerous improvements for the EnVar methods are now available: enhanced localisation, faster distribution of members, hybrid background error covariance matrix, block-Krylov methods, etc. The implementation of all observation operators in the OOPS framework has been recently completed, enabling deeper investigations about the EnVar behaviour in an operational-like setup.

A crucial aspect of ensemble methods, and especially of EnVar algorithms, is the need for an efficient and well-tuned localisation. Affordable methods are now available to diagnose localisation functions objectively, using the ensemble only. A detailed presentation of both mathematical background and practical implementation of these methods will be given, showing consistent results for several atmospheric and oceanic models. Extensions of the theory towards hybrid covariance matrices and 4D localisation have been developed recently. All these diagnostics are publicly available in an open-source, easily operable code that works for any kind of model grid.

 

5 May
at 10:30

Room: LT

Improved Climate Forecasting Service of Australian Bureau of Meteorology

Speaker:  William Wang (Bureau of Meteorology, Australia)

Abstract

 

Australia Bureau of Meteorology has started seasonal climate outlook service since 1989 based purely on statistical methods using tropical SST conditions as predictors. In May 2013 the Bureau of Meteorology Climate Information Services has started issuing climate outlooks based on the Predictive Climate Ocean Atmosphere Model for Australia (POAMA), a dynamical climate model developed by the Bureau of Meteorology and CSIRO Marine and Atmospheric research division. Since 2016, we have been developing our next generation of climate forecasting system and products based on UKMO Glosea5 or ACCESS-S. This talk is to give an overall picture of the climate prediction services in Australia through the following aspects:

  1. A very brief introduction of our Bureau’s climate forecast service history; why we did what we did, experiences and lessons we obtained from the past;
  2. The current dynamic climate forecasting system
    1. A very brief introduction of (POAMA)
    2. The revolutionary transition from statistic to dynamic and how did we make such a CHANGE;
    3. What we investigate when we develop a new climate forecasting system and how, including verification metrics, such as percent consistence, LEPS, ROC skill, Brier skill score etc. and some major results as the benchmarks of a system;
    4. The current services and the online products and relevant issues.
  3. The development of the next generation of Bureau’s climate forecasting system
  4. The new ACCESS-S/UK Met Office Glosea5 model and why it was chosen;
    1. Development of new forecasting system and products– the concept of seamless forecasting from weather to annual climate prediction.
    2. Primary results of this development
    3. A brief introduction of a new verification metric, the so called weighted percent consistence

 

25 April
at 14:00

Room: LT

Non-oscillatory Forward-in-Time Unstructured-Mesh Models for Fluid Flows

Speaker: Joanna Szmelter (Loughborough Univ., UK)

Abstract

The presentation will summarize the development of a fully unstructured (and hybrid) mesh class of models for multi-physics applications with emphasis on simulating inertia gravity waves. Global and limited area atmospheric models will be discussed. The methodology employs an edge-based, finite volume discretisation within the non-oscillatory forward-in-time (NFT) framework. The edge-based data structure allows integration of the generic physical form of the governing PDE over arbitrarily-shaped cells. Aspects of unstructured meshes flexibility, such as optimal point distribution, adaptivity and multigrid will be discussed together with different options of applying the general NFT framework based on the unstructured mesh Multidimensional Positive Definite Advection Transport Algorithm (MPDATA) and elliptic (Krylov) solvers. This approach will be evaluated for a range of models from soundproof and compressible nonhydrostatic to engineering flow solvers. Simulations of stratified orographic flows and the associated gravity-wave phenomena in media with uniform and variable dispersive properties will be presented. They verify the developments and demonstrate the efficacy of the implicit large eddy simulation operating on unstructured meshes for study of stratified turbulent flows.  Numerical examples for engineering problems will include coastal flows, gas dynamics at Mach numbers from zero to supersonic, and solutions from Arbitrary Lagrangian Eulerian (ALE) codes using conservative MPDATA based remapping techniques.

 

21 April
at 10:30

Room: LT

SODA3 (Simple Ocean Data Assimilation ocean/sea ice reanalysis) and a step toward a coupled reanalysis?

Speaker: James Carton (Univ. Maryland, USA)

Abstract

Atmospheric reanalyses produce surface fluxes as a residual of their update cycle. These fluxes should be consistent with estimates of ocean heat and freshwater storage and divergence of transports.  Here we compare the ERA-Int, MERRA2, and JRA-55 fluxes with the imbalances apparent in the increments produced by the SODA3 ocean reanalysis system during the data-rich eight year period 2007-2014.   The heat flux comparison reveals that the regional imbalance falls in the range of 10-30 W/m2 in time mean with even larger imbalances on seasonal time scales.  In the vertical the corresponding temperature imbalances are concentrated in the mixed layer.  We argue that in the interior gyres these imbalances are the result of seasonal errors in the atmospheric reanalysis representation of surface heat and freshwater fluxes (the problem is the atmosphere).  In other regions such as the Gulf Stream we show evidence of substantial ocean model error. Elsewhere errors in momentum fluxes appear to dominate the temperature and salinity increments.  We also examine the impact on fluxes of the choice of bulk parameterization used to calculate surface fluxes from atmospheric state variables.

In the second part of the talk we present a strategy for correcting surface fluxes based on the ocean observations and we then discuss the results of experiments testing this strategy.  Examination of the results confirms that in the interior gyres our bias-correction strategy reduces the seasonal error in surface heat flux to less than 5 W/m2.  We conclude with a discussion of the changes that are needed to address biases at high latitude.

11 April
at 14:00

Room:LT

Dataflow for Geocomputing

Speaker: Georgi Gaydadjiev (Maxeler, London, UK)

Abstract

Ever since Von Neuman, predicting the weather has been one of humanities top computing challenges. The challenge can be split into two parts: (1) writing the software to predict the weather and (2) running the software to predict the weather. Running the software includes deciding on discretisation, data sizes, and building custom configurations of computing, memory and storage components. 

A key question arises. Should we build computers that are easy to program or should we build computers that are efficient in running the largest models we can conceive. Of course in the initial decades, software, math and models need to be developed. However in the steady state, we will need to transition to optimising the operational aspects of running complex models. 

Maxeler Multiscale Dataflow Computing addresses the steady-state challenge of optimising operational efficiency in Space, Time and Value (STV). We re-evaluate choices in discretisation of space, time and value, and build optimal computational arithmetic pipelines to compute PDE solutions in minimal time and with minimal cost. 

Maxeler Dataflow computing has been demonstrated in Italy to compute a local weather model 60x faster and in China, to be 14x faster than a Tianhe 1a (dual GPU node) on a node-2-node basis. The UK government purchased a significant Maxeler machine for Daresbury labs. Still programming challenges remain and we are looking forward to continuing the dialogue with the geoscientists on how this new computational paradigm can be deployed with minimal impact on programming and code maintenance.

23 March
at 10:30

Room: MR1

Uncertainties in simulated evapotranspiration from land surface models over a 15-year Mediterranean crop succession

Speaker: Sebastien Garrigues (CEH/UoR/INRA)

Abstract

PDF icon

6 March
at 14:00

Room: LT

Water Information in Australia - The Post Millennium Drought Experience

Speaker: Dr Dasarath Jayasuriya (BoM, Australia)

Abstract

Water data and information are needed for planning, managing and meeting the sustainable development goals related to water resources taking into account climate changes, population growth and other drivers impacting supply security. The Australian National Plan for Water Security (2007) postulated that better water information was essential for providing water security during Australia's periodic droughts. After the Millennium Drought which lasted from 1996 to 2007, the Government and the policy bureaucracy understood that they were facing a water crisis, but had no nationally consistent and accessible data to make informed decisions. Water data were held by approximately 200 organisations across state and local governments with no standardisation, transparency, and accessibility. Parochial interests often led to ‘gaming’.

The presentation lays out the Australian experience and its response to the water data and information challenge and shares the Government’s Water Information Program initiative over the last 10 years. Highlights relate to socialising the universal availability of water data, water information products and water forecasting services. Collectively they provide water intelligence to Bureau’s stakeholders including Government, industry and community.

Dr Dasarath Jayasuriya

23 February
at 10:30

Room: LT

Subseasonal forecasting – a battle between damped persistence and tropical forcing?

Speaker: Warwick Norton (CUMULUS, UK)

Abstract

Cumulus
 
We review some of the forecast performance across the 2016/17 winter where the seasonal forecasts expected a negative NAO state yet this not really played out. Rather there has been significant forecast volatility, with no forecast skill past week 2 for Europe (or in forecasting the NAO). In the southern hemisphere Australia has experienced an extremely hot summer which was also poorly forecast in the subseasonal range. We discuss sources of forecast error particularly associated with underestimating maritime continent convection but also other factors that may have led to poor forecast skill this year such as lack of MJO activity. 
 
We examine what could be predicted with perfect knowledge of the tropics from relaxation experiments. We compare this year to other recent years where skill in predicting the NAO has been higher. Results based on Rossby wave source analysis suggest that tropical teleconnections are too weak in the ECMWF model - this can lead to periods of under confidence in forecasting predictable extratropical signals. We conjecture that current subseasonal models contain too much damped persistence of the initial conditions and not enough forcing from the tropics.

15 February
at 10:30

Room: MR1

Flood forecast sensitivity to temperature using ECMWF ensembles for 145 catchments in Norway

Speaker: Trine Jahr Hegdahl (NVE, Norway)

Abstract

The Norwegian flood forecasting service is based on a flood-forecasting model run on 145 basins. The basins are located all across Norway and differ in both size and hydrological regime. Current flood forecasting system is based on deterministic meteorological forecasts, and uses an auto-regressive procedure to achieve probabilistic forecasts. An alternative approach is to use meteorological and hydrological ensemble forecasts to quantify the uncertainty in forecasted streamflow. The aim of our study is to establish and assess the performance of both meteorological and hydrological ensembles for 145 catchments in Norway, which differ in size, elevation and hydrological regime. We identify regional differences and improvements in performance for preprocessed meteorological forecasts. A separate study further investigates the sensitivity to forecasted temperature for specific snowmelt induced floods. In Norway, snowmelt and combined rain and snowmelt floods are frequent. Hence, temperature is important for correct calculations of snowmelt. Temperature and precipitation ensembles are derived from ECMWF covering a period of nearly three years (01.03.2013 to 31.12.2015). To improve the spread and reduce bias we used standard methods provided by the Norwegian Meteorological Institute. Precipitation is corrected applying a zero-adjusted gamma distribution method (correcting the spread), and temperature is bias corrected using a quantile-quantile mapping (using Hirlam (RCM) 5 km temperature grid as a reference). Observed temperature and precipitation data are station data for all of Norway, interpolated to a 1×1 km2 grid (SeNorge.no). Streamflow observations are available from the NVE database. The hydrological model is the flood-forecasting operational HBV model, run with daily catchment average values. The results show that the methods applied to meteorological ensemble data reduce the cold bias present in the ECMWF temperature ensembles. Catchments on the western coast, having a lower initial performance, show the highest improvement by the temperature corrections, whereas some inland catchments in southeastern Norway show reduced performance. Ensemble spread for precipitation improves, but is not recognized in the discharge performance measures. Both precipitation and temperature show an east-west divide in performance. Corrected temperature ensemble lead to improved performance in discharge for some western catchments. Overall, the regional analyzes including all data, show that catchments have different sensitivity to temperature correction and will benefit from regional or catchment specific bias correction. Spring flood events, in catchments located west and southeast, showed different discharge response to temperature correction (more than 2°C). For the western catchment the increased temperature, led to higher discharge, whereas there were minor change for the southeastern catchment.

 

12 January
at 14:00

Room: LT

The global ICON Ensemble at DWD

Speakers: Michael Denhard and Cristina Primo (DWD, Germany)

Abstract

Since October 2015 DWD runs an experimental ICON ensemble suite with 40 members and approx. 40km horizontal resolution on the global scale up to +168h lead time twice a day (00/12UTC). The global grid contains a 20km two-way nested area over Europe. The ensemble is initialized by analyses from our ensemble data assimilation system (ICON EDA) which is a combination of a Local Ensemble Transform Kalman Filter (LETKF) with a hybrid ensemble/3D-Var variational system for the high-resolution deterministic model. At the time there is no stochastic physics implemented and the error growth properties of the ensemble are determined by the diverse co-variance inflation techniques in the LETKF such as multiplicative inflation factors, relaxation to the prior and stochastic SST perturbations.  Moreover, the static NMC Background error co-variances are added to the flow dependent ensemble co-variances to rescale the innovations. In the first part we show verification results for the ICON-EPS forecasts in comparison to the ECMWF-EPS and analyze the spread skill relation for both ensembles. The second part introduces techniques for predicting the error growth properties along trajectories in the state space of a model. We use the "Broyden family" methods to iterate a Broyden matrix in state space of the Lorenz63 and 95 models. During iteration the Broyden matrix gains information on the error growth properties of the dynamical system. We discuss, if the information in the Broyden matrices along a trajectory can be used as an approximation of the singular vector approach.

11 January
at 10:30

Room: LT

How good (or bad) is the circulation of the stratosphere and mesosphere in the IFS?

Speaker: Inna Polichtchouk (University of Reading, UK)

Abstract

Accurate representation of the stratospheric circulation is important for tropospheric predictability on intraseasonal timescales, because of the downward influence of the stratosphere on the troposphere.  The “downward control” principle states that the stratospheric Brewer Dobson circulation (BDC) is primarily driven by the wave breaking/saturation aloft. Thus, the stratospheric circulation in turn depends on the representation of the mesospheric momentum budget. This talk reviews the state of the middle atmosphere in the IFS, with a focus on the BDC and the semi-annual oscillation. I will compare the middle-atmosphere circulation to reference datasets and assess the impact of 1) the parametrized non-orographic gravity wave drag; 2) treatment of the sponge layer; 3) the cubic octahedral discretization; and, 4) stochastic physics.

LT = Lecture Theatre, LCR = Large Committee Room, MZR = Mezzanine Committee Room,
CC = Council Chamber