Workshop on Predictability, dynamics and applications research using the TIGGE and S2S ensembles

Europe/London
ECMWF

ECMWF

Reading
Description

This workshop provided an opportunity to review the main scientific advances in predictability, dynamical process studies and applications of ensemble forecasts across the medium and S2S forecast ranges. Examples of sectors rapidly developing in ensemble applications included energy, retail and agriculture, as well as disaster risk mitigation worldwide. The emphasis was on the utilisation of the TIGGE and S2S databases in research and contributions on seamless prediction, multi-model prediction and ensemble post-processing are particularly welcome. One session was dedicated to the technical development of ensembles, the TIGGE and S2S data bases and proposals for future development.

Poster presentation schedule
Events team
    • 13:15 13:45
      Registration Weather Room

      Weather Room

    • 13:45 14:15
      Introduction 30m Lecture Theatre

      Lecture Theatre

      ECMWF

      Speaker: Florian Pappenberger (ECMWF)
    • 14:15 15:30
      Predictability and dynamics: Chair - John Methven (University of Reading) Lecture Theatre

      Lecture Theatre

      ECMWF

      • 14:15
        Flow-dependent predictability of wintertime Euro-Atlantic weather regimes in medium-range forecasts 30m Lecture Theatre

        Lecture Theatre

        ECMWF

        This study assesses the medium-range flow-dependent forecast skill of Euro-Atlantic weather regimes: the positive and negative phases of the North Atlantic Oscillation (NAO+ and NAO−), Atlantic ridge (ATLR), and Euro-Atlantic blocking (EABL), for extended winters (November–March) in the periods 2006/2007–2013/2014 and 1985/1986–2013/2014 using The Interactive Grand Global Ensemble (TIGGE) and the National Oceanic and Atmospheric Administration (NOAA)’s Global Ensemble Forecasting System (GEFS) reforecast datasets, respectively. The models show greater-than-observed (smaller-than-observed) frequencies of NAO− and ATLR (NAO+) with forecast lead time. The increased frequency of NAO− is not due to its excess persistence but due to more frequent transitions mainly from ATLR, but also from NAO+. In turn, NAO+ is under-persistent. The models show the highest probabilistic skill for forecasts initialised on NAO− and the NAO− forecasts during the TIGGE period. However, the GEFS reforecast during the period 1985/1986–2013/2014 revealed that these recent high skills reflect the occurrence of four long-lasting (>30 days) NAO− events in 2009/2010–2013/2014 and that the skill for forecasts initialised on NAO− before 2009/2010 (the longest duration was 22 days and the second-longest 16 days)was the lowest. The longer theNAO− events persist, the higher the skill of forecasts initialised on NAO−. The skill dependency on regime duration is less clearly observed for the other regimes. In addition, the GEFS reforecast also revealed that the highest skill of the NAO− forecasts during the period 1985/1986–2013/2014 is attributed to the higher skill of the NAO− forecasts during the active NAO− periods. The EABL forecasts initialised on ATLR show the lowest skill, followed by the NAO− (EABL) forecasts initialised on NAO+ or ATLR (NAO+). These results suggest that the recent models still have difficulties in predicting the onset of blocking.

        Speaker: Mio Matsueda (Center for Computational Sciences, University of Tsukuba)
      • 14:45
        The role of stratosphere-troposphere coupling in sub-seasonal to seasonal prediction using the S2S database 30m Lecture Theatre

        Lecture Theatre

        ECMWF

        A major source of sub-seasonal predictability for the mid-latitudes during boreal winter and spring and austral spring is variability of the stratospheric polar vortex. While a number of studies have now demonstrated that surface predictability is enhanced during both sudden stratospheric warming (SSW) and strong vortex events, there has been little comparison of model performance beyond a small number of case studies of individual extreme events. The S2S database represents a step change in the availability of data to interrogate and understand sub-seasonal skill in the stratosphere and links to surface predictability. The Stratospheric Network for the Assessment of Predictability is a joint project between WCRP/SPARC and the S2S project which brings together a small community of scientists interested in working on this problem. In this talk, analysis from a recent, first inter-comparison of stratospheric predictability and stratosphere-troposphere coupling in the S2S models will be presented. Analysis of a wide variety of dynamical events will be considered, including SSW and strong vortex events, the spring time final stratospheric warming and coupling linked to reflection of planetary-waves. Finally, the extent to which the ability to capture these processes leads to enhanced tropospheric forecast skill will be discussed.

        Speaker: Andrew Charlton-Perez (University of Reading)
      • 15:15
        Stratospheric influences on subseasonal predictability of European energy-industry-relevant parameters 15m Lecture Theatre

        Lecture Theatre

        ECMWF

        Surface weather variability on subseasonal timescales influences energy production, demand, and prices. Improving the skill of subseasonal predictions of surface weather is thus of high interest for the energy industry. This is particularly the case for near-surface wind due to the ongoing shift toward more wind power generation. Anomalous states of the stratospheric polar vortex during winter are one important source of enhanced subseasonal predictability, because they are typically followed by persistent large-scale flow patterns over Europe. We thus investigate how the state of the stratospheric polar vortex affects month-ahead wind power generation in Europe during winter and how this effect influences the skill of subseasonal numerical weather forecasts of various energy-relevant surface parameters. In a first step, we use the ERA-Interim reanalysis and the wind power dataset “Renewables.ninja” to demonstrate a strong correlation between the strength of the lower-stratospheric circulation and month-ahead wind power generation in different regions of Europe. This relationship exists due to episodes of troposphere–stratosphere coupling, which lead to prolonged periods resembling positive or negative phases of the North Atlantic Oscillation (NAO). Since these persistent NAO-like periods are associated with strong near-surface wind anomalies, they have an important impact on wind power generation, in particular in Northern Europe. Motivated by this empirical relationship, we develop a simple statistical forecasting approach based on the strength of the lower-stratospheric circulation, which provides skillful forecasts of month-ahead wind power generation in Europe. In a second step, we investigate the skill of the ECMWF subseasonal model from the S2S database in predicting month-ahead 10-m wind speed, 2-m temperature, and precipitation as three energy-industry-relevant model parameters. The skill of the ECMWF model is generally higher than the skill of the simple statistical forecast for wind power generation, particularly for short lead times. It is substantially driven by the strength of the lower-stratospheric circulation at initialization time and the associated NAO-like phases throughout the forecast, which reflects the empirical relationship from the reanalysis data also in the model. However, there are substantial differences in the skill between different European regions as well as parameters, with implications for both the energy industry and the numerical modeling community.

        Speaker: Dominik Büeler (Karlsruhe Institute of Technology, Institute of Meteorology and Climate Research + Institute for Atmospheric and Climate Science, ETH Zurich)
    • 15:30 16:00
      Coffee break 30m Weather Room

      Weather Room

    • 16:00 17:15
      Predictability and dynamics: Chair - Frederic Vitart (ECMWF) Lecture Theatre

      Lecture Theatre

      ECMWF

      • 16:00
        Understanding predictability of the MJO in S2S enesemble 15m

        Predictability of the MJO, especially the initiation and eastward propagation of the large-scale convection and precipitation over the Indian Ocean and West Pacific warm pool across the Maritime Continue, is extensively investigated using the S2S ensemble. The model forecasts are compared with the Large-scale Precipitation Tracking (LPT) of the MJP using the 20-years TRMM-GPM precipitation data from 1998-2018 base on the method described in Kerns and Chen (2016, JRG-Atmosphere). In this study we first examine the model predictions of the overall global tropical convection statistics including the monsoons, ITCZ, and the MJO from 19980-2018. There is a wide range of seasonal and interannual variability of the MJO. The predictability of the MJO convective initiation and eastward propagation is assessed using model predictions of over 200 MJO events are evaluated using the LPT data at leadtimes from 4-45 days. The predictability of the MJO as measured by the RMM index is relatively longer and invariant than the MJO convective initiation and eastward propagation in terms of precipitation. The model predicative skill decrease with increasing in leadtime, especially after 10 days. The predictive skill is particularly low across the Maritime Continent. The model captures the stronger MJO during the boreal winter DJF better than the summer JJA. We are currently extending the analysis to the surface wind using the satellite based CCMP data from 1998-2018. A more comprehensive analysis will be presented at the workshop.

        Speaker: Shuyi Chen (University of Washington)
      • 16:15
        MJO Impact on Temperature Extremes over Australia during Austral Spring 15m

        As a potential source of multiweek predictability, we investigate the MJO’s impact on temperature extremes during Austral spring. We find a significant MJO influence on weekly mean temperature extremes (defined here as exceeding the upper and lower quintiles) over southeastern Australia when the MJO is in phases 2, 3, and 6 and 7. During these phases, the occurrence of maximum and minimum temperature exceeding the weekly top quintile or falling below the bottom quintile can increase by more than 50% in large areas over southeastern Australia.

        The physical mechanism for MJO influence on the temperature extreme is via a Rossby wave train forced by the MJO convection. This wave train results in a persistent local circulation anomaly centered around southeastern Australia, thus promoting warm northerly or cool southerly flow across southern Australia. Excitation of this Rossby wave train is strongly influenced by the presence of the subtropical westerly jet. The large positive meridional gradient of mean absolute vorticity (denoted as beta) along the core of the subtropical jet acts to provide a strong localized Rossby wave source as the MJO convection traverses west top east to the north of Australia. On the poleward side of the tropical jet beta weakens and together with the strong mean zonal winds results in an undefined stationary Rossby wave number (Ks=sqrt(beta*/Ubar)), and so Rossby wave propagation is prevented across the subtropical jet. However, in a small region at the longitude of Australia, there is a local maximum in Ks, so allowing Rossby waves to disperse poleward and influence south eastern Australia temperatures. .

        The depiction of this wave train and the predictability of the MJO impacts on extreme temperatures are also examined using the Bureau of Meteorology’s new S2S prediction system.

        Speaker: Harry Hendon (Bureau of Meteorology)
      • 16:30
        Extratropical predictability from the Quasi-Biennial Oscillation and the MJO in S2S models 15m

        The effect of the Madden-Julian Oscillation (MJO) on the Northern Hemisphere wintertime stratospheric polar vortex is evaluated in operational subseasonal forecasting models. Reforecasts which simulate stronger MJO-related convection in the Tropical West Pacific also simulate enhanced heat flux in the lowermost
        stratosphere and a more realistic vortex evolution. The time scale on which vortex predictability is enhanced lies between 2 and 4 weeks for nearly all cases. Those stratospheric sudden warmings that were preceded by a
        strong MJO event are more predictable at ~20 day leads than stratospheric sudden warmings not preceded by a MJO event. Hence, knowledge of the MJO can contribute to enhanced predictability, at least in a probabilistic sense, of the Northern Hemisphere polar stratosphere. One model (NCEP model) succeeds in capturing the observed relationship that the MJO->stratosphere-> Europe is more important than the direct impact of the MJO on Europe for lags longer than 3 weeks.

        The effect of the Quasi Biennial Oscillation on the vortex in these same models is also evaluated. The UK Met Office model, the ECMWF model, and the NCEP model all show a Holton Tan effect that is similar though weaker to that observed. Theres is a hint of downward propagation to the surface.

        Garfinkel C.I., C. Schwartz, D. I. P. Domeisen, S-W Son, A. H. Butler, I. P. White (2018), Extratropical stratospheric predictability from the Quasi-Biennial Oscillation in subseasonal forecast models , JGR, doi: 10.1029/2018JD028724.

        Garfinkel, C. I., & Schwartz, C. (2017). MJO-related tropical convection anomalies lead to more accurate
        stratospheric vortex variability in subseasonal forecast models. Geophysical Research Letters, 44, 10,054–
        10,062. https://doi.org/10.1002/2017GL074470

        Speaker: Chaim Garfinkel (Hebrew University)
      • 16:45
        Intra-seasonal and Seasonal Variability of the Northern Hemisphere Extra-tropics 15m

        The natural variability of the extra-tropics is studied at seasonal and intra-seasonal time scales. Nonlinear oscillations in the extra-tropics are extracted from daily anomalies of 500-hPa geopotential height for the period 1979-2012 using a data-adaptive method. Three propagating oscillations with broad-band spectra centered at 120, 45, and 28 days are found. When combined, the oscillations explain up to 30% of the natural variability of the extra-tropics on the intra-seasonal to seasonal time scales. These oscillations share some features with the circumglobal wave guide and in some phases of their lifecycles they project onto the canonical teleconnection patterns. When used as predictors in a simple linear regression model with 2-meter temperature as predictand, the mid-latitude oscillations extend the potential predictability of dependent variable to about 20 days.

        For the 120-day oscillation, the S2S models show forecast skill beyond 4 weeks lead time.

        Speaker: Cristiana Stan (GMU)
      • 17:00
        Subseasonal Forecast Skill over the Northern Polar Region in Three Operational S2S Systems 15m

        Pentad forecast skill over the Northern polar region in boreal winter is evaluated for the subseasonal to seasonal prediction (S2S) systems from three operational centers: the European Centre for Medium-Range Weather Forecasts (ECMWF), the U.S. National Centers for Environmental Prediction (NCEP) and Environment and Climate Change Canada (ECCC). The former two systems are running with air-sea coupled models, whereas the latter with an atmospheric-only model. One objective of this study is to assess the impact of air-sea coupling on polar subseasonal forecast skill. Previous studies have reported that the ECMWF system has a better Madden-Julian Oscillation (MJO) forecast skill than the other systems. Whether the MJO skill translates to polar forecast skill is of great interest.
        The results indicate that for a lead time longer than about 10 days the forecast skill of 2-meter air temperature and 500-hPa geopotential height in the polar area is low comparing to the tropical and middle latitude regions. The three S2S systems have comparable forecast skill in the northern polar region. Relatively high skill is observed in the Arctic sector north of the Bering Strait in pentads 4-6. The polar temperature forecast skill is found to be dependent on the tropical MJO phase in the initial condition. Forecasts starting from MJO phases 3 and 6, which correspond to enhanced (suppressed) convection in the equatorial eastern Indian Ocean and suppressed (enhanced) convection in the tropical western Pacific, tend to be more skillful than those initialized from other MJO phases. To improve the polar prediction on the subseasonal time scale, it is important to have well represented tropical MJO and its associated teleconnections in the model.

        Speaker: Hai Lin (Environment and Climate Change Canada)
    • 17:15 19:15
      Posters and drinks reception 2h Lobby/Weather Room

      Lobby/Weather Room

    • 09:00 10:45
      Database Technical Development: Chair - Richard Mladek (ECMWF) Lecture Theatre

      Lecture Theatre

      ECMWF

      • 09:00
        The technical development of the TIGGE and S2S databases 25m

        The TIGGE database consists of medium-range ensemble forecasts from eleven global NWP centres up to 15 days ahead. It started in October 2006 and it has currently reached 3 PiB of data. The S2S database consists of real-time and reforecast outputs from eleven global models producing sub-seasonal to seasonal forecasts up to 60 days ahead. Although the S2S database was established only in 2015, it has already attracted more than one thousand users who have downloaded more than 400 TiB of data.
        This presentation will focus on the technical aspects related to both databases, from their design and development to routine daily operations. It will be shown that good data design, user friendly interfaces and efficient ways to deliver the data are necessary for success of similar. Last but not least it will be stressed the importance of the effective long term cooperation with all stakeholders, providing routinely their model outputs, filling gaps when necessary and allowing for the evolution of the models in changing operational environment.

        Speaker: Manuel Fuentes (ECMWF)
      • 09:25
        TIGGE and S2S status and developments at CMA 20m

        CMA's latest progress in the development of S2S Numerical Model database Management and Services Including the following aspects (Xing Hu):
        1. Preprocessing of S2S model data. The original big data files sent by the ECMWF is received and disassembled by preprocessing to be easy for the users to obtain the download.
        2. Convenient and flexible data retrieval. Provide online access based on HTTP and OPeNDAP protocols, and data sets including historical averages, maps etc. 2D and 3D Visualization for user’s easy visual experience.
        3. Service statistics. Data volume statistics, user access and usage statistics.
        4. Data archiving. Online hard disk for data service, hard disks and tapes for backup.

        Current Situation and Future of TIGGE in CMA (FeiFei Yang):
        TIGGE forecasting data is a valuable resource of global ensemble forecasting. As CMA is one of the two archive centers in the world, TIGGE data is very important for CMA to improve the accuracy and reliability of weather and climate forecasting on all time scales. The presentation is mainly carried out in the following three aspects.

        1. CMA and ECWMF together completed the design and adjustment of TIGGE data format and the establishment of quality control process in non-LDM environment in 2018.

        2. About the total amount of TIGGE historical data in CMA, the way of archiving, data management and service.

        3. CMA's plan for TIGGE in 2019 mainly includes the upgrade of TIGGE portal website, the establishment of the subsequent filing process of TIGGE data after the change of TIGGE data format, and the discussion on data management and service.

        Speakers: Xing Hu (China Meteorological Administration ), FeiFei Yang (China Meteorological Administration)
      • 09:45
        The S2S Data Base in IRI Data Library: Maprooms and online analysis tools 30m

        The International Research Institute for Climate and Society Data Library (IRIDL) is a powerful and freely accessible online data repository and analysis web-service that allows a user to view, analyze, and download hundreds of terabytes of climate-related data (including sub-seasonal data) through a standard web browser in a computer or a smartphone. A wide variety of operations, from simple anomaly calculations to more complex analysis such as empirical orthogonal function (EOF), canonical correlation analysis or cluster analyses can be performed with just a few clicks. The IRIDL provides a flexible and fully online interface for easy subsetting, analysis & visualization, and download in a variety of formats, including NetCDF, Google Earth’s KML and GIS-compatible layers. Furthermore, the IRIDL is an OpenDAP server, which means local client programs --e.g., written in Python, R or Matlab-- can read the desired data online, avoiding the need to download it explicitly, saving disk space and increasing efficiency. IRIDL, conceived in the 1980s/90s, is perhaps the first example in the climate community of the data-server based processing paradigm in which users bring their calculations to the data, rather than simply downloading the data from the library.

        Over 50 TB of the S2S Database forecasts and reforecasts, including indices used for evaluating the Madden and Julian Oscillation, recognized by the research community as a key phenomenon acting as source of predictability on the sub-seasonal timescale, are presently available in the IRIDL . All these data are obtained from the ECMWF server and kept up to date in the IRIDL as new forecasts & reforecasts are made.

        This talk will introduce the S2S data base in IRIDL and present some examples of online maprooms and analysis tools.

        Speaker: Andrew Robertson (International Research Institute for Climate and Society)
      • 10:15
        Ensemble forecasting at ECMWF 15m

        An overview of the methodologies for representing initial uncertainties and model uncertainties in the ECMWF ensemble forecasts will be given. Initial uncertainties are represented with perturbations from an ensemble of 4D-Vars with perturbed observations and with perturbations based on singular vectors. Model uncertainties are represented with the Stochastically Perturbed Parametrization Tendency scheme (SPPT). The recent upgrade of SPPT will be described.

        In 2019, exchangeable initial conditions for the atmosphere are planned to be introduced. Their benefits for ensemble development will be discussed. Last but not least, future directions for the representation of model uncertainties will be summarized.

        Speaker: Martin Leutbecher (ECMWF)
      • 10:30
        The global ICON-EPS: a contribution to TIGGE? 15m

        Since January 2018 DWD runs a global ICON ensemble suite with 40 members and approx. 40km horizontal resolution including a grid refinement for Europe of 20km. Forecasts are generated at 00/12UTC up to +180h and at 06/18UTC up to +120h. To improve the boundary conditions for the COSMO-D2-EPS four additional runs take place at 03/09/15/21 with a limited forecast time of +30h. The ICON-EPS initializes from analysis states generated by the Local Ensemble Transform Kalman Filter (LETKF) data assimilation system running at DWD. Random perturbations of some physical model parameters are selected at the beginning of each forecast. We provide operational ensemble products on our open-data servers according to the WMC (World Meteorological Center) requirements and would appreciate contributing to other international projects like TIGGE. In the talk we report the actual configuration of the system and give an outlook on our future plans.
        When introducing the ICON-EPS to the forecasters at DWD the most important step was to evaluate the system against the well established ECMWF-EPS. We will show that the ICON-EPS adds value especially in the case of extreme events. On this basis we started the development of a global risk index called EWI (Extreme Weather Index). It combines ECMWF ensemble products (EFI, SOT, quantiles) with the forecasted quantiles of the ICON-EPS probability distributions.

        Speaker: Michael Denhard (Deutscher Wetterdienst)
    • 10:45 11:15
      Coffee break 30m Weather Room

      Weather Room

    • 11:15 11:45
      Database Technical Development: Chair - Manuel Fuentes (ECMWF) Lecture Theatre

      Lecture Theatre

      ECMWF

      • 11:15
        An Assessment of Predictability and Prediction of NCEP GEFS for Subseasonal Forecast 15m

        The demand for forecast information beyond week-2 has increased significantly in recent years. This information provides valuable guidance for various users who use it to guide public safety, quality of life, and business decisions that drive economic growth. Discussions of predictability, current numerical model capability and applications are greatly enhancing our understanding for prediction beyond week-2, which is known as sub-seasonal to seasonal (S2S) prediction. The National Centers for Environmental Predictions (NCEP) Global Ensemble Forecast System (GEFS) has demonstrated great success for weather and week-2 forecast in past decades by providing reliable probabilistic forecast to general public and now its application the S2S prediction is being explored.

        To assess the predictability and capability of the subseasonal forecast, a modified ensemble version of the NCEP spectral model (GEFS) was applied in support NOAA SubX (Subseasonal multi-model Experiments). An 18-year reforecast with this modified version of the GEFS is used as a reference system and is compared to the newly adopted Finite-Volume dynamics (FV3)-based GEFS which includes major model dynamical changes, different horizontal resolution, different microphysics, etc. The improvement of predictability and predictions will be investigated for all these changes through various measurements across tropical to extratropical areas in terms of deterministic (ensemble mean) and probabilistic (ensemble distribution) forecasts.

        Speaker: Yuejian Zhu (EMC/NCEP/NWS/NOAA)
      • 11:30
        Using the S2S Database to Evaluate the Performance of the Navy Earth System Prediction Capability (ESPC) Ensemble 15m

        The Navy Earth System Prediction Capability (ESPC) is a coupled global model consisting of the Navy Global Environmental Model (NAVGEM) atmospheric model coupled to the Global Ocean Forecast System (GOFS), which consists of the Hybrid Coordinate Ocean Model (HYCOM) and the Los Alamos Community Ice Code (CICE). The performance of a 15-member Navy ESPC ensemble during 2017 is compared to coupled ensemble systems from the Subseasonal-to-Seasonal (S2S) database. S2S database forecasts served as a valuable benchmark as the Navy ESPC was being evaluated prior to its transition into operations.

        Comparisons between the Navy ESPC and archived S2S ensembles focus on the relationship between key tropical modes of variability such as the Madden-Julian Oscillation (MJO) and El Nino Southern Oscillation (ENSO) and weekly-averaged conditions in the atmosphere from the tropics to the arctic. The MJO in the Navy ESPC is shown to be too intense and fast; in contrast most models have an MJO which is too weak and slow. Process-based diagnostics shown to have a strong relationship with MJO behavior, such as the relationship between rain rate and outgoing longwave radiation and mean biases in mid-level humidity, are used to explain differences in MJO behavior in the Navy ESPC and S2S models. Composites of MJO behavior in each of the models and their global teleconnections are used to explain differences in the predictive skill of the models. Metrics used to evaluate the Navy ESPC and S2S ensembles include measures of predictive skill, mean biases, and ensemble performance.

        Speaker: Matthew Janiga (Naval Research Laboratory)
    • 11:45 13:00
      Prediction and verification: Chair - Manuel Fuentes (ECMWF) Lecture Theatre

      Lecture Theatre

      ECMWF

      • 11:45
        Receiver Operating Characteristic (ROC) curves 30m

        Receiver Operating Characteristic (ROC) curves

        Tilmann Gneiting, Peter Vogel, and Eva-Maria Walz

        Heidelberg Institute for Theoretical Studies and Karlsruhe Institute of Technology, Germany

        Relative or Receiver Operating Characteristic (ROC) curves are used in a very wide range of settings where covariates, features, markers, or probability forecasts for binary events are to be evaluated. In meteorology, the WMO mandates the use of the Area Under the ROC curve (AUC) measure for evaluating skill in long-range forecasts, and an ECMWF supplementary headline score uses the AUC to assess extreme forecast index (EFI) skill for wind speed over Europe. All these uses require the conversion of a continuous, real-valued weather quantity to a binary event.

        I will review and examine the construction and usage of ROC curves from a general scientific perspective, with emphasis on current and future applications to ensemble weather forecasts, and illustrated on probabilistic quantitative precipitation forecasts over the tropics. Technically, a ROC curve is simply a plot of the hit rate vs. the false alarm rate as the cut-off value of a real-valued predictor variable for a binary event ranges over all possible thresholds. We distinguish raw ROC diagnostics and ROC curves, elucidate the special role of concavity in interpreting and modelling ROC curves, and establish an equivalence between ROC curves and cumulative distribution functions (CDFs).

        In the final part of the talk, I will hint at very recent developments, in which we seek generalizations of ROC curves that apply to predictors of real-valued quantities directly, thereby eliminating the need for the conversion to a binary event.

        Speaker: Tilmann Gneiting (Heidelberg Institute for Theoretical Studies)
      • 12:15
        A verification framework for South American sub-seasonal precipitation predictions 30m

        Caio A. S. Coelho, Mári A. F. Firpo, Felipe M. de Andrade

        Affiliation:
        Centro de Previsão de Tempo e Estudos Climáticos (CPTEC), Instituto Nacional de Pesquisas Espaciais (INPE)

        We propose a verification framework for South American sub-seasonal (weekly accumulated) precipitation predictions produced one to four weeks in advance. The proposed framework assesses both hindcast and near real time forecast quality focusing on a selection of the most fundamental attributes (association, discrimination, reliability and resolution). These attributes are measured using various deterministic and probabilistic verification scores. Such an attribute-based framework allows the production of verification information in three levels according to the availability of sub-seasonal hindcasts and near real time forecasts samples. The framework is also useful for supporting future routine sub-seasonal prediction practice by helping forecasters to indentify model forecast merits and deficiencies and regions where to best trust the model guidance information. The three information levels of the proposed framework are defined according to the verification sampling strategy and are referred to as target week hindcast verification, all season hindcast verification, all season near real time forecast verification. The framework is illustrated using ECMWF sub-seasonal precipitation predictions. For the investigated period (austral autumn), reasonable accordance was indentified between hindcasts and near real time forecast quality across the three levels of verification information produced. Overall, sub-seasonal precipitation predictions produced one to two weeks in advance presented better performance than those produced three to four weeks in advance. The northeast region of Brazil consistently presented favorable sub-seasonal precipitation prediction performance through the computed verification scores, particularly in terms of association and discrimination attributes. This region was therefore identified as a region where sub-seasonal predictions produced one to four weeks in advance with the ECMWF model are most likely to be successful in South America. When aggregating all predictions over the South American continent the probabilistic assessment showed modest discrimination ability, with predictions clearly requiring calibration for improving reliability and possibly combination with predictions produced by other models for improving resolution. The proposed framework is also useful for providing feedback to model developers in identifying strengths and weaknesses for future sub-seasonal predictions systems improvements.

        Speaker: Caio Coelho (CPTEC/INPE)
      • 12:45
        Spread of global 2-meter temperature analyses: disentangling forecast systematic errors from mis-estimation of ensemble spread 15m

        Global 2-meter temperature analyses in 2018 from ECMWF, JMA, and the UK Met Office were downloaded for 2018 from the TIGGE database. Using the multi-model analysis mean as a surrogate for truth, the daily spreads were decomposed into an estimate of the systematic error component (from time-mean differences between analyses) and a random component (the perturbation magnitude with respect to each system's time-mean systematic component). This error decomposition permits a tentative evaluation of the magnitude and location of systematic errors in the analyses and an evaluation of whether the initial spread estimates are properly set. This diagnostic approach will be applied to the ECMWF system, with results presented orally.

        Speaker: Dr Tom Hamill (NOAA ESRL PSD)
    • 13:00 14:00
      Lunch break 1h
    • 14:00 15:15
      Prediction and verification: Chair - Laura Ferranti (ECMWF) Lecture Theatre

      Lecture Theatre

      ECMWF

      • 14:00
        Use of TIGGE/Global Ensembles in Tropical Cyclone Research and Operational Forecasts 15m

        Current status on the use of TIGGE/Global ensembles in tropical cyclone (TC) research and operational forecasts is presented. First, the current status on the use of global ensembles in TC track, intensity and genesis forecasting at operational tropical cyclone forecasting centres over the world is presented based on a questionnaire survey conducted by the World Meteorological Organization (WMO) HIWeather Project and the WMO World Weather Research Programme (WWRP) Predictability, Dynamics and Ensemble Forecasting (PDEF) working group in 2018. The result of the questionnaire show that ensemble forecasts are seen to be particularly important in track and genesis forecasting, but that there are still hurdles limiting the pull-through of the use of probabilistic ensemble forecast information in to operational warnings. Second, the North Western Pacific Tropical Cyclone Ensemble Forecast Project (NWP-TCEFP), which is a Research and Development Project (RDP) of WWRP, is introduced as an example of having achieved a research to operation (R2O) transfer using TIGGE. With this successful R2O transfer, TC track and genesis forecast products by multiple global ensembles are provided to the Typhoon Committee Members in real time. Third, the result of literature search for papers using TIGGE, which has been conducted under a PDEF activity every year, is presented. This survey confirmed that TC is the most studied research area with TIGGE. Finally, recent TC studies using TIGGE are introduced. Those include the optimization of TC track uncertainty cone with multiple global ensembles, development of TC genesis guidance with TIGGE and initiatives toward improving TC intensity and structure forecasts.

        Speaker: Helen Titley (Met Office)
      • 14:15
        Achieving seamless verification across sub-seasonal time scales from weather to climate 15m

        Sub-seasonal time scales lie in the middle of the range of "seamless" prediction that spans from deterministic instantaneous weather prediction to the probabilistic time-averaged conditions of seasonal weather forecasts. However, attaining true seamlessness across these disparate types of forecasts is not easy. We propose a flexible approach to model validation that blends day-to-day weather forecast verification and time-averaged predictions for longer time scales in a truly seamless manor. The method combines two weighting methods to transition daily forecast data into multi-day time means whose averaging period increases with forecast lead.

        The Poisson distribution is nearly ideally suited to be such a weighting function, as it is narrow and positively skewed for low parameter values, corresponding to short forecast lead times of one to a few days, becoming a Gaussian distribution approximating weekly or monthly means at progressively longer lead times. However, it cannot represent the character of a deterministic forecast, even at very short time scales. Deterministic forecasts are effectively time-weighted forecasts whose weighting function takes the form of a Kronecker delta. Applied to daily forecast model output, Poisson and Kronecker distributions along the time axis share two key characteristics: they both peak at the value of their sole parameter, and the sum of weights over all times is one. This allows a linear combination of the two functions to be well-behaved, so that a transition from Kronecker to Poisson weighting with increasing forecast lead time makes for a seamless transition. The 2-parameter Hill equation is used to define the transition, which can be tuned to specify the lead time at which the Poisson distribution becomes dominant as well as the steepness of the transition with lead time. This allows the approach to be tailored to specific applications, different forecast variables, locations, and seasons as appropriate. Examples will be shown and caveats discussed.

        Speaker: Paul Dirmeyer (George Mason University)
      • 14:30
        Uncertainties in Extended-Range Precipitation Forecasts: Model Biases or Predictability Limits 15m

        In this study, we analyze the reason for low prediction skill in extended-range precipitation forecasts over the US west coast during boreal winter from the NCEP Climate Forecast System version 2 (CFSv2). Our assessment focuses on whether low skill is due to the biases in CFSv2 or is consistent with the possibility of low inherent predictability over that region. The analysis is based on large dataset of ensemble forecasts from CFSv2 and other six dynamical models. The results strongly indicate that low prediction skill for the precipitation over the US west coast is not because of model biases, but may be due to low underlying signal-to-noise ratio (SNR), i.e., the low inherent predictability. Consistent with low SNR, there are large inconsistencies in the precipitation ENSO responses across different El Niños and across different models. The linkage of the low prediction skill over the US west coast and its possible dynamical reasons are also studied.

        Speaker: Mingyue Chen (Climate Prediction Center/NCEP/NWS/NOAA)
      • 14:45
        Ensemble Prediction and Predictability of Extreme Weather via Circulation Regimes 15m

        We use S2S multi-model ensemble forecasts of circulation regimes to assess the changes in the probability of storminess over Europe and the US in the sub-seasonal forecast range.
        The existence of preferred planetary scale flow structures (circulation regimes) in the Pacific - North American (PNA) and Euro-Atlantic (EA) sectors on sub-seasonal time scales has been well validated in the literature. The shifts in extreme US weather associated with each circulation regime were established by recent published work (Amini and Straus, 2018): each boreal winter circulation regime is associated with dramatic shifts in the storminess metrics: regional likelihood of extreme precipitation (flood or drought), atmospheric river occurrences and moisture flux, and the storm track configurations.
        We evaluate S2S ensemble forecasts in the circulation regime context. We identify those times for which the verification of 500 hPa geopotential height (Z500) is strongly associated with a particular regime in reanalyses. The regimes are identified separately in the PNA and EA sectors using machine-learning techniques (e.g. k-means cluster analysis), but strong association of a reanalysis state with a regime composite is further assessed via pattern correlation. The S2S forecasts are evaluated by determining the fraction of ensemble members strongly associated with the observed regime using Z500.
        We apply these ideas to create a forecast tool for storminess. When one or more circulation regimes are identified in the ensemble forecasts for a particular forecast range, the storminess metrics associated (in observations) with that (those) regime(s) provide a forecast for the likelihood of extreme precipitation, shifts in atmospheric rivers, and changes in storm track configurations. This method is independent of the model statistics of e.g. precipitation, relying solely on the models’ large-scale features.

        Speaker: Kathleen Pegion (George Mason University)
      • 15:00
        Ensemble forecasts for the midlatitudes on sub-seasonal time scales (10-60 days): exploring new products for predicting Atlantic-European weather regimes 15m

        On sub-seasonal time scales the large-scale extratropical circulation often has a better predictability than surface weather at a specific location and time. However, classical ensemble forecast products focussing on the ensemble mean and spread are often indistinguishable from the underlying model climatology. Therefore novel approaches are needed to reveal the predictable components of the atmosphere on sub-seasonal time scales.
        In the Atlantic-European region variability in the large-scale circulation is in first order dominated by the bi-modal North Atlantic Oscillation (NAO). Still it is difficult to describe multi-day variability in surface weather accurately for all of Europe based on the NAO. Weather regimes (e.g. Vautard 1990, Michel and Rivière 2011), provide an alternate concept of 4 to 8 different flow patterns, that account for most of the multi-day variability of the large-scale flow and associated surface weather on sub-seasonal time scales (Grams et al. 2017, Zubiate et al. 2017).
        In this presentation we explore novel ensemble forecast products for 7 year-round Atlantic-European weather regimes and associated surface weather. These regimes are based on a k-means clustering of normalized 5-day low-pass filtered 500 hPa geopotential height anomalies (Z500’). The likelihood of the 7 regimes in the ensemble is explored based on the weather regime index of Michel and Rivière (2011), which describes how well each regime is established.
        We suggest an intriguing visualisation of the regime behaviour that allows a forecaster to assess very quickly and easily the large-scale flow evolution in medium- and extended-range ensemble forecasts. Complemented by maps of other variables describing the large-scale flow or surface weather these products provide a sharper view on the forecast evolution than classical approaches based, for example, on the ensemble mean and spread. Currently we assess forecast skill of the suggested regime products compared to other approaches.

        Further reading:
        Grams, C. M., R. Beerli, S. Pfenninger, I. Staffell, and H. Wernli, 2017: Balancing Europe’s wind-power output through spatial deployment informed by weather regimes. Nature Climate Change, 7, 557–562, doi:10.1038/nclimate3338.
        Michel, C., and G. Rivière, 2011: The Link between Rossby Wave Breakings and Weather Regime Transitions. J. Atmos. Sci., 68, 1730–1748, doi:10.1175/2011JAS3635.1.
        Vautard, R., 1990: Multiple weather regimes over the North Atlantic: analysis of precursors and successors. Mon. Wea. Rev., 118, 2056–2081, doi:10.1175/1520-0493(1990)118<2056:MWROTN>2.0.CO;2.
        Zubiate, L., F. McDermott, C. Sweeney, and M. O’Malley, 2017: Spatial variability in winter NAO–wind speed relationships in western Europe linked to concomitant states of the East Atlantic and Scandinavian patterns. Q.J.R. Meteorol. Soc., 143, 552–562, doi:10.1002/qj.2943.

        Speaker: Christian M. Grams (IMK-TRO, Karlsruhe Institute of Technology (KIT))
    • 15:15 15:45
      Coffee break 30m Weather Room

      Weather Room

    • 15:45 17:00
      Prediction and verification: Chair - David Richardson (ECMWF) Lecture Theatre

      Lecture Theatre

      ECMWF

      • 15:45
        Prospects for subseasonal sea ice prediction at both poles 15m Lecture Theatre

        Lecture Theatre

        ECMWF

        With retreating sea ice and increasing human activities comes a growing need for reliable sea ice forecasts up to months ahead. We exploit the subseasonal-to-seasonal (S2S) prediction database and provide a thorough assessment of the skill of operational forecast systems in predicting the location of the Arctic and Antarctic sea ice edges on these time scales. We find large differences in skill between the systems, with some showing a lack of predictive skill even at short weather time scales, and the best producing skillful Arctic forecasts more than 1 1/2 months ahead. We assess the forecast skill in both hemispheres, thereby showing that prospects for subseasonal sea ice predictions are promising, especially for Arctic late summer forecasts. To fully exploit this potential, it will be imperative to reduce systematic model errors and develop advanced data assimilation capacity.

        Speaker: Lorenzo Zampieri (Alfred Wegener Institute)
      • 16:00
        2014 Indo-Pak’s cataclysmic flood: Can potential future plights could be alleviated with currently available forecasting skill ? 15m Lecture Theatre

        Lecture Theatre

        ECMWF

        In September 2014, a devastating flood wrought havoc in the Indian state of Jammu and Kashmir and adjoining areas of the Pakistan. The very heavy rainfall during last stage of monsoon has resulted in a devastating flood which caused an estimated death toll well above 2000 peoples over the region. After a disaster on this scale and its associated implications in various sectors, the question arises whether strategies of S2S prediction that have proved useful elsewhere can they be adapted to the complex terrain of Himalayas and adjoining areas as well? The aim of this study is in two-folds. Firstly, it attempts to assess the predictive skill of a set of S2S model products and identify forecast windows of opportunity. Secondly, an attempt has been also made to simulate the above mentioned event at higher resolution using a convective permitting model. Finally, the plausible reasons of model failure, potential sources of predictability and how S2S framework may play a key role in addressing such issues is highlighted.

        Key words: Flood, S2S, predictability.

        Speaker: Pushp Raj Tiwari (Centre for Atmospheric and Climate Physics Research, University of Hertfordshire)
      • 16:15
        Assessment of prediction skill for sub-seasonal rainfall variability over Brazil in ensemble-based prediction systems 15m Lecture Theatre

        Lecture Theatre

        ECMWF

        Accurate forecasting of sub-seasonal to seasonal (S2S) variations in rainfall can help mitigate hydrological hazards throughout the tropics. It is therefore essential to analyse the skill of contemporary S2S forecasts systems and investigate whether there are windows of opportunity within which these systems may be more skilful, based on the regional- or large-scale atmospheric circulation. As part of DUBSTEP (Diagnosing and Understanding Brazilian Subseasonal Tropical and Extratropical Processes) project, we examine prediction skill for sub-seasonal rainfall variability over Brazil and surrounding regions in all seasons and analyse conditional skill during the El Niño-Southern Oscillation (ENSO) and Madden-Julian Oscillation (MJO) during the austral summer season (December-January-February). We evaluate hindcasts from three global models; the Met Office (UKMO) Global Seasonal Forecasting System (GloSea5) Global Configuration 2.0 (GC2), the National Centers for Environmental Prediction (NCEP) Climate Forecast System (CFSv2) and the European Centre for Medium Range Weather Forecasts (ECMWF) Integrated Forecasting System (IFS), for the common period of 1999-2010, against observed precipitation estimates from the Global Precipitation Climatology Project (GPCP). Our results highlight possible model deficiencies over the wettest regions and potentially weaker ENSO teleconnections than observed.

        Speaker: Amulya Chevuturi (NCAS)
      • 16:30
        Predicting Sudden Stratospheric Warming 2018 and its Climate Impacts with a Multi-Model Ensemble 15m Lecture Theatre

        Lecture Theatre

        ECMWF

        Sudden Stratospheric Warmings (SSWs) are significant source of enhanced sub-seasonal predictability but whether this source is untapped in operational models remains an open question. Here we report on the prediction of the SSW on 12 February 2018, its dynamical precursors, and surface climate impacts by an ensemble of dynamical forecast models. The ensemble forecast from 1 February predicted 3 times increased odds of an SSW compared to climatology, although the lead-time for SSW prediction varied among individual models. Errors in the forecast location of an Ural high and underestimated magnitude of upward wave activity flux reduced SSW forecast skill. Although the SSW’s downward influence was not well forecasted, the observed northern Eurasia cold anomaly following SSW was predicted, albeit with a weaker magnitude, due to persistent tropospheric anomalies. The ensemble forecast from 8 February predicted the SSW, its subsequent downward influence and a long-lasting cold anomaly at the surface.

        Speaker: Alexey Karpechko (Finnish Meteorological Institute)
      • 16:45
        A zonal component of monsoons and the variability in the strength of the Madden-Julian Oscillation events 15m

        Understanding the variations in the strength of the Madden-Julian Oscillation events as they propagate across the Indo-Pacific Maritime Continent (MC) has been an important challenge. In this study a method of estimating moisture sources associated with sustenance of MJO strength directly from precipitation observation is introduced. The method is used to show the existence of slow eastward propagating zonal moisture flux convergence due to the difference between the longitudes of the Asian and Australian monsoon convergence centers. The variability in the strength of individual MJO events is related their propagation across this moisture convergence signal. Specifically, November, December and January events are likely stay within this convergence signal and therefore may sustain their strength. February, March and April events tend to start weaker and strengthen as they catch up with it. May, June and July events are most likely to weaken for they leave the convergence signal behind.

        Speaker: Samson Hagos (Pacific Northwest National Laboratory)
    • 17:00 19:00
      Poster session 2h Lobby/Weather Room

      Lobby/Weather Room

    • 19:00 21:00
      Workshop dinner 2h ECMWF Restaurant

      ECMWF Restaurant

    • 09:00 10:45
      Prediction and Verification Multi-model approaches to prediction: Chair - Craig Bishop (NRL) Lecture Theatre

      Lecture Theatre

      ECMWF

      • 09:00
        Ensemble Tropical Cyclone Forecast Performance and Prediction of Ensemble Forecast Error 30m

        The primary variables forecast by the U.S. tropical cyclone (TC) forecast centers, National Hurricane Center (NHC) and Joint Typhoon Warning Center (JTWC), are TC track, intensity, and radius of gale-force winds. For all three variables the primary forecast guidance products used by the forecasters are multi-model ensemble mean or consensus forecasts derived using the forecasts from global and regional numerical weather prediction (NWP) models and from a number of statistical models. The consensus forecast guidance for TC track is derived entirely from NWP model forecasts while that for TC intensity and radius of gale-force winds use a combination of NWP and statistical model forecasts.
        Predicted consensus error (GPCE) products for TC track, intensity, and radius of gale-force winds have been developed and installed on the Automated Tropical Cyclone Forecasting system (ATCF) to provide “guidance on guidance” for the forecasters at both NHC and JTWC.

        In this presentation the TC track forecast performance of the multi-model consensus guidance used by NHC is shown along with that for the best of the TIGGE single-model ensembles (ECMWF, MetOffice, and NCEP) for the 2017 Atlantic hurricane season. The TC track forecast errors for the various ensemble means and members are displayed along with the degree of independence of the members of the multi-model and single-model ensembles. Next, the GPCE products for TC track, intensity, and radius of gale-force winds used by the forecasters at NHC and JTWC are illustrated along with their performance. Finally, the techniques used to derive the multi-model consensus TC track GPCE are applied to the ECMWF single-model ensemble for the 2017 Atlantic hurricane season. The results are shown and compared with those for the multi-model consensus.

        Speaker: James Goerss (SAIC, NRL Monterey)
      • 09:30
        Multi-model Prediction on Subseasonal Timescales at the US NOAA Climate Prediction Center: Approaches to Calibration and the Identification of Forecasts of Opportunity 30m

        Forecasting temperature and precipitation on subseasonal timescales beyond two weeks lead-time is at the limits of predictability and modeling capabilities. The NOAA Climate Prediction Center (CPC) relies on both dynamical and statistical models to make operational and experimental, above and below median, temperature and precipitation forecasts for weeks 3 and 4. Both statistical and dynamical forecast models attempt to utilize the enhanced predictability during active climate events, related to modes of climate variability, such as ENSO and the Madden-Julian Oscillation. Other than the predictability due to decadal climate change, much of the skill of subseasonal forecasts is related to these drivers of climate variability. Furthermore, much of the utility of subseasonal forecasts lies in the forecast of extremes in temperature and precipitation, which by their nature are intermittent and often associated with high amplitude climate drivers. Ensemble models allow for the generation of probabilistic forecasts, while calibration assures reliability of forecast probabilities and correction of model biases. Various methods of calibration and combination to create multi-model ensemble (MME) forecasts have been considered. While a calibrated multi-model ensemble (MME) of dynamical model forecasts has proven to be one of the most skilful tools in CPC operational, subseasonal forecasts, skill remains low for precipitation forecasts and at times, near zero for all forecasts. Therefore, identification of forecasts of opportunity, when predictability is enhanced, could greatly improve the utility of forecasts on this timescale. To analyse the potential to identify forecasts of opportunity, we examine the skill of subseasonal forecasts (including extremes relative to the past climatological distribution) when the signal magnitude is greater, using hindcasts from the S2S and SubX databases. In this way, we examine if intermittent forecasts of larger magnitude signals represent the best opportunity to obtain information for extended-lead subseasonal forecasts (weeks 3-4), including extremes, and determine appropriate metrics of forecasts of opportunity. A multi-model ensemble significantly improves the capacity to identify forecasts of opportunity over individual models.

        Speaker: Daniel Collins (NOAA Climate Prediction Center)
      • 10:00
        Isotonic Distributional Regression (IDR): A powerful nonparametric calibration technique 15m

        We introduce isotonic distributional regression (IDR), a nonparametric technique for the generation of calibrated probabilistic forecasts from numerical weather prediction (NWP) model output. IDR learns calibrated predictive distributions directly from training data, subject to natural monotonicity constraints, without invoking any parametric distributional assumptions.

        In a nutshell, IDR solves a quadratic programming problem to find distributional forecasts that are simultaneously optimal in terms of a wide class of scoring rules, including but not limited to the continuous ranked probability score (CRPS). For output from a single NWP model, IDR honors the compelling constraint that the probabilistic forecasts ought to be ordered in the same way as the model output. When using output from a single- or multi-model ensemble, IDR honors a suitable partial order. For example, we reduce the ECMWF ensemble output to the triple (HRES, CNT, MPM) consisting of the high-resolution member (HRES), the control member (CNT), and the mean of the 50 perturbed members (MPM), and we define a forecast to be larger than another if it is larger in all three components. The method applies to all types of univariate variables, including binary, discrete, continuous and mixed discrete-continuous weather quantities.

        We apply IDR to obtain probabilistic quantitative precipitation forecasts based on the ECMWF ensemble at the airports in Brussels, Frankfurt, London (Heathrow), and Zurich in 2007 to 2017, using rolling training periods. The IDR forecasts outperform the raw ensemble as well as Bayesian model averaging (BMA) and ensemble model output statistics (EMOS) calibrated forecasts. We sketch adaptations to the TIGGE and S2S multi-model ensembles, for which we anticipate
        major gain in predictive performance.

        Speaker: Johanna Ziegel (University of Bern)
      • 10:15
        A Bayesian framework for postprocessing multi-ensemble weather forecasts 15m

        A Bayesian framework for postprocessing multi-ensemble weather forecasts

        Clair Barnes (1), Richard E. Chandler (1) & Christopher M. Brierley (2)
        (1) Department of Statistical Science, University College London
        (2) Department of Geography, University College London

        Ensemble weather forecasts often under-represent uncertainty, leading to overconfidence in their predictions. Multi-model forecasts combining several individual ensembles have been shown to display greater skill than single-ensemble forecasts in predicting temperatures, but tend to retain some bias in their joint predictions. Established postprocessing techniques are able to correct bias and calibration issues in univariate forecasts, but are generally not designed to handle multivariate forecasts (of several variables or at several locations, say) without separate specification of the structure of the inter-variable dependence.

        We propose a flexible multivariate Bayesian postprocessing framework, developed around a directed acyclic graph representing the relationships between the ensembles and the observed weather. The posterior forecast is inferred from the ensemble forecasts and an estimate of their shared discrepancy, which is obtained from a collection of past forecast-observation pairs. The approach is illustrated with an application to forecasts of UK surface temperatures during the winter period from 2007-2013.

        Speaker: Clair Barnes (Department of Statistical Science, University College London)
      • 10:30
        Benefits of a multimodel approach for forecasting precipitation over New Caledonia (SW Pacific) at S2S timescales 15m

        Located in the tropical Southwest Pacific, New Caledonia is prone to heavy rainfall events that may be related either to tropical or to mid-latitude perturbations. Precipitation over New Caledonia should therefore exhibit some subseasonal predictability owing to large-scale drivers, such as ENSO and the Madden-Julian Oscillation. This study aims to assess the skill in predicting precipitation over New Caledonia at the subseasonal range in GCM forecasts from the S2S project through multimodel ensembling. Gridded precipitation data from six S2S hindcasts (BoM, CMA, ECCC, ECMWF, Météo-France and UKMO) are combined on a common period for 12 startdates in the austral summer season December-January-February 1996-2013, which corresponds to the rainiest season. Given that the models do not share a common ensemble size, a subset of 4 members per model is selected so as to build a balanced 24-member ensemble multimodel. Verification is carried out on weekly periods so as to consider time spans for which predictability should be more relevant at intraseasonal lead times (week 1, week 2…). Hindcasts are verified against both gridded precipitation datasets and station data. The skill of the multimodel ensemble is compared to the skill of each individual model using deterministic metrics such as the spatial average of grid-point correlation and a probabilistic approach with the ROC Skill Score based on discrete rainfall categories (e.g occurrence vs non-occurrence of a day with more than 25 mm rain rate within a given week). Results show the multimodel performs alongside the best individual models for most lead times and scores considered. Moreover, the construction of a multimodel ensemble helps assess precipitation skill more robustly than with individual models with a small ensemble size. These first results therefore suggest that using a multimodel approach for intraseasonal rainfall prediction bears some added value compared to using a single model.

        Speaker: Damien Specq (Météo-France)
    • 10:45 11:15
      Coffee break 30m Weather Room

      Weather Room

    • 11:15 12:00
      Prediction and Verification Multi-model approaches to prediction: Chair - Mark Rodwell (ECMWF) Lecture Theatre

      Lecture Theatre

      ECMWF

      • 11:15
        Subseasonal Prediction of European Summer Heat Waves in the S2S Hindcast Ensembles 15m

        Subseasonal Prediction of European Summer Heat Waves in the S2S Hindcast Ensembles

        Authors: Ole Wulff and Daniela Domeisen
        Affiliation: ETH Zurich, Switzerland

        Due to their devastating impact, the prediction of heat waves beyond the weather forecasting range is of great significance to society. The potential for successful subseasonal predictions of summer heat waves stems from e.g. the fact that they are most often related to persistent anticyclonic atmospheric circulation conditions. Especially when these systems are embedded in large-scale teleconnection patterns, it is possible to extend predictable lead times before the heat event. Furthermore, it has been shown that the likelihood of reaching extremely high temperatures increases strongly if dry soils are observed in advance. As soil moisture acts as a memory of the near-surface atmospheric conditions, the coupling between the land surface and the atmosphere could provide further predictability in regions where it is strong.

        In our study, we investigate the probabilistic skill of a subset of forecasting systems from the S2S database in predicting 11 European summer heat events in the period from 1999 to 2010 on the subseasonal time scale. The skill analysis is complemented by an assessment of the drivers of specific heat events and how these are represented in the S2S forecasting systems. To evaluate the ability of the models to simulate the driving mechanisms, we split the hindcast ensembles retaining those members that perform best in forecasting predictor fields such as geopotential height, sea surface temperatures, soil moisture and surface heat fluxes. The skill of the thus created sub-ensembles in predicting 2m temperatures is then compared to the forecast skill of the full ensemble. From this, we assess whether the chosen variable acts as a predictor of 2m temperatures in the different models. Our analysis reveals large differences between the models and single heat events regarding these relationships.

        Speaker: Ole Wulff (ETH Zurich)
      • 11:30
        Experimental subseasonal forecasting of atmospheric river variations for western N. America during Winters 2017-2018 and 2018-2019 15m

        We utilize the Guan and Waliser (2015) atmospheric river (AR) detection algorithm and DeFlorio et al. (2018) AR activity forecast approach on three operational subseasonal forecast systems (ECMWF, NCEP, and ECCC) to predict AR activity over the eastern Pacific Ocean and western N. America, with a focus on subseasonal variations, and in particular the week-3 lead time. The predictand for these subseasonal forecasts is the likelihood of an AR occurring at any time during a given week, with the primary focus being the week-3 window. Forecast verification statistics for these subseasonal AR forecasts will be presented, as well as case studies demonstrating periods where the AR activity appeared to exhibit enhanced/diminished predictability and how well and consistently the models performed during these periods. This is a collaborative activity between the Jet Propulsion Laboratory/NASA, the Center for Western Weather and Water Extremes of the University of California at San Diego, and the Joint Institute for Regional Earth System Science (JIFRESSE) at the University of California, Los Angeles, with sponsorship from the California Department of Water Resources. This activity leverages the ECMWF, NCEP and ECCC hindcasts of the Subseasonal to Seasonal (S2S) Prediction Project, along with their real-time data stream counterparts, and represents one of the S2S Prediction Project’s Pilot Projects for applications use. We will also discuss the “operational” framework for the collaboration, the interferences from the 2017-18 winter, and changes and plans for the 2018-19 winter.

        Speaker: Michael DeFlorio (Center for Western Weather and Water Extremes)
      • 11:45
        Use of S2S forecasts for humanitarian decision making in Kenya 15m

        Many people living in East Africa are significantly exposed to risks arising from climate variability. Droughts and floods in the region are common; poor performance of the rainy season in 2011 and 2016 led to widescale threats to food security and flooding in 2018 led to significant impacts on human life.

        The project ForPAc (Toward Forecast-Based Preparedness Action, funded by NERC & DFID) is working together with humanitarian institutions in Kenya to explore the possibility for taking preventative action based on subseasonal and seasonal forecasts. Such actions may range from early cash-transfers based on months-ahead warnings of drought, through to planning for flood response based on anticipation provided by subseasonal forecasts.

        The work of the project will be highlighted in this presentation, particularly focusing on the 2018 March-May season; the wettest ever recorded in the region when over 300 people lost their lives. The representation of processes leading to predictability at subseasonal and seasonal timescales will be assessed, alongside an evaluation of the performance of S2S models throughout this season as well as a discussion of the potential actions which could have been triggered by these forecasts.

        Speaker: Dave MacLeod (University of Oxford)
    • 12:00 13:00
      Application studies: Chair - Mark Rodwell (ECMWF) Lecture Theatre

      Lecture Theatre

      ECMWF

      • 12:00
        Quantifying and attributing predictable signals on sub-seasonal timescales 30m

        As a user of subseasonal weather forecasts we are particularly interested in quantifying and attributing drivers of enhanced sub seasonal predictability. Here we will discuss methods we have developed to address this question including various approaches to quantity both the signal but also systematic model errors in processes such as regime transitions and tropical-extratropical teleconnections via analysis from empirical models and modelling experiments involving tropical relaxation.

        Speaker: Dan Rowlands (Citadel)
      • 12:30
        Digiscape: A one-platform solution for seasonal climate integration into Agriculture. 15m

        Weather and Climate information lies at the heart of managing agriculture. Decisions that need to be made on farm are wide-ranging from hourly spraying conditions to longer term crop choices or stocking management. Beyond the farm gate weather and climate influence shipping strategies and supply chains, insurance premiums and commodity markets. Australia’s CSIRO has invested in a Future Science Platform known as Digiscape to build the infrastructure common to all these applications. While each domain has specific needs, they all require a bank of weather and climate data that can be manipulated easily and fed to appropriate agronomy models. Here we discuss the value in taking a one-platform approach to delivering actionable knowledge to many different agricultural applications, and the value that can bring. One focus is combining sub-seasonal to seasonal climate forecasts with sensor data for a model-data fusion approach to agricultural forecasting. Examples are given for generating crop yield estimates for farmers and shipping handlers, irrigation decision tools, pasture prediction, and tactical bespoke data feeds for smaller horticultural industries.

        Speaker: Jaclyn Brown (CSIRO Agriculture and Food)
      • 12:45
        A flood alert system for Switzerland based on integrated water vapor fluxes 15m

        Floods in the Alpine region can be destructive and lead to large losses. Many rivers and lakes in Switzerland, however, are regulated and flood damage can therefore be mitigated through an optimal management of lake levels and runoff. To support planning and prevention, high-quality forecasts of atmospheric flood precursors extending beyond short-range predictions are expected to be useful. One such flood precursor is the integrated vapor transport (IVT), which has been shown to be causally related to flooding in the Alpine region when transport is perpendicular to the main orography. We therefore develop a flood alert system based on medium-range forecasts of IVT.

        We present the verification of probabilistic medium-range forecasts of IVT and precipitation by the European Centre for Medium-Range Forecasts (ECMWF) Integrated Forecasting System (IFS). Based on 20 years of reforecasts, we show that both regular and extreme IVT is better predictable than precipitation and IVT predictions are skillful out to day eight. As the direction of IVT is of central importance for flood risk in Switzerland, we develop an index summarizing probabilistic information on both direction and magnitude of the IVT. Together with compact visualizations this forms the basis of an operational flood alert system extending beyond the range of traditional weather forecasts using only atmospheric flood precursors.

        Speaker: Jonas Bhend (MeteoSwiss)
    • 13:00 14:00
      Lunch break 1h
    • 14:00 14:15
      Description of working groups Lecture Theatre

      Lecture Theatre

      ECMWF

    • 14:15 15:30
      Working groups
    • 15:30 16:30
      Posters and coffee break 1h Lobby/Weather Room

      Lobby/Weather Room

    • 16:30 18:00
      Working groups
    • 09:00 10:30
      Application studies: Chair - Andrew Robertson (IRI) Lecture Theatre

      Lecture Theatre

      ECMWF

      • 09:00
        Transmuting S2S forecasts into applications 30m

        Considering lessons learned from experiences at seasonal timescale, this talk discusses some concrete S2S applications using both calibrated and uncalibrated forecasts from the S2S Database and the SubX project.

        First, we illustrate how a Python interface for IRI’s Climate Predictability Tool —PyCPT— can be employed to assess skill and calibrate sub-seasonal forecasts in ways that are useful for the development of S2S societal applications.

        Then we present how a combination of seasonal and sub-seasonal forecasts can be used to identify onset and demise of rainfall seasons and the mid-summer drought in Central America, and how that information is being used locally to make decisions for the food security sector in the region.

        Finally, the talk will discuss how S2S forecasts are being used to co-develop drought Forecast-based Early Warning and Forecast-based Financing systems in developing countries.

        Speaker: Ángel G. Muñoz (IRI - Columbia University)
      • 09:30
        The S2S4E project, sub-seasonal to seasonal climate predictions for energy 15m

        The S2S4E project aims to bring sub-seasonal to seasonal climate predictions to the renewable energy sector. Raw climate predictions come with a set of challenges which require the deployment of a climate service in order to produce valuable information for users. This involves the development of robust methodologies to calibrate predictions, a quality assessment, and the effective communication of the prediction products, as well as their expected added value. This work is done within the S2S4E project with focus in different areas of the energy sector. The main outcome of the project will be the provision of real-time forecasts of essential climate variables as well as energy indicators through a decision support tool that is being co-developed with users.
        To illustrate the potential benefits of S2S predictions several case studies have been analysed, i.e. periods pointed out by the energy companies as having an unusual climate behaviour that affected the energy market. Two of these case studies will be presented to analyse how the climate predictions issued several weeks ahead of each event would have helped the stakeholders in decision making. In the first case, a cold wave over France and Germany in January 2017 increased the electricity demand while low wind speeds limited the renewable energy production. In the second case, a heat wave affecting Spain at the beginning of September 2016 increased the electricity demand. Sub-seasonal predictions from ECMWF monthly system issued from 4 weeks to 1 week prior to these events were calibrated with a variance inflation method using the corresponding 20 year hindcast and presented as a probability distribution with associated skill. Results show S2S predictions have potential to anticipate episodes of high electricity demand a few weeks in advance although there is still limited confidence in predicting the energy supply beyond week 2.

        Speaker: Andrea Manrique-Suñén (Barcelona Supercomputing Center)
      • 09:45
        Drought Monitoring and Prediction Using Sub-Seasonal Predictions 15m

        Accurate real-time monitoring and prediction of drought on a sub-seasonal timescale enhance our capability of drought risk assessment and management. This study investigated the feasibility of the sub-seasonal drought prediction using dynamical model outputs. To this aim, we adopted the Keetch-Byram Drought Index (KBDI) to estimate soil moisture deficits and used real-time and historical satellite-based measurement of precipitation, GSMaP, provided by JAXA, and historical and real-time climate analysis, JRA-55, and sub-seasonal ensemble prediction data provided by the WWRP/WCRP Sub-seasonal to Seasonal Prediction Project (S2S). We first verified the sub-seasonal predictive skills for precipitation using reforecast data. We found that the skills are generally higher over the ocean than land; the skills are moderately high but in numerous terrestrial regions. Then we assessed the predictive performance of KBDI and found much higher predictive capability (Kendall rank correlation coefficient) for KBDI than that for precipitation due to slow variability and the predictability originated from a memory of soil moisture. We identified potential regions where the dynamical model outputs can enhance the predictive skill of KBDI. In boreal summer, for instance, these regions include Southeast Asia, western United States, northern South America, Australia and Central Africa. Some of these regions are susceptible to droughts and wildfires, thus results suggest that the new products developed in this study are potentially useful for decision making of wildfires and agriculture. We further investigated Indonesian drought. Remarkable events are dry and wet conditions in 2006 (Eastern-Pacific El Nino and positive Indian Ocean Dipole (IOD) case) and 2010 (negative IOD case) and these events were well captured by the ECMWF model, suggesting that El Nino and IOD provide the subseasonal predictability in the Indonesian archipelago.

        Speaker: Yuhei Takaya (MRI/JMA)
      • 10:00
        Developing capacity of Southeast Asian countries to apply subseasonal-to-seasonal forecasts in impact forecasting tools 15m

        The skill of subseasonal to seasonal forecasts models is relatively high for Southeast Asia; however, uptake of such information by the disaster risk reduction community is still in its infancy. As part of the implementation of the ASEAN-UN Joint Strategic Plan of Action on Disaster Management (2016-2020), recommendations were made to build capacities for using S2S products at the timescales of 2 weeks to 1 month. By incorporating weather and climate information at the S2S timescale into the decision making processes, it is hoped that this will allow for pre-emptive action and mid-course corrections in tandem with long-range forecasting and near-real time monitoring, as well as other sources of information.

        As part of an effort to demonstrate applicability of S2S products in Southeast Asia, this work revisits three recent disasters related to heavy rainfall and drought, exploring how such forecasts could have potentially been used to support decision makers. Along with assessing the forecast from various models in the S2S Database, risk-sensitive decisions on the S2S timescale are also identified, exploring how S2S forecasts could have potentially strengthen these decisions, along with risk assessments and early warning services in Southeast Asia in general.

        The results from this study are expected to be incorporated into the S2S-Southeast Asia (S2S-SEA) Capability Building Project (2017-2020), which aims to familiarise Southeast Asian National Meteorological and Hydrological Services with S2S products, as well as to equip them with skills to support end users in their own countries and regions.

        Speaker: Thea Turkington (Meteorological Service Singapore)
      • 10:15
        Subseasonal forecasting for the telecommunication network 15m

        Telecommunication networks are integral part of secure and competitive societies where commercial enterprise and essential services depend on low-cost and reliable communications. In the UK, an estimated net economic contribution of £33bn/year (or 1.5% of GDP) is attributable to telecommunications infrastructure, however, as with many other aspects of infrastructure, the exposed nature of the network leads to weather risk. The quantification, prediction and management of weather-related line fault rates is therefore an important problem, with each aspect – quantification, prediction and management – presenting distinct challenges.

        With unique access to observational records for the UK telecommunications infrastructure, this presentation will address all three aspects outlined above, providing an end-to-end demonstration of how subseasonal meteorological forecasts might be used in an important practical setting, assessing forecast value to the end user on both short term ‘operational’ (days, weeks) and longer term ‘planning’ timeframes (months, years). The user applications discussed – the latter of which involves simulating operational decision-making as well as the weather - shares similarities with other end-user applications in energy-meteorology. On this basis, it is suggested that there is a need for new techniques in assessing the quality of s2s forecasts going beyond the traditional “cost-loss” framework.

        Speaker: David Brayshaw (University of Reading)
    • 10:30 11:00
      Coffee break 30m Weather Room

      Weather Room

    • 11:00 12:00
      Working groups
    • 12:00 13:00
      Plenary session Lecture Theatre

      Lecture Theatre

      ECMWF