Assessment of selected global models in short range forecasting over West Africa: Case study of Senegal
Sadibou Ba (ANACIM)
Over the past decade, flooding has been increasingly frequent in the Sahel, especially during the rainy seasons of 2005, 2008, 2009, 2010. More recently during the summer season of 2012, devastating flood events have been recorded in Senegal, causing widespread severe damages including the loss of life and property, and the displacement of several people. It became increasingly apparent that the frequency of extreme events is associated with global warming. Severe floods in West Africa are associated mainly with meso-scale thunderstorms and squall lines. Forecasting such severe weather to reduce the risk of hazards is one of the challenges faced by many National Meteorological and Hydrological Services (NMHSs). However, recent progress in numerical weather prediction, the acquisition of new equipments and collaboration with the international community through capacity building and training led to improvement of the quality of the forecasts and to development of early warning systems. This presentation contains inter-comparison studies using selected Global models (ECWMF, GFS, UKMET) and attempts to evaluate the strengths and weaknesses of each model with respect to three rainfall events. Thus, the results of this preliminary work indicate the importance of the use of ensemble forecast system in West Africa. We will focus also the role of the ensemble forecast for the success of the SWFDP-West Africa (Severe Weather Forecasting Demonstration Project).
The different application of the ECMWF Short-Term Forecast model in the management of renewable energies
Lucia Benito Capa (Iberdrola Renovables)
The social conscience and the irrefutable fact of the negative effect that has the use of non-renewable energies in climate change make renewable energies the present and the future.
Iberdrola, committed to the protocols to reduce greenhouse gas emissions, focus its day a day on the management of wind, solar and hydraulic energy, in short, renewable energy.
For efficient energy management, Iberdrola bases a large part of its R + D + i on quality weather forecast based largely on the ECMWF short-term forecast models.
A quality weather forecast is essential for:
• Good planning of the maintenance of the different facilities
• A secure interaction in the energy market
• A reliable diagnosis of possible incidents
To facilitate these tasks, Iberdrola presents the available information in a graphic way, making it possible to streamline the work
The forecast value of global flood forecasts in Uganda
Leonore Boelee (HR Wallingford & Open University)
There is an increasing interest in mitigating the effects of fluvial flooding using global forecasts for disaster risk management in developing countries with low data availability. In order for the forecasts to have impact and value they need to be skilful enough for mitigating action to be successfully triggered. However, the application of global forecast products to mitigate the effects severe weather remains limited. To increase the impact of forecasts products in mitigating severe weather in these data scares regions a greater understanding of the influence of model skill and uncertainty on the forecast value is required. Coughlan de Perez et al (2015) have developed a method to link together forecast and appropriate humanitarian action in data scares regions using a pilot project which focuses on two locations in North Eastern Uganda (Magoro and Ngariam), working with the Uganda Red Cross Society using forecast from the Global Flood Awareness System (GloFAS). However, there is a lack of understanding of how uncertainties impact the forecast value for a wider range of flood events, locations, catchments sizes and forecast lead times for detailed area and application, like the Uganda forecasts. Forecast uncertainty is expected to increase with increasing lead time and decreasing catchments size. Therefore, forecast uncertainty and skill will be variable for different lead times and locations within the Ugandan case study region and also wider across the globe. The understanding of these uncertainties can contribute to being able to find the optimal balance where increasing forecast uncertainties have not decreased the forecast value. This poster aims to answer the question: ‘How do the model skill and uncertainty affect the forecast value in North Easter Uganda?’
Efforts for Mitigation Weather Hazard Impacts in Portugal
Sandra Correia (IPMA)
Nowadays, there is an increasing need to raise population awareness to weather warnings and therefore to weather impacts. For a particular national meteorological service, this is best achieved through cooperation with regional, national and international institutions.
At the Portuguese Meteorological Institute, information interchange is performed with Civil Protection – ANPC (extreme weather events on land and ocean), Navy (issues on warnings in ocean and coastal areas), National Health Institute – DGS (heat and cold waves) and the National Forest Institute – ICNF (during forest fire critical seasons).
Information is currently disseminated on the IPMA website, radio, social networks and television and can be used to address impacts, such as forest fires, dry seasons, basin management due to flooding, and coastal zones hazards.
There are also requests on information by private companies due to extreme weather events, namely those affecting energy and telecommunications networks.
At IPMA, derived products from ECMWF fields are being produced, such as lightning probability, total accumulated precipitation in basins, soil water content, standardized precipitation index, and other that will be presented.
Some impact statistics will be shown regarding, reports that have been requested to IPMA after storms have affected mainland Portugal, essentially due to impacts of wind, precipitation and lightning, but occasionally other phenomena have been included in the request, such as forest fires and freezing rain.
Portugal is also participating in the WGCEF (Working Group on the Cooperation between European Forecasters of EUMETNET) task team on storm naming in Europe since late 2017.
IPMA has recently finished a 2 year participation in ARISTOTLE – All Risk Integrated System TOwards Trans-boundary hoListic Early-warning project, that provided multi-hazard expert information to support the Emergency Response Coordination Centre (ERCC) of the European Commission.
Portugal cooperates with the Tsunami Early Warning Service Providers in the NEAM regions, which will issue alert messages in case of a tsunami on or nearby Portuguese shores. This new service will considerably increase Europe's capacity to issue tsunami alerts to its citizens, and is in operation since late 2017.
Direct Model Outputs Versus Statistically Postprocessed Outputs to Forecast Low Visibility
Sebastian J. Dietz (University of Innsbruck)
Low visibility conditions at airports require special procedures that reduce the operational flight capacity. The capacity reductions are defined with the low-visibility procedure (lvp) states which are determined by combinations of visibility and ceiling thresholds. This study compares the performance of direct model output, statistically postprocessed model output, and statistically processed observations for lvp state forecasts at Vienna Airport with lead times from 1 hour up to 15 days.
The models used for statistical processing are boosting trees, which work quiet well for forecasts of the lvp state. For lead times shorter than 6 hours forecasts from statistically processed observations provide highest benefit. With longer lead times models based on NWP output perform better. Therefore postprocessed HRES information has higher benefit for lead times up to 1 day and afterwards models with ENS output perform more accurate. Direct output of the ENS, however, has only a small difference to the performance of the statistical models for lead times between 2 and 5 days. The difference in forecast performance between the best models and climatology vanishes after a lead time of approximately 8 days. NWP outputs contributing most to the skill of the statistical models are dew point depression, boundary layer height, evaporation, and sensible heat flux.
Assessment of statistical post-processing of monthly forecast
Nicole Girardot (Météo-France)
A set of 2 years of statistical post-processing of 2m-temperature monthly forecasts over France has been studied. This set correspond to current version of monthly forecast up to 46 days. The goal is to explore the ability of this system to forecast temperature anomalies. With basic scores (detection rate, false alarms), different aspects are tested : impact of space or time averaging, impact of using quantiles rather than ensemble mean, significance of large forecast anomalies. The results show some systematic defaults of the forecasts but also what are the best capacities of the system. This study gives also the opportunity to list occasional bad medium-range forecasts.
Forecast Error Analysis of a Persistent Heavy Rainfall Event
Yue Guan (CMA)
Errors occur for persistent heavy rainfall over Yangtze river in China from 22 to 27 June 2017 by ECMWF ensemble model. By using conventional data, ECMWF ensemble model and NCEP ensemble forecast data, forecasts of cumulative rainfall and cause of the forecast error is analyzed. The results reveal that (1) The location of the rain belt by NCEP ensemble model forecast is closer to the observed rain belt than ECMWF. However，after 25 June EC model adjust the location of precipitation southward，thus closer to the observed precipitation. (2) Compare wind at 850hPa with NCEP model，the wind shear resulting in heavy rainfall from ECMWF model is obviously located more north than the observed wind and NCEP forecast. (3) The north edge of the Western Subtropical High from EC’s ensemble mean at 500hPa is also predicted more north than NCEP, but the error is slighter compared with low level wind. Thus, the forecaster at 8:00 BT 22 JUNE could take advantage of the middle-level and high-level circulation to adjust the location of precipitation.
Applying new 3-D jet core visualisation techniques to the study of extreme cyclones
Tim Hewson (ECMWF)
Wide-ranging scientific advances are making real-time use of 3D visualisation in meteorology increasingly attractive. Improved computer technology, mainly derived from the gaming industry, and encapsulated in modern-day graphics cards, now allows for transparency, fly through, re-orientation and animation of complex 3D model atmosphere scenes in real time. Meanwhile one can adapt relatively recent mathematical techniques to represent the salient atmospheric features (e.g. cyclones, jet cores, trough lines, fronts) as points, strings or surfaces in 3D space. These algorithms compress large volumes of synoptically relevant gridded data into a very compact yet meaningful form, and likewise dramatically reduce obscuration of other aspects of interest. Together these advances can allow researchers (and indeed forecasters) to quickly establish a clear 3D picture of the key features of the model atmosphere, and their evolution in time. For some aspects one can even visualise the 4D behaviour of multiple ensemble members. ECMWF is beginning to integrate these approaches into its investigations of forecast performance.
This poster illustrates how the above developments are brought together in the open-source, interactive 3D meteorological visualisation tool “Met.3D” (http://met3d.readthedocs.org), showing in particular how a new algorithm for identifying jet cores in 3D as “strings” performs in this environment. The jet core algorithm extends and adapts 2D jet core mathematics to 3D. It derives from a momentum-based definition of a jet core line; the algorithm is described.
A case of an extreme extra-tropical cyclones (Xavier, 5 October 2017) is used to illustrate applications, and the benefits of 3D versus 2D visualisation. This shows in a clearer way than hitherto the dynamical links between extreme cyclogenesis events and “vertically stacked” jet cores. The plots can also provide helpful pointers to the mechanisms of downward momentum transfer in sting jet regions. There is also an illustration of how jet stream core ‘bundles’ can be derived from the 51-member ECMWF ensemble and portrayed in a meaningful way.
Performance of forecast deep convection objects as inferred from ECMWF simulated infrared radiances
K. K. Hon (Hong Kong Observatory)
Deep convection, with its associated hazards of lightning, convective turbulence and possibly hail, is one of the aviation-impact weather phenomena for which SIGMET warnings are issued operationally. Due to a lack of radar observations over vast ocean areas, a commonly-accepted means of identifying regions of significant convection on a regional/global scale would be through combination of sensitive frequency channels of infrared sounders on board geostationary satellites.
This study verifies the forecast deep convection “objects” as inferred from IR1 and IR3 channels of the ECMWF-IFS simulated radiance output against corresponding observations from the Advanced Himawari Imager (AHI) of the Himawari-8 satellite over the East Asia and western North Pacific regions.
Hydrometeorological drivers of 2017 Flood in Bangladesh and associated forecasting skill
Sazzad Hossain (University of Reading)
Flood is the most common natural disaster in Bangladesh which occurs almost every year and causes huge economic losses. There are several reasons that cause flood-geographical location, topography, monsoon climate etc. Transboundary flow is the major source of flood water. The characteristics of monsoon river flood of 2017 is a representative severe flood in terms of duration and magnitude. Several parts of the country experienced flooding for different duration. The Brahmaputra river basin (northwest region) experienced flood in two times- July and August and the Meghna basin (northeast region) experienced flash flood in early April in 2017. Several rivers of the country exceeded their previous historical flood level.
The study identified several meteorological and hydrological characteristics which were responsible for severity of flood. The excess rainfall in pre-monsoon (April-May) provided additional soil moisture, monsoon onset at the beginning of June and two extreme rainfall events northwest region of Bangladesh as well as adjacent catchments to the international border are some major meteorological drivers. The unusual flood water rise was recorded in the rivers of Brahmaputra basin and rivers were reached to their danger level within short period of time. The abnormal rise of water in the Brahmaputra is a new dimension of flood in Bangladesh. Bangladesh has adopted both structural and nonstructural measures for flood management. Flood forecasting is considered as a major non-structural flood management approach. Flood Forecasting and Warning Centre (FFWC) in Bangladesh is responsible for flood warning services, and it provides 3 to 5 days deterministic forecast. Extended range forecast provided by GLOFAS is also available for the major river basins in Bangladesh and FFWC took the advantage of this forecast in flood warning dissemination. Forecast skill has been assessed by mean absolute error (MAE) and coefficient. The present research findings show that extended range forecast provided by GLOFAS was very consistent. Deterministic forecast shows very good correlation with the observed data. However, performance of the deterministic forecast depends on the correct boundary estimation of the model. The objectives of the present study are to investigate different hydrometeorological drivers which causes flood, and its forecasting skill.
Evaluation of IFS and AROME day-ahead and very short term GHI, DNI and DIF forecasting skill
Sophie Martinoni Lapierre (Météo-France)
In the current context of energy transition, the amount of power produced by photovoltaic farms is rapidly increasing and raises the question of the integration in the grid. The electricity production and consumption need to be fairly equal at any time. However, unlike other energy sources, PV production depends substantially on meteorological conditions (irradiance and temperature) which means that PV farms generate only intermittent energy. To reconcile these two constraints it is becoming essential to precisely forecast the PV power production in order to anticipate its variations and better monitor the grid.
The energy division at Météo-France is developing, in connection with the French National Meteorological Research Center (CNRM), a PV power production forecast product that uses mainly numerical weather prediction model outputs. For this purpose it is a major stake to evaluate the errors in the models and especially in the solar irradiance forecasts.
The quality of the day ahead and very short term (<6h range) global, direct and diffuse irradiances from IFS and AROME (the high-resolution model from Météo-France) have been evaluated. Scores have been computed over France using two years of data from Météo-France’s ground stations network. Discrimination of cloudy and clear sky simulation is applied. The clear sky situations have been filtered to quantify the impact of the errors of the cloud forecasts on irradiance forecasts and to study the behaviour of the radiation scheme without interactions with clouds.
This study shows for instance the large influence of cloud positioning errors on the quality of the irradiance forecasts and a bias in the distribution between the direct and diffuse components in clear sky conditions (underestimation of the direct component).
Ensemble subsetting for dynamical downscaling of global seasonal climate predictions: Evaluation for selected semi-arid regions
Patrick Laux (Karlsruhe Institute of Technology (KIT) Institute of Meteorology and Climate Research)
Until 2025 approximately 1.8 billion people are expected to suffer from absolute water scarcity. It is known that an improved water management, with certainly the highest potential in semi-arid regions, can help to mitigate water scarcity. Seasonal climate forecasts may provide crucial information for water management, but globally available products suffer from a too coarse resolution to support decision making on local scales. Since initial condition are not perfectly known, perturbed simulations are conducted to derive uncertainties, resulting in forecasts consisting of relatively large ensembles. Dynamical downscaling provides a suitable approach to bridge the scale gap from global to local, and is preferential over statistical approaches if subsequent impact models are being applied. The reason is that dynamical downscaling is providing physically consistent hydrometeorological input variables. On the other hand, it requires large computational resources, which may permit the evaluation of the full downscaled forecast ensemble.
Embedded in the Seasonal Water Resources Management (SaWaM) project, this study evaluates whether or not single members from the seasonal ensemble forecast can be omitted. Based on the assumption, that a poor performance in the global forecasts leads to a poor performance in the downscaled forecasts, an ensemble subsetting approach is applied and evaluated for semi-arid regions in Brazil, Iran, and West Africa. For this reason, the hindcasts of ECMWF system 4 and 5 data and the interpolated precipitation observation products GPCC and CRU are used and analyzed for the wet season of period 2001 to 2010. The procedure and results of the subsetting approach will be presented and discussed in the context of decision support in water management.
Use of ECMWF IFS forecast for the provision of flight specific turbulence forecast
Jeffrey Chi Wai Lee (Hong Kong Observatory)
High impact weather including turbulence, significant convection and high altitude Ice Crystal (HAIC) may bring hazards to aircraft operations in en-route phase as ascend and descending phases. Under the modernization of the International Civil Aviation Organization (ICAO) global aviation navigation plan (GANP), airlines and air traffic management requires more flight specific information in preparing better flight plans, tactical/pre-tactical flight operations as well as better air space, airport flight management. This paper presents the trial use of ECMWF IFS data to provide flight specific forecast turbulence products based on the flight route specified in the flight plans of the commercial flights. The performance of the forecast product is verified against the in-situ measurement recorded by the quick access recorder (QAR) on board the commercial aircraft. More than 170 selected flight datasets are used in the study for tuning the forecast product as well as for the verification. The calibrated turbulence indices might contribute to the development of next generation of the global turbulence guidance product. The methodology and preliminary results will be presented in the paper.
Ensemble-based predictability and diagnostics of tropical cyclone outflow and structure change during the 2017 Atlantic hurricane season
Sharanya Majumdar (University of Miami)
A focus of the United States Office of Naval Research (ONR) Tropical Cyclone Intensity (TCI) initiative is to better understand the role of outflow in tropical cyclone (TC) structure and intensity change. An objective of this study is to use ECMWF ensemble data to diagnose the establishment of outflow channels for 2017 Hurricanes Harvey, Irma, Jose and Maria, via interactions with the large-scale environment. The influence of these interactions on TC structure change will be quantified, together with their predictability. Conclusions from an earlier study on Hurricane Joaquin (2015) suggested that the predictability of the TC motion and intensification was largely dependent on the initial vortex structure, and that the interaction with an upstream trough was responsible for enhancing outflow in the north-western quadrant and intensifying the hurricane. It remains to be determined whether the outflow plays an active (causal) role in TC intensification, and whether the conclusions can be generalized. The 2017 hurricanes offer an opportunity to explore these questions in more breadth and depth.
Application of ECMWF products in winter weather situations at OMSZ
Andras Mesterhazy (OMSZ)
To forecast weather situations in winter - especially concerning the precipitation type around 0°C - is one of the most difficult tasks in meteorology. This is the case in the Carpathian Basin, and in Hungary as well. There is a very big difference in the expected weather character and impact depending on snow, sleet or rain. In the Hungarian Meteorological Service (OMSZ), a number of products based mainly on the ECMWF model have been developed and introduced to operational work in the recent years, hence these weather situations can be better identified by public and our customers as well. These new developments will primarily help to predict the precipitation type, and to specify the areas affected by snowdrift with higher precision.
How the uncertainty of a forecast or warning could be communicated
Thomas Schumann (Deutscher Wetterdienst)
Uncertainty of warnings and forecasts and how to communicate that.
Early heat warnings for European workers
Jan Rajczak (MeteoSwiss, Analysis and Forecasting)
The European HEAT-SHIELD project aims at increasing the thermal resilience of European workers in the context of global warming. As part of technical solutions to counter the heat-induced risk to workers’ health and productivity we aim at providing robust meteorological heat predictions on different temporal scales and tailor the forecasts to specific needs of key European industries. We here present a prototype system of a European early-warning system for heat stress episodes several weeks ahead based on the IFS extended range forecasts of ECMWF. The wet bulb globe temperature (WBGT) is applied as primary heat stress indicator. Ensemble forecasts of WBGT are used to derive daily probabilities of WBGT exceeding certain thresholds. The choice of appropriate thresholds allow to produce tailored heat risk predictions for different working sectors in consideration of different work intensities. Based on forecasts of the past 20 years, we analyze the performance of the predictions against an extensive European-wide observation data set. The results demonstrate that appropriate post-processing of forecast model output is essential for achieving skillful heat event predictions beyond 10 days lead time. Best forecast performance is found around the Baltic Sea and in Eastern Europe where skill (referenced against climatological forecasts) extends to lead times of about 20 days.
Impact of ECMWF products in Energy forecasting
Andrei Steau (NEAS Energy)
Compared forecast between HRES and global model Maps indicating the real production and the impact of the maps need to know and need to do.
ECMWF products at MET Norway
Vibeke Thyness (Norwegian Meteorological Institute)
The poster gives a brief overview of the main uses of ECMWF products at MET Norway
AutoAutomatic gale warning proposals for Swiss lakes and airports
Christophe Voisard (Federal Office of Meteorology and Climatology MeteoSwiss)
The Swiss national meteorological service has the duty to issue warnings for wind gusts exceeding 25 knots 30 minutes to 3 hours before such event occurs at regional airports and relevant lakes. In order to support the forecasters duty an automatic system combining model output and continuous measurements of atmospheric data has been set up. The system is composed of a set of algorithms developed with passed data, combined with logistic regression. Overall the algorithms obtain significantly higher hit rate than model output or forecaster. However the large proportion of false alarms remains an unsolved issue. In daily operations the automatic system issues proposals which the forecaster on duty can accept or reject. This procedure combines the high hit rate of the algorithms with the expertise of the forecaster. This strategy helps improve the warning system.
Use of ECMWF Model Data in support of Seamless Forecast of High Impact Weather at the Hong Kong Observatory
Wai Kin Wong (Hong Kong Observatory)
The Hong Kong Observatory (HKO) makes use of ECMWF model data to support its weather forecasting and warning service. With advances in data assimilation, model dynamics, physical processes and ensemble forecasting, the ECMWF deterministic model and ensemble prediction system (EPS) demonstrate increasing level of skills in short-range to medium-range forecasts of high impact weather such as that of tropical cyclone (TC). This presentation will outline the applications of ECMWF data, post-processing techniques of deterministic and EPS products, and verification of TC track and intensity forecasts.
In recent years, new forecast products such as the Extreme Forecast Index (EFI) and the Shift-of-Tail (SoT) are found to provide useful reference about the chance of extreme cold and hot days, or extreme rain episodes. In supporting forecast assessment of significant convective weather, post-processing techniques are developed for EPS data to generate thunderstorm potential (PoTS) and convective diagnostics using ingredient-based approach, which enable forecasters to better assess the possible scenarios and chance of severe convection.
Recently, HKO has launched a new experimental extended outlook to provide probabilistic forecasts of the daily minimum and maximum temperatures out to the next 14 days and the TC track probability for the next 9 days, paving way for the development of probabilistic and high-impact weather services in the forthcoming years.
New research development on ECMWF model products will also be discussed. For instance, the use of clustering technique to select EPS members with better similarity to the actual rainfall is found to give a more skillful quantitative precipitation forecast in short-range and possibly useful for blending with radar or satellite nowcast products. Application of new verification methodology to understand model forecast performance of TC wind structure will be illustrated as well.
ECMWF IFS Bias Analysis on Vortex Shear Line Rainstorm in Southern China in 2016-2017
Jun Xu (National Meteorological Center of China)
Vortex shear line rainstorm in southern china, as the most important precipitation system in CMA forecast operation, often leads to thousands of sufferers and huge economic loss. Since ECMWF IFS resolution was updated to 9km in 2016, bias statistics revealed that most forecast rain bands above 50mm per day laid north to the observation by tens of kilometers even in 36h forecast time, which troubled forecaster. Analysis was conducted on aspects of mesoscale convective system initiation, maintenance and propagation. Results showed that temperature and moisture deficiency in low level troposphere gave rise to the low level CAPE deficiency. Hence either stronger wind convergence or longer way of lifting was needed to trigger convection. Then precipitation initiation location was behind the observation and started later. Low level CAPE deficiency could also make mesoscale convective precipitation in the warm sector ahead of the front missed. This problem often occurred in southwest part of China in the outset of rainstorms, which could be seemed as the source of the bias. Convection was weaker in the model on account of CAPE deficiency and convection parameterization that could not remove excessive CAPE and moisture. Then excessive grid-scale precipitation leading to mesoscale cyclogenesis and stronger vortex formed. In forecast, mesoscale convective system structure was not clear on account of weak cold pool and downdraft, which also made the precipitation system move slowly. Finally, forecast rain bands fell behind. In conclusion, convection strength and location estimation is important in operation meanwhile simulated infrared radiation product is efficient. More ECMWF products showing convections such as simulated radar echo are necessary.