Hydrological Services for Business workshop - 8-9 May 2018

Workshop summary

The workshop, attended by 55 participants from 17 countries, offered the opportunity for developers and users of GloFAS to meet, exchange ideas and influence the future directions of GloFAS and of hydrological services relying on it.

The 1.5-day workshop was structured as a mix of oral and poster presentations, interactive sessions, products training and discussion forums, to maximize the interactions between participants and the collection of feedback about GloFAS products (see agenda here). All presentations are publicly available online at https://www.ecmwf.int/en/learning/workshops/workshop-hydrological-services-business.

Session 1: Hydrological services and GloFAS

GloFAS (the Global Flood Awareness System) is an operational hydrological service of the Copernicus Emergency Management Service of the European Commission. It provides quick overviews of upcoming flood events and river outlooks up to 4 months ahead, served through an interactive forecast viewer. GloFAS products and services are available to all without restrictive use, except for Europe, through dedicated ftps tailored to each user need.

GloFAS is co-developed by the European Commission Joint research centre, ECMWF and the University of Reading, with support from downstream users such as the Red Cross Red Crescent Climate Centre (RCRCCC), the Regional Integrated Multi-Hazard Early Warning System (RIMES) and the National Centre of Natural Hazard Monitoring and Waring in Brazil (Cemaden), among many others.

For more information and registration for access to GLOFAS services, please visit: http://globalfloods.eu

Session 2: Defining business needs

Users’ needs must be at the centre of any hydrological service. But what are those needs? Following presentations by two different sectors (humanitarian aid and consultancies), an interactive session took place, with participants divided according to their activities.

Interactive session: hydrological services for business

The exercise was designed to find out important criteria defining a good Global Hydrological Service (GHS) through five leading questions, and ranking of answers by participants. The exercise was conducted by five self-governing groups defined by sectors: Users (19 participants); service providers (16 participants); Academia (8 participants); GloFAS Operational team (8 participants); GloFAS Development team (4).

Headline message

3 sectors or more 2 sectors
What is a GHS used for? Local decision making Provision of information in regions lacking data Scientific and retrospective analysis Knowledge exchange
How to select a GHS? Quality of forecast Documentation Service reliability Cost Availability of training material Data accessibility Resolution of forecasts Contextual information
What are the barriers of use of a GHS Access to observations Lack of awareness of the system Multiplication of sources Cost and resource for use Lack of local knowledge Perceived insufficient skill of forecasts Reliability of the service Data format
Elements missing from current GHS Access to observations Local information integration Data assimilation techniques
Criteria to define a good GHS Overall forecast skill Forecast accuracy Forecast reliability Reliability of service

Session 3: Shaping the future of GloFAS

Co-design and co-development of their services is at the heart of the Copernicus Emergency Management Service for floods. This is achieved through an independent advisory board composed of academics and users, regular meetings and workshops, and consultations. Suggestions from users inform the possible evolution of the service.

After presentations from a service-design point of view and from the GloFAS operational team, four main functionalities were discussed in self-organised break-out groups through a set -of table top exercises designed to understand the workflow associated with different products and how the associated service could be made more efficient.

GloFAS 30-day forecasts feedback

The GloFAS 30-day flood forecast product was the first to be implemented operationally in GloFAS, and was designed during the pre-operational phase around 2011. The main aim of the exercise was to gather feedback on the usability of these products, what works, what is missing or incomplete, how to improve current products and what new products would help better support decision making.

The exercise required participants to advise on the planning or cancellation of a ministerial visit to India’s Gujarat state, based on 30-day GloFAS forecast products with lead time of 2 weeks, 9, 5 and 1 day. The criteria for cancelling the visit was the exceedance of the 1-in-20-year flood return flood period threshold.

All groups Two groups One group
Current performance Good probability 2 weeks ahead Forecast accuracy greater at shorter lead times
Map viewer Animated forecasts missing Colours not intuitive
Information type More / flexible reporting points Forecasts should be displayed at catchment scale
Missing Information on confidence / uncertainty Time-to-peak maps Daily rainfall forecast
New product Summary bulletins for decision makers Multiple thresholds / composite maps Risk maps Vulnerability indices Pluvial flooding

GloFAS Seasonal

GloFAS Seasonal was released in a quasi operational mode in November 2018, after two years of development. The design of GloFAS Seasonal was based largely on feedback obtained on EFAS Seasonal, the hydrological seasonal outlook operational on the European domain from the CEMS-Floods EFAS service. The aim of the exercise, organised as an open discussion, was to gather feedback on the current products and how they could evolve.

Headline message

Can the GloFAS Seasonal products help reach an actionable decision?
Issues Solutions
Product not yet trusted and used because lack of skill Information on skill / how to make decisions
Temporal aggregation (outlooks averaged by weeks) not compatible with known level of skill Volume forecasts defined as flow over 1 to 16 consecutive weeks
Spatial aggregation too coarse (mixed hydrological characteristics) Work on regions based on climate categories and topography
Crossing the 80% / 20%ile thresholds does not reflect hydrological extremes Use 90-95%ile / 10-5%ile thresholds

What is currently missing from GloFAS Sesonal?

  • Drivers of outlooks, i.e. initiation conditions (snow, soil moisture, …) and seasonal atmospheric forecasts
  • Outlook of water balance components, i.e. snow storage, soil moisture etc… for downstream applications (e.g. reservoir management etc…)
  • Simple hydrograph plots for non-specialists. More complex information to be provided for advanced users
  • Information on skill from analysis of reforecasts
  • Technology for users to assess the skill themselves
  • User-defined temporal windows for the forecast (e.g. crop cycle)
  • Dynamic spatial aggregation

GloFAS Rapid Risk Assessment

A new Rapid Risk Assessment product is envisaged for GloFAS and the aim of the exercise was to get feedback from a similar operational product already implemented in the European domain (EFAS) and gain insights on how to better shape the global version.

The exercise used the Paris flood of January 2018 as a case study, with participants being given snap shots of the EFAS Rapid Risk Assessment products.

Headline message

Familiarity with current risk products

Risk assessment products are known (e.g. from Environment Agencies, USGS …), but there is some uncertainty regarding how the EFAS risk matrix is defined (selection of spatial aggregation and risk thresholds).

Background information to be considered

Three main background information types were highlighted to build a useful risk matrix:

Exposure Hazard
Social Physical
Critical Population including density and vulnerable population Source: GHSL, WorldPop, local data where available Infrastructure including critical (road, rail, hospitals, powerplants, water treatment) Source: World Bank, OpenStreetMap
  • Real-time hydrological hazard magnitude (flow, river levels) Source: River gauges, satellites, models, social media / crowdsourcing Source: modelled return period, Media reports (Floodlist) Flood maps (e.g. JBA)
  • Probabilistic information (ensemble) summarised as graphs or tables Source: GloFAS ensembles / probability tables
  • Important Land Use and Land Cover Source: USGS, Copernicus
    Could have
  • Internet / communication access; Social media engagement in the region; Source: Development banks, CAT model companies
  • Socio-economic and cultural value
  • Biodiversity information
  • Defence / risk mitigation features Source: Flowpros database, local data, dynamic data on performance during flood Forecasts tendency Source: GloFAS ensembles / probability tables
    Feedback on current EFAS Rapid Risk Assessment product

    The mind map below summarises the main feedback received regarding the existing EFAS Rapid Risk Assessment products.

    Shaping the future GloFAS Rapid Risk Assessment product - what is currently missing?

    Presentation Information content
    Qualitative descriptions for users Division of affected population by severity / depth Historical records of floods and duration - MODIS / VIIRS
    Statement on uncertainty Vulnerable population - temporal population data e.g. refugees - from mobile phone data, Facebook or nightlights imagery Gauge data (where available)
    Evaluation of risk estimates Critical infrastructure Flood duration
    Administrative and political boundaries (consistent e.g. GADM data)

    GloFAS weekly forecast summary

    As part of GloFAS evolution, a new product providing a GloFAS weekly forecast summary is under development. The aim of the exercise was to gather early feedback and ideas to help shape the product development, including preferred content, length, presentation, and delivery mechanism(s).

    Headline message

    Summarising forecasts to respond to varying users’ needs would be challenging, as interests differ on lead time (short, medium, long range), forecast aspect (flood vs flow volume forecasts), spatial scale (catchment vs continental scale), and application (emergency response, business sector etc…). A customised dash-board would be more effective in delivering products responding to users’ needs (regional zoom, combination of products) and should include information on skill.

    Session 4: Service provision

    Successful hydrological services need to be a user-designed, with efficient service provision, so that downstream applications can benefit most from the offered information. The final interactive session of the workshop aimed to identify the type of practical applications users might undertake with data from a hydrological service. It followed presentation of two practical examples of downstream application (the Copernicus Climate Data Store and Tool box, and the ‘Rapid’ streamflow forecast).

    Interactive session: GloFAS workflow and user stories

    Following the format of user stories, participants were invited to describe the workflow and applications they would develop from a global hydrological service. Results were analysed in terms of functionalities (service, content).

    Headline messages

    Data access

    What Why
    Local river flow forecasts (time series simulations), both in real-time through web services and archiving facilities, Downstream applications
    Forecasts from the terrestrial water cycle (soil moisture, snow water equivalent, evaporation etc…) Downstream applications
    River flow past forecasts User verification
    Historic and real-time gauged river flow, or flood event data User own research analysis
    Weather forcing data (real-time/ forecast) User to develop own hydrological forecasting system

    Information content

    What

    Why

    Tutorial and case study datasets

    Training

    Documentation (detailed and for non-expert)

    Better understanding of limitations of the system

    Dynamic graphic facilities to extract summary information for user-defined points/ regions

    Better assessment of areas affected for flood events

    Summary forecast bulletins

    Easy assessment of the situation by non-experts

    Forecast verification metrics

    User’s own evaluation of forecast value in their area of interest

    Impact forecasts including surge floods and inundated areas

    Downstream actions

    Forecasts of the full river flow regime, including droughts

    Downstream actions

    Modelling framework

    What

    Why

    Interactive calibration tool

    Improved performance on user’s area of interest

    On-demand real-time data assimilation

    Improved forecasting performance

    Ability to run own experiments on global forecast modelling chain

    Quantifying modelling uncertainty and comparing with other forecasting systems

    High-resolution forecasts (<100m)

    Downstream applications