Skip to main content

Identifying weather extremes from ensemble weather forecasting or stochastic weather generators

Interested: send an email with CV to Jaap.Kwadijk@deltares.nl,  m.j.booij@utwente.nl,  ferdinand.diermanse@deltares.nl , philip.ward@vu.nl, bruno.merz@gfz-potsdam.de 

Participating institutes: Deltares, University of Twente, VU Amsterdam, KU Leuven, GFZ Potsdam

Areas under investigation: e.g. De Dommel, Geul, Vecht, Rur, Erft, Ahr river basins

Rationale and objective

Over the last years, we have faced extreme weather-related events world wide of such magnitude/intensity that they can beconsidered (at least) rare, even under changed climate conditions. Some events do not seem to appear in the climate projections, even the more extreme projections. It is scientifically widely acknowledged that extremes will intensify with climate change. 

In water management practice we are preparing for extreme events that, just like the climate, are gradually changing. Preparation will become much more complicated if those events would not gradually grow, but if the frequency and magnitude change stepwise. In that case we may not consider the events of last year to be such extreme events, but events (of an unknown probability) that can be expected to occur more frequently in a world of ~+ 1.5 C warmer than the pre-industrial climate.  The question to be answered would be: suppose this is true, what can we do? When stationarity can no longer be assumed, classical approaches are no longer reliable for estimating the probabilities of extreme events. And if changes in extremes do not change gradually this will have severe consequences for all engineering designs.

 In the scientific world these thoughts are not new. But the recent events are signals that make both scientists and decisionmakers aware that alternative approaches should rapidly be considered. To prove or disprove that extremes do not change gradually, but stepwise, using the observational record may take many decades (Diermanse et al, 2010). Relying on this approach could mean that over a long period the hazard from extreme weather may be much larger than we assume. Alternative approaches for assessing return periods for extreme events include the use of ensemble weather forecasts for extreme value estimation (van den Brink et al. 2004, van den Brink et al. 2005, Alfieri et al. 2019, Bottema et al. 2018).

The objective of this research is to explore the suitability of the ensemble forecasting archives for use in extreme weather hazards analysis.

Approach

With a group of MSc students of different universities we will explore the potential of using the ensemble forecasts archives for extreme events assessment. The approach uses real weather forecasts which can be analysed on the occurrence of extreme events in those forecasts (whether they really occurred or not).  Ensemble forecasting is a method used in or within numerical weather prediction. Instead of making a single forecast of the most likely weather, a set (or ensemble) of forecasts is produced. This set of forecasts aims to give an indication of the range of possible future states of the atmosphere. We can extract this information from the European Centre for Medium-Range Weather Forecasts (ECMWF) archives. The data used for this project is data from the ERA5 model from ECMWF. The archive comprises two weekly ensemble forecasts (11 members, 0-15 days forecast lead time) between 1979 and 2021, having a spatial resolution of 31km.  Alternatively, we could extract information from the seasonal forecasting archive by pooling ensemble members and lead times from the ECMWF seasonal prediction system SEAS5.  The atmospheric data from the ERA5 ensembles have a resolution of 63 kilometers and a 3-hourly output. From this total of ten predictions twice weekly over 40 years we can determine events of large x-hourly rainfall total for the areas relevant for the case studies. The following applications will be explored:

  • Trends in the upper and lower tail members and compare these with the central ones on the appearance of trends.
  • Estimation of Storm-Centred Areal Reduction Factors and intensity-frequency-duration relations
  • Assessment of maximum plausible precipitation events
  • Selection of extreme weather events as boundary condition for multiscale stress tests of water management systems
  • Applicability for assessing return periods of unprecedented extreme events
  • Scalability of the use of ensemble archives for the purposes mentioned for extreme wet and dry events
  • Comparison with alternative approaches (synthetic series, GEV approaches)

 Your (research) activities will include:

  • Learn about ensemble weather prediction for the literature
  • Collect the ensemble series and project rainfall series over the basins under consideration
  • Carry out simulations with hydrological models the ensemble series to identify the effect on river flow and analyse the resulting flow series on impact
  • Analyse the series for one or more purposes in the basins under consideration.
  • Report in form of a thesis

Deliverables

Next to a series of MSc thesis, the results will be summarized in form of a guideline to use the ensemble forecast archives in the practice of flood management.

Requirements

Successfully completed courses on hydrology and/or hydrological modellling and/or water resources management,  and a reasonable acquaintance with Python or other script languages for programming.