Fictitious Forcings have Macondo in Trouble: A Hackathon Story

Contributed by Georgy Ayzel. 

Meteorological ensemble forecasts form the core of every system for hydrological ensemble prediction. There are multiple data sources where our meteorological ensembles could come from: historical observations, numerical weather prediction and climate models, or stochastic weather generators. Different sources serve different ways we want to communicate and deliver the information of possible runoff responses to various meteorological conditions. However, there is a source that is typically ignored in the modern landscape of ensemble runoff prediction – fictitious forcings.

The #OpenDataHack2018 on “Innovate with Open Climate Data” organized by @CopernicusECMWF over the weekend of 9th – 10th June 2018 in Reading, UK, provided us (Georgy Ayzel and Shaun Harrigan) with the unique possibility to come together and fill this significant gap in the HEPEX research domain.

Figure 1: Hackathon volunteers and participants busy coding, and keeping the caffeine intake up (source: Shaun Harrigan).

With 24 hours on the clock we decided to estimate the runoff response of 314 UK river basins to one of the most extreme weather scenarios, as mentioned in the iconic fiction novel One Hundred Years of Solitude written by Gabriel Garcia Marquez:

The rains last for four years, eleven months, and two days (or 4Y11M2D, for short).

Following the critical direction of making hydrological modeling studies reproducible we used only open data and software to perform our research (Figure 2). Our fictitious forcings were extracted from the ERA5 reanalysis for 4Y11M2D for three hypothetical rainfall scenarios characterized by daily intensities of:

  1. The observed maximum for 4Y11M2D,
  2. The observed mean for 4Y11M2D,
  3. random value between mean and maximum for 4Y11M2D.

The three rainfall scenarios of 1,796 consecutive wet days were used to force the GR4J open-source hydrological model to produce the runoff for each catchment.

Figure 2: Schematic of our workflow

To communicate the obtained results, we created the web-service Macondo in trouble where you can have a look at the modeled basin response in the approximation that every basin will have the future of the Buendia family’s native town (Figure 3).

Figure 3: Landing page of Macondo in trouble

Our results reveal the apparent pattern of UK river basin behavior: it is possible to avoid water scarcity caused by everyday rain with a mean observed intensity only, migrating from England and Wales to Scotland and Northern Ireland (where evaporation does not take over). However, with random or maximum rain scenarios the very long range forecast would be for fictitiously large floods.

Take home messages:

  • ECMWF’s #OpenDataHack is an excellent place to meet colleagues and attack exciting problems during a weekend,
  • Doing reproducible research is much harder than tweeting about its importance in modern computational intensive science,
  • Simple tools lead to insights about complex phenomena. The catchment response to different rainfall scenarios in, for example, flashy vs slow responding catchments or regions with higher/lower evaporation, revealed interesting findings and can be used as a useful teaching or research tool for exploring the limit of our hydrological models.

You can read more about the #OpenDataHack challenges in Georgy’s blog post, especially if you ever wanted to know what ERA5 data sound like!


Thanks to Florian Pappenberger and Calum Baugh for brainstorming these ideas during the hackathon.


  1. What a spectacular exercise! and above all, it evoked one of the best books I’ve read in my life! a Macondo a little far from Reading, in Colombia my beloved country, but with real severe flood problems.

    … and the post of sound generation with Deep Learning and ERA temperature, definitely inspiring to approach the musical intelligence profiles according to Gardner.

    Thank you very much for sharing this info with us!

    1. Thanks a lot for your response!

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.