Versions Compared

Old Version 4

changes.mady.by.user Corentin Carton de Wiart

Saved on

compared with

New Version 5

changes.mady.by.user Christel Prudhomme

Saved on

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

...

The reporting points are created using the following process. For each NWP-driven forecast, a flood probability is computed for each return period flood threshold (2-year, 5-year and 20-year) using the maximum forecast discharge value over the 10-day forecast horizon. This flood probability is only computed on the large rivers with an upstream area > 250 km2. Very dry areas are also excluded by filtering the rivers with a 1.5-year return period < 1 m3/s. A total probability of exceedance is computed from these maximum forecast probabilities (from DWD-HRES, ECMWF-HRES, ECMWF-ENS and COSMO-LEPS). The weights of the forecasts in the total probability are equals but, in order to avoid the predominance of deterministic forecasts in the region where COSMO-LEPS is not available, we impose a ratio of 50% deterministic, 50% probabilistic in the computation of the total probability. This gives the two cases (DWD-HRES, ECMWF-HRES,ECMWF-ENS, COSMO-LEPS):

  • Inside COSMO-LEPS domain: (25%, 25%, 25%, 25%)
  • Outside COSMO-LEPS domain: (25%, 25%, 50%, 0%)

The persistence probability is then calculated by averaging the total probabilities from the current and the previous forecast. Reporting points are finally created (and their alert levels are computed) based on the following rules:

  • 20-year return period: persistence probability > 30% and upstream area > 1000 km2km2.
  • 5-year return period: persistence probability > 30% and upstream area > 1000 km2km2.

  • 2-year return period: persistence probability > 50% and upstream area > 1000 km2.

In regions where no static points (hydrological stations) are available from the EFAS hydrological database, an algorithm is used to generate dynamic points. The algorithm is based on the following properties:

  • A dynamic point will be created at the most downstream location and at the peak signal location of a flood signal of at least 5 pixels.
  • Once a dynamic point is created, it stays on the map as long as a flood signal exceeding at least the 2-year return period is detected.
  • Points to too close to each other are removed (points from previous forecasts will remain).
  • If two points are close to each other but are on different rivers (based on the upstream area > 1000 km2 network), the two points are kept.

...