The ERIC flash flood forecasting is done by comparing the forecasted surface runoff accumulated over the upstream catchment with a reference threshold. It is based on the 20-member COSMO-LEPS ensemble precipitation and soil moisture forecasts from the LISFLOOD hydrological model and provides indicators for the next 5 days for catchments smaller than 2,000km2.
The ERIC flash flood forecasting modelling chain
Step 1: Surface Runoff Calculation
Surface runoff is calculated by combining the 6 hourly COSMO-LEPS precipitation forecasts with the soil moisture forecasts produced by the LISFLOOD hydrological model when forced with COSMO-LEPS. Firstly both datasets are downscaled to 1 km using nearest neighbour. Next snow is removed from the COSMO-LEPS precipitation forecasts by removing all precipitation where the temperature is less than or equal to 0 degrees. Surface runoff is then calculated through the following equation:
Surface runoff = (a+b*soil_moisture^c)*rainfall
where a, b and c are regression coefficients which are found by fitting the relationship between soil moisture and the runoff coefficient from the 20 year EFAS re-analysis simulation at every 1 km location (Figure 1)
Figure 1. Example of the fitting between soil moisture and runoff coefficient using data from the 20 year EFAS re-analysis for a forest location (Raynaud et al., 2014)
Step 2: Accumulating the Surface Runoff
The 6 hourly surface runoff time series are accumulated on the 1 km channel network for all cells with an upstream area less than or equal to 2000km2. This means that for each 6 hourly timestep, at each 1 km channel cell, the total surface runoff which occurs upstream of the cell is calculated. These 6 hourly accumulations are then further accumulated over time periods of 12 and 24 hours.
Step 3: Comparison against Climatological Thresholds
The accumulated surface runoff at 6, 12 and 24 hour time periods are divided by the mean annual maximum surface runoff at each respective time period to compute the ERIC index value at each 1 km channel cell. If the accumulated surface runoff for a given time period exceeds the mean annual maximum, then the ERIC index will be greater than 1. The mean annual maximum values are derived from a 19 year COSMO-LEPS reforecast series. At each 1 km channel cell for each 6 hourly timestep, the maximum ERIC index value across the three time periods is then selected. Then the maximum ERIC index value across all the 6 hour timesteps in the 5 day forecast period is selected.
Step 4: Generating the Reporting Points
The above three steps are repeated for all 20 COSMO-LEPS ensemble members. The next step calculates the exceedance probability of ERIC Index = 1 i.e. the probability of being above the mean annual maximum which is equivalent to the bankfull height of the river channel. Only 1 km channel cells where at least 4 ensemble members have an ERIC index >= 1.0 are retained. The retained cells are clumped together, the most downstream cell in each clump is chosen as the location for where the ERIC reporting point is generated. At these reporting point locations the ERIC index for each ensemble member at each 6 hourly timestep is converted into a return period using gumbel parameters obtained from a 19 year COSMO-LEPS forecast series. Then the maximum exceedance probability of the 2, 5 and 20 year return periods are computed across the 5 day forecast period.
This information is then used to generate the ERIC Reporting Points and ERIC Affected Area layers shown on the EFAS website (Figure 2).
Figure 2. Example of the ERIC flash flood products available on the EFAS website.
Products
The EFAS ERIC flash flood forecasts are available as:
- Reporting points: ERIC reporting points
- Gridded affected area upstream of each point: ERIC Affected Area
References:
Raynaud, D., Thielen, J., Salamon, P., Burek, P., Anquetin, S. and Alfieri, L. (2015), A dynamic runoff co‐efficient to improve flash flood early warning in Europe: evaluation on the 2013 central European floods in Germany. Met. Apps, 22: 410-418. doi: 10.1002/met.1469