Flood impact forecasts are especially important in the preparedness phase, to support the planning and allocation of rescue assets, and to get a first estimate of the forecasted flood event’s potential socio-economic consequences.

In CEMS-Flood, the flood impact forecasts are based on three components: 1) medium-range/ 30 days flood forecasts, 2) event-based rapid flood mapping, and 3) impact assessment. The original method to generate these products is given in Dottori et al., 2017.

  1. Medium-range flood forecast: every time a flood event greater than the 10 year return period is forecasted in CEMS-Flood, the return period of the maximum discharge (based on the ECMWF-ENS ensemble forecast median in both GloFAS and EFAS) over the forecast period (30 days in GloFAS, 10 days in EFAS) is computed in each grid cell (shown by the coarse cells in figure above). These values do not consider the possible role of local flood defences, therefore they represent an unprotected scenario. A second scenario which considers flood defences is computed by comparing the forecasted return period values against estimated flood protection levels from FLOPROS (Scussolini et al., 2016). River grid cells where the protection levels are exceeded are selected and form the protected scenario.
  2. Rapid flood mapping: for each CEMS-Flood river cell identified in step 1, flood prone areas are delineated, using a catalogue of higher resolution flood inundation maps. The obtained event-based inundation map has a spatial resolution of 3 arc seconds (approximately 90 m) (up to GloFAS version 3, the resolution was 30 arc seconds (approximately 1 km), and up to EFAS version 4 the resolution was 100 m). This is produced for both the unprotected and protected scenarios, the former is shown on the “Rapid Flood Mapping” layer on the webviewer. The flood inundation maps catalogue is generated using the LISFLOOD-FP hydraulic model (CA2D hydraulic model for GloFAS until version 3).
  3. Rapid impact assessment: the event-based inundation maps for both the unprotected and protected scenarios are combined with exposure information to assess regional impacts (shown on the “Rapid Impact Assessment” layer). Considered exposure includes population, critical infrastructure and land cover. The result is summarised on administration units sourced from NUTS in Europe and GADM in the rest of the world.

The EFAS and GloFAS flood impact forecasts are provided to the users as a set of dedicated layers described in EFAS Rapid Flood Mapping and Rapid Impact Assessment and GloFAS Rapid Flood Mapping and Rapid Impact Assessment.

Rapid Flood Mapping

This procedure creates an estimate of the potentially inundated area by using a lookup between the GloFAS/EFAS grid cells and the high resolution flood maps generated for different return period scenarios. For each EFAS/GloFAS grid cell where the maximum median forecasted discharge is expected to exceed the 10 year return period, a lookup table is used to find the closest matching high resolution flood map. The forecasted return period in the GloFAS/EFAS cell is matched to the return period scenarios for which flood maps have been generated (10, 20, 50, 75, 100, 200, 500 years), this is done by rounding down to the nearest scenario, for example if a return period of 17 years was forecast, the flood map for the 10 year return period scenario would be selected.

Once all the high resolution flood maps have been selected, they are mosaicked into a single gridded layer, where multiple flood maps overlap the maximum depth value is selected. The mosaicked layer is then converted into a binary flood mask, all depth values >=0.099m are set to 'flooded', all other values are set to 'not flooded'. This masked layer is shown on the EFAS/GloFAS webviewers as the 'Rapid Flood Mapping' layer.

The above procedure does not consider the role of flood defences, therefore it represents an unprotected scenario. To account for flood defences, a protected scenario is  generated using the estimated flood protection levels from the FLOPROS dataset (Scussolini et al., 2016) available at administration region level. This dataset does not predict the exact locations of local flood defences, instead it predicts the likely level of protection if flood defences were present within an administration region. For all the GloFAS/EFAS grid cells which were identified above as exceeding the 10 year return period, these are compared to the FLOPROS estimated level of defence protection, GloFAS/EFAS grid cells which do not exceed this protection level are discarded. The remaining GloFAS/EFAS grid cells which exceed the estimated protection level are used to generate a new estimate of the potentially inundated area. 

The Rapid Flood Mapping layer shown on the GloFAS and EFAS webviewers is from the unprotected scenario, it does not consider the role of flood defences.

An example of the Rapid Flood Mapping layer in the northern Philippines. This represents the unprotected scenario which does not consider the role of flood defences.

Rapid Impact Assessment

The Rapid Impact Assessment layer, is created by overlaying the Rapid Flood Mapping layer, created previously, with exposure information regarding population, landcover and critical infrastructure, and summarised by administration regions to estimate the possible impacts related to a forecasted flood event.

Example of the Rapid Impact Assessment layer, shaded administration regions show where flood impacts are possible. Further information is available in a pop-out window when the a shaded region is clicked.

Exposure Information

The following exposure information is used:

  • Population: Global Human Settlement Layer - Population (GHS-POP; Schiavina et al., 2023) v.2023A for the 2020 epoch at 3 arc second spatial resolution using the WGS84 projection. Download link
  • Landcover: Copernicus global land cover version 3 (Buchhort et al., 2020)
    • The original landcover classes were summarised into 6 categories:
      • Artificial surfaces
      • Agriculture
      • Forest and semi-natural
      • Coastal
      • Water
      • Bare surfaces
  • Critical infrastructure: Hospitals, education, airport and power generation facilities, downloaded from OpenStreetMap. Note, for GloFAS version 4.0 these critical infrastructure data are not currently available, this will be resolved in a subsequent update
  • Cities: Downloaded from OpenStreetMap. Note, for GloFAS version 4.0 this dataset is not currently available, this will be resolved in a subsequent update

For each of these exposure types, they are overlaid with the flood inundation extent predicted by the Rapid Flood Mapping layer to identify the population, landcover and critical infrastructure which lies within the flooded area. These datasets which have been masked by the flood extent are then summarised over administration regions defined by the NUTS (in Europe) and GADM (outside Europe) datasets, typically administration region level 1 or 2 units are used. This procedure is done for both the protected and unprotected Rapid Flood Mapping scenarios, which respectively consider and do not consider the potential role of flood defences.  These summarised results are presented in a table which appears in a pop-out window when an administration region, highlighted in the Rapid Impact Assessment  layer, is clicked. The table shows the total population, the total area of each landcover type and the total number of assets for each infrastructure type within the forecasted flood extent within that administration region, there is one column each for the results from the protected and unprotected Rapid Flood Mapping scenarios.

Exposure information table which is shown in the pop-out window when the Rapid Impact Assessment layer is clicked. It summarises within an administration region the population, cities, critical infrastructure and landcover types which lie within the flood extent forecasted in the Rapid Flood Mapping layer.

Shading the Administration Regions using an Impact Matrix

The colour shading of each administration region shown in the Rapid Impact Assessment layer, is chosen using an impact matrix which summarises the lead time to the event and the total population potentially exposed, both of which come from the protected scenario which considers the role of flood defences. The lead time represents the earliest time within the administration region when the forecasted median river discharge exceeds the 5 year return period, note that this is only calculated in grid cells where the maximum forecasted median streamflow is expected to exceed the estimated flood defences protection level. The 5 year return period is considered here as river discharge at this level can still be associated with some flood impacts and the aim is to highlight the earliest time when flood impacts are possible. The population represents the total number of people within the administration region who live within the forecasted flood inundation area. 

The impact matrix is shown at the top of the pop-out window when an administration region in the Rapid Impact Assessment layer is clicked, a tick will show where within the impact matrix that administration region has been categorised.

Grey shaded administration regions signify regions where 0 population is expected to live within the forecasted flood inundation area according to the protected scenario, however the unprotected scenario predicts >0 population living within the forecasted flood inundation area. 

The impact matrix which is used to decide the colour shading of the administration regions in the Rapid Impact Assessment layer. In this example, the region would be shaded red (note the location of the tick).

Flood Event Information

A second table in the pop-out window of the Rapid Impact Assessment layer summarises information about the timing and magnitude of the flood event within the administration region. Each of the following are computed separately for the protected and unprotected scenarios.

  • Estimated mean return period: this is the mean return period of all GloFAS/EFAS grid cells within the administration region where the return period of the maximum forecasted ensemble median river discharge exceeds 1) 10 year return period for the unprotected scenario, 2) the estimated flood defence protection level for the protected scenario
  • Estimated protection levels: the estimated level of protection (return period) from flood defences within the same administration region
  • Estimated flooding start date: earliest date when the forecast ensemble median river discharge exceeds the 5 year return period threshold
  • Estimated flooding end date: latest date when the forecast ensemble median river discharge exceeds the 5 year return period threshold
  • Estimated flooding duration (days): number of days between the Estimated flooding start date and Estimated flooding end date
  • Estimated peak date: date of maximum discharge value 
  • Estimated flooded area (km2): total area of the forecasted inundation extent from the Rapid Flood Mapping layer
  • Mean probability of exceeding the 2-years threshold: for all grid cells in which the maximum forecasted ensemble median river discharge 1) exceeds the 10 year return period (unprotected scenario) or 2) exceeds the estimated flood defence protection level (protected scenario), compute the maximum exceedance probability of the 2 year return period threshold and then calculate the mean over all the grid cells within the administration regions.
  • Mean probability of exceeding the 5-years threshold: for all grid cells in which the maximum forecasted ensemble median river discharge 1) exceeds the 10 year return period (unprotected scenario) or 2) exceeds the estimated flood defence protection level (protected scenario), compute the maximum exceedance probability of the 5 year return period threshold and then calculate the mean over all the grid cells within the administration regions.
  • Mean probability of exceeding the 20-years threshold: for all grid cells in which the maximum forecasted ensemble median river discharge 1) exceeds the 10 year return period (unprotected scenario) or 2) exceeds the estimated flood defence protection level (protected scenario), compute the maximum exceedance probability of the 20 year return period threshold and then calculate the mean over all the grid cells within the administration regions.

Flood event information table which appears in the pop-out window when the Rapid Impact Assessment layer is clicked.


Product Limitations

The forecasted flood inundation in the Rapid Flood Mapping layer only considers flooding from riverine processes. It does not consider flooding from coastal, pluvial or any other processes. This means that for compound events where riverine flooding coincides with pluvial and/or coastal flooding, the forecasted flood inundation will likely be under-estimated.

The forecasted flood inundation will be prone to biases/errors in the GloFAS/EFAS re-analysis river discharge datasets, as these were used as inputs to generate the flood inundation maps. For example, if the GloFAS/EFAS discharge re-analysis contains a negative bias at a given location, the resulting simulated flood inundation will likely be under-estimated as insufficient water was input into the simulation. The opposite will be true if there is a positive bias. Further information about biases/errors in the GloFAS/EFAS simulations can be found in the Hydrological Model Performance layer

The performance of the forecasted flood maps may be degraded in areas where there is significant management of river flows, for example diversion through artificial channels. The flood maps were generated assuming natural river flows and therefore do not consider potential actions which may mitigate flood impacts, such as the diversion of water or the retention within dams.

Channel bathymetry is not represented within the simulations used to generate the flood inundation maps, as no global or pan-European dataset was available. To compensate for this the river discharge associated with the 1.5 year return period, which is associated with bankfull discharge, was subtracted from the flow inputs used within the flood inundation simulation allowing for the simulation of floodplain flow only. This means however that flood maps were only generated down to a minimum of the 10 year return period, below this the importance of representing channel bathymetry increases.

The 3 arc second (~90 m) spatial resolution of the flood inundation maps will be insufficient to represent local topographic features such as levees. This will also cause difficulties to predict flooding within urban areas. Therefore a user should take care when interpreting the results in these areas. 

Flood defence information from FLOPROS (Scussolini et al., 2016) is only estimated, it does not give the precise location and protection level of individual assets. Instead, it estimates the likely protection level of any defences which might be present within an administration region. A user should consult local information about the location and protection level of any flood defences within their area of interest, if this information is available to them.

References

Buchhorn, M.; Smets, B.; Bertels, L.; De Roo, B.; Lesiv, M.; Tsendbazar, N.E., Linlin, L., Tarko, A. 2020. Copernicus Global Land Service: Land Cover 100m: Version 3 Globe 2015-2019: Product User Manual; Zenodo, Geneve, Switzerland, September 2020; https://doi.org/10.5281/zenodo.3938963 

Dottori, F., Kalas, M., Salamon, P., Bianchi, A., Alfieri, L., and Feyen, L., 2017. An operational procedure for rapid flood risk assessment in Europe, Nat. Hazards Earth Syst. Sci., 17, 1111-1126, https://doi.org/10.5194/nhess-17-1111-2017

Schiavina, Marcello; Freire, Sergio; Alessandra Carioli; MacManus, Kytt. 2023. GHS-POP R2023A - GHS population grid multitemporal (1975-2030). European Commission, Joint Research Centre (JRC) [Dataset] doi: 10.2905/2FF68A52-5B5B-4A22-8F40-C41DA8332CFE

Scussolini, P., Aerts, J. C. J. H., Jongman, B., Bouwer, L. M., Winsemius, H. C., de Moel, H., and Ward, P. J.. 2016. FLOPROS: an evolving global database of flood protection standards, Nat. Hazards Earth Syst. Sci., 16, 1049–1061, https://doi.org/10.5194/nhess-16-1049-2016.