Contributors: A. Reder (CMCC), A. Essenfelder (CMCC), P. Mercogliano (CMCC), G. Rianna (CMCC), M. Raffa (CMCC), M. Mancini (CMCC), R. Padulano (CMCC), S. Bagli (GECOSISTEMA), P. Mazzoli (GECOSISTEMA)
Acronyms
Introduction
This document serves as Product User Guide (PUG) for Flood risk indicators for European cities from 1989 to 2018 dataset, as part of the C3S Sectoral Information System (SIS) to support Disaster Risk Reduction (DRR) concerning Pluvial Flood Risk Assessment in Urban Areas. More information about the project can be found at https://climate.copernicus.eu/pluvial-flood-risk-assessment-urban-areas.
Executive Summary
The dataset presents indicators to evaluate spatial distribution of flood risk in terms of hazard and direct damages. These are provided as a focused, high-resolution, product for 20 cities located in Europe.
#The dataset combines high-resolution, probabilistic description of extreme precipitation, exposure datasets and damage/vulnerability models into a comprehensive pluvial flood risk assessment for cities across Europe for the current climate. It allows Users to exploit flood risk analysis over the city. Operatively, the dataset moves from data available on the Climate Data Store (CDS) and Land Monitoring Service (LMS). ERA5 reanalysis data are dynamically downscaled over a 2kmx2km grid for the 20 European cities to reach a spatial and temporal resolution more suitable for pluvial flood analysis. Indeed, such a downscaled product is used for deriving hourly precipitation input at prescribed recurrence intervals that, in combination with supporting DEM (such as EU-DEM and Lidar- Derived DEM), feed hazard and damage models.
The basis for this dataset is the ERA5 reanalysis, provided under the C3S umbrella. This dataset was produced on behalf of the Copernicus Climate Change Service.
Figure 1: Spatial distribution of 20 European Cities, selected according to User requirements as vulnerable to urban pluvial flooding.
Scope of Documentation
The PUG promotes the Product from several perspectives by:
- drawing identified gaps and Users' requirements, and how the dataset may address these gaps and needs;
- providing an overview as for key dataset characteristics as for computed variables/indicators;
- describing used input data providing synthetic informative features;
- providing an overview about the methodologies driving the dataset development with a special focus on how methods and resulting products may be able to address the identified gaps, and on results of tailored evaluations, accounting for also benefits, limitations, and modelling assumption made in the production of the dataset.
Product Description
Product Target Requirements
The analysis of gaps and Users' requirements have been carried out exploiting jointed ways:
- identification of key users and mobilization of the network of DRR practitioners and experts;
- desk review of surveys of users' requirements produced in the context of C3S and other projects relevant for the purpose of the Sectoral Information System concerning DRR;
- participation in relevant DRR events to inform people about the project (oral presentations, posters, side-sessions) and to initiate discussions;
- consultations via interviews, webinars, and a workshop, engaging DRR experts and policy networks.
Based on insights and caveats collected through such initiatives, four main gaps are identified:
- GAP1: frameworks and support permitting reliable but expeditious assessments about the areas potentially interested by pluvial flooding in urban contexts under reference precipitation events. They include outputs supporting hazard/risk zonation required for civil protection purposes (e.g., affected areas, water heights) and socio-economic estimations (e.g., expected costs and losses under reference precipitation events with fixed estimated return period).
- GAP2: a clearer understanding about the added value of very high-resolution dynamical downscaling from ERA5 reanalysis in terms of localization and magnitude of precipitation events at urban context scale. There is a growing interest in assessing the added value of information returned by high resolution dynamical downscaling of reanalysis; it could be highly relevant for extreme atmospheric events (as heavy precipitations) usually requiring a detailed representation of geomorphological features. In this regard, a substantial increase in performances (better characterization of spatial distribution and cumulative values) due, for example, to a better representation of the orography or urban environments could justify the required investment in time and computational resources.
Product Overview
This product relies on the concept of climate data integration, whose main goal is to develop a strong connection between data available in CDS and data ad-hoc provided to improve spatial and temporal requirements resolution making them more suitable for pluvial flood analysis.
According to this definition, the Flood risk indicators for European cities from 1989 to 2018 dataset represents a potential pillar enabling the Users to investigate (extreme) precipitation features and their effects thanks to a suite of tailored basic and advanced information provided for a pool of 20 European Cities, selected according to User requirements as vulnerable to urban pluvial flooding.
This goal is addressed by performing a series of scenario analyses at city scale, able to turn fixed precipitation inputs into variables related to flood hazard and damages.
To this aim, an additional information layer is generated by dynamically downscaling ERA5 reanalysis at about 2km (ERA5-2km) on the 20 selected cities, by using the COSMO-CLM model (including specific parameterizations for urban areas). In compliance with the targets of the pluvial flood risk analysis, this new generation of hourly precipitation data are processed to define rainfall events with
specific return times representing the inputs for the scenario analyses. These inputs are turned into a sub-set of variables (e.g., water depth and damage suffered by the existing assets) exploiting specific models for hazard and damage assessments.
Data Description
Table 1: Overview of key characteristics of the Flood risk indicators for European cities from 1989 to 2018 dataset.
Data Description | |
Dataset title | Flood risk indicators for European cities from 1989 to 2018 |
Data type | Indicators |
Topic category | Disaster and Risk Reduction (DRR) |
Sector | Civil Protection, Insurance, Land Use Planning |
Keyword | Extreme Precipitation |
Dataset language | eng |
Domain | European cities: Antwerp and Brussels (Belgium); Frankfurt am Main and Koln (Germany); Paris (France); Amersfoort (the Netherlands); Birmingham and London (UK); Vienna (Austria); Prague (Czech Republic); Budapest (Hungary); Riga (Latvia); Stockholm (Sweden); Milan and Rimini (Italy); Bilbao and Pamplona (Spain); Amadora (Portugal); Athens (Greece); Bucharest (Romania) |
Horizontal resolution | EU-DEM: 25m x 25m |
Temporal coverage | 1989-2018 (30-years) |
Temporal resolution | 30-year |
Vertical coverage | Single level |
Update frequency | None |
Version | v1 |
Provider | Centro Euro-Mediterraneo sui Cambiamenti Climatici (CMCC), GECOSistema |
Terms of Use | For datasets available in the Climate Data Store, the licence to use Copernicus products applies. For the meteorological station dataset provided by the European Climate Assessment (ECA&D), an agreement between the European Environment Agency EEA, acting as the in-situ coordinator for Copernicus, and EUMETNET, the grouping of 31 European Meteorological Services (which are the main suppliers of data to ECA&D), exists which indicates that the derived data based on ECA&D and provided through this contract C3S_430, are made available to Copernicus services under an Open |
Figure 2: Overview image of the dataset: water depth for the city of Vienna (Austria) resulting from the pluvial hazard analysis performed by as input an hourly precipitation event with return time = 100 y.
Variable Description
Table 2: Overview (long name, short name, and unit) and description of variables included in the Dataset. Specifically, the long name represents what the Users find in the catalogue entry whilst the short name is what the variables are called in the NetCDF.
Variables | |||
Long Name | Short Name | Unit | Description |
Water Depth | wd | m | Pluvial Flooding Water Depth induced by hourly precipitation events with fixed return periods computed over 30-year (1989-2018) for 20 European cities |
Water Mask | wm | dimensionless | Binary mask of flooded areas. The values are between 0 (not affected) and 1 (affected) and are function of hourly precipitation events with fixed return periods computed over 30-year (1989- |
Urban depression | dep | dimensionless | Topographic depressions which are defined as areas without an outlet and often referred to as sinks or pits |
Expected damage | dam | €/m2 | Direct damage suffered by the existing assets due to pluvial flooding induced by hourly precipitation events with fixed return periods computed over 30-year (1989-2018) for 20 European cities |
Input Data
Table 3. Overview of climate model data for input to Flood risk indicators for European cities from 1989 to 2018, summarizing the model properties.
Input Data | ||||||
Dataset name | Type | Spatial coverage | Spatial resolution | Temporal coverage | Temporal resolution | Source (link) |
ERA5-2km | Dynamical downscaling of ERA5 reanalysis | 20 | 0.02° x 0.02° | 1989- | Hourly |
ERA5-2km
ERA5-2km represents an additional hourly dataset at horizontal resolution of ~ 2.2 km for a pool of 20 user-selected cities over 1989-2018. It is developed by dynamically downscaling ERA5 with the regional climate model COSMO-CLM (Rockel et al., 2008) switching on the module TERRA-URB for accounting the urban parameterizations (Wouters et al., 2016). The downscaling activity relies on a one-step nesting strategy, in which the simulation at 2.2km is directly "one-way nested" in ERA5 (1:15 resolution jump). In ERA5-2km, observations are indirectly accounted through the atmospheric forcing of ERA5.
Method
Background
- GAP 1 & GAP2: dynamical downscaling of ERA5 reanalysis at very high resolution (about 2 km) for 20 cities allows an improvement of spatial and temporal characterization of precipitation events. Such data are used as input for pluvial flooding risk analysis. In this regard, hazard analysis (potentially affected areas and water depth) is carried out by means of a rapid DEM based rainfall-runoff modelling method permitting to simulate the dynamic filling-spilling overland flow processes (Samela et al., 2020). The outputs of hazard simulations permit pluvial flooding risk analysis, by combining it with harmonised damage records and advanced micro-economic risk assessment models. This framework exploits fragmented and incomplete records of past flood losses, complements these with high resolution exposure and impact data by state-of-the art hazard, loss, and recovery modelling. The built dataset provides for the 20 cities several information about the hazard (affected areas and water height) or in terms of economic impacts (e.g., expected losses). Selected by the stakeholders, five reference scenarios have been considered (expected precipitation with return periods of 5,10, 25, 50 and 100 years) permitting a characterization of events able to induce very different impacts over the urban areas.
Model / Algorithm
General Description
The conceptual framework adopted to derive Flood risk indicators at city scale for 20 cities situated across Europe is shown in Figure 3.
Figure 3: Conceptual framework adopted for flood risk analysis at city scale.
The framework moves from data available on the CDS and Land Monitoring Service (LMS). Firstly, ERA5 reanalysis data are dynamically downscaled at ~ 2 km to reach a spatial and temporal resolution suitable for pluvial flood analysis up to city scale. Such a downscaled product is then used for deriving hourly precipitation input at prescribed recurrence intervals that, in combination with supporting DEM (such as EU-DEM and Lidar-Derived DEM) and other external data sources, feed a hazard model to derive flood extent and water depth. These outcomes represent flood indicators related to hazards included in this Dataset; the water depth also serves as input, in combination with external data sources, for a damage model to derive the indicator related to damage (i.e., direct losses to the urban fabric) included in this Dataset.
Intensity-Duration-Frequency (IDF) curves at city scale from ERA5-2km
To interpret extreme precipitation values and to obtain annual maximum hourly precipitation at prescribed recurrence intervals, the storm index method (Viglione et al., 2007; Padulano et al., 2019) is adopted.
According to the storm index method, the rainfall depth of an extreme precipitation event x with return period T and rainfall duration d can be estimated (Eq. 1) as the product of a scale parameter (μ) only depending on duration d (i.e., deterministic part of Eq. 1), and a frequency parameter or “growth factor” (kT) only depending on the return period T (i.e., probabilistic part of Eq. 1):
\[ x(d,T) = \mu[x(d)] \cdot k_{T}(T), (1) \]Eq. 1 is known to ensure rainfall consistency, as it preserves the increasing dependence of precipitation depth on both duration and return period. Moreover, in practical applications, Eq. 1 is often subject to a regionalization process whose aim is to identify homogeneous areas (usually related to significant hydrographic units such as watersheds) where the statistical behaviour of extreme rainfall can be considered the same, and relations are calibrated basing on pooled samples. In this perspective, within a homogeneous region only one probability model is calibrated, whereas mean rainfall is considered spatially distributed (implying a dependence on elevation z as well as on duration d).
In the following, the procedure applied at city scale to calibrate Eq. 1 is described:
- for each grid point, the hourly precipitation amount provided by ERA5-2km over 1989-2018 are firstly aggregated with a fixed-width moving window for six reference durations d (1, 2, 3, 6, 12 and 24 hours); from these six samples of precipitation, the annual maximum rainfall depths (AMR) are then extracted to obtain six AMR samples;
- the different AMR samples are subjected to some hypothesis tests1 to check whether extreme data related to different points of a domain of interest can be considered extracted from the same population (in terms of mean values, growth factors, or both) and then pooled together to build robust calibration of Eq. 1; in case data pass these tests, the city can be considered as a unique homogeneous area for which the mean rainfall (i.e., deterministic part of Eq. 1) can be considered as spatially distributed and only one probability model will be calibrated (i.e., probabilistic part of Eq. 1);
for the scale parameter μ (depending on elevation z as well as on duration d), the model proposed by Sherman (Chow et al., 1988) is adopted and calibrated:
\[ \mu[x(d,z)] = \frac{A \cdot d}{(C + d)^{B+D \cdot z}} \]the goal of such a calibration procedure is to determine the set of parameters (i.e., A, B, C, D). By assuming z = z* (mean city elevation) and d = 1 h, it is possible to derive the scale parameter μ*;
for the growth factor kT (only depending on the return period T), the Generalized Extreme Value (GEV) probability distribution model, whose Probability Density Function (PDF) can be expressed by the following equation, is adopted and calibrated:
\[ f(x|k, \sigma, \mu) = \left( \frac{1}{\sigma} \right) \cdot exp \left \{ - \left[ 1+k \cdot \frac{(x-\mu)^{- \frac{1}{k}}}{\sigma} \right] \right \} \left[ 1+k \cdot \frac{(x-\mu)^{-1 - \frac{1}{k}}}{\sigma} \right] \] \[ f(x|0, \sigma, \mu)=\left( \frac{1}{\sigma} \right) \cdot exp \left \{ -exp \left[ - \frac{(x-\mu)}{\sigma} \right] - \frac{(x-\mu)}{\sigma} \right\} \]the goal of such a calibration procedure is to determine the set of GEV parameters (i.e., shape k, scale σ, location μ). The calibration procedure is applied on the pooled sample carried out by normalising for each duration the AMR sample by its mean value. In this way, a pooled dimensionless sample is obtained. By assuming the prescribed values T* (i.e., 5, 10, 25, 50, and 100 y) as return period T, it is possible to derive the growth factors kT*;
extreme hourly (d* = 1 h) precipitations at prescribed recurrence intervals T* at city scale are finally obtained as the product of μ* and kT* in accordance with Eq. 1. These values represent for each city the precipitation input that will feed the hazard model described in §2.4.2.3.
1 One-way ANOVA test (Kottegoda and Rosso, 2008) and the k-sample Anderson-Darling test (Anderson and Darling, 1954)
Flood Hazard Modelling
Indicators related to flood hazard (i.e., Water Depth, Binary Water Masks and Depression Id) at city scale are carried out by using a HFSA (Hierarchical filling and spilling algorithm), acknowledged as Safer_RAIN (Samela et al., 2020). Safer_RAIN is a fast-processing pluvial hazard model which identifies pluvial-flooded areas on the basis of nested surface depressions extracted from high-resolution DEMs and the rainfall depth associated with the considered event (possibly accounting for spatially distributed rainfall input and infiltration processes). As main simplifying assumption, Safer_RAIN neglects overland flow dynamics: net-rainfall volume accumulates into the system of nested depressions according to their capacity and hierarchical structure.
The operational workflow adopted to perform flood hazard simulations from data collection to hazard modelling is shown in Figure 4; a list of collected external input data is instead reported in Table 4.
Figure 4: Workflow from data collection to hazard modelling.
Table 4: List of collected input data with the relative sources
Data type | Source |
City area – vector | Copernicus Urban Atlas LCLU 2018 |
DEM – Geotiff | Copernicus EU Digital Elevation Model (EU-DEM) @25m, version 1.1 |
LiDAR@1m Koln | Official Geoportal NRW2 |
LiDAR@1m Milan | Lombardy regional geoportal3 |
LiDAR@2m Pamplona | Spanish National Geoportal (Centro Nacional de Información Geográfica)4 |
Buildings – vector | Open Street Maps |
River network – vector | Copernicus EU-Hydro - River Network Database |
In the following, some modelling remarks are listed:
- Extension of the urban study area is identified through the Copernicus Land Monitoring Service European Urban Atlas 2018 dataset5, which provides reliable, inter-comparable, high- resolution land use and land cover data for 785 Functional Urban Area (FUA) for the 2018 reference year in EEA39 countries. Extension of the core urban area for each city is used to define the analysis domain.
- Base digital elevation model (DEM) describing surface topography for the 20 cities is the European Digital Elevation Model (EU-DEM, v. 1.1) at 25 m ground resolution from Copernicus Land monitoring Service6. For each city, EU-DEM is resampled at 5 m to enable the buildings extrusion process. Moreover, EU-DEM is hydrologically conditioned through a stream burning operation (a common flow enforcement technique used to correct surface drainage patterns derived from digital elevation models). In this sense, the EU-Hydro - River Network Database from Copernicus Land Monitoring Service is used to depress by – 10 m elevation along river network.
- For the city of Koln, Milan and Pamplona, the analyses are duplicated also using high- resolution DEM. Specifically, it has been adopted a LiDAR-Derived DEM at 1 m for Koln and Milan, and a LiDAR-Derived at 2 m for Pamplona. For these additional elaborations, the analysis domain is the same of LiDAR. Regarding a potential stream burning operation, the high resolution of LiDAR does not require any stream burning operation, as the river network is well represented in the dataset.
- To consider presence of obstacles (buildings) that reduces effective storage capacity over DTM surface, Open Street Map 7 (OSM) vector buildings database is retrieved for the area of interest. Building polygons have been used to extrude a conventional elevation of + 10 m over original elevation inside each building to simulate the obstacle to water storage.
- Regarding Safer_RAIN assumptions, such a model is adopted by considering, on the safe side, impermeable surface (neglecting soil infiltration capacity that reduces rainfall runoff volumes). It is also neglected the sewage network capacity to collect and drain rainfall out of the domain, although this assumption is closer to reality the more pluvial events exhibit short duration and high intensity.
Direct Damage Model
Indicators related to the direct losses to the urban fabric due to a pluvial flooding event at city scale (i.e., Direct Damage) are assessed by using a methodology (Figure 5) based on a customary flood risk assessment approach (e.g., Huizinga et al. 2017).
Figure 5: Workflow showing the approach for the assessment of direct economic impact from flood events.
Such a methodology, originally developed for fluvial inundation, can be adapted to pluvial flooding assuming that the dynamic of impact from long-setting floods depends on the same factors, namely "hazard magnitude" and "size and value of exposed assets".
- The hazard magnitude is defined by assuming the Water Depth as damage predictor for urban areas. Such a variable derives from the flood hazard analyses (§2.4.2.2).
- The size and value of exposed assets are derived from external sources by combining geometric information about the exposed physical assets (OSM buildings), land cover (Corine Land Cover 20188, Copernicus Land Monitoring Service), and socio-economic evaluation of reconstruction costs extracted from cadastral estimates (Construction Cost 2010 EUR, EC Harris, 2010).
- Three different types of buildings are considered, namely "residential", "commercial", and "industrial"; each building polygon is categorized by means of construction material, where the construction materials considered are masonry, concrete, wood, mud, or informal (e.g., slum). The classification of each building type and material is done using the information provided by OSM in the open-access building vector file. In case the information provided by OSM is insufficient to classify a certain building polygon (e.g., when a building is simply classified as "yes"), the Corine Land Cover 2018 is utilized for performing the building type classification, while the material considered for residential and commercial building types is masonry, and wood for agricultural buildings.
- A depth-damage function is then applied to translate the hazard magnitude into expected economic damages in relation to the exposed values (Huizinga et al., 2017). Figure 6 depict the pluvial flood depth-damage function for each type of building considered in the analysis, while the constructions costs for each case study are shown in Table 5.
Figure 6: Depth-damage function considered in this report, for the three building types (adapted from Huizinga et al., 2017).
Table 5: Estimated construction costs for each city (adapted from EC Harris 2010 and Huizinga et al., 2017; monetary values are in EUR 2010).
|
|
| Construction Cost (€/m2 – 2010) | Depreciated Value (factor) | Undamageable Fraction (factor) | Max Damage Content/Inventory (fraction) | Maximum Damage (€/m2 – 2010) |
Portugal | Amadora | Residential | 837 | 0.63 | 0.40 | 0.50 | 475 |
Commercial | 898 | 0.63 | 0.40 | 1.00 | 679 | ||
Industrial | 487 | 0.63 | 0.40 | 1.50 | 460 | ||
The Netherlands | Amersfoort | Residential | 1015 | 0.63 | 0.40 | 0.50 | 576 |
Commercial | 1365 | 0.63 | 0.40 | 1.00 | 1032 | ||
Industrial | 958 | 0.63 | 0.40 | 1.50 | 905 | ||
Belgium | Antwerp Brussels | Residential | 1431 | 0.63 | 0.40 | 0.50 | 811 |
Commercial | 1478 | 0.63 | 0.40 | 1.00 | 1117 | ||
Industrial | 983 | 0.63 | 0.40 | 1.50 | 929 | ||
Greece | Athens | Residential | 1108 | 0.63 | 0.40 | 0.50 | 628 |
Commercial | 989 | 0.63 | 0.40 | 1.00 | 748 | ||
Industrial | 716 | 0.63 | 0.40 | 1.50 | 677 | ||
Spain | Bilbao Pamplona | Residential | 1099 | 0.63 | 0.40 | 0.50 | 623 |
Commercial | 1134 | 0.63 | 0.40 | 1.00 | 857 | ||
Industrial | 503 | 0.63 | 0.40 | 1.50 | 475 | ||
United Kingdom | Birmingham London | Residential | 1600 | 0.63 | 0.40 | 0.50 | 907 |
Commercial | 1557 | 0.63 | 0.40 | 1.00 | 1177 | ||
Industrial | 875 | 0.63 | 0.40 | 1.50 | 827 | ||
|
| Residential | 816 | 0.63 | 0.40 | 0.50 | 463 |
Commercial | 832 | 0.63 | 0.40 | 1.00 | 629 | ||
Industrial | 563 | 0.63 | 0.40 | 1.50 | 532 | ||
Hungary | Budapest | Residential | 750 | 0.63 | 0.40 | 0.50 | 425 |
Commercial | 897 | 0.63 | 0.40 | 1.00 | 678 | ||
Industrial | 705 | 0.63 | 0.40 | 1.50 | 666 | ||
| Cologne Frankfurt am Main | Residential | 2159 | 0.63 | 0.40 | 0.50 | 1224 |
Commercial | 1494 | 0.63 | 0.40 | 1.00 | 1129 | ||
Industrial | 978 | 0.63 | 0.40 | 1.50 | 924 | ||
Italy | Milan Rimini | Residential | 1365 | 0.63 | 0.40 | 0.50 | 774 |
Commercial | 1607 | 0.63 | 0.40 | 1.00 | 1215 | ||
Industrial | 740 | 0.63 | 0.40 | 1.50 | 699 | ||
France | Paris | Residential | 1621 | 0.63 | 0.40 | 0.50 | 919 |
Commercial | 1534 | 0.63 | 0.40 | 1.00 | 1160 | ||
Industrial | 1070 | 0.63 | 0.40 | 1.50 | 1011 | ||
Czech Republic | Prague | Residential | 1065 | 0.63 | 0.40 | 0.50 | 604 |
Commercial | 922 | 0.63 | 0.40 | 1.00 | 697 | ||
Industrial | 534 | 0.63 | 0.40 | 1.50 | 505 | ||
Latvia | Riga | Residential | 889 | 0.63 | 0.40 | 0.50 | 504 |
Commercial | 812 | 0.63 | 0.40 | 1.00 | 614 | ||
Industrial | 697 | 0.63 | 0.40 | 1.50 | 659 | ||
Sweden | Stockholm | Residential | 1688 | 0.63 | 0.40 | 0.50 | 957 |
Commercial | 1695 | 0.63 | 0.40 | 1.00 | 1281 | ||
Industrial | 1319 | 0.63 | 0.40 | 1.50 | 1246 | ||
Austria | Vienna | Residential | 1485 | 0.63 | 0.40 | 0.50 | 842 |
Commercial | 1540 | 0.63 | 0.40 | 1.00 | 1164 | ||
Industrial | 915 | 0.63 | 0.40 | 1.50 | 865 |
Validation
Comparing Intensity-Duration-Frequency (IDF) curves at city scale from different data sources
This Section compares (Figure 7) the IDF curves carried out by processing ERA5 and ERA5-2km data (over 1989-2018) with those emerging from a desk-review and interactions with potential local Users for a sub-set of the 20 selected cities (Table 6).
Table 6: Overview of the local IDF reference curves emerged from a desk-review and interactions with potential local Users.
Country | City | Reference | Notes |
|
|
| Homogeneous area; precipitation intensity correlated to the duration for prescribed return periods (2, 5, 10, 50, 100 y); |
the Netherlands | Amersfoort |
| Curve provided by KNMI for the whole country |
Greece | Athens | Koutsoyiannis et al. (1998) | Derived from data recorded at the Helliniko airport (Athens) |
Belgium | Brussels | Van de Vyver (2018) | Related to the Uccle station |
|
| Olsson et al. (2018) | Homogeneous area; table with values for hourly precipitation at prescribed return levels (2, 5, 10, |
Austria | Vienna | Obtained from two gridded precipitation datasets (MaxModN and ÖKOSTRA) at 6 km | |
| Cologne |
|
|
Italy | Milan Rimini | VAPI10 (Flood Evaluation) project | Italian initiative for regional frequency analysis of extreme rainfall and flood data |
United Kingdom | Birmingham London |
| Gridded estimates of hourly areal rainfall (1990—2014) at 1km interpreted with procedure of §2.4.2.2 |
Figure 7: Comparison at city scale of rainfall depth – return period for hourly precipitation carried out by using ERA5 and ERA5-2km with local references retrieved from desk review and Users' interactions.
The comparison points out how the rainfall depth – return period curve for hourly precipitation provided by ERA5-2km represent a reliable tool for pluvial flood risk analysis and design purposes in comparison with local reference data. With respect to ERA5, ERA5-2km returns an evident added value due to the enhancement horizontal resolution, ensuring the ability of the additional layer in reproducing maxima and extreme values.
Examples of pluvial flood risk analysis at city scale
This Section shows the Pluvial Flood Hazard Maps with Return Time 10 and 100 years and the probability damage curve for three selected cities: Cologne in Germany, Milan in Italy, and Pamplona in Spain. For the city of Pamplona, the flood hazard maps, and the probability damage curves are reported both processing EU-DEM at 25 m and LiDAR at 2 m.
The extension of flooded areas and the mean water depth related to different rainfall intensity events characterized by different Return Time (RT) shows a linear behaviour. This linear relationship has been assessed for all the twenty cities under evaluation (see Appendix I) and allows user to evaluate the response of the city's hydraulic system to precipitation input also for values different from those analysed. In this sense, it permits to investigate how the errors in precipitation input from ERA5-2km compared to local institutional references can affect the results and, at the same time, to provide insights about the potential climate change impacts. Of course, the findings should be account for all the assumptions related to the modelling chains and it should be viewed as hardly generalisable to further test cases.
Cologne (Germany) with EU-DEM
Figure 8: Resulting flooded areas and depth applying SAFER_RAIN algorithms to EU-DEM at 25 m burned for Cologne (Germany) – 10 [y] RT.
Figure 9: Resulting flooded areas and depth applying SAFER_RAIN algorithms to EU-DEM at 25 m burned for Cologne (Germany) – 100 [y] RT.
Figure 10: Probability damage curve and expected direct damages [in million EUR 2010] for the commercial, industrial, and residential building classes – Cologne (Germany).
Figure 11: Regression plots Rainfall Intensity versus Flooded Area Extension (left) and Mean Water Depth (right) for Cologne (Germany).
Milan (Italy) with EU-DEM
Figure 12: Resulting flooded areas and depth applying SAFER_RAIN algorithms to EU-DEM at 25 m burned for Milan municipality (Italy) – 10 [y] RT.
Figure 13: Resulting flooded areas and depth applying SAFER_RAIN algorithms to EU-DEM at 25 m burned for Milan municipality (Italy) – 100 [y] RT.
Figure 14: Probability damage curve and expected direct damages [in million EUR 2010] for the commercial, industrial, and residential building classes – Milan (Italy).
Figure 15: Regression plots Rainfall Intensity versus Flooded Area Extension (left) and Mean Water Depth (right) for Cologne (Germany).
Pamplona (Spain) with EU-DEM
Figure 16: Resulting flooded areas and depth applying SAFER_RAIN algorithms to EU-DEM at 25 m burned for Pamplona (Spain) – 10 [y] RT.
Figure 17: Resulting flooded areas and depth applying SAFER_RAIN algorithms to EU-DEM at 25 m burned for Pamplona municipality – 100 [y] RT.
Figure 18. Probability damage curve and expected direct damages [in million EUR 2010] for the commercial, industrial, and residential building classes – Pamplona (Spain).
Figure 19: Regression plots Rainfall Intensity versus Flooded Area Extension (left) and Mean Water Depth (right) for Pamplona (Spain).
Pamplona (Spain) with LiDAR
Figure 20: Resulting flooded areas and depth applying SAFER_RAIN algorithms to LiDAR at 2 m burned for Pamplona municipality (Spain) – 10 [y] RT.
Figure 21: Resulting flooded areas and depth applying SAFER_RAIN algorithms to LIDAR at 2 m burned for Pamplona municipality (Spain) – 100 [y] RT.
Figure 22: Probability damage curve and expected direct damages [in million EUR 2010] for the commercial, industrial, and residential building classes – Pamplona (Spain), using LiDAR data.
Concluding Remarks
Flood Hazard Model Limitations and Outcomes
- Safer_RAIN is a fast-processing pluvial hazard model which identifies pluvial-flooded areas on the basis of nested surface depressions extracted from high-resolution DEMs and the rainfall depth associated with the considered event (possibly accounting for spatially distributed rainfall input and infiltration processes). As main simplifying assumption, Safer_RAIN neglects overland flow dynamics: net-rainfall volume accumulates into the system of nested depressions according to their capacity and hierarchical structure.
- Safer_RAIN flood hazard model is applicable wherever static water depth is the main source of flood hazard (and risk), this is a typical situation occurring for example in land reclamation managed flatlands or in urban districts characterized by deeply altered topographic surfaces and frequent and large flat and sub-horizontal areas. In such contexts most severe conditions are well represented by accumulation of pluvial runoff right at the end an intense event before soil drainage and sewage system drainage can gradually recover them.
- DTM plays a crucial role both in identifying real depressions; specifically, EU-DEM at 25 m, although able to quickly identify macroscopic depression along the surface, suffers from poor hydrological connectivity with real surface channel -river network. This leads to artificial sink and to identification of fake flooded areas. The problem can be partially overcome using hydrological conditioning via stream burning along existing river-channel network, accepting local errors this may introduce along the network itself. EU-DEM, improved through stream burning, is suitable option for the analysis across European selected cities due to the availability of the required datasets.
- The extension of flooded areas and total water volume or mean water depth has been observed to have a linear regression with rainfall intensity representative of RT from 10 to 200 years.
Pluvial Flood Risk Limitations and Outcomes
- The considered pluvial flood risk assessment methodology builds upon the fast-processing pluvial flood hazard model Safer_RAIN by promoting an expeditive methodology capable of being replicated across different urban centres in Europe while maintaining a good compromise between providing direct damage estimates at the building scale and computational requirements.
- The pluvial flood risk and the expected annual damages due to pluvial flooding strongly depends on the pluvial flood hazard characterisation, which in turn is strictly related to the identification of terrain depressions on the urban fabric. In this aspect, pluvial flood risk assessments are expected to improve significantly as high-resolution DEM are made available and those are validated and corrected for specific urban-scale applications.
- Data supporting the identification of exposed elements is also crucial for the correct identification and classification of buildings. OpenStreetMap (OSM) provides a very useful, open-access spatial database that can support this kind of analysis. However, data quality and contents are subject to availability to the case study area, which in turn is subject to contributions from the OSM community. As such, the characterisation of exposed elements can vary significantly from case study to case study.
- The utilisation of a same functional form to describe the depth-damage relationship across different urban centres in Europe can be a limiting factor. As such, for urban and local scale applications, it is recommended the utilisation of locally-calibrated flood damage functions.
Appendix I: Flood Hazard Regression Plots at city scale
This section provides the extension of flooded areas and the mean water depth related to different rainfall intensity events characterized by different Return Time (RT) computed for all the twenty cities under evaluation with EU-DEM as supporting DEM.
Figure 23: Regression plots Rainfall Intensity versus Flooded Area Extension (left) and Mean Water Depth (right) for Amadora (Portugal).
Figure 24: Regression plots Rainfall Intensity versus Flooded Area Extension (left) and Mean Water Depth (right) for Amersfoort (the Netherlands).
Figure 25: Regression plots Rainfall Intensity versus Flooded Area Extension (left) and Mean Water Depth (right) for Antwerp (Belgium).
Figure 26: Regression plots Rainfall Intensity versus Flooded Area Extension (left) and Mean Water Depth (right) for Athens (Greece).
Figure 27: Regression plots Rainfall Intensity versus Flooded Area Extension (left) and Mean Water Depth (right) for Bilbao (Spain).
Figure 28: Regression plots Rainfall Intensity versus Flooded Area Extension (left) and Mean Water Depth (right) for Birmingham (United Kingdom).
Figure 29: Regression plots Rainfall Intensity versus Flooded Area Extension (left) and Mean Water Depth (right) for Brussels (Belgium).
Figure 30: Regression plots Rainfall Intensity versus Flooded Area Extension (left) and Mean Water Depth (right) for Bucharest (Romania).
Figure 31: Regression plots Rainfall Intensity versus Flooded Area Extension (left) and Mean Water Depth (right) for Budapest (Hungary).
Figure 32: Regression plots Rainfall Intensity versus Flooded Area Extension (left) and Mean Water Depth (right) for Cologne (Germany).
Figure 33: Regression plots Rainfall Intensity versus Flooded Area Extension (left) and Mean Water Depth (right) for Frankfurt Am Main (Germany).
Figure 34: Regression plots Rainfall Intensity versus Flooded Area Extension (left) and Mean Water Depth (right) for London (United Kingdom).
Figure 35: Regression plots Rainfall Intensity versus Flooded Area Extension (left) and Mean Water Depth (right) for Milan (Italy).
Figure 36: Regression plots Rainfall Intensity versus Flooded Area Extension (left) and Mean Water Depth (right) for Pamplona (Spain).
Figure 37: Regression plots Rainfall Intensity versus Flooded Area Extension (left) and Mean Water Depth (right) for Paris (France).
Figure 38: Regression plots Rainfall Intensity versus Flooded Area Extension (left) and Mean Water Depth (right) for Prague (Czech Republic).
Figure 39: Regression plots Rainfall Intensity versus Flooded Area Extension (left) and Mean Water Depth (right) for Riga (Latvia).
Figure 40: Regression plots Rainfall Intensity versus Flooded Area Extension (left) and Mean Water Depth (right) for Rimini (Italy).
Figure 41: Regression plots Rainfall Intensity versus Flooded Area Extension (left) and Mean Water Depth (right) for Stockholm (Sweden).
Figure 42: Regression plots Rainfall Intensity versus Flooded Area Extension (left) and Mean Water Depth (right) for Vienna (Austria).
Appendix II: Variable Description
This section provides a more detailed analysis for each indicator, including variable in this section and include information on the valid and real range of values of the variable, the flag values for missing data, and further descriptive statistics.
Water depth (wd)
- Description: Water Depth
- Units: m
- Definition: Pluvial Flooding Water Depth induced by hourly precipitation events with fixed return periods computed over 30-year (1989-2018) for 20 European cities
- Flag values for missing data: Not Expected
Table 7: Mean, median, and standard deviation carried out by merging for each city data derived from different return periods (data < 0.1 m are removed from the sample)
City | DEM | Mean (m) | Median (m) | St. Dev. (m) |
Amadora | EU-DEM | 0.68 | 0.46 | 0.70 |
Amersfoort | EU-DEM | 0.37 | 0.30 | 0.26 |
Antwerp | EU-DEM | 0.42 | 0.30 | 0.41 |
Athens | EU-DEM | 0.69 | 0.47 | 0.70 |
Bilbao | EU-DEM | 0.82 | 0.53 | 0.91 |
Birmingham | EU-DEM | 0.58 | 0.37 | 0.69 |
Brussels | EU-DEM | 0.58 | 0.40 | 0.53 |
Bucharest | EU-DEM | 0.47 | 0.36 | 0.44 |
Budapest | EU-DEM | 0.52 | 0.39 | 0.44 |
Frankfurt am Main | EU-DEM | 0.52 | 0.39 | 0.47 |
Koln | EU-DEM | 0.40 | 0.31 | 0.31 |
Koln | LiDAR | 0.32 | 0.21 | 0.34 |
London | EU-DEM | 0.45 | 0.32 | 0.45 |
Milan | EU-DEM | 0.43 | 0.34 | 0.34 |
Milan | LiDAR | 0.44 | 0.24 | 0.60 |
Pamplona | EU-DEM | 0.60 | 0.40 | 0.67 |
Pamplona | LiDAR | 0.41 | 0.27 | 0.41 |
Paris | EU-DEM | 0.49 | 0.35 | 0.47 |
Prague | EU-DEM | 0.66 | 0.42 | 0.76 |
Riga | EU-DEM | 0.39 | 0.30 | 0.31 |
Rimini | EU-DEM | 0.45 | 0.35 | 0.36 |
Stockholm | EU-DEM | 0.64 | 0.41 | 0.83 |
Vienna | EU-DEM | 0.52 | 0.34 | 0.61 |
Water Masks
- Description: Binary
- Units: Dimensionless
- Definition: Mask of flooded areas. The values are between 0 (not affected) and 1 (affected) and are function of hourly precipitation events with fixed return periods computed over 30-year (1989-2018) for 20 European cities
- Flag values for missing data: Not Expected
Urban depression
- Description: Depression Id
- Units: Dimensionless
- Definition: Topographic depressions which are defined as areas without an outlet and often referred to as sinks or pits
- Flag values for missing data: Not Expected
Expected damage
- Description: Direct Damage
- Units: €/m2
- Definition: Direct damage suffered by the existing assets due to pluvial flooding induced by hourly precipitation events with fixed return periods computed over 30-year (1989-2018) for 20 European cities
- Flag values for missing data: Not Expected
Table 8: Mean, median, and standard deviation carried out by merging for each city data derived from different return periods (data < 0.1 €/m2 are removed from the sample)
City | DEM | Mean (€/m2) | Median (€/m2) | St. Dev. (€/m2) |
Amadora | EU-DEM | 64.03 | 31.92 | 100.88 |
Amersfoort | EU-DEM | 104.85 | 54.54 | 128.82 |
Antwerp | EU-DEM | 85.59 | 48.11 | 103.95 |
Athens | EU-DEM | 34.14 | 16.83 | 54.44 |
Bilbao | EU-DEM | 65.14 | 38.21 | 92.05 |
Birmingham | EU-DEM | 101.65 | 42.05 | 185.52 |
Brussels | EU-DEM | 124.92 | 63.85 | 179.38 |
Bucharest | EU-DEM | 26.29 | 11.72 | 38.05 |
Budapest | EU-DEM | 26.06 | 13.45 | 35.88 |
Frankfurt am Main | EU-DEM | 47.13 | 19.23 | 76.76 |
Koln | EU-DEM | 49.07 | 21.15 | 71.44 |
Koln | LiDAR | 37.52 | 17.74 | 53.17 |
London | EU-DEM | 95.90 | 44.47 | 142.31 |
Milan | EU-DEM | 33.64 | 17.48 | 49.59 |
Milan | LiDAR | 38.83 | 21.96 | 51.83 |
Pamplona | EU-DEM | 51.51 | 33.88 | 66.05 |
Pamplona | LiDAR | 47.73 | 40.44 | 36.65 |
Paris | EU-DEM | 43.91 | 21.56 | 64.49 |
Prague | EU-DEM | 29.49 | 13.67 | 49.97 |
Riga | EU-DEM | 20.83 | 10.49 | 28.71 |
Rimini | EU-DEM | 40.48 | 15.35 | 58.48 |
Stockholm | EU-DEM | 44.49 | 20.02 | 78.01 |
Vienna | EU-DEM | 39.34 | 20.59 | 61.61 |
Appendix III: Input Data Description
Input Data 1: ERA5-2km
Table 9: Overview of key characteristics of ERA5-2km Reanalysis
Data Description | |
Main Variables | Total Precipitation (mm) |
Domain | Europe (20 selected cities) (variable domain sizes) |
Horizontal resolution | 0.02° x 0.02° |
Temporal coverage | 1989-01-01 00:00/to/2018-12-31 23:00 |
Temporal resolution | Hourly |
Vertical coverage | Near surface |
Update frequency | None |
Model | Rockel et al., 2008 (DOI: 10.1127/0941-2948/2008/0309) |
Experiment | Evaluation |
Provider | Centro Euro-Mediterraneo sui Cambiamenti Climatici (CMCC) |
References
Beersma, J., Versteeg, R., and Hakvoort, H.: Neerslagstatistieken voord korte duren, Tech. Rep., STOWA, Amersfoort, Netherlands, 58 pp., 2018.
Chow, V.T., Maidment, D.R. & Mays, L.W. (1988) Applied Hydrology, McGraw-Hill
EC Harris, 2010. International buildings costs worldwide | may 2010. EC Harris Built Asset Consultancy.
Huizinga J, De Moel H, Szewczyk W (2017) Methodology and the database with guidelines Global flood depth-damage functions
Kottegoda, N. T., & Rosso, R. (2008). Applied statistics for civil and environmental engineers (2nd ed.). UK: John Wiley and Sons
Koutsoyiannis, D., Kozonis, D., Manetas, A. (1998). A mathematical framework for studying rainfall intensity-duration-frequency relationships. Journal of Hydrology, 206 (1–2), 118-135. https://doi.org/10.1016/S0022-1694(98)00097-3
Lewis, E., Quinn, N., Blenkinsop, S., Fowler, H. J., Freer, J., Tanguy, M., Hitt, O., Coxon, G., Bates, P., and Woods, R.: Gridded estimates of hourly areal rainfall for Great Britain (1990-2014) [CEH- GEAR1hr], NERC Environmental Information Data Centre, doi:10.5285/d4ddc781-25f3-423a-bba0- 747cc82dc6fa, 2019.
Olsson, J., Berg, P., Eronn, A., Simonsson, L., Södling, J., Wern, L., and Yang, W.: Extremregn i nuvarande och framtida klimat: analyser av observationer och framtidsscenarier, Klimatologi 47, SMHI, 83 pp., 2018.
Padulano, R., Reder, A. and Rianna, G. (2019) An ensemble approach for the analysis of extreme rainfall under climate change in Naples (Italy). Hydrological Processes 33.14 (2019): 2020-2036, https://doi.org/10.1002/hyp.13449
Rockel, B.; Will, A.; Hence, A. The regional climate model COSMO-CLM (CCLM). Meteorol Z 2008, 17, 347–348, doi: 10.1127/0941-2948/2008/0309.
Rossi, F., Villani, P. (1994). Valutazione delle Piene in Campania [Flood evaluation in Campania]. CNR‐GNDCI, Publ. No. 1472, Grafica Matelliana & C., Cava de’ Tirreni (SA), Italy (in Italian).
Samela, C.; Persiano, S.; Bagli, S.; Luzzi, V.; Mazzoli, P.; Humer, G.; Reithofer, A.; Essenfelder, A.; Amadio, M.; Mysiak, J.; Castellarin, A. Safer_RAIN: A DEM-Based Hierarchical Filling-&-Spilling Algorithm for Pluvial Flood Hazard Assessment and Mapping across Large Urban Areas. Water 2020, 12, 1514.
Van de Vyver, H. Spatial regression models for extreme precipitation in Belgium. Water Resour. Res. 2012, 48.
Viglione, A., Laio, F., & Claps, P. (2007). A comparison of homogeneity tests for regional frequency analysis. Water Resources Research, 43, W03428
Wouters, H.; Demuzere, M.; Blahak, U.; Fortuniak, K.; Maiheu, B.; Camps, J.; Tielemans, D.; van Lipzig,
N.P.M. The efficient urban canopy dependency parametrization (SURY) v1.0 for atmospheric modelling: description and application with the COSMO-CLM model for a Belgian summer. Geosci Model Dev 2016, 9(9), 3027-3054, doi:10.5194/gmd-9-3027-2016.