Earth Syst. Sci.

Data, 16, 375–385, 2024

https://doi.org/10.5194/essd-16-375-2024
© Author(s) 2024. This work is distributed under
the Creative Commons Attribution 4.0 License.

Flood simulation with the RiverCure approach: the open

dataset of the 2016 Águeda flood event
Ana M. Ricardo1 , Rui M. L. Ferreira3 , Alberto Rodrigues da Silva2 , Jacinto Estima2,4 , Jorge Marques2 ,
Ivo Gamito2 , and Alexandre Serra2
1 CERIS – Civil Engineering Research and Innovation for Sustainability, 1049-001 Lisbon, Portugal
2 INESC-ID, Instituto Superior Técnico, Universidade de Lisboa, 1049-001 Lisbon, Portugal
3 CERIS, Instituto Superior Técnico, Universidade de Lisboa, 1049-001 Lisbon, Portugal
4 CISUC, Departamento de Engenharia Informática, Universidade de Coimbra, 3030-790 Coimbra, Portugal

Correspondence: Ana M. Ricardo (ana.ricardo@tecnico.ulisboa.pt)

Received: 20 March 2023 – Discussion started: 12 April 2023

Revised: 16 September 2023 – Accepted: 17 November 2023 – Published: 16 January 2024

Abstract. Floods are among the most common natural disasters responsible for severe damages and human
losses. Numerically produced data, managed by user-friendly tools for geographically referenced data, have been
adopted to increase preparedness and reduce vulnerabilities. This paper describes the locally sensed and numer-
ically produced data that characterize a flood event that occurred in February 2016 in the Portuguese Águeda
River, referred to as Agueda.2016Flood for short. The data were managed through the RiverCure portal, a collab-
orative web platform connected to a validated shallow-water model featuring modelled dynamic bed geometries
and sediment transport. The dataset provides a synthesis of topo-bathymetric, hydrometric and numerically pro-
duced data from a calibrated hydrodynamic model. Due to the lack of measured hydrometric data near the city,
the numerically produced data are crucial for a complete description of the flood event. The Agueda.2016Flood
dataset constitutes a complete validation test for flood forecasting models and a tool to better mitigate floods in
this river and in similar rivers. Thus, Agueda.2016Flood is a relevant dataset for Águeda River stakeholders as
well as for the community of flood modellers, as it provides a well-documented validation event for forecasting
tools. The dataset can be accessed at https://doi.org/10.4211/hs.937927473a3a4e66a07a2e2fdd9d581e (Ricardo
et al., 2024).

1 Introduction While planning for flood resilience differs from nowcast-

ing (Yuan et al., 2022), both activities rely on analysing avail-
Floods cause widespread damages and human losses. For in- able hydrological data and on hydraulic modelling, even at
stance, according to the Centre for Research on the Epidemi- different timescales. Crucially, both require decision sup-
ology of Disasters, floods were the most frequent type of port systems that channel available input data, that conduct
natural disaster between 1998 and 2017 (Wallemacq et al., hydraulic modelling (possibly combined with the assimila-
2018). There have been changes in the frequency and magni- tion of input data) and that collect and organize results into
tude of floods (Amponsah et al., 2018), including an increase datasets ready to be interpreted by decision-makers. Foster-
in their severity in central and northern Europe (Blöschl ing the Digital Earth concept, these benefit from current re-
et al., 2019; Hall et al., 2015). Recently, severe floods were search efforts in designing and developing digital earth twin
observed in Germany and Belgium in the summer of 2021 hydrology systems (DARTHs), which are not isolated mod-
(Fekete and Sandholz, 2021). Accurate forecasting tools and els but are services or components that can be assembled to-
early warning systems have been proposed to increase pre- gether (Rigon et al., 2022).
paredness and reduce systemic vulnerabilities (Yuan et al.,
2022).

Published by Copernicus Publications.

376 A. M. Ricardo et al.: Flood simulation with the RiverCure approach

This paper describes a synthesis of data that allows the 2 Supported software tools
spatial and temporal characterization of a flood that occurred
in the Portuguese Águeda River in February 2016. It also The software tools used in the scope of this research are the
describes the main features of the RiverCure portal (RCP), RiverCure portal and the HiSTAV (High Performance Com-
the information system employed to manage and articulate puting version of the original Strong Transients in Alluvial
the locally sensed data and the numerically produced data. Valleys) numerical model, which run in an integrated way.
The RiverCure portal (http://rivercure.inesc-id.pt/, last ac- The RCP is a collaborative web application, publicly avail-
cess: 20 December 2023) is a web platform aligned with able at http://rivercure.inesc-id.pt/ (last access: 20 Decem-
the DARTHs paradigm and developed to integrate hydrody- ber 2023), that allows different organizations to set up and
namic and morphodynamic modelling tools and input data manage their geographic contexts, sensors and events in a
(Rodrigues da Silva et al., 2023). controlled way. The RCP has been developed with Web and
The Águeda River runs through the city of Águeda, which GIS technologies, such as Django, GeoDjango, PostgreSQL
has registered many flood occurrences causing significant and PostGIS. The RCP conveys geographical and hydrolog-
damages. As an area prone to flooding, the municipality of ical data to a numerical model as a processing service to
Águeda has made considerable investments in flood protec- provide simulation results. A detailed description of the con-
tion: in 2015, it invested around EUR 2 million towards con- cepts, design and technologies behind the RCP is presented
structing a secondary river channel to divert the river flow. by Rodrigues da Silva et al. (2023). Herein, only a brief de-
However, those efforts were not enough to prevent a severe scription of the main features is provided.
flood event in February 2016, which was the most significant The structure of the RCP consists of two main top-level
flood of the previous 15 years (https://www.tsf.pt/sociedade/ concepts: contexts and sensors. Context allows definition of
maiores-cheias-dos-ultimos-anos-em-agueda-5027639. the geometry of the area and management of the events to
html, last access: 7 January 2024). That flood revealed weak- be modelled and simulated. The geometrical details consist
nesses in the flood defence infrastructure of the municipality, of the set of georeferenced features that are visualized and
although the water level of the river, in the city’s vicinity, edited over a basemap from OpenStreetMaps (Haklay and
was, at its peak, 10 to 20 cm below the protection wall Weber, 2008) through the Leaflet (https://leafletjs.com/, last
crest. That flood event was the result of heavy precipitation access: 16 September 2023) library (Peterson, 2014; Edler
associated with strong instabilities in the North Atlantic, and Vetter, 2019). The polygons and polylines can be either
causing significant disturbances in the city of Águeda and uploaded (i.e. when they are developed offline) or directly
affecting its mobility, public services and infrastructures. defined in the RCP. The raster files containing the digital el-
Although flooding is a recurring problem of the city of evation model (DEM) and the roughness coefficients should
Águeda, there are no curated datasets that allow for a descrip- be uploaded. The geographic features of the context include
tion of its flooding events. This paper aims to fulfil gap as- the mesh generation process, as the mesh is a required fea-
sembling of the key input geographical and hydrologic data, ture of the numerical modelling. The mesh is generated from
the pre-processing steps and the rationale for establishing the the geometrical and raster input data and employs the open-
parameters of the hydrodynamic modelling tools that ulti- source 3D finite-element grid generator Gmsh (Geuzaine and
mately produce the hydrometric datasets of the aforemen- Remacle, 2009). The connection between the RCP and the
tioned February 2016 flood event. To have a complete flood numerical model is established by defining an event to be
description, numerically produced hydrometric data (flow modelled. The event features allow the user to define the re-
depths, discharges and flow velocities) are required due to sults’ writing rates and the simulation’s starting and ending
the lack of reliable physical data as the hydrometric sensor times.
installed at the Águeda bridge was malfunctioning during the The sensors represent cyber-physical devices that load and
flood event. Thus, the presented dataset is a unique synthesis store the environmental data in the RCP. Generically, a sen-
of topo-bathymetric, hydrological and numerically produced sor reads and collects data in a given location, usually a ge-
hydrometric data. It is very relevant locally as it documents ographical point, with a given frequency. The sensor obser-
an event relevant to Águeda River stakeholders. It is also rel- vations (data values) are automatically obtained or manually
evant to a wider community of flood modellers as it provides introduced in the RCP by users, either on a value-by-value
a well-documented validation event for flood models, partic- basis or in sets of values by a given spreadsheet. The map-
ularly shallow-water models. The community can also use ping between a sensor and a context is defined by associating
this dataset as a starting point to design independent experi- a sensor with a boundary point or an entire boundary of a
ments and tools and as a training example for more general context.
usage of the RiverCure portal. The numerical model currently linked to the RCP is HiS-
TAV; however, it could use other existing simulation tools.
HiSTAV is an in-house shallow-water model featuring dy-
namic bed geometries and sediment transport modelled via
the difference between capacity bedload discharge and local

Earth Syst. Sci. Data, 16, 375–385, 2024 https://doi.org/10.5194/essd-16-375-2024

A. M. Ricardo et al.: Flood simulation with the RiverCure approach 377

Figure 1. Map of the modelled area (domain) including the geometrical features for mesh generation (alignment and refinements) and the
location of the inlet and outlet boundaries. The orange triangles show the locations of the hydrometric stations (Ribeiro and Ponte Redonda).
The map was produced with QGIS employing the basemap from © OpenStreetMap contributors 2023. Distributed under the Open Data
Commons Open Database License (ODbL) v1.0.

solid discharge (Ferreira et al., 2009; Canelas et al., 2013; with two nodes, each with two Xeon(R) E5-2650 Octa-core
Conde et al., 2020; Conde, 2018). The model comprises CPUs and four NVIDIA GPUs with 128 GB ECC RAM in-
the total mass conservation equation, the momentum conser- stalled and a 10 K rpm hard disk with a capacity of 1 TB
vation equation and the sediment mass conservation equa- (TYAN FT77A-B7059).
tion. The conservation equation system is hyperbolic and is The data produced by HiSTAV are stored in .vtk files
discretized with a finite-volume method employing a flux- created by the Visualization Toolkit (VTK), state-of-the-art
splitting technique and a reviewed Roe–Riemann solver with open-source software for manipulating and displaying scien-
appropriate source-term formulations to ensure full conser- tific data (https://vtk.org/, last access: 16 September 2023).
vativeness (Conde et al., 2020). The sediment mass conser-
vation expresses the interaction between the bed and the flow
in the bed. 3 Data records
In HiSTAV, the closed conservation equations are dis-
cretized by a finite-volume approach and are solved explic- 3.1 Overview of the dataset
itly, obeying a Courant–Friedrichs–Lewy (CFL) condition The simulation of the 2016 Águeda flood event involved a ge-
that is conventional for restricting the computational time ographical extent corresponding to an area of 560 ha crossed
step. The implementation is entirely cross-compatible be- by a 9.8 km long stretch of the Águeda River, including the
tween CPUs and GPUs through an intuitive object-oriented riverfront part of Águeda as represented in Fig. 1. Regarding
approach (Conde et al., 2020). It thus supports distributed the time span, this flood event was defined between 9 Febru-
and heterogeneous computing of significant problems at very ary 2016, 00:00:00 WET and 16 February 2016, 23:00:00.
high resolutions. Therefore, there are no special requirements Figure 2 presents an overview of the Agueda.2016Flood
for the machines that may run HiSTAV; it can spread its load dataset. The dataset includes the geometrical features of the
through available GPUs or CPUs. The current simulations simulated context and two main subsets of data: the field
were run on a multi-CPU or multi-GPU computational server data, including topo-bathymetric, land use and hydrologic

https://doi.org/10.5194/essd-16-375-2024 Earth Syst. Sci. Data, 16, 375–385, 2024

378 A. M. Ricardo et al.: Flood simulation with the RiverCure approach

Figure 2. Schematic representation of the dataset organization with the identification of the files included in each part.

data collected to feed the flood event simulation, and the nu- refinement zones as well as of the polyline defining the main
merically produced hydrometric data. Details on each part of channel alignment. Furthermore, to fulfil the mesh generator
the dataset are provided in the following subsections. requirements, the geometries cannot intersect. That is, any
The Agueda.2016Flood dataset (Ricardo et al., 2024) is larger polygon must completely contain a given polygon or
publicly available in the HydroShare repository. The specifi- polyline.
cation of the employed data entities uses the controlled nat- The geometrical features subset also includes the
ural language RSL as discussed in Rodrigues da Silva and Agueda_mesh.vtk file containing the mesh used in the sim-
Savić (2021) (see Appendix A for details on the main con- ulation of the flood event. The mesh consists of a set of un-
cepts and the properties related to the context definition). structured triangular cells containing the following data: cell
ID (named “Address”), bed level (m), roughness coefficient
3.2 Geometrical features (m1/3 s−1 ), a physical tag to identify the cells corresponding
to boundary cells and the length of the largest cell size (m).
The subset named Geometrical features contains auxiliary
elements rather than actual data. These elements include the 3.3 Field data
features to geometrically define the simulation and its mesh.
Concerning the geometrical characteristics of the simula- The field data subset corresponds, in a broad sense, to
tion, the numerical model requires the following set of lines data locally collected. It includes local topography, river
and polygons in GeoJSON format. bathymetry, land use data and hydrologic data obtained from
two hydrometric stations.
– Domain: a polygon delimitating the area to be modelled

– Refinements: polygons delimitating areas where spe- 3.3.1 Hydrological data

cific mesh refinements are desired The hydrological data employed as model input data con-
sist of hourly discharge time series obtained from two hy-
– Alignments: polylines identifying the river centreline
drometric stations managed by the Portuguese Environmen-
– Boundaries: polylines aligned with domain limits defin- tal Agency (APA, https://snirh.apambiente.pt/, last access:
ing the inlet and outlet boundaries of the domain. 16 September 2023), Ribeiro and Ponte Redonda stations
(Table 1). Ribeiro station is located within the simulated
The boundaries are characterized by the type of boundary area on the Alfusqueiro River, close to its confluence with
condition (input, known output, critical or transmissive) and, the Águeda River. Ponte Redonda station is located on the
when applicable, the type of data to be given as input to the Águeda River, 5 km upstream of the domain boundary. Both
model (e.g. depth, discharge, velocity or elevation). These stations are equipped with automatic sensors that measure
features can be set or updated through the user interface of the hydrometric level at every hour, except when a critical
the RCP. The polylines defining the main channel alignment threshold is reached. In that case, the measurement frequency
and the refinement polygons are not mandatory: they are used is increased (Pereira, 2021). The measured flow levels (h) are
for custom-made improvements of the mesh size definition. converted to discharges (Q) by the rating curves defined and
A feature named “CL”, storing an indicative value for the updated by the national authorities. Further details on the hy-
length of the largest cell size in metres, must be an attribute drometric stations and the corresponding rating curves can
of the polygons defining the domain and the desired mesh be obtained from the APA website or from Pereira (2021).

Earth Syst. Sci. Data, 16, 375–385, 2024 https://doi.org/10.5194/essd-16-375-2024

A. M. Ricardo et al.: Flood simulation with the RiverCure approach 379

Table 1. Information on the name of the file containing the discharge data, the location and the rating curves of each station.

Name Data file Location Rating curve

(ETRS89/Portugal TM06)
coordinate system
Latitude Longitude
Ribeiro Hydro_Ribeiro.xlsx 40.566 −8.398 Q = 16.222 × (h − 0.805)2.39 for 0.80 m < h < 1.83 m
Q = 4.632 × h2.168 for 1.83 m < h < 4.5 m
Ponte Redonda Hydro_PonteRedonda.xlsx 40.546 −8.378 Q = 11.079 × (h − 0.877)2.35 for 0.877 m < h < 2.22 m
Q = 1.124 × (h + 0.64)2.84 for 2.22 m < h < 5 m

A synthetic and constant time series, with 20 h in length cant effort was made to identify adequate and accurate data
and a discharge equal to the value obtained for 9 Febru- from elements provided by the municipality of Águeda and
ary 2016, 00:00, was added to the event data series to ensure to collect field data regarding the elevation of the protec-
the model warm-up. Table 1 summarizes the information re- tion walls of the riverfront of Águeda and regarding the river
garding the two hydrometric stations. bathymetry. Since the city was flooded not because of the
Another hydrometric station exists within the simulated protection walls overtopping but because of vulnerabilities
area, Ponte de Águeda station, located on the bridge that in these walls, these had to be investigated to be incorpo-
crosses the river in downtown Águeda. The data from that rated into the DEM, reproducing the geometry of the under-
station proved to be inaccurate due to possible malfunction- passages of the Águeda bridge.
ing of the sensor as well as the lack of an update of the rating Although based on the relatively poor resolution of the
curves after the regulation works carried out in 2015 on the ASTER GDEM, the produced DEM ensured the accuracy
river banks in the vicinity of Águeda. and adequacy of the spatial resolution regarding the river bot-
tom and the city in the river’s vicinity.

3.3.2 Topo-bathymetric data 3.3.3 Land use data

The ASTER Global Digital Elevation Model (GDEM) The spatial distribution of the roughness coefficient over the
Version 3 (ASTGTM), publicly provided by NASA modelled area is also a required model input. It was obtained
(https://search.earthdata.nasa.gov/), was the base for the to- from COS2018, a spatial dataset produced by the Portuguese
pographic characterization of the modelled area. The ASTER governmental office for territorial development, DGT (Di-
platform has three subsystems: a visible and near-infrared reção Geral do Território – https://www.dgterritorio.gov.pt/
radiometer, a shortwave-infrared radiometer and a thermal- dados-abertos, last access: 16 September 2023), represent-
infrared radiometer (Yamaguchi et al., 1998; Fujisada et al., ing the thematic map of land use and land cover for main-
2018). Version 3 counts with a technical methodology to im- land Portugal for the year 2018. COS2018, made available
prove the initial tile-based waterbody data (Fujisada et al., as Linked Open Data (LOD) in RDF format, is a map of
2018). The geographic coverage of the ASTER GDEM ex- polygons with a defined minimum cartographic unit (1 ha)
tends from 83◦ N to 83◦ S, divided into 22 912 tiles with a with a distance between lines equal to or greater than 20 m.
spatial resolution of 1 arcsec (approximately 30 m horizon- It is based on photointerpretation and has a terminology with
tal posting at the Equator). Each tile is distributed in Geo- more than 80 classes. Manning’s coefficient values (n) used
TIFF format and is projected onto the 1984 World Geodetic for specific land cover are based on the well-established
System (WGS84)/1996 Earth Gravitational Model (EGM96) values presented in Chow (1959) and van der Sande et al.
geoid. The ASTGTMV003_N40W009 tile, which includes (2003). The Agueda_FrictionCoef.tif raster file includes the
the littoral centre and the north of Portugal, was downloaded, spatial distribution of the Manning–Strickler friction coeffi-
and the region of interest was cropped and employed as a cient, Ks = 1/n (m1/3 s−1 ).
starting point for the DEM provided by the Agueda_DEM.tif
raster file. The base raster was edited for accuracy and reso-
3.4 Hydrometric data
lution improvement by adding detail to the river bottom and
banks, particularly in the vicinity of Águeda. The hydrometric data to characterize the flood event were
The river centreline and width were defined through produced numerically by HiSTAV. This subset of the dataset
Google Earth. The Águeda River in front of Águeda has includes the spatial distribution of the maximum values of the
suffered many interventions that resulted in a channel with hydrometric variables and instantaneous maps for a selected
a rectangular cross section with protection walls. A signifi- set of time instants.

https://doi.org/10.5194/essd-16-375-2024 Earth Syst. Sci. Data, 16, 375–385, 2024

380 A. M. Ricardo et al.: Flood simulation with the RiverCure approach

Figure 3. Photo of the water level of the river (a) and © Google Earth Pro view (street view) from the location where the photo was taken.

Figure 4. Photo of the flood in a square (Praça da Repúlica) and a street (Rua Vasco da Gama) in downtown Águeda and © Google Earth
Pro view (street view) from the location where the photo was taken.

3.4.1 Envelope of maxima 4 Data validation

The Agueda_maxima.vtk file corresponds to the envelope of
maxima for each variable, i.e. stores the maximum values of The accuracy of the discharge time series computed from the
each mesh cell for the modelled flow variables. The variables data collected by the Ribeiro and Ponte Redonda hydromet-
included are flow depth, water level, flow velocity, hazard ric stations was confirmed by Pereira (2021) by hydrological
index (defined as h × (v + 0.5), where h is the flow depth and and hydraulic modelling employing rain data from radar and
v is the flow velocity according to Penning-Rowsell et al., from rain gauges.
2005), the time to reach the maximum hazard index of each The fixed bed version of HiSTAV requires the empirical
cell and the time to reach the wet state of each cell. quantification of one parameter to complete the roughness
The rate of update of maximum values during a given sim- closure. In this case, it is the conveyance coefficient (the in-
ulation can be set by the user on the RCP. In the present sim- verse of a roughness coefficient) of the Manning–Strickler
ulations, the maxima were updated every second. equation. The model was given maps of the Manning–
Strickler coefficients that were merged with the mesh at the
3.4.2 Time lapse pre-processing stage. No formal calibration was necessary. A
bulk verification of the roughness parameters and the topo-
The spatial distribution of bed level (m), water level (m) and bathymetry was undertaken: it verified that the roughness
flow velocity (m s−1 ) for a set of time instants is provided by and topo-bathymetric data led to the observed river water
the hydro-*.vtk files. The numbers in the filename indicate level in the vicinity of Águeda at t = 110 h (12 February
the simulation time step. 2016, 18:00:00 WET), i.e. 10–20 cm below the protection
In the present case, the time-dependent result files were wall crest.
written hourly but were sampled every 3 h as a compromise The validation tests guarantee the quality of the numeri-
between a detailed time characterization and storage man- cally produced data shown by the HiSTAV tool. The HiS-
agement. The hourly dataset will be shared directly with the TAV model has been compared with theoretical solutions
authors upon request. and experimental data for both resistance and solid transport
(Canelas et al., 2013) and has been tested in a benchmark text
of 2D dam-break flows over a sand bed (Soares-Frazão et al.,
2012). HiSTAV has also been employed to model a tsunami

Earth Syst. Sci. Data, 16, 375–385, 2024 https://doi.org/10.5194/essd-16-375-2024

A. M. Ricardo et al.: Flood simulation with the RiverCure approach 381

Figure 5. Distribution of the maximum values of the computed water depth overlapped on the map. The colour scale is in metres (m). The
map was produced with QGIS employing the basemap from © OpenStreetMap contributors 2023. Distributed under the Open Data Commons
Open Database License (ODbL) v1.0.

in the Tagus estuary (Lisbon), demonstrating a good perfor- Another way to verify the reliability of the produced
mance on the Monai Valley benchmark and in comparison dataset is to compare the input discharge data with the nu-
with historical data (Conde et al., 2015). merically produced data, as shown in Fig. 6. The “upstream”
The 2016 Águeda flood event was largely reported in the hydrograph corresponds to the numerically produced dis-
Portuguese media, allowing us to gather informal but accu- charge on the cross section at approximately 400 m down-
rate data regarding the water depth and extent of the flooded stream of the domain limit at the intersection of the two in-
areas. The efforts in improving the topo-bathymetric data al- let channels. The discharge computed in that section coin-
lowed us to obtain good agreement between the modelled cides with the summation of the two input hydrographs. The
and observed flood extent in downtown Águeda. “downstream” hydrograph was obtained at the domain outlet
The 2016 flood event is of particular interest due to the boundary. The comparison of the two computed hydrographs
large media coverage allowing us to gather informal but shows the expected delay and damping of the flood from the
accurate data regarding the water depth and extent of the entrance to the outlet of the domain.
flooded areas. In this event, the water depth of the river near
the bridge was, at its peak, 10 to 20 cm below the protection 5 Code and data availability
wall crest. This has been observed by local authorities and
registered in photos and videos between 04:00 and 20:00 on All data are stored in the Agueda.2016Flood dataset
12 February. These photos were used to verify the extent of (https://doi.org/10.4211/hs.937927473a3a4e66a07a2e2fdd9d
the inundation along Rua Luís de Camões and Rua Vasco 581e, Ricardo et al., 2024), which is publicly available in
da Gama. Examples of the photos collected and analysed for the HydroShare repository. HydroShare is an open-source
comparison with the numerically produced data are shown in web-based system developed, mainly, for water-related pro-
Figs. 3 and 4. fessionals to easily share, collaborate and publish all types
Figure 5 presents a 2D distribution of the maximum wa- of scientific data. HydroShare adheres to the FAIR (findable,
ter depth values on the modelled domain overlapped on the accessible, interoperable, and reusable) archival standard and
terrain model. provides a representational state transfer (REST) application

https://doi.org/10.5194/essd-16-375-2024 Earth Syst. Sci. Data, 16, 375–385, 2024

382 A. M. Ricardo et al.: Flood simulation with the RiverCure approach

6 Conclusions

This work describes the Agueda.2016Flood dataset. It

presents a unique and highly relevant dataset to fully char-
acterize the flood event that occurred in February 2016 in the
Portuguese Águeda River, a river for which there are no cu-
rated datasets of flooding events, despite their relatively high
frequency. The dataset was managed through the RiverCure
portal, a collaborative web platform connected to HiSTAV, a
validated shallow-water model featuring modelled dynamic
bed geometries and sediment transport.
The dataset includes hydrological, topo-bathymetric and
land use data together with numerically produced hydromet-
ric data from a calibrated model. Per se, none of these data
subsets would be sufficient: it is the synthesis that allows
the complete description of the flood event. Thus, the dataset
constitutes a relevant and complete validation test for other
flood forecasting models and a tool for better understanding
and mitigating floods in this river and in similar rivers.

Appendix A
Figure 6. Time series of flow discharge at the inlet boundaries
Below is illustrated the data model supported by the River-
(Ponte Redonda and Ribeiro), at the confluence of the two inlet
streams (“Upstream”) and at the downstream boundary. The up- Cure portal that shows its main concepts and properties re-
stream and downstream hydrographs were computed from the nu- lated to context definition. This specification is defined ac-
merically produced database. The time instant t = 0 h corresponds cording to the ITLingo RSL language (Rodrigues da Silva
to 04:00 of 8 February 2016. and Savić, 2021).
// Contexts
DataEntity e_Context ”Context”: Master [
program interface (API) that allows third-party applications attribute id ”Id”: Integer [constraints (PrimaryKey)]
to interact with HydroShare resources. HydroShare is a large attribute code ”Code”: String [constraints (NotNull
Python and Django application with some extra features and Unique)]
technologies added on. All data uploaded to HydroShare attribute Name ”Name”: String [constraints (NotNull
become part of a “resource” which provides the ability Unique)]
to group together multiple files of different types in one attribute organization ”Organization”: Integer [constraints
location. The dataset presented herein is the resource named (ForeignKey(e_Organization))]
Agueda.2016Flood. The resource contents are organized attribute creator ”Creator”: Integer [constraints (For-
into three folders following the schematic representation of eignKey(e_User))]
Fig. 2. Download of the files can be done by scrolling down attribute createDate ”Create Date”: Datetime
the resource’s landing page until the “Content” section. attribute isPublic ”Is public” : Boolean [defaultValue
This can be done by downloading each file individually, by ”False”]
downloading a zip file or by downloading all the contents as attribute hydrofeature ”Hydro Feature”: Integer [con-
a zipped Bagit archive. Further details are provided at https: straints(NotNull ForeignKey(e_HydroFeature))]
//help.hydroshare.org/creating-and-managing-resources/ attribute geomExternalBoundary ”Geometry External
view-and-download-a-resource/ (last access: 10 Jan- Boundary”: GeoPolyline // domain ?
uary 2024). attribute CLExternalBoundary ”CL External Boundary”:
The RiverCure portal is publicly available at http:// Decimal
rivercure.inesc-id.pt/ (last access: 20 December 2023) and attribute hasMesh ”Has mesh” : Boolean [defaultValue
can be used to run the presented case study or any other ”False”]
case upon free registration. The HiSTAV numerical model attribute taskId ”Task id”: String
can be explored through the RiverCure portal. Conde et al. attribute requester ”Requester”: Integer [constraints (For-
(2020) provide further details about the numerical model, eignKey(e_User))] ]
its discretization and its implementation within the paradigm // Context’s Geo Features
of high-performance computing (HPC) for GPU- and CPU- DataEntity e_ContextDTM ”Context DTM”: Parameter [
based computer architectures. attribute id ”Id”: Integer [constraints (PrimaryKey)]

Earth Syst. Sci. Data, 16, 375–385, 2024 https://doi.org/10.5194/essd-16-375-2024

A. M. Ricardo et al.: Flood simulation with the RiverCure approach 383

attribute context ”Context”: Integer [constraints(NotNull attribute sensor ”Sensor”: Integer [con-
ForeignKey(e_Context))] straints(ForeignKey(e_Sensor))]
attribute rasterFile ”Raster File”: FilePath attribute Description ”Description”: String [constraints
attribute raster ”Raster”: GeoRaster ] (NotNull)]
DataEntity e_ContextFrictionCoeff ”Context Friction Co- attribute associateDateTime ”Associate Date Time”: Date-
efficient”: Parameter [ time ]
attribute id ”Id”: Integer [constraints (PrimaryKey)] // Context’s Events
attribute context ”Context”: Integer [constraints(NotNull DataEntity e_ContextEvent ”Context Event”: Docu-
ForeignKey(e_Context))] ment:Regular [
attribute geometry ”Geometry”: GeoPoint [constraints attribute id ”Id”: Integer [constraints (PrimaryKey)]
(multiplicity ”*”)] // LineString attribute context ”Context”: Integer [constraints(NotNull
attribute fileName ”Geometry File Name”: FilePath ] ForeignKey(e_Context))]
DataEntity e_ContextRefinement ”Context Refinement”: attribute Name ”Name”: String [constraints (NotNull
Parameter [ Unique)]
attribute id ”Id”: Integer [constraints (PrimaryKey)] attribute type ”Type”: DataEnumeration ContextEvent-
attribute context ”Context”: Integer [constraints(NotNull TypeChoices [constraints (NotNull)]
ForeignKey(e_Context))] attribute subType ”Subtype”: DataEnumeration Contex-
attribute geometry ”Geometry”: GeoPolygon tEventSubTypeChoices [constraints (NotNull)]
attribute fileName ”Geometry File Name”: FilePath attribute State ”State”: DataEnumeration Contex-
attribute cl ”CL”: Decimal ] tEventStateChoices [constraints (NotNull)]
DataEntity e_ContextAlignment ”Context Alignment”: attribute startDateTime ”Start Datetime”: Datetime [con-
Parameter [ straints (NotNull)]
attribute id ”Id”: Integer [constraints (PrimaryKey)] attribute endDateTime ”End Datetime”: Datetime [con-
attribute context ”Context”: Integer [constraints(NotNull straints (NotNull)]
ForeignKey(e_Context))] attribute Description ”Description”: String
attribute geometry ”Geometry”: GeoPolyline [constraints attribute returnPeriod ”Return Period”: Integer [default-
(multiplicity ”*”)] // LineString Value ”1 ”]
attribute fileName ”Geometry File Name”: FilePath attribute warmUp ”Warm Up”: Boolean [defaultValue
attribute cl ”CL”: Decimal ] ”False”]
DataEntity e_ContextBoundaryLine ”Context Boundary attribute writingPeriodicity ”Writing Periodicity”: Deci-
Line”: Parameter [ mal [defaultValue ”1.0”]
attribute id ”Id”: Integer [constraints (PrimaryKey)] attribute writingPeriodicityUnit ”Writing Periodicity
attribute context ”Context”: Integer [constraints(NotNull Unit”: String
ForeignKey(e_Context))] attribute updateMaximumValue ”Update Maximum
attribute geometry ”Geometry”: GeoPolyline //GeoPoint Value”: Decimal [defaultValue ”1.0”]
[constraints (multiplicity ”*”)] //FileField attribute updateMaximumValueUnit ”Update Maximum
attribute fileName ”Geometry File Name”: FilePath Value Unit”: String
attribute type ”Type”: DataEnumeration ContextBound- attribute hasSimulation ”Has Simulation”: Boolean [de-
aryChoices [constraints (NotNull)] faultValue ”False”]
attribute dataType ”Data type”: DataEnumeration Con- attribute taskId ”Task id”: String
textBoundaryLineDataKindChoices [constraints (NotNull)] attribute requester ”Requester”: Integer [constraints (For-
] eignKey(e_User))] ]
DataEntity e_ContextBoundaryPoint ”Context Boundary DataEntity e_ContextEventResult ”Context Event Re-
Point”: Parameter [ sult”: Document:Weak [
attribute id ”Id”: Integer [constraints (PrimaryKey)] attribute id ”Id”: Integer [constraints (PrimaryKey)]
attribute contextBoundaryLine ”Context Boundary Line”: attribute contextEvent ”Context Event”: Integer [con-
Integer [constraints(ForeignKey(e_Context))] straints(ForeignKey(e_ContextEvent))]
attribute geometry ”Geometry”: GeoPoint attribute maxDepth ”Maximum Depth”: GeoRaster
attribute fileName ”Geometry File Name”: FilePath ] attribute maxLevel ”Maximum Level”: GeoRaster
// Context’s Sensors attribute maxQ ”Maximum Q”: GeoRaster
DataEntity e_ContextSensor ”Context Sensor”: Master [ attribute maxVel ”Maximum Velocity”: GeoRaster
attribute id ”Id”: Integer [constraints (PrimaryKey)] attribute time ”Time”: Datetime [constraints(NotNull)] ]
attribute boundaryPoint ”Boundary Point”: Integer [con-
straints(ForeignKey(e_ContextBoundaryPoint))]

https://doi.org/10.5194/essd-16-375-2024 Earth Syst. Sci. Data, 16, 375–385, 2024

384 A. M. Ricardo et al.: Flood simulation with the RiverCure approach

Author contributions. All the authors contributed to the arti- Conde, D.: Mathematical Modelling of Tsunami Impacts on Built
cle’s conception and design. The RiverCure portal was designed Environments, PhD thesis, Instituto Superior Técnico, Universi-
by ARdS, RMLF and JE and developed by AS, JM and IG. AMR dade de Lisboa, Lisbon, 2018.
performed material preparation and data collection. The manuscript Conde, D., Canelas, R. B., and Ferreira, R. M. L.: A
was written by AMR, ARdS and RMLF. All the authors read and unified object-oriented framework for CPU+GPU ex-
approved the final manuscript. plicit hyperbolic solvers, Adv. Eng. Softw., 148, 102802,
https://doi.org/10.1016/j.advengsoft.2020.102802, 2020.
Conde, D. A., Telhado, M. J., Viana Baptista, M. A., and Ferreira,
Competing interests. The contact author has declared that none R. M.: Severity and exposure associated with tsunami actions in
of the authors has any competing interests. urban waterfronts: the case of Lisbon, Portugal, Nat. Hazards, 79,
2125–2144, https://doi.org/10.1007/s11069-015-1951-z, 2015.
Edler, D. and Vetter, M.: The simplicity of modern audio-
Disclaimer. Publisher’s note: Copernicus Publications remains visual web cartography: An example with the open-source
neutral with regard to jurisdictional claims made in the text, pub- javascript library leaflet.js, KN J. Cartogr. Geogr. Inf.. 69, 51–
lished maps, institutional affiliations, or any other geographical rep- 62, https://doi.org/10.1007/s42489-019-00006-2, 2019.
resentation in this paper. While Copernicus Publications makes ev- Fekete, A. and Sandholz, S.: Here comes the flood, but
ery effort to include appropriate place names, the final responsibility not failure? Lessons to learn after the heavy rain and
lies with the authors. pluvial floods in Germany 2021, Water, 13, 3016,
https://doi.org/10.3390/w13213016, 2021.
Ferreira, R. M. L., Franca, M. J., Leal, J. G. A. B., and Car-
doso, A. H.: Mathematical modelling of shallow flows: Closure
Acknowledgements. This research has been supported
models drawn from grain-scale mechanics of sediment transport
by the Portuguese Foundation for Science and Technology
and flow hydrodynamics, Can. J. Civil Eng., 36, 1605–1621,
through national and European funds within the COMPETE
https://doi.org/10.1139/L09-033, 2009.
2020 and PORL-FEDER programs through projects RiverCure
Fujisada, H., Urai, M., and Iwasaki, A.: Technical Methodology for
PTDC/CTA-OHR/29360/2017 and DikesFPro PTDC/ECI-
ASTER Global Water Body Data Base, Remote Sens.-Basel, 10,
EGC/7739/2020; by 02/SAICT/2017/29360, UIDB/50021/2020,
1860, https://doi.org/10.3390/rs10121860, 2018.
and UIDB/00326/2020; by CISUC funding UIDB/00326/2020; and
Geuzaine, C. and Remacle, J. F.: Gmsh: A 3-D finite ele-
by CERIS funding UIDB/04625/2020.
ment mesh generator with built-in pre-and post-processing
facilities, Int. J. Numer. Meth. Eng., 79, 1309–1331,
https://doi.org/10.1002/nme.2579, 2009.
Financial support. This research has been supported by the Fun- Haklay, M. and Weber, P.: Openstreetmap: User-
dação para a Ciência e a Tecnologia (grant nos. PTDC/CTA- generated street maps, IEEE Pervas. Comput., 7, 12–18,
OHR/29360/2017 and PTDC/ECI-EGC/7739/2020). https://doi.org/10.1109/MPRV.2008.80, 2008.
Hall, J., Arheimer, B., Aronica, G. T., Bilibashi, A., Boháč, M.,
Bonacci, O., Borga, M., Burlando, P., Castellarin, A., Chirico,
Review statement. This paper was edited by Alexander Gelfan G. B., Claps, P., Fiala, K., Gaál, L., Gorbachova, L., Gül, A.,
and reviewed by four anonymous referees. Hannaford, J., Kiss, A., Kjeldsen, T., Kohnová, S., Koskela, J. J.,
Macdonald, N., Mavrova-Guirguinova, M., Ledvinka, O., Me-
diero, L., Merz, B., Merz, R., Molnar, P., Montanari, A., Os-
References uch, M., Parajka, J., Perdigão, R. A. P., Radevski, I., Renard,
B., Rogger, M., Salinas, J. L., Sauquet, E., Šraj, M., Szolgay, J.,
Amponsah, W., Ayral, P.-A., Boudevillain, B., Bouvier, C., Braud, Viglione, A., Volpi, E., Wilson, D., Zaimi, K., and Blöschl, G.: A
I., Brunet, P., Delrieu, G., Didon-Lescot, J.-F., Gaume, E., European Flood Database: facilitating comprehensive flood re-
Lebouc, L., Marchi, L., Marra, F., Morin, E., Nord, G., search beyond administrative boundaries, Proc. IAHS, 370, 89–
Payrastre, O., Zoccatelli, D., and Borga, M.: Integrated high- 95, https://doi.org/10.5194/piahs-370-89-2015, 2015.
resolution dataset of high-intensity European and Mediter- Penning-Rowsell, E. C., Floyd, P., Ramsbottom, D., and
ranean flash floods, Earth Syst. Sci. Data, 10, 1783–1794, Surendran, S.: Estimating Injury and Loss of Life in
https://doi.org/10.5194/essd-10-1783-2018, 2018. Floods: A Deterministic Framework, Nat. Hazards, 36, 43–64,
Blöschl, G., Hall, J., Viglione, A., Perdigão, R. A., Para- https://doi.org/10.1007/s11069-004-4538-7, 2005.
jka, J., Merz, B., Lun, D., Arheimer, B., Aronica, G. T., Pereira, A.: On the use of a weather radar for flood forecast-
Bilibashi, A., and Boháč, M.: Changing climate both increases ing in the Centre of Portugal, Master thesis, Instituto Superior
and decreases European river floods, Nature, 573, 108–111, Técnico, Lisbon, https://fenix.tecnico.ulisboa.pt/cursos/meambi/
https://doi.org/10.1038/s41586-019-1495-6, 2019. dissertacao/846778572212913 (last access: 16 September 2023),
Canelas, R., Murillo, J., and Ferreira, R. M. L.: Two- 2021.
Dimensional Depth-Averaged Modelling Of Dambreak Peterson, G.: GIS Cartography: A Guide To Effective Map Design,
Flows Over Mobile Beds, J. Hydraul. Res., 51, 392–407, 2nd edn., CRC Press, Boca Raton, ISBN 978-1-4200-8213-5,
https://doi.org/10.1080/00221686.2013.798891, 2013. 2014.
Chow, V. T.: Open-Channel Hydraulics, McGraw-Hill, New York,
ISBN 007085906X, 1959.

Earth Syst. Sci. Data, 16, 375–385, 2024 https://doi.org/10.5194/essd-16-375-2024

A. M. Ricardo et al.: Flood simulation with the RiverCure approach 385

Ricardo, A. M., Silva, A., Estima, J., Ferreira, R. M., Marques, van der Sande, C. J., De Jong, S. M., and De Roo, A. P. J.:
J., Gamito, I., and Serra, A.: Águeda.2016Flood – The open A segmentation and classification approach of IKONOS-2 im-
dataset of the 2016 Águeda flood event, HydroShare [data set], agery for land cover mapping to assist flood risk and flood
https://doi.org/10.4211/hs.937927473a3a4e66a07a2e2fdd9d581e, damage assessment, Int. J. Appl. Earth Obs., 4, 217–229,
2024. https://doi.org/10.1016/S0303-2434(03)00003-5, 2003.
Rigon, R., Formetta, G., Bancheri, M., Tubini, N., D’Amato, C., Wallemacq, P., Below, R., and McLean, D.: UNISDR and CRED
David, O., and Massari, C.: HESS Opinions: Participatory Dig- report: Economic Losses, Poverty & Disasters (1998–2017), Re-
ital eARth Twin Hydrology systems (DARTHs) for everyone – port of Centre for Research on the Epidemiology of Disasters
a blueprint for hydrologists, Hydrol. Earth Syst. Sci., 26, 4773– (CRED) and United Nations Office for Disaster Risk Reduction
4800, https://doi.org/10.5194/hess-26-4773-2022, 2022. (UNISDR), CRED, Belgium, 2018.
Rodrigues da Silva, A. and Savić, D.: Linguistic Patterns and Lin- Yamaguchi, Y., Kahle, A. B., Tsu, H., Kawakami, T., and Pniel,
guistic Styles for Requirements Specification: Focus on Data En- M.: Overview of Advanced Spaceborne Thermal Emission and
tities. Appl. Sci. 11, 4119, https://doi.org/10.3390/app11094119, Reflection Radiometer (ASTER), IEEE T. Geosci. Remote, 36,
2021. 1062–1071, https://doi.org/10.1109/36.700991, 1998.
Rodrigues da Silva, A., Estima, J., Marques, J., Gamito, I., Serra, A., Yuan, F., Fan, C., Farahmand, H., Coleman, N., Esmalian, A., Lee,
Moura, L., Ricardo, A. M., Mendes, L., and Ferreira, R. M. L.: C. C., Patrascu, F. I., Zhang, C., Dong, S., and Mostafavi, A.:
A Web GIS Platform to Modeling, Simulate and Analyze Flood Smart flood resilience: Harnessing community-scale big data
Events: The RiverCure Portal, ISPRS Int. J. Geo-Inf., 12, 268, for predictive flood risk monitoring, rapid impact assessment,
https://doi.org/10.3390/ijgi12070268, 2023. and situational awareness, Environ. Res. Infrastruct. Sustain., 2,
Soares-Frazão, S., Canelas, R., Cao, Z., Cea, L., Chaudhry, H. 025006, https://doi.org/10.1088/2634-4505/ac7251, 2022.
M., Die Moran, A., El Kadi, K., Ferreira, R., Cadórniga,
I. F., Gonzalez-Ramirez, N., Greco, M., Huang, W., Imran,
J., Le Coz, J., Marsooli, R., Paquier, A., Pender, G., Pon-
tillo, M., Puertas, J., Spinewine, B., Swartenbroekx, C., Tsub-
aki, R., Villaret, C., Wu, W., Yue, Z., and Zech, Y.: Dam-
break flows over mobile beds: experiments and benchmark
tests for numerical models, J. Hydraul. Res., 50, 364–375,
https://doi.org/10.1080/00221686.2012.689682, 2012.

https://doi.org/10.5194/essd-16-375-2024 Earth Syst. Sci. Data, 16, 375–385, 2024