FloodVis: Visualization of Climate Ensemble Flood Projections in Virtual Reality

Abstract

Anthropogenic greenhouse gas emissions are leading to accelerating climate change, forcing politicians and administrations to take actions to mitigate climate change and adapt to its impacts, such as changes in flood regimes. For European countries, an increasing frequency and severity of extreme rainfall and flood events is expected. However, studies on future flood risks caused by climate change are associated with various uncertainties. The risk simulations are elaborate as they consider (i) climate data ensembles (temperature, precipitation), (ii) hydrological modeling (flood generation), (iii) hydrodynamic modeling (flood conveyance), and (iv) vulnerability modeling (damage assessment) involving a huge amount of data and their handling with Big Data methods. The results are difficult to understand for decision makers. Therefore, FloodVis offers a means of visualizing possible future flood risks in Virtual Reality (VR). The presentation of the results in a VR especially supports the user in understanding the complexity of the dynamics of the risk system enabling the feeling of presence. In FloodVis the user enters into a virtual surrounding to interact with the data, examine the temporal evolution, and compare alternative development pathways. Critical structures that require improved protection can be identified. The user can follow the inundation process in hourly resolution. We evaluated FloodVis through an online and offline user study on the context of whether VR can provide a better visualization of ensemble flood risk data and whether the sense of presence in VR can influence the decision making and help to raise awareness.

Full text

Workshop on Visualisation in Environmental Sciences (EnvirVis) (2022)
S. Dutta and K. Feige and K. Rink and D. Zeckzer (Editors)
FloodVis: Visualization of Climate Ensemble Flood Projections in
Virtual Reality
M. T. Oyshi1,2, V. Maleska2,3,4, J. Schanze2,3,4, F. Bormann1, R. Dachselt2,5,6,7and S.Gumhold 1,2,6,7
1Chair for Computer Graphics and Visualization, TU Dresden, Germany
2ScaDS.AI, Germany
3Leibniz Institute of Ecological Urban and Regional Development (IOER)
4Chair of Environmental Development and Risk Management, TU Dresden, Germany
5Interactive Media Lab Dresden, Cluster of Excellence Physics of Life, TU Dresden, Germany
6Centre for Tactile Internet with Human-in-the-Loop (CeTI), TU Dresden, Germany
7Cluster of Excellence Physics of Life, TU Dresden (PoL), Germany
Abstract
Anthropogenic greenhouse gas emissions are leading to accelerating climate change, forcing politicians and administrations to
take actions to mitigate climate change and adapt to its impacts, such as changes in flood regimes. For European countries, an
increasing frequency and severity of extreme rainfall and flood events is expected. However, studies on future flood risks caused
by climate change are associated with various uncertainties. The risk simulations are elaborate as they consider (i) climate
data ensembles (temperature, precipitation), (ii) hydrological modeling (flood generation), (iii) hydrodynamic modeling (flood
conveyance), and (iv) vulnerability modeling (damage assessment) involving a huge amount of data and their handling with
Big Data methods. The results are difficult to understand for decision makers. Therefore, FloodVis offers a means of visualizing
possible future flood risks in Virtual Reality (VR). The presentation of the results in a VR especially supports the user in
understanding the complexity of the dynamics of the risk system enabling the feeling of presence. In FloodVis the user enters
into a virtual surrounding to interact with the data, examine the temporal evolution, and compare alternative development
pathways. Critical structures that require improved protection can be identified. The user can follow the inundation process in
hourly resolution. We evaluated FloodVis through an online and offline user study on the context of whether VR can provide a
better visualization of ensemble flood risk data and whether the sense of presence in VR can influence the decision making and
help to raise awareness.
Categories and Subject Descriptors (according to ACM CCS): I.3.3 [Computer Graphics]: Visualization—Virtual Reality, Immer-
sive visualization, natural interaction, sense of presence I.3.3 [Environment]: Climate change—Flood risks
1. Introduction
Climate change causes a shift in worldwide weather conditions in a
potentially hazardous and disruptive manner. Anthropogenic green-
house gas emissions and global warming are playing a vital role in
this shift. With the rising temperatures, extreme weather events will
become more serious. Flood events occur more frequently and in-
tensely in the future resulting in multi-facetting and long-lasting
impacts [IPC21]. The consequences of flooding are particularly
disruptive if urban areas are inundated resulting in infrastructural
damage, damage to houses including loss of human life. According
to the report on financial management of flood risks published by
OECD in 2016, flood damages exceed $40 billion worldwide annu-
ally [OEC16]. In July 2021, Germany experienced the worst flash
flood event for decades after Western Germany had been affected
by heavy rainfall. At the same time, other European countries, es-
pecially Austria, Switzerland, France, Luxembourg, Belgium, the
Netherlands, Czech Republic, Croatia, Italy, Romania, and United
Kingdom, suffered numerous flood events with an overall damage
of $43 billion and more than 190 deaths [DW21], [EGU21]. These
major impacts occurred despite the availability of forecasting and
warning of extreme rainfall and flooding [Cor21].
Researchers are working on the projections of possible future
flood risks under climate change conditions to derive adaptive risk
reduction measures. It is assumed that increasing temperature will
change future precipitation patterns where additional atmospheric
moisture can be held by warmer air with rising potential for heavy
rainfall and flood events. Flood damage might cost around C48 bil-
lion instead of current C7.8 billion per year on the European conti-
nent itself [Cor21]. According to Huber et al. [HG11], it might be
possible to limit the increasing probability and severity of extreme
© 2022 The Author(s)
Eurographics Proceedings © 2022 The Eurographics Association.

Marzan Tasnim Oyshi / FloodVis: Visualization of Climate Ensemble Flood Projections in Virtual Reality
weather events by stopping anthropogenic climate change as it is
playing a fundamental role for extreme events. Besides mitigating
climate change common measures to reduce flood risks exist. The
basis for the derivation of those measures are risk assessments. The
spectrum of measures ranges from flood prevention due to dams
and dikes to behavioural training of rescue teams and the people at
risk.
Flood simulations are crucial for risk reduction. Simulations of
alternative flood scenarios support a better understanding of risks
with their probabilities of occurrence and the impacts they cause.
However, even if there are valuable flood simulations available,
they are not always used to their full potential since resulting risk
curves with their uncertainty bands are difficult to understand for
decision-makers and the people affected.
This paper presents the novel method FloodVis for the visualiza-
tion of ensemble flood events and their impacts in Virtual Reality
(VR) to increase the awareness for future change by making flood
simulation data accessible in an immersive and playful way. VR
is deployed here as it enhances user experience and enables im-
mersive visualization. The ability of creating a sense of presence
is considered as the main factor of VR where individuals can have
a feeling of being somewhere else even though physically they are
not there [Jer15], [BBL∗05], [SVS05]. FloodVis allows users to un-
derstand flood data without prior knowledge. It was evaluated by an
online and offline user study.
2. Related Work
Scientists started focusing on ensemble flood forecasting over the
past decade due to the expansion of numerical weather prediction
and climate projection and growth of high performance computing
for a risk-based decision making advancement [WED∗20]. Instead
of a single forecast, ensemble forecasting and projection generate
a bandwidth of data applying varying initial conditions, parame-
terizations as well as model approaches. The visualization of flood
simulation data has been a topic of research for years to under-
stand the flood risks in a better and easier way. To reduce climate
change impacts such as more frequent and intense floods, adapta-
tion occurred to be a key management strategy of decision makers.
Hereby, uncertainties play a major role for both climate projections
and impact assessment. [JRP01].
At the science-policy interface between the experts of forecast-
ings and projections and decision makers, it has become obvi-
ous that decision making does not just follow rational choices
but rather additionally involves emotions. The influence of feel-
ings has been researched in the field of virtual hospital experience
[GCL∗21], stress reduction [GRB21], consumer decision making
[ASAD16], stock investments decisions [GNG16], ethical decision
making [SSNO22] and many more. Seo et al. [SFB07] present an
investigation on whether subjective experience of emotions is func-
tional or dysfunctional. They studied 101 stock investors’ decisions
for 20 consecutive business days and the result shows individuals
experiencing more intense feelings accomplish stronger decision
making.
There is an ongoing debate on whether VR is important for en-
vironmental data visualization or decision making, where studies
show that VR allows the sense of presence and guides to a positive
attitude towards the surrounding environment [TJD18] [CWT∗11].
VR amplifies the human intelligence and establishes the human-
machine simulation system which can support decision making in,
e.g. emergency management [BWW95]. Berg et al. [BV17] present
a case study that investigates VR as a decision making tool for early
design making. For this research, a group of design and manufac-
turing engineers was invited to test an immersive environment for
a new product development project. According to the results, indi-
viduals could identify design flaws and potential solutions which
were not identified using non-immersive tools.
VR allows users to view a 3D dataset in a 3D/360 environment
instead of a 2D display by connecting dots between data and real-
ity [SS18]. Mol et al. [MBB22] examined flood simulation in im-
mersive reality to check whether simulated disasters can influence
people for investing in risk reducing measures. According to them,
understanding and preparation for low-probability/high-impact risk
is often difficult for individuals without experiencing respective
events in person. The lack of understanding and preparation usually
leads to additional damages. According to their study, individuals,
participating in virtual flood visualizations are more likely to invest
significantly in flood risk reduction.
Simpson et al. [SPKK22] evaluated storm surge flooding in an
immersive environment for risk analysis. According to them, it is
challenging to communicate storm surge flood risks using tradi-
tional methods like maps. At the same time, virtual flooding visu-
alizations can potentially reduce physical harm as individuals can
explore external aid and own behaviour in immersive virtual reality
(iVR). Understanding the various impacts of different heights also
improves in iVR. The high potential of immersive visualization for
user experience is discussed by Kraus et al. [KKF∗21].
Toshio et al. [TF20] tested public interventions in VR for flash
flood to encourage evacuation decision as people blow hot and cold
while leaving their home due to natural disasters. In their research
they examined whether the social and environmental change in VR
caused by flash flood can promote early evacuation decision. 103
students participated in this study from Kumamoto University in
Japan and at the end they could effectively take evacuation deci-
sions earlier. Enes et al. [EY19] present a web based framework
for estimating flood loss using HAZUS. Regions that require assis-
tance to support resource allocation and mitigate planning can be
prioritized using their framework. However, this solution does not
ensure the sense of presence also they have limitations of the tech-
nical requirements, number of inundation scenarios as well as data
options.
Several traditional flood risk visualization approaches exist in-
cluding 2D maps, animations and other methods but none of them
can ensure the sense of presence. Grottel et al. [GSH∗15] presents
a real-time flood visualization in 3D targeting non-professionals.
The paper discusses the implementation details, data concept and
results. But there are no details about the interaction of the users
and their feedbacks. The paper does not include VR and the discus-
sion of the sense of presence.
The goal of FloodVis is to visualize ensemble flood data in im-
mersive reality to enhance the user’s experience referring to the dif-
ferences between alternative scenarios including uncertainty, iden-
© 2022 The Author(s)
Eurographics Proceedings © 2022 The Eurographics Association.

Marzan Tasnim Oyshi / FloodVis: Visualization of Climate Ensemble Flood Projections in Virtual Reality
Module
“Hydrologic
Modelling”
Module
“Climate Data
Ensembles”
Module
“Hydrodynamic
Modelling”
Module
“Damage
Modelling “
CDE HdM HdrM DM
Figure 1: Representation of the applied model chain comprising
the modules (i) climate data ensemble, (ii) hydrologic modeling,
(iii) hydrodynamic modeling, and (iv) damage modeling [Mal21]
tification of houses under water and to see the flood propaga-
tion over time. The principal aim of this paper is to determine
whether VR supports flood projection data visualization and to
know whether the sense of presence influences the decision making
regarding the consideration of flood risk reduction measures.
3. Applied Data
The data used in FloodVis were generated in a project that investi-
gates the dynamics and uncertainties of flood risks in the context of
climate change [Mal21]. An ensemble-based model chain includ-
ing the modules (i) climate data ensemble, (ii) hydrologic model-
ing, (iii) hydrodynamic modeling, and (iv) damage modeling was
set up for this purpose. Figure 1shows the model chain which is
applied for the Mulde River catchment with a focus on inundation
areas at the community of Bennewitz (a municipality in Saxony,
Germany).
The climate data ensemble includes results from global and re-
gional climate modeling of the variables temperature, precipitation,
air pressure, humidity, wind, and global radiation. The simulation
results visualized in FloodVis are based on the global climate mod-
els CCCma3 and EC-Earth in combination with COSMO-CLM
(KIT, Institute of Meteorology and Climate Research). The data
set "CCCma3" employs the climate scenario "A1B" where "EC-
EARTH" is based on the scenario "RCP8.5". Scenario A1B rep-
resents rapid economic growth and scenario RCP8.5 represents a
high-emissions scenario showing the likely outcome if society does
not foster the mitigation of greenhouse gas emissions.
The hydrologic model applies climate data as input and provides
the discharge generated at a specific location in the river. The sim-
ulation data comprise 25 flood events each representing one hy-
drologic model parametrisation. Parameters relevant for hydrologic
modeling are for example properties of the soil as saturated wa-
ter conductivity to calculate how much water is absorbed by the
ground, gets drained, or does evaporate. The resulting flood dis-
charge is used as input for the hydrodynamic modeling input.
The hydrodynamic modeling determines the flow characteristics
such as inundation area, water depth, flow velocity and also the
flood propagation. 2D models are applied based on a digital ele-
vation model (DEM) of the terrain. The DEM includes buildings
due to their impact on the flow of the water as flow obstacles. The
output of this module is an elevation value of the water surface for
every cell of the simulation grid. If there is no water in one grid cell
a no-data value is returned.
In the damage modeling module, the types and footprints of the
buildings in Bennewitz are used in combination with the water
height around the buildings to estimate the inventory and construc-
tion damage at every house.
The spatial and temporal resolution of the simulation, 4 m and
1 h, is extremely high in comparison to other flood simulations. The
diversity in climate models, scenarios and the use of different hy-
drologic model parametrizations allows to get a better understand-
ing for the uncertainties of the simulation results. If there are, for
instance, many similar flood events across all the different simu-
lation runs then this suggests that the uncertainties involved in the
data are relatively low. On the other hand, if there are major vari-
ations in the results of each model run, then this indicates that the
uncertainties in the simulation results are much higher [Mal21].
The data set used in this research includes a digital elevation
model (DEM) with a resolution of about 1 pixel/m2, aerial images
with a resolution about 1 pixel /0.2m2taken between 20.05.2018 -
01.08.2020, data describing buildings with their footprint, height,
and roof type sourced as CityGML (LOD1) files from the Free
State of Saxony (Staatsbetrieb Geobasisinformation und Vermes-
sung Sachsen, GeoSN). All of the three geospatial data sets are
published as tiles with dimensions of 2x2 km. The input data was
given as ascii files in the CRS "EPSG:32633". The simulation data
resolution is 1 pixel /4m2due to hydrodynamic model applica-
tion. To visualize the development of flood events over time, time-
series data was provided for the events CCCma3_A1b_1_2 and
EC −E ART H_rcp85_1_4 with a time interval of 1 hour.
4. Traditional Visualization Techniques in Risk Assessment
Flood hazards are usually visualised by color-graded water depth
maps and road maps or aerial photos in the background. The map
visualizes the water depth with a blue color gradient. The dark color
indicates a high water depth, while a light blue represents a lower
water depth. In figure 2, the water depth of a simulated extreme
flood event is shown. The background employs aerial pictures as
provided by Google or Open Street Map. The 2D visualization ap-
proach is sufficient for experts in the field of flood hazard analy-
sis as they are used to analyze flood events by interpreting colored
maps.
However, this traditional method cannot ensure the sense of pres-
ence and according to published research and also the user study
performed in the context of FloodVis which is represented in the
results and evaluation section. It suggests that, in some cases non-
experts better understand the impact of extreme flood events when
they can experience them in a 3D environment. Another shortcom-
ing of 2D maps that is resolved by FloodVis is that only one time
step is displayed on each map, usually the maximum water depth,
and the temporal evolution of the floodplain is neglected. This can
be relevant for identifying critical locations during the flood event.
© 2022 The Author(s)
Eurographics Proceedings © 2022 The Eurographics Association.

Marzan Tasnim Oyshi / FloodVis: Visualization of Climate Ensemble Flood Projections in Virtual Reality
max water depth CCCma3 A1b
> 0,00 m
- 0,50 m
- 1,00 m
- 1,50 m
> 1,50 m
Figure 2: Color-graded 2D hazard map for water depth; dark blue
indicating a relatively high water surface elevation, [Mal21]
Figure 3: Data preparation and conversion workflow for creating
the 3D model of Bennewitz in Unity 3D
5. Methods
FloodVis aims to provide users with a digital twin for flood risk
assessments tested for the municipality of Bennewitz in 3D and
allows for interaction to have a VR experience. The game engine
Unity is used as it is known to be one of the most beginner-friendly
development platforms. Also the usage of existing resources from
the asset store can lead to time savings in the development process.
5.1. Data Preparation and Conversion
To create a 3D model from the supplied data they had to be con-
verted into 3D objects that can be imported in the Unity develop-
ment platform. The data preparation workflow is summarized in
figure 3.
The DEM of the study area is supplied in nine tiles covering the
entire study area. Accordingly, the nine tiles in the XYZ file format
are merged using QGIS and cropped to the extent of the study area.
The resulting DEM is saved into a new ASC file. Afterwards, the
Figure 4: Artifacts along shoreline before filling of missing values
(left) and after filling (right)
DEM is rendered into a grayscale TIF file. The last step is to scale
the resulting TIF file to a square format using the software Adobe
Photoshop and to export it as 16-bit grayscale RAW (uncompressed
and unprocessed image data) file because Unity only accepts this
format as heightmaps for terrains. The tiles of the aerial images
are merged into a single file and cropped to the study area’s extent
as already explained for the DEM. The resulting TIF file is down-
sampled using Adobe Photoshop to reduce the file size of 10 GB.
To create meshes from the simulation results the Geographic In-
formation System (GIS) tool QGIS is used to fill the cells that do
not contain water height values with the values of the cells next to
it. This is necessary to avoid artifacts along the edge of the water
surface in the visualization. Figure 4shows an image of the arti-
facts that are caused by the lower resolution of the simulation data
(4 m). The new copies of the original simulation data are trans-
formed into the CRS "EPSG:25833" to ensure that they properly fit
together with all other data sourced from the federal state‘s web-
site. After the filling and transformation process, meshes are cre-
ated using the addon Blender GIS for Blender. All the meshes are
exported into one FBX file (a format to exchange 3D geometry and
animation data). Before exporting, 20% of the original mesh face
count is reduced using the decimate feature of Blender to reduce
the complexity. For time-series data the meshes of one time-series
are split into multiple FBX files to reduce the file size of each FBX
file as Blender tends to run out of memory when too many meshes
are edited and exported at once.
The information about the buildings in the study area is acces-
sible as CityGML files. This data format is an open data model
and based on the XML (an extensible markup language file) for-
mat. It allows the storage and exchange of virtual 3D city models.
CityGML files contain information like the footprints of buildings,
along with the buildings’ height and their roof type. The CityGML
data can be visualized using a 3D view in the QGIS software as
shown in Figure 5. To convert the building data into 3D meshes
the Blender add-on Up3date is used, which only allows importing
CityJSON files - a JSON-based exchange format for the CityGML
data model [LAOK∗19] that was developed to increase the models’
usability for developers, compared to the XML-based version. To
convert the given CityGML files to CityJSON files the open-source
citygml-tools are used.
5.2. Assembling of the model in Unity
After the data preparation and conversion, all data are imported
into Unity 3d. The XR Interaction Toolkit is used to develop the
© 2022 The Author(s)
Eurographics Proceedings © 2022 The Eurographics Association.

Marzan Tasnim Oyshi / FloodVis: Visualization of Climate Ensemble Flood Projections in Virtual Reality
Figure 5: CityGML data visualized in the 3D view of QGIS
3D model of Bennewitz, a terrain layer, water surface, teleporta-
tion areas, building and skybox. The terrain layer is generated us-
ing the previously created heightmap in form of a 16-bit grayscale
RAW file. The aerial image is applied as a texture. For the proper
dimensions of the terrain layer the width, depth, and height are set
to match the values shown in QGIS using the Unity editor. As for
terrain height, the difference between the highest and lowest point
of the elevation model is used.
For the water surface, a plane is chosen. The scale of the plane
is set to the extent of the simulation data. By switching the meshes
of the plane, all different flood events can be visualized. To make
the water surface appear like water the basic water shader from the
Unity Standard Assets Version 1.1.610 is applied to the plane.
To allow teleportation on the terrain and the water surface, the
teleportation area component from the SteamVR plugin is applied
to each of those game objects. The buildings are imported from the
previously created FBX file. They are then placed on the terrain us-
ing a script. Some adjustments have to be made by hand to properly
align the buildings with the terrain. To add a sky box that fits to the
dramatic flood events shown in the flood visualization, a skybox
showing a cloudy gray sky from the asset AllSky Skybox Set on the
unity asset store is applied to enhance immersion and visual appeal.
5.3. Navigation and Interaction
The navigation and interaction of FloodVis includes selection, con-
troller input using laser-pointer and teleportation. To interact with
the 3D view of Bennewitz in VR, users need to wear the HTC VIVE
headset and need controllers for the interaction. The user can sit
comfortably or stand while using FloodVis. For exploring around
users can move their head or use the snap turn feature by press-
ing the east or west button of the controller. The snap turn feature
allows users to turn left and right with an 45° angel without mov-
ing head/body. For navigation, the metaphor of teleportation was
used to avoid motion sickness as much as possible. Users can use
teleportation by pressing the north, south or center button of the
right controller and choose a potential teleportation point using a
ray of the virtual laser pointer coming out of the right controller.
The green color of the ray indicates teleportation possibility and
red color indicates teleportation is not possible. Users can increase
or decrease elevation by choosing the north and south buttons on
the trackpad of the left controller. This action instantly changes the
camera position by 10 m along the vertical axis. Users can pop up
Figure 6: Navigation concept mapping the actions to the buttons
of the HTC VIVE controllers
the menu by selecting the menu button of the right controller and
choose anything from the menu by pulling the trigger of the right
controller. Turning on the playback mode that cycles through all the
flood events is possible by choosing the grip button. The interaction
mapping is shown in figure 6.
5.4. System Requirements
As already mentioned, FloodVis is developed using Unity but this
is not required to run the program. Users need to have a HTC
VIVE base station, headset and controllers for the interaction and
SteamVR installed. A minimum OS - Windows 7 SP1, Windows
8.1 or later, Windows 10; processor - Intel Core i5-4590/AMD FX
8350 equivalent or better and memory of at least 4 GB RAM is re-
quired. For graphics - NVIDIA GeForce GTX 1060 AMD Radeon
RX 480 equivalent is recommended.
6. Results and Evaluation
6.1. Results
Figure 7represents the non-flooded 3D model of Bennewitz from
the birds’ eye view (left) and the flooded ground from ground view
(right). These views can be achieved by teleporting to the positions
or varying the elevation.
Figure 8illustrates the main menu of FloodVis. Choosing Data
Set Selection brings up a new menu dialog where users can choose
between four different datasets including simulation data for 25 hy-
drologic parametrizations and time-series data with an interval of
1 hour. From the main menu, users can adjust the interval length
of every flood event and visualize how each scenario is inundat-
ing Bennewitz by enabling the playback mode. It is possible to
check the controller mapping by selecting help through the [?]-
button from the top-left corner of the main menu. Users can de-
cide whether they want to teleport to the ground or if they want
to maintain their current elevation during teleportation by check-
ing the teleport to ground option. Additionally, the user can switch
off the visibility of the water level diagram showing the inundation
depth of all scenarios of a specific location indicating uncertainty
(figure 9) within the selected dataset. Choosing a point of interest
and pulling the trigger button on the right controller, updates the
diagram to visualize all water levels at this point. Saving and load-
ing allows the opportunity to save specific inundation depths of the
point of interest. Quit Visualization closes the entire application.
Figure 9also shows two different views from above the study
area. In the top most image, users placed labels in the virtual world
© 2022 The Author(s)
Eurographics Proceedings © 2022 The Eurographics Association.

Marzan Tasnim Oyshi / FloodVis: Visualization of Climate Ensemble Flood Projections in Virtual Reality
Figure 7: 3D model of Bennewitz from different perspectives; (top)
view without flood, (bottom) submerged ground view
Figure 8: Main menu of the FloodVis User Interface; the selection
tool using laser ray pointer from right controller supports to choose
anything from the menu
to measure elevations and water levels. All labels are updated when
the flood event is switched by the left and right buttons on the left
controller or by enabling the playback mode using the grip button
on the left controller. To make sure that the user can read the values
shown in the labels, they always face towards the user and scale
depending on the distance to the user. User can have an idea about
the current scenario using the label attached to the left controller. It
is also possible to identify fully or partially flooded houses easily.
6.2. Evaluation
In general people are under-prepared for floods [MK18] due to less
frequent flood experiences. But the sense of presence can influence
Figure 9: A scenario is attached to the left controller. The diagram
shows water depths of 25 flood events with low uncertainty for dif-
ferent data sets (please check applied data for more details about
different data set). Labels indicate elevation and water level (top).
the preparation for rare frequent floods. To evaluate FloodVis, we
performed an extensive user study with 28 participants both online
(15) and offline (13) by (i) comparing the traditional flood maps,
animations, FloodVis as a desktop application and (ii) FloodVis in-
cluding VR experience.
Along with traditional flood maps and animations, FloodVis as
desktop app without VR interaction was presented to the partici-
pants to determine whether the desktop app can provide a sense
of presence and influence the decision making. Participants were
asked to identify critical locations and submerged houses. At the
end all offline participants were surveyed concerning the sense of
presence and whether this sense of presence can influence decision
making (directly - verbal, indirectly - questions that indicate influ-
ence). Additionally, this solution was presented in the inauguration
of a ScaDS.AI’s "Living Lab" as a demonstrator where politicians
and decision makers were present and the feedback was extremely
positive.
To evaluate the sense of presence in the survey, our question-
naire shown in table 1was inspired from the WS Presence Ques-
tionnaire [WS94]. For system evaluation, we used the System Us-
ability Scale (SUS) [JTWM96] combining positive and negative
questions where the user needs to think more before answering.
To measure whether users felt simulation sickness, we applied the
Simulation Sickness Questionnaires [KLBL93]. The evaluation of
© 2022 The Author(s)
Eurographics Proceedings © 2022 The Eurographics Association.

Marzan Tasnim Oyshi / FloodVis: Visualization of Climate Ensemble Flood Projections in Virtual Reality
FloodVis resulted from concrete project based questions. A choice
of the project focused questions are shown in table 2and simula-
tion sickness questions depicts table 3. In total there are 49 ques-
tions including 24 project focused questions, 10 system usability
questions, 14 simulation sickness questions and 11 sense of pres-
ence questions. Except the questions, users were also open to give
feedback or highlight if they liked/disliked something significantly.
All the questions used for the user study as well as the results
can be found in the supplementary materials as Q1 (general, com-
parison with non-VR, FloodVis, system usability and simulation
sickness) and Q2 (sense of presence).
As introduction of the user study, users were given an overview
of the data, VR experience and risks of motion sickness were also
described to the participants as 90% of the participants did not have
prior experience with VR. Most of the participants have a civil
engineering background with at least a Bachelor’s degree. Some
participants come from directions of BigData, Renewable Electric-
ity Production, Spatial Analysis, Geoinformation Systems and Re-
mote Sensing, Political Science, and Scientific Visualization. Par-
ticipants were from the age group of 25-42 years, most of them
were students from Technische Universität Dresden, researchers
from ScaDS.AI and countries from Bangladesh, Ethiopia, France,
Germany, India, Italy, Mozambique and Pakistan.
Due to the restrictions of the Covid pandemic regarding personal
meetings, an online user study was implemented for the partici-
pants who really wanted to experience the system but could not at-
tend personally. Although it is difficult to ensure the sense of pres-
ence online, videos of FloodVis with and without VR interaction
were used to provide a general understanding of FloodVis.
15 users participated in the online user study and most of them ex-
pressed their interest to test the system in person and almost all of
them agreed to have a better visualization in VR comparing to non-
VR mode shown in figure 10. One participant actually participated
the user study in person (BigData researcher) later and agreed that
the sense of presence influences the decision making in a positive
manner.
Another survey was performed to understand the sense of pres-
ence including 13 participants in a scale from 1 to 7 where the
scaling is different for each question, in general with 1=not at all
and 7=very important except Q9. The average results are shown
in table 1. Users were asked whether this immersive inspection is
important for the people living in flood prone areas and majority of
this group highly agreed that it is important (Q1[sense of presence]:
Mean=6.85, SD=0.38).
For most of the users, being able to identify submerged houses was
interesting and according to them if the residents of Bennewitz as
well as policy makers could experience this application and iden-
tify their own submerged houses, they would be much more con-
cerned about the flood risks. None of the participants complained
to have long delays between their actions and expected outcomes
(Q9[sense of presence]: Mean=2.08, SD=0.64) scaled as 1 (no de-
lays) to 7 (very long delays).
To understand the usability and in-general idea about Flood-
Vis, another user study was performed with 10 participants in a
scale from 1 (strongly disagree) to 5 (strongly agree). In general,
FloodVis received a positive feedback. Most of the users liked the
Table 1: Questions and users’ input in a scale of 1 to 7 to under-
stand the sense of presence
Mean SD
Please state your opinion on how important the
immersive inspection is for people owning a building
in a flood affected area?
6.8 0.4
Please state your opinion on how important the
immersive inspection is for people that want to buy
or build a building in a flood affected area?
6.9 0.3
How much did the visual aspects of the environment
involve you? 6.1 0.8
How involved were you in the virtual environment
experience? 6.4 0.8
How well could you examine objects (houses/flood
events) from multiple viewpoints? 6.2 0.7
How closely were you able to examine objects
(houses/flood events)? 6.5 0.5
How well could you place lables (elevation & water
depth) in the virtual environment? 6.5 0.7
How helpful were the labels while playback (both in
ground view and bird eyes view)? 6.5 0.5
How much delay did you experience between your
actions and expected outcomes? 2.1 0.6
How natural did your interactions with the
environment seem? 5.5 0.5
How proficient in moving and interacting with the
virtual environment did you feel at the end of the
experience?
6.5 0.5
overall experience (Q1: Mean=4.6; SD=0.70), and most of the par-
ticipants agreed or strongly agreed to have navigation efficiency
for teleportation (Q9: Mean=4.50; SD=0.53). Almost all of the
users agreed or strongly agreed that presenting complex data in
VR is possible (Q18: Mean=4.8; SD=0.42), whereas the major-
ity of them agreed or strongly agreed that VR supports informa-
tion presentation for policy and public (Q21: Mean=4.7; SD=0.48)
and almost all of them agreed or strongly agreed to have a smooth
navigation in VR, with a significance of the elevation labels (Q7:
Mean=4.3; SD=0.48, Q23: Mean=4.6; SD=0.70). The average SUS
score achieved 79.0 with a rank B, meaning a good system with
room for improvement.
An overview of the user study results are shown in figure 11. The
same group was questioned regarding simulation sickness in a scale
of 1 (not at all) to 5 (very strong) and none of them complained to
have any mild to major simulation sickness like general discom-
fort, fatigue, headache, difficulty concentrating (Q35; Mean=1.2;
SD=0.4, Q36: Mean=1.0; SD=0.0, Q37: Mean=1.2; SD=0.4, Q43:
Mean=1.0; SD=0.0) except one who stated to have difficulty on fo-
cusing (Q39: Mean=1.6; SD=1.1).
The applied non-VR materials for this user study, e.g. flood risk
maps and simulations, as well as the video demonstration of Flood-
Vis can be found in supplementary material.
© 2022 The Author(s)
Eurographics Proceedings © 2022 The Eurographics Association.

Marzan Tasnim Oyshi / FloodVis: Visualization of Climate Ensemble Flood Projections in Virtual Reality
Table 2: Sample of project-based questions and users’ input in a
scale of 1 (strongly disagree) to 5 (strongly agree)
Mean SD
I like the overall experience 4.6 0.7
Possibility of smooth navigation inside the VR scene
of Bennewitz 4.3 0.5
The teleportation feature allowed efficient navigation
in VR 4.5 0.5
Presenting complex data in a Virtual Reality
environment is possible 4.8 0.4
The temporal evolution of the flood event supports to
detect critical locations 4.6 0.5
VR supports information presentation for policy and
public 4.8 0.4
Table 3: Sample of simulation sickness questions and users’ input
in a scale of 1 (not at all) to 5 (very strong)
Mean SD
General Discomfort 1.2 0.4
Fatigue (an overall feeling of tiredness or lack of
energy) 1 0.0
Headache 1.2 0.4
Difficulty Focusing 1.6 1.1
Difficulty Concentrating 1.0 0.0
7. Current Limitations and Future work
Although FloodVis received a positive feedback from most of the
users, the user study identified some limitations. Currently, it is not
possible to see any map to have a proper understanding of the area
which misleads the user initially. Although after applying FloodVis
for a while, the user gets familiar with the environment. Our next
step is to add some 2D map to guide users as well as a guided
tour to make this visualization more exciting and easy. Later, we
wish to extend this solution to ScaDS.AI’s "Living Lab" to make it
accessible to public users.
Figure 10: Pie chart from the online survey with 15 participants
showing the majority agreeing on better visualization of FloodVis
comparing to non-VR visualization
P1 P2 P3 P4 P5 P6 P7 P8 P9 P10
SUS Score
0 20 40 60 80 100
Figure 11: SUS results of 10 individual participants
The approach of FloodVis is based on open source data from
Germany and software tools which makes the generalization of
FloodVis to other areas possible. A limitation arises from the lack
of high temporal and spatial resolution modeling results for flood
events. Scalability cannot be assessed at this time, but will result
from the future application of FloodVis in other areas. We will ad-
ditionally extend the project to the visualization of constructions
damages at single houses. In future, we plan to extend this VR vi-
sualization to other climate impacts like heavy rainfall, heatwaves
and drought events.
8. Conclusions
The goal of this research was to visualize ensemble flood risk for
climate change in a way that is easier, exciting and significant for
everyone and to know whether VR visualization justifies flood en-
semble visualization and to know whether the sense of presence
can influence decision making. This paper can argue about the pos-
itive feedback received from different groups of participants attend-
ing the evaluation virtually as well as physically. According to one
user, "we have been watching weather related disaster news and
videos for years but we do not feel the outcome as we are not being
effected personally. But this type of visualization can actually en-
able the sense of presence and this might be an asset helping us to
prepare for the worst." None of the participant disagrees with this
statement. This paper can also argue about smooth navigation, fast
response, techniques to avoid simulation sickness and the overall
experience. With further development, FloodVis can be an asset to
visualize weather and climate-related hazard and risk data and pre-
pare for the worst. We, therefore, plan to continue this research by
working on the limitations and extensions.
9. Acknowledgement
This work was supported by the German Federal Ministry of Edu-
cation and Research (BMBF, 01/S18026A-F) by funding the Cen-
ter for Scalable Data Analytics and Artificial Intelligance ScaDS.AI
Dresden/Leipzig. The authors gratefully acknowledge for providing
computing time through the Center for Information Services and
HPC (ZIH) at TU Dresden. This work also received funding from
Deutsche Forschungsgemeinschaft through DFG grant 389792660
© 2022 The Author(s)
Eurographics Proceedings © 2022 The Eurographics Association.

Marzan Tasnim Oyshi / FloodVis: Visualization of Climate Ensemble Flood Projections in Virtual Reality
as part of TRR 248 and the two Clusters of Excellence CeTI (EXC
2050/1, grant 390696704) and PoL (EXC-2068, grant 390729961).
Further funding was provided by the Foundation for Environment
and Loss Prevention (Stiftung Umwelt und Schadenvorsorge) of the
SparkassenVersicherung Gebäudeversicherung, Stuttgart.
Finally, the authors acknowledge all the participants for taking part
in the user study and for sharing their valuable feedback. Spe-
cial thanks to Apurv Deepak Kulkarni and Neringa Jurenaite from
ScaDS.AI for the fruitful discussions and the responses.
References
[ASAD16] ACH AR C., SOJ., AGRAWAL N., DU HACH EK A.: What we
feel and why we buy: the influence of emotions on consumer decision-
making. https://doi.org/10.1016/j.copsyc.2016.01.
009, August 2016. (Accessed on 02/16/2022). 2
[BBL∗05] BAÑ OS R., BOTELLA C.A ND AL CA ÑI Z M., LI AÑ O V.,
GUERRERO B., REY B.: Immersion and emotion: Their impact on the
sense of presence. https://doi.org/10.1089/cpb.2004.7.
734, February 2005. 2
[BV17] BERG L. P., VANCE J. M.: An industry case study: Investigat-
ing early design decision making in virtual reality. https://rb.gy/
sryrbv, March 2017. 2
[BWW95] BEROG GI G. E., WAIS EL L., WAL LACE W. A.: Employing
virtual reality to support decision making in emergency management.
https://rb.gy/fpff7v, July 1995. 2
[Cor21] CORNWALL W.: Europe’s deadly floods leave scientists stunned.
https://rb.gy/atdsrp, July 2021. (Accessed on 02/10/2022). 1
[CWT∗11] CHEN J. F., WARD EN C. A., TAI D. W.-S., CHE N
F.-S., CHAO C .- Y.: Level of abstraction and feelings of pres-
ence in virtual space: Business english negotiation in open won-
derland. https://doi.org/10.1016/j.compedu.2011.05.
017, November 2011. 2
[DW21] DW: Climate change: Cost of weather disasters surged in 2021 |
news | dw | 27.12.2021. https://rb.gy/ki4iel, December 2021.
(Accessed on 02/09/2022). 1
[EGU21] EGU: Egu - news & press - flooding in western europe: what
happened in july 2021? https://rb.gy/cop8j9, July 2021. (Ac-
cessed on 02/09/2022). 1
[EY19] ENES YILDIR IM I. D.: An integrated web framework for hazus-
mh flood loss estimation analysis, August 2019. doi:https://doi.
org/10.1007/s11069-019- 03738-6.2
[GCL∗21] GORINI A., CAP ID EV IL LE C. S. , LE O G. D ., MANTOVANI
F., RI VA G.: The role of immersion and narrative in mediated presence:
The virtual hospital experience. https://doi.org/10.1089/
cyber.2010.0100, March 2021. (Accessed on 02/16/2022). 2
[GNG16] GHELICHI M. A., NAK HJAVAN B., GH AR EH DAG HI M.: Im-
pact of psychological factors on investment decision making in stock
exchange market. https://rb.gy/9pjuid, July 2016. (Accessed
on 02/16/2022). 2
[GRB21] GONZALEZ D. A. Z., RIC AR DS D., B IL GI N A. A.: Making it
real: A study of augmented virtuality on presence and enhanced benefits
of study stress reduction sessions. https://doi.org/10.1016/
j.ijhcs.2020.102579, March 2021. (Accessed on 02/16/2022). 2
[GSH∗15] GROT TE L S. , STAIB J., HEY ER T., VE TT ER B., GUM HO LD
S.: Real-time visualization of urban flood simulation data for non-
professionals. https://rb.gy/y8wc1d, 2015. 2
[HG11] HUBER D. G., GUL LE DG E J.: Extreme weather and climate
change: Understanding the link and managing the risk. 13. 1
[IPC21] IPCC: Climate change 2021: The Physical Science Basis. Con-
tribution of Working Group I to the Sixth Assessment Report of the Inter-
governmental Panel on Climate Change. Cambridge University Press,
2021. 1
[Jer15] JERALD J.: The vr book: Human-centered design for virtual re-
ality. https://dl.acm.org/doi/book/10.1145/2792790,
October 2015. 2
[JRP01] JOHN REILLY PETE R H. STO NE C. E. F. M. D. W. H. D. J.,
PRINN R. G.: Uncertainty and climate change assessments. https:
//doi.org/10.1126/science.1062001, July 2001. (Accessed
on 03/10/2022). 2
[JTWM96] JORDA N P. W., THO MA S B., WE ER DM EESTER B. A., MC-
CLELLAND I. L.: Usability evaluation in industry - google books.
https://rb.gy/juemzj, 1996. (Accessed on 03/02/2022). 6
[KKF∗21] KRAUS M., KLE IN K., FUC HS J., K EI M D. A., SCHR EI BE R
F., SE DL MA IR M.: The value of immersive visualization. https:
//rb.gy/nkdvie, July 2021. 2
[KLBL93] KENNEDY R. S., LA NE N. E., BE RB AUM K. S., LI LI EN -
THAL M. G.: Simulator sickness questionnaire: An enhanced method
for quantifying simulator sickness. The International Journal of Aviation
Psychology 3, 3 (July 1993), 203–220. (Accessed on 03/02/2022). URL:
https://doi.org/10.1207/s15327108ijap0303_3.6
[LAOK∗19] LEDOUX H., ARROYO OHOR I K., KU MA R K., DU KA I B. ,
LABETSKI A., VI TALI S S.: Cityjson: A compact and easy-to-use en-
coding of the citygml data model. https://rb.gy/qyxdu7, 2019.
(Accessed on 02/25/2022). 4
[Mal21] MALESKA V.: A methodology for investigating the dynamics
and uncertainties of flood risks. https://rb.gy/wcocjw, Novem-
ber 2021. (Accessed on 02/21/2022). 3,4
[MBB22] MOL J., BOT ZE N W. W. J., BLASCH J. E.: After the vir-
tual flood: Risk perceptions and flood preparedness after virtual reality
risk communication. https://rb.gy/k2tx9n, January 2022. (Ac-
cessed on 02/21/2022). 2
[MK18] MEYER R., KU NR EU TH ER H.: The ostrich paradox: Why we
underprepare for disasters. https://rb.gy/cuekdl, May 2018.
(Accessed on 02/21/2022). 6
[OEC16] OECD: Financial management of flood risk. https://rb.
gy/wzbj2n, July 2016. (Accessed on 02/09/2022). 1
[SFB07] SEO M.-G., FE LD MA N BAR RE TT L.: Being emotional
during decision making—good or bad? an empirical investiga-
tion. https://www.ncbi.nlm.nih.gov/pmc/articles/
PMC2361392/, August 2007. 2
[SPKK22] SIMPSON M., PADI LL A L. , KELLER K., KL IP PE L A.: Im-
mersive storm surge flooding: Scale and risk perception in virtual reality.
https://rb.gy/veigz5, April 2022. (Accessed on 02/21/2022). 2
[SS18] SKO LN IK S., S IL COTT A.: Connecting the dots between data
and reality: Using virtual reality to visualize flooding impacts. https:
//rb.gy/lx2ykw, December 2018. (Accessed on 02/18/2022). 2
[SSNO22] SMITH I. H., SOD ER BE RG A. T., NET CH AE VA E., OKH UY-
SEN G. A.: An examination of mind perception and moral reason-
ing in ethical decision-making: A mixed-methods approach. DOI:
10.1007/s10551-021- 05022-9, January 2022. 2
[SVS05] SANCHEZ-VI VE S M. V., SLATE R M.: From presence to con-
sciousness through virtual reality. https://www.nature.com/
articles/nrn1651, April 2005. (Accessed on 02/18/2022). 2
[TF20] TOS HI O FUJ IMI K.: Testing public interventions for flash flood
evacuation through environmental and social cues: The merit of virtual
reality experiments, November 2020. doi:https://doi.org/10.
1016/j.ijdrr.2020.101690.2
[TJD18] TUSSYADI AH II S P.AND WANG D., JUNG T. H., DIE CK M . T.:
Virtual reality, presence, and attitude change: Empirical evidence from
tourism. https://rb.gy/pvboju, June 2018. 2
[WED∗20] WUW., EME RTON R., DUA N Q., WO OD A. W., WETTER-
HALL F., RO BE RTSON D. E.: Ensemble flood forecasting: Current status
and future opportunities. https://rb.gy/pfmwfr, March 2020. 2
[WS94] WITMER, SI NG ER: Presence questionnaire. https://rb.
gy/sf9d9h, November 1994. (Accessed on 02/16/2022). 6
© 2022 The Author(s)
Eurographics Proceedings © 2022 The Eurographics Association.