Conference PaperPDF Available




Content may be subject to copyright.
1ICCS, Athens, Greece
2IHE, Delft, The Netherlands
3U-Hopper, Trento, Italy
4Xteam, Rovigo, Italy
5IBM, Haifa, Israel
Flood risk prediction has been traditionally based on models that are developed from time-series of data
collected over long periods of time from expensive and hard to maintain in-situ sensors available only
in specific areas. The climate change has made the monitoring of the flood events imperative and has
raised the question of whether the development of flood models can be disengaged from the in-situ
sensors. The Scent toolbox is based on smart collaborative and innovating technologies that augment
costly in-situ infrastructure, enabling citizens to become the eyesof the policy makers by monitoring
Land Cover/ Land Use (LC/LU) changes in their everyday activities and environmental phenomena
like floods by crowdsourcing relevant information. Experts in the field of flood models define areas of
interest through a specifically designed tool and ask volunteers to collect specific data needed at these
areas using engaging gaming applications. These data may include images that are processed through
an Intelligent engine and classified based on a LC/LU taxonomy, video of floating objects and images
of water level indicators that allow the automated extraction of the water velocity and the water level
and sensor measurements with low-cost portable environmental sensors. It will be described in detail
how the volunteers are engaged to collect these data, how the data are validated, and how they are used
to create improved LC/LU maps and contribute to the development of improved flood models reducing
the cost and infrastructure needed.
Keywords: scent toolbox, crowdsourcing observations, OGC-compliant data handling.
Europe has invested a lot in infrastructure to achieve an accurate Earth observation capacity.
Initiatives such as Copernicus provide a mapping of forest areas, wetlands or artificial
surfaces; yet, the burden of investing in new equipment or maintaining the current
infrastructure is unsustainable. Ways of complementing the in-situ infrastructure with
citizen-sourced data at a low cost are currently investigated. Recognising that citizen
participation in environmental policy making is in its infancy and that citizens feel unable to
influence environmental policies the Scent toolbox aims to alleviate this barrier. Through a
constellation of smart technologies, it enables citizens to support the policy makers by
monitoring LC/LU changes as part of their everyday activities augmenting the in-situ
infrastructure with a people-generated observation web.
In order to achieve its goals, the Scent toolbox is based on an innovative architectural
design that is presented in Fig. 1, where key components dedicated to specific tasks are
connected in a way that allows the flow of information from local authorities, to volunteers,
to environmental experts and back to the authorities that can now have improved monitoring
of the phenomena of interest allowing them to make educated decisions that can help and
support the areas. The key components of Fig. 1 are briefly presented here., ISSN 1743-3509 (on-line)
WIT Transactions on The Built Environment, Vol 184, ©2018 WIT Press
Urban Water Systems & Floods II 121
A crowd-sourcing platform that provides a series of tools and applications that allow the
flow of information between the components of the toolbox as well as the creation of
information from policy makers and contributions from the volunteers. It includes a dedicated
tool (Authoring tool) for policy makers that allows them to identify areas of interest, create
campaigns and Points of Interest (PoIs) and access the collected and extracted information in
a user-friendly way as map overlays. Part of the crowdsourcing platform is also a series of
gaming applications that aim to engage volunteers to collect images and sensor
measurements as defined in the Authoring tool and to contribute to environmental monitoring
by providing, qualifying and interpreting information about LC/LU.
An intelligence engine, which uses innovative machine learning techniques to classify and
annotate images from citizens and open platforms. The classification is done with respect to
the Scent taxonomy, which is based on the CORINE taxonomy but is also enriched to include
elements needed for the flood models. To fully utilize the collected and extracted
information, the validated data are used to create improved LC/LU maps. In addition, in order
to support the collection of river measurements in a uniform way, useful to the flood models
and invariant of the experience of the volunteers to collect scientific data two tools have been
developed. The tools use state of the art image recognition algorithms that extract from
multimedia (video and image) water level and water surface velocity measurements.
A collection of environmental sensors that are going to support the collection of the data
needed for the flood models. The sensors will range from in-situ sensors that are available in
the areas of interest, portable sensors that the volunteers use to collect additional
measurements as well as sensors mounted on drones flying over the areas of interest.
A Harmonisation platform that collects all the crowdsourced observations, transforms
them to OGC compliant observations, stores them with respect to time and space as such and
offers them to the Scent toolbox components as needed as well as to GEOSS as web services
making the collected data findable and re-usable. Part of the Harmonisation platform is also
the Data Quality Module that assigns a trust level to every registered user and sensor.
The Scent toolbox will be evaluated for a year during a series of campaigns starting the
summer of 2018 in two large scale pilots of great environmental impact; the urban site of
Attica Kifisos River and the rural site of Danube Delta. The two pilot areas have been chosen
carefully for their specific characteristics as well as their different needs and topology.
Finally, the collected data along with the extracted information from the different
components will be used to improve the flood models of the areas, quantify the impact of
LC/LU changes to flood maps and spatio-temporal flooding patterns enabling more effective
flood-related planning and management by policy makers.
The Crowdsourcing platform includes four main modules of the Scent Toolbox:
The Crowdsourcing backend, which handles communication and interaction of
crowdsourced content among the platform components.
The Authoring tool, which represents the entry point for local authorities to (i)
define and customise citizen engagement campaigns on LC/LU data collection (ii)
access crowdsourced images and citizen notifications (iii) view and explore the
extracted information from the crowdsourced data.
The Open Image tool, which crawls open image repository for relevant content, in
order to augment the crowdsourced data with already available information.
The gamification applications that aim to engage citizens into contributing data., ISSN 1743-3509 (on-line)
WIT Transactions on The Built Environment, Vol 184, ©2018 WIT Press
122 Urban Water Systems & Floods II
Figure 1: Architectural overview of Scent toolbox.
2.1 Crowdsourcing backend
The Crowdsourcing backend caters for the upload and management of citizen-generated data
such as images, videos, annotations and questionnaires. It supports the flow of information
[1] from the policy makers (Authoring tool) to the volunteers (gaming applications), to the
Intelligent engine and to the Harmonisation platform. In order to ensure that the produced
data are of high quality a functionality that validates the annotations (given either from the
volunteers or the Intelligent engine) is implemented allowing the data that are sent to the
Harmonisation platform for storage to be considered final. The implementation supports the
following set of functionalities:
Data quality management: the quality of the classifications for each image is
checked and it is decided whether further annotations (from the dedicated gaming
applications) are required in order to meet the overall Scent quality assurance
standards, or whether the classifications provided can be considered finalised.
Provide images to annotate to the various frontends: when the available annotations
are not satisfactory and should be complemented by additional ones the content (and
metadata) become available for annotation to the gaming applications.
Collects (annotated) images/videos from various sources: all user-generated content
is sent to the Crowdsourcing backend, which handles and manages it, making sure
it is forwarded and/or made available to the relevant Scent toolbox components.
Provide campaign configuration data to the gaming applications (received from the
Authoring tool): campaigns are configured by policy makers at the Authoring tool,
stored at the Crowdsourcing backend and forwarded to the gaming applications., ISSN 1743-3509 (on-line)
WIT Transactions on The Built Environment, Vol 184, ©2018 WIT Press
Urban Water Systems & Floods II 123
Provide raw crowdsourced data to the Harmonisation platform: crowdsourced data
uploaded to the Crowdsourcing backend are made available to the Harmonisation
platform through a purposeful interface.
2.2 Open Image tool
The Open Image tool provides crawling functionality by querying a predefined set of open
image platforms when policy makers decide to use images provided by third parties under an
open data license in order to complement Scent-generated data. Configuration data (bounding
box for position and time range for timestamp) are provided by the policy makers through
the Authoring tool. The Open Image tool provides the functionality of querying open image
platforms for relevant images based on the configuration data.
2.3 Authoring tool
The Authoring tool provides a user-friendly web interface, as shown in Fig. 2, for policy
makers, enabling them to create, view, edit or delete content related to Scent Campaigns as
well as PoIs. It is responsible for managing the policy makers’ user accounts, their personal
settings as well as for notifying them of any relevant reported events. It also keeps the users
informed of important events (account changes) through logs. In addition, it gives to the
policy makers access to crowdsourced images and sensor measurements as well as maps of
the areas of interest with information regarding LC/LU, flood models, Campaigns and/or
questionnaires. The Authoring tool implementation therefore supports the following set of
Manage user accounts: The users of the Authoring tool are separated in two
categories, policy makers and premium users, with different privileges.
oThe policy makers can create and view Campaigns and PoIs, they can also edit
and delete any Campaign and PoI that is in the system. They can add/delete a
premium user and give administrative privileges to a premium user. The policy
makers can also create a questionnaire and reviewed the collected information
in a structured way.
oThe premium users can create and view Campaigns and PoIs but they can only
edit Campaigns and PoIs that they have created.
No unregistered user is allowed access to any of the available information.
Manage personal settings: The Authoring tool allows the users to save their
personal settings such as language and place of interest for their accounts.
Manage Campaigns and PoIs: The Authoring tool gives the users the ability to
handle their own Campaigns through their accounts. In addition, provides users the
ability to create, view, edit or delete both PoIs and Campaigns using the map of the
area of interest. Users can also choose to use a PoI in order to provide critical or
important information for the area rather than request information.
Access user-generated data and sensor information: The users can access
crowdsourced images, sensor measurements and taxonomy metadata through a
purposeful web interface.
Visualise flood risk maps and LC/LU maps: The users can have an overview of a
place of interest through a map. The Authoring tool shows (as layers over the
original map) the flood risk map and/or the LC/LU map., ISSN 1743-3509 (on-line)
WIT Transactions on The Built Environment, Vol 184, ©2018 WIT Press
124 Urban Water Systems & Floods II
Inform users of questionnaire results: The users can see the information collected
as answers to the questionnaires that have been set by any user in a structured way.
Initiate the Open Image tool: The users may require an updated LC/LU map for an
area of interest. Through the Authoring tool the bounding box of the area and the
time of interest are defined. The configuration data are sent to the Crowdsourcing
backend to be handled accordingly.
2.4 Gamification apps
A series of data and actions have been defined as very important for the monitoring of the
LC/LU changes and the improvement of the flood models, these data have been mapped to
straightforward, time-effective and meaningful actions that the volunteers can carry out to
collect them [2], [3]. These are actions can be described as follows. (i) Annotate, with tags
from the Scent taxonomy, images collected by the Open Image tool. (ii) Take an image of
specific objects in a predefined area (e.g., vegetation at the river bank, waste and brought
materials in the manholes, tree banks/branches, dustbins, cars and vehicles along the river
bank or in smaller streams connected to the main river). (iii) Take an image or video in a very
specific point of interest (e.g, image of a water level indicator, video of a predefined floating
object). (iv) Answer to questionnaires at areas of interest to support the collection of
structured data. (v) Use portable sensors to take some environmental measurements.
Based on the above identified actions four independent applications have been designed
to simplify the process, boost the user engagement and collect properly formatted data of
high quality.
2.4.1 Explore app
Scent Explore is a mobile application that the volunteers use to locate specific areas of
interest, find each point where specific tasks should be carried out, and gain rewards when
successfully concluding them. Two reward mechanisms (points and badges [4]) are used to
boost the user engagement. The tasks that the users are asked to carry out and for which they
are rewarded are: (i) Take a picture containing a specific object from the Scent taxonomy;
(ii) Go to a specific location and finding the Little Animals”.
Figure 2: Scent Authoring tool., ISSN 1743-3509 (on-line)
WIT Transactions on The Built Environment, Vol 184, ©2018 WIT Press
Urban Water Systems & Floods II 125
This functionality will allow the user to spot the exact place where a water level indicator
is located, or where a sensor measurement should be carried out. (iii) Answer questionnaires
regarding the area.
The user can either play as guest or as registered user; while it has been decided to have a
single database for user profiles. This means that the users who register through the app, can
also use the same account for the other gaming applications having a consistent user
experience. Another functionality of the Explore app is that when a user is near a little
animal/PoI, even if she is not looking at the map and/or the phone is in standby mode, the
device will vibrate, the camera will automatically turn on and the user will be invited to take
a picture of the area. The user will then be asked to tag the picture with elements from the
Scent taxonomy. The user can also take and tag pictures of areas and objects which do not
correspond to predetermined points of interest; allowing the user to contribute at any given
moment data that she feels are of interest. When the user looks into the camera while
searching for a little animal, the GPS position will be integrated with the vector data of the
gyroscope or with data coming from the accelerometer guiding the user to locate the ‘little
animal’ and as a result take images with specific orientation that will include information of
interest. The user interface of the application is presented in Fig. 3.
2.4.2 Collaborate
The Scent Collaborate is a browser-based crowdsourcing platform that allows users to
annotate existing pictures choosing elements from the Scent taxonomy that exist in them.
The same picture can be annotated by more than one user so as to support the validation of
the annotations from the Crowdsourcing backend. Here too the same two reward mechanisms
are applied as the user is rewarded with points based on the number of annotated pictures.
Once more the points are connected to badges that acknowledge the achievements of the
users. Scent Collaborate is oriented toward a broad audience interested in joining the Scent
movement even if they are not living close to an area of interest. The platform targets citizens
interested in environmental issues, contributing to science and supporting the policy makers.
2.4.3 Captcha
This is a browser-based plugin to be integrated in third-party websites when there is a need
to verify that the user is actually a human and not a robot. The user is asked to spot which of
the images presented to her include a tag from the Scent taxonomy, the response is used to
support the validation mechanism for the annotations.
2.4.4 Sensor measurement collection app
This is a mobile application that is used in addition to the Explore App to communicate with
the portable sensors available at the area of interest, record the measurements and send them
to the backend. The users can login using their available account and collect rewards for their
contributions. The same two reward mechanisms are applied here as well with the user
getting points based on the number of sensor measurements. Once more the points are
connected to badges that celebrate the achievements of the users.
The Scent Intelligent engine uses state of the art machine learning techniques in order to
ensure that all the available information included in the multimedia collected by the
volunteers is fully utilised., ISSN 1743-3509 (on-line)
WIT Transactions on The Built Environment, Vol 184, ©2018 WIT Press
126 Urban Water Systems & Floods II
Figure 3: Scent Explore app.
3.1 Image Analysis tool
The Image Analysis tool (IAT) will analyse the input image and classify items seen in the
image as belonging to the elements of the Scent taxonomy. The classification process will be
handled by a specifically developed Deep Learning (DL) network optimized for Scent
oriented taxonomy classifications. The DL network will be trained with representative
images collected from open image platforms and manually annotated with respect to the
Scent taxonomy. Use of other existing broad-category image classification tools (e.g. IBM
Watson Visual Recognition service) will be investigated as a means for improving the
accuracy and confidence level of the classification. In addition, the IAT will receive the
information collected from the Crowdsourcing backend validation mechanism for
incremental learning purposes.
3.2 Water level and water velocity measurements from multimedia
In order to get the river measurements with a consistent and accurate way from volunteers
that have no technical knowledge two tools have been designed.
The first one is the Water Level Measurement tool which uses state of the art image
recognition techniques in order to extract the water level from images containing a water
level indicator that is half-submerged into water. The initial goal of the tool is to “read” the
indicator and extract the number that is closer to the water lever. In order to achieve that the
tool extracts features of the input images and match them with pre-calculated model’s
invariant of rotation, luminance variation, image noise or scale that are stored in a model
database. If the quality of the image does not allow for the recognition of the numbers of the
indicator, calibration techniques based on the patent of the indicator and tis bounding box are
exploit resulting in a less trusted estimation of the water level.
The Water Velocity Calculation tool uses state of the art video processing algorithms in
order to extract the water surface velocity from a video containing a pre-defined floating, ISSN 1743-3509 (on-line)
WIT Transactions on The Built Environment, Vol 184, ©2018 WIT Press
Urban Water Systems & Floods II 127
object moving on the surface of a water body [5]. The Kalman filter is used for the object
localisation, while the object displacement is estimated based on feature matching algorithms
and calibrated using the known dimensions of the floating object. The tool is resistant to noise
introduced to the video by intentional or unintentional movement of the camera as long as
the floating object is included in the video frames. The more stable elements from the
surrounding environment (such as trees and rocks) the video includes the more accurate
calculation of the estimation of the displacement can be achieved.
Each measurement estimation extracted by the tools is accompanied by a degree of trust
that highly depends on the quality of the input data. The tools have been designed so that a
high degree of trust can be achieved from images and videos taken from regular smartphones.
3.3 LC/LU map overlays
There are many approaches and algorithmic tools that are adequate for effectively performing
map segmentation regarding LC/LU; their applicability however is closely related to the
classification problem and the characteristics of the study area. After a study of the available
aerial images of the areas of interest and preliminary data exploration, it has been concluded
that the preferable way of performing the map segmentation is to handle it as an aerial scene
classification task. Accordingly, the map under consideration is split into overlapped tiles of
predefined size, which are slid across both vertical and horizontal directions, aiming to build
one or more classifiers that are capable of classifying all the tiles according to the Scent
Environmental sensors aim to collect the detailed set of environmental and operational
parameters that are important to both policy makers as well as hydrology experts. Important
measurements collected for further use and exploitation include soil moisture, air and water
temperature. In this context innovative sensors are deployed and adapted as part of the Scent
toolbox. Furthermore, environmental sensors and digital cameras are integrated on
multicopters based on open source hardware platforms for obtaining spatially continuous
data, including elevation data for quantitative mapping and monitoring parameters with an
increased accuracy.
In-situ sensors: The majority of environmental data is traditionally collected through in-
situ monitoring stations. Not wanting to exclude the information that might be available from
such stations in areas of interest water level and water velocity monitor station that are
available to the pilot areas are integrated with the Harmonisation platform providing
measurements which are stored as OGC compliant observations. These measurements when
available serve two very important roles. On the one had they provide the ground truth for
the Data Quality Control module allowing for better and with higher degree of trust definition
of the thresholds used to judge a measurement. On the other hand they provide continuous
data that can be incorporated to the flood models using traditional techniques.
Portable sensors: An off-the-shelf sensor has been chosen as the portable sensor that the
volunteer will use for the sensor measurements. A list with some very important and critical
for the user engagement criteria was composed to assist on this choice. The criteria included
restrictions regarding portability, reliability, ease in use, low-cost and connectivity. The
sensors are currently available and integrated with the Measurement Collection App ready to
be used in the first pilot activities.
Drone sensors: Ready-to-fly digital camera and environmental sensors mounted on a
multicopter will be deployed during the pilot activities for a quantitative mapping of, ISSN 1743-3509 (on-line)
WIT Transactions on The Built Environment, Vol 184, ©2018 WIT Press
128 Urban Water Systems & Floods II
environmental parameters that will cover air, land and water parameters. The multicopter will
take photos of the regions in orthomosaic and DEM files as well as in near-IR. Parameters
will be collected as digital aerial photography that spatially complements ground-based data.
The UAV will measure also temperature by using a thermal camera. Images in near-IR and
orthomosaic will be used in the improvement of the flood models. In order to improve the
location accuracy a differential GPS will be used.
The harmonisation platform is a scalable web-based platform that stores IoT as OGC
compliant observations and offers them as a web service while at the same time verifying the
quality of the data and extracting a trust level for the sensors and users involved.
5.1 OGC compliance
The choice for storing the data as OGC observations was made based on the importance of
making the collected information findable and reusable. Based on that the Harmonisation
platform has been designed to receive as input IoT-data, which are sensor measurements,
custom metadata from images and videos, drone-generated images/measurements and map
overlays. The platform translates the aforementioned data and metadata to OGC compliant
observations using the applicable standard, such as SOS, SES, WMS, WFS and SensorML.
The OGC compliant observations are then stored in a distributed and secure environment,
indexed with respect to time and space and are accessible using CQL (Common Query
Language) always with respect to the demands of the OGC standard [6].
5.2 Data Quality Control
The Data Quality Control system is aiming to cope with the following main challenges:
correctness/accuracy of the sensor measurements, user reliability and protection of the
system from malicious contributions. In order to handle the volume of the data, a filtering
process flags as invalid a percentage of the input data that do not fall within the expected
range. During this step, it is preferable to flag a valid result than allowing an invalid one to
enter the system. Depending on the value measured, the area, the availability of historical
measurements and the volume of the data, different solutions can be applied to deal with this
issue. In the context of Scent, two basic approaches are implemented:
Gaussian probability distribution: Provides data plausibility tests (range test, sigma
test, delta test) employing different statistical methods (standard deviation, variance,
etc.). For these tests, the calculation of a threshold parameter poses the most critical
challenge, being determined by statistical distributions based on existing data over
a period sufficiently long to capture the full suite of variability [7], [8].
Fuzzy logic: Enables the possibility to encode linguistic rules and heuristics,
reducing the solution time since the expert’s knowledge can be built in directly. In
addition, its qualitative representation form makes fuzzy interpretations of data
natural and an intuitively plausible way to formulate and solve problems [9], [10].
The next step is to scrutinize the flagged results to see if they correspond to invalid or
marginal data. Extra logic rules are implemented to distinguish the two cases. These rules try
to connect the available information based simply questions such as: Is this the only
measurement from the area/time flagged? Was this measurement from a trusted user/ reliable
sensor? Was the sensor calibrated/maintained recently?, ISSN 1743-3509 (on-line)
WIT Transactions on The Built Environment, Vol 184, ©2018 WIT Press
Urban Water Systems & Floods II 129
User-wise, an ELO rating scheme is implemented, allowing the system to judge the users
regarding the quality of the data that they produce. By keeping an internal score of trust
towards each user, the system can be protected from harmful intents, whilst keeping the data
without noise. For instance, wins against the same ELO rating all the time will result in the
user’s rating reaching a maximum value and become invariable. Moreover, the ratings for
rewards/ punishments are chosen in a way that the novice user will be rewarded a lot when
he does something right and lose a little when he does something wrong. This adapts as the
expertise level increases protecting the system from malice while not punishing inexperience.
Last but not least, detecting malice contributions mixed within valid ones can be achieved by
a periodic check about ELO ratings that are stable while the user contributes a lot as this
means that the user is introducing a lot of noise to the system.
The two regions where the Scent toolbox is going to be test are the Danube Delta in Romania
and the Kifisos River basin in Athens, Greece. The two pilot areas were chosen carefully in
order to test the toolbox in two fundamental different topologies of areas with different needs
and challenges. In the Danube Delta in Romania, the flooding is a necessary and welcome
part of maintaining the ecosystem, as the unique ecosystem of the area with the vibrant
wildlife and the remarkable plant life, is sustained by these waters. There it is important to
understand how these flood waters work, how the climate changes and the changes in the
LC/LU of the area affect the phenomena and what measures should be taken for the
sustainability of the ecosystem before it is too late.
In the Kifisos River the challenges are completely different given the urban environment.
The landscape is changing fast and dramatically as streams are covered with concrete, forests
are torn down and the natural course of the river is modified. There it is very important to
understand the river, its course and the LC/LU changes. Measures should be taken to avoid
any overflows of the river that affect the communities living close by.
Preparatory activities for the execution of the two pilot demonstrations have already been
initiated. From August 2018 onwards, several thematic campaigns, focused on the collection
of LC/LU images, sensor measurements and river data, will take place in each pilot area,
where volunteers will be engaged with specific test cases and tools. The pilot activities will
run for a period of approximately 10 months, foreseeing at least 6 different campaigns to take
place in each pilot area. The concept of the thematic campaigns was carefully chosen, to
allow the organization of a dedicated workshop at the start of each campaign where
volunteers will not only be informed about the project and but also be trained in the use of
the Scent toolbox used during the campaign.
When floods are the problem to be addressed, models can serve as a base for discussion and
evaluation of possible adaptive measures to control their impact on the environment. In this
context, the Scent toolbox includes models representing hydrology and/or hydrodynamics of
flow, in relation to flood events, such that data used for their simulation will be extended with
contributions from citizens (crowdsourced data). Data collected by citizens will be integrated
into the models and tested for its validity.
Initially, flow models of two selected pilot study areas are built with existing data,
acquired without any contributions from citizens. These two models are compared, with the
models enriched with data collected by citizens. The two pilot case studies under
consideration have different characteristics, hence they require different modelling
approaches. The tools used for model building are part of a non-commercial modelling suite, ISSN 1743-3509 (on-line)
WIT Transactions on The Built Environment, Vol 184, ©2018 WIT Press
130 Urban Water Systems & Floods II
developed by the Hydrologic Engineering Centre (HEC) of United States Army Corps of
Engineers (USACE); the river analysis system (HEC-RAS) and the hydrological modelling
system (HEC-HMS). These two software packages are freely available for download and
were chosen based on their capabilities to represent complex river network systems (such as
Danube Delta) and fast occurring floods (such as Kifisos catchment).
The emphasis in the Danube Delta [11] pilot is on determining the spatio-temporal
flooding patterns which are important for the ecosystem of the delta. This requires a detailed
1D/2D hydrodynamic model, which represents the flow of water through channels, lakes and
over the land. For the whole area of the Danube Delta, the set of built models based on
combinations of different geometry and/or boundary and initial conditions. Three types of
hydrological regimes are taken into consideration: dry, average and wet year, with
corresponding flows on the main Danube delta branches.
In Kifisos pilot [12] the emphasis is related to flood as a hazard. Based on flood behavior
in this area, combined with data availability a hydrological model that can generate flood
runoff is considered for the upstream part of the catchment. A smaller hydrodynamic model
is also considered for the last 3 km of the catchment. The hydrological and hydrodynamic
models were developed and tested separately, however the input for the upstream boundary
condition of the hydrodynamic model is received from the hydrological model. The
connection of the two developed types of models is shown in Fig. 4.
The developed models have been developed and tested with existing data. However, they
still have significant uncertainties that will be addressed with the crowd-sourcing campaigns
of Scent. In Danube Delta, the additional LC/LU data will be used to provide improved
roughness parameter, and other data will validate the model in terms of spatial distribution
of flood extent, flood levels and flood velocities and discharges, calculated by the 1D-2D
HEC-RAS model. In Kifisos the LC/LU data contributed by citizens will be used for updating
infiltration and runoff parameters of the rainfall runoff model built by HEC-HMS and
improved cross section data and updated measurements of water levels and
velocities/discharges in the river will be used to better calibrate and validate the HEC-RAS
hydraulic model of the downstream part of the catchment.
Figure 4: Connection of HEC-HMS and HEC-RAS model., ISSN 1743-3509 (on-line)
WIT Transactions on The Built Environment, Vol 184, ©2018 WIT Press
Urban Water Systems & Floods II 131
It has been presented above that careful planning along with an innovative architectural
design and a constellation of smart technologies can contribute to better decision-making
through the empowerment and active role of citizens in environmental monitoring. The Scent
toolbox is allowing citizens to contribute the proper scientific data as needed for the
researchers, resulting in the production of improved flood models. These models become
invaluable tools for the policy makers in order to make educated decisions regarding local
preparations for the wet season.
This paper is supported by European Union’s Horizon 2020 research and innovation
programme under grant agreement no 688930, project SCENT (Smart Toolbox for Engaging
Citizens into a People-Centric Observation Web).
[1] Carreras, I. et al., Crowd-sensing: Why context matters. Pervasive Computing and
Communications Workshops (PERCOM Workshops), 2013 IEEE International
Conference on IEEE, 2013.
[2] Fucks, M., Fizek, S., Ruffino P. & Schrape N., Rethinking Gamification, Meson Press,
[3] Dörner, R., Göbel, S., Effelsberg, W. & Wiemeyer, J., Serious Games: Foundations,
Concepts and Practice, Springer, 2016.
[4] Chou, Y.-K., Actionable Gamification - Beyond Points, Badges, and Leader Boards,
Octalysis Media, 2015.
[5] SelvaBalan, M., Sharma, N., Deshpande, G., Kankariya, C. & Naik, A.A., Surface
water velocity measurement using video processing: A survey. 2014 International
Conference on Electronics and Communication Systems (ICECS), Coimbatore, 2014.
[6] Open Geospatial Consortium (OGC), Online.
[7] Taylor, J.R. & Loescher, H.L., Automated quality control methods for sensor data: A
novel observatory approach. Biogeosciences, 10, pp. 49574971, 2013.
[8] Fiebrich, C.A., Morgan, C.R., McCombs, A.G., Hall, P.K. Jr., & McPherson, R.A.,
Quality assurance procedures for mesoscale meteorological data. J. Atmos. Oceanic
Technol., 27, pp. 15651582, 2010.
[9] Mahajan, A., Wang, K. & Ray, P.K., Multisensor integration and fusion model that
uses a fuzzy inference system. IEEE/ASME Trans. Mechatron., 6(2), pp. 188196,
[10] Timms, G.P., de Souza, P.A., Reznik, L. & Smith, D.V., Automated data quality
assessment of marine sensors. Sensors, 11(10), pp. 95899602, 2011.
[11] Giosan, L., Constantinescu, S., Filip, F. & Deng, B., Maintenance of large deltas
through channelization: Nature vs. humans in the Danube delta. Anthropocene, 1, 35
45. DOI: 10.1016/j.ancene.2013.09.001, 2013.
[12] SelvaBalan, M., Sharma, N., Deshpande, G., Kankariya, C. & Naik, A.A., Surface
water velocity measurement using video processing: A survey. International
Conference on Electronics and Communication Systems (ICECS), Coimbatore, 2014., ISSN 1743-3509 (on-line)
WIT Transactions on The Built Environment, Vol 184, ©2018 WIT Press
132 Urban Water Systems & Floods II
As the Internet of Things (IoT) systems gain in popularity, an increasing number of Big Data sources are available. Ranging from small sensor networks designed for household use to large fully automated industrial environments, the IoT systems create billions of measurements each second making traditional storage and indexing solutions obsolete. While research around Big Data has focused on scalable solutions that can support the datasets produced by these systems, the focus has been mainly on managing the volume and velocity of these data, rather than providing efficient solutions for their retrieval and analysis. A key characteristic of these data, which is, more often than not, overlooked, is the spatial information that can be used to integrate data from multiple sources and conduct multi-dimensional analysis of the collected information. We present here the solutions currently available for the storage and indexing of spatial datasets produced by the IoT systems and we discuss their applicability in real-world scenarios.
Full-text available
National and international networks and observatories of terrestrial-based sensors are emerging rapidly. As such, there is demand for a standardized approach to data quality control, as well as interoperability of data among sensor networks. The National Ecological Observatory Network (NEON) has begun constructing their first terrestrial observing sites, with 60 locations expected to be distributed across the US by 2017. This will result in over 14 000 automated sensors recording more than > 100 Tb of data per year. These data are then used to create other datasets and subsequent "higher-level" data products. In anticipation of this challenge, an overall data quality assurance plan has been developed and the first suite of data quality control measures defined. This data-driven approach focuses on automated methods for defining a suite of plausibility test parameter thresholds. Specifically, these plausibility tests scrutinize the data range and variance of each measurement type by employing a suite of binary checks. The statistical basis for each of these tests is developed, and the methods for calculating test parameter thresholds are explored here. While these tests have been used elsewhere, we apply them in a novel approach by calculating their relevant test parameter thresholds. Finally, implementing automated quality control is demonstrated with preliminary data from a NEON prototype site.
Full-text available
Mesoscale meteorological data present their own challenges and advantages during the quality assurance (QA) process because of their variability in both space and time. To ensure data quality, it is important to perform quality control at many different stages (e.g., sensor calibrations, automated tests, and manual assessment). As part of an ongoing refinement of quality assurance procedures, meteorologists with the Oklahoma Mesonet continually review advancements and techniques employed by other networks. This article's aim is to share those reviews and resources with scientists beginning or enhancing their own QA program. General QA considerations, general automated tests, and variable-specific tests and methods are discussed.
Full-text available
The automated collection of data (e.g., through sensor networks) has led to a massive increase in the quantity of environmental and other data available. The sheer quantity of data and growing need for real-time ingestion of sensor data (e.g., alerts and forecasts from physical models) means that automated Quality Assurance/Quality Control (QA/QC) is necessary to ensure that the data collected is fit for purpose. Current automated QA/QC approaches provide assessments based upon hard classifications of the gathered data; often as a binary decision of good or bad data that fails to quantify our confidence in the data for use in different applications. We propose a novel framework for automated data quality assessments that uses Fuzzy Logic to provide a continuous scale of data quality. This continuous quality scale is then used to compute error bars upon the data, which quantify the data uncertainty and provide a more meaningful measure of the data's fitness for purpose in a particular application compared with hard quality classifications. The design principles of the framework are presented and enable both data statistics and expert knowledge to be incorporated into the uncertainty assessment. We have implemented and tested the framework upon a real time platform of temperature and conductivity sensors that have been deployed to monitor the Derwent Estuary in Hobart, Australia. Results indicate that the error bars generated from the Fuzzy QA/QC implementation are in good agreement with the error bars manually encoded by a domain expert.
Full-text available
An intelligent multisensor integration and fusion model that uses fuzzy logic is developed. Measurement data from different types of sensors with different resolutions are integrated and fused based on the confidence in them derived from information not usually used in data fusion, such as operating temperature, frequency range, fatigue cycles, etc. These are fed as additional inputs to a fuzzy inference system (FIS) that has predefined membership functions for each of these variables. The output of the FIS are weights that are assigned to the different sensor measurement data that reflect the confidence in the sensor's behavior and performance. A modular approach is adopted. It allows adding or deleting a sensor, along with its fuzzy logic controller (FLC), anytime without affecting the entire data fusion system. This paper presents a preliminary model that fuses the data from three different types of sensors that monitor the strain at a single location in a cantilever beam. This will be later extended to sensors that will be fixed at different locations on the same beam. The results from the proposed work are a stepping stone toward the development of generic autonomous sensor models that are capable of data interpretation, self-calibration, data fusion from other sources, and even learning so as to improve their performance with time. This work is aimed at the development of smart structural health monitoring systems, but has applications in diverse fields such as robotics, controls, target tracking, and biomedical imaging
Conference Paper
Crowd-sensing is becoming a popular computing and sensing paradigm for enclosing humans in the sensing loop. The underlying idea is that people, together with their mobile device, can act as mobile and pervasive sensors, gathering information about the surrounding environment and potentially providing direct input. In this work we focus on how to embed context-awareness in a crowd-sensing system in order preserve the battery of user's mobile device, while maximizing the user participation to crowd-sensing campaigns. We present the design and implementation of the Matador platform, and a preliminary evaluation obtained through a small-scale pilot study.
Rethinking Gamification
  • M Fucks
  • S Fizek
  • P Ruffino
  • N Schrape
Fucks, M., Fizek, S., Ruffino P. & Schrape N., Rethinking Gamification, Meson Press, 2014.
Surface water velocity measurement using video processing: A survey
  • M Selvabalan
  • N Sharma
  • G Deshpande
  • C Kankariya
  • A A Naik
SelvaBalan, M., Sharma, N., Deshpande, G., Kankariya, C. & Naik, A.A., Surface water velocity measurement using video processing: A survey. 2014 International Conference on Electronics and Communication Systems (ICECS), Coimbatore, 2014.
Maintenance of large deltas through channelization: Nature vs. humans in the Danube delta
  • L Giosan
  • S Constantinescu
  • F Filip
  • B Deng
Giosan, L., Constantinescu, S., Filip, F. & Deng, B., Maintenance of large deltas through channelization: Nature vs. humans in the Danube delta. Anthropocene, 1, 3545. DOI: 10.1016/j.ancene.2013.09.001, 2013.