IEEE Consumer Electronics Magazine - November/December 2021 - 90

Special Section on Advances in Smart Cities
continuously increasing global warming. Unfortunately,
many regions have already experienced
these consequences in terms of loss of human
lives, natural ecosystems, and economic losses,
including damage to housing, infrastructure, primary
production systems, and tourism, not to
mention food, insecurity, and the significant
impact on health. In the context of the Sendai
Framework for Disaster Risk Reduction (SFDRR-
https://www.undrr.org) 2015-2030, preventing
and reducing disaster risks involves understanding
them, strengthening disaster-management
policies, increasing investments to reduce the
impacts of these events, and improving the preparedness
for effective disaster response.1
The early identification of water-based disasters
can be lifesaving; hence, the challenge is to
identify appropriate and timely warning measures
in a continuously changing environment.2
New methodologies for obtaining observed flow
data are currently being explored through crowdsensing
tools.35 In this regard, mobile crowdsensing
(MCS) is a priority research line, which
relies on data collection from many mobile sensing
devices to empower ordinary citizens to
extract intelligence from crowds and deliver
human-centric services thorough the aggregation
and fusion of these data in the cloud.6 MCS is
inexpensive, since there is no need for network
deployment, and its spatiotemporal coverage is
outstanding. Following Guo et al.,7 two data-generation
modes in MCS can be distinguished: 1)
mobile sensing, which leverages raw data generated
from the hardware sensors that are embedded
in mobile devices (e.g., accelerometer, GPS,
camera, or microphone, among others), and 2)
social sensing (or social networking), which leverages
user-contributed data from social media.
The latter mode considers participants as " social
sensors, " i.e., agents that provide information
about their environment through social-media
services after the interaction with other agents.
In the context of MCS, this article explores the
applicability of social sensing to achieve the early
detection of water-related issues through an analysis
of people's posts and opinions in social networks.
Indeed, the ubiquitous use of social
networks, such as Twitter or Facebook, establishes
a generic framework for having a sufficiently
large social critical mass to provide
90
citizens with a real-time communication channel
for reporting problems related to our environment.
Several studies have employed social
media to automatically identify natural hazards.
Data collected from Twitter have been analyzed
for the detection/location of earthquakes and
typhoons,8 wildfires9 and heat waves.10 Moreover,
Spielhofer et al.11 provided interesting
insights into a flooding crisis in U.K. by leveraging
data from Twitter, where they concluded that
this type of techniques could be used in real time
to provide actionable intelligence for emergency
services. Similarly, Shi et al.12 analyzed and classified
a batch of tweets generated during a flooding
episode in the Kinu River Basin, Japan. They
proved that the information contained in these
tweets actually provided valuable information
for disaster management in the target area.
The previous studies used Twitter in an offline
batch-based manner, downloading information
for a posteriori analysis following a streaming
approach. With respect to real-time applications
to track social-media information for waterrelated
issues, Bruijin et al.4 developed a database
for detecting floods in real time on a global
scale using Twitter. Their tools are useful to
obtain datasets for an offline analysis but, they
do not provide actionable information, i.e., they
do not identify issues related to floods.
In contrast, our application WATERSensing
takes real-time input from different social-media
sources, which is then analyzed with natural language
processing (NLP) techniques to identify
water-related issues in real time. Several metrics
are also used to categorize microtexts and provide
a problem-relatedness perception index
(PPI). Moreover, location, person and organization
entities are extracted through named-entity
recognition (NER) techniques to visualize these
issues in real time. WATERSensing is validated in
a real natural disaster scenario, i.e., Storm Gloria,
which heavily affected eastern Spain during
January 2020. Our results show that the identification
of issues performed by our application
matches with the timing of a natural disaster,
and it is correlated with actual events.
SOFTWARE ARCHITECTURE
The overall software architecture of WATERSensing
is based on three main modules: 1) the
IEEE Consumer Electronics Magazine
https://www.undrr.org
IEEE Consumer Electronics Magazine - November/December 2021

Table of Contents for the Digital Edition of IEEE Consumer Electronics Magazine - November/December 2021
