ArticlePDF Available

Towards resilient and smart cities: A real-time urban analytical and geo-visual system for social media streaming data

Authors:

Abstract and Figures

Cities worldwide are vulnerable to unpredictable extreme events such as disasters and public health crises. Urban big data and data-driven technologies have played an increasingly important role in building smart and resilient cities that can respond rapidly to these perturbations. However, many existing approaches had limited capabilities for processing big data, which has led to time-consuming and costly decision-making. Thus, we develop a real-time data-driven analytical and geo-visual system to enable smart and rapid responses to urban extreme events. The system is built on ArcGIS’s GeoEvent Server and Apache Spark and processes streaming data from social media with high speed, massive volume, and multiple modalities. The system employs online topic modeling and domain-adaptive sentiment analysis to track small-scale, undefined events, visualizes their spatial and semantic dynamics, and provides early alerts for crises and emergencies via an interactive online GIS platform. The proposed system has been applied during a large-scale hurricane and demonstrated effectiveness and agility in tracking and reporting emerging small-scale crises. The developed system can be applied in various urban scenarios to enable timely situation awareness and rapid response. This research contributes to the smart city safety and building rapidity of resilient cities.
Content may be subject to copyright.
Urban Agility and Resilience Lab
Page 1/31
Citation: Yao, F., & Wang, Y. (2020). Towards resilient and smart cities: A real-time urban
analytical and geo-visual system for social media streaming data. Sustainable Cities and Society,
102448. DOI: https://doi.org/10.1016/j.scs.2020.102448
Towards Resilient and Smart Cities: A Real-Time Urban Analytical and Geo-Visual
System for Social Media Streaming Data
FANG YAO 1, YAN WANG, PhD 2*
1. PhD Candidate, Department of Urban and Regional Planning, College of Design, Construction
and Planning, University of Florida, 1480 Inner Road, Gainesville, FL, 32601, USA; E-mail:
fang.yao@ufl.edu; ORCID: 0000-0002-5516-5318.
2* Assistant Professor, Department of Urban and Regional Planning and Florida Institute for Built
Environment Resilience, University of Florida, P.O. Box 115706, Gainesville, FL 32611, U.S.A.
(corresponding author); Tel: +1(352) 294-1484; E-mail: yanw@ufl.edu; ORCID: 0000-0002-
3946-9418.
Abstract
Cities worldwide are vulnerable to unpredictable extreme events such as disasters and public health
crises. Urban big data and data-driven technologies have played an increasingly important role in
building smart and resilient cities that can respond rapidly to these perturbations. However, many
existing approaches had limited capabilities for processing big data, which has led to time-
consuming and costly decision-making. Thus, we develop a real-time data-driven analytical and
geo-visual system to enable smart and rapid responses to urban extreme events. The system is built
on ArcGIS’s GeoEvent Server and Apache Spark and processes streaming data from social media
with high speed, massive volume, and multiple modalities. The system employs online topic
modeling and domain-adaptive sentiment analysis to track small-scale, undefined events,
visualizes their spatial and semantic dynamics, and provides early alerts for crises and emergencies
via an interactive online GIS platform. The proposed system has been applied during a large-scale
hurricane and demonstrated effectiveness and agility in tracking and reporting emerging small-
scale crises. The developed system can be applied in various urban scenarios to enable timely
situation awareness and rapid response. This research contributes to the smart city safety and
building rapidity of resilient cities.
Keywords: Geo-Visualization, Real-Time, Resilience, Smart City, Social Media, Urban Analytics
Urban Agility and Resilience Lab
Page 2/31
1. Introduction
Cities are complex and dynamic systems that contain infrastructures, information, and innovation
and house the majority of the world’s population (Batty, 2008). Cities are also exposed to a variety
of unforeseeable extreme events, such as disasters and infectious diseases, which have at times
caused tremendous economic and social losses (Arafah & Winarso, 2017; Zhu et al. 2019).
Disasters have affected more than a third of the world’s population (1.5 billion) and cost more than
US$1.3 trillion in economic losses (UN DESA, 2018). In recent years, influenza epidemics have
caused up to 56,000 deaths annually in the United States and have had substantial financial costs
(McGowan et al., 2019). To respond actively to such events, researchers and practitioners from
multiple disciplines develop theories and approaches to help cities prepare for unexpected
perturbations (Woetzel et al., 2018; Zhang & Li, 2018).
In light of the need to building resilient and smart cities in this context, urban studies and
practices strive to maintain cities’ essential functionality while reducing the adverse effects when
disruptions happen (Allam & Newman, 2018; Angelidou et al., 2018; Desouza & Flanery, 2013;
Hatuka et al., 2018; Leichenko, 2011; Wang et al., 2019). Existing literature body has discussed
four critical aspects of resilience: robustness (the ability to withstand stress without suffering
degradation or loss of function), redundancy (the extent to which components can be substituted
for to recover reduced or lost functionality), resourcefulness (the capacity to identify problems,
establish priorities, and allocate resources), and rapidity (the ability to meet priorities and achieve
goals promptly) (Bruneau et al., 2003; Godschalk, 2003; Zobel, 2011). However, as cities grow
and gain complexity, conventional approaches that treat resilience as a conceptual process and use
static data can become ineffective for achieving the conditions outlined above (Meerow et al.
2016). Thus, it is necessary to implement new methods and technologies to address the challenges
of extreme events and promote resilience.
In the context of burgeoning big data and advanced information and communication
technologies (ICTs), more “smart” solutions have also been proposed to help cities survive and
function under extreme stresses (Palmieri et al., 2016; Soyata et al., 2019; Yang et al. 2017). A
recent article proposed the smart robustness, smart redundancy, smart resourcefulness, and smart
rapidity to leverage resilience by embedding smart technologies and systems in the fabric of cities
(Desroches & Taylor, 2018). Although rapidity (e.g., speed) in responding to risks is essential to
resilient cities (Al Nuaimi et al., 2015; Desouza & Flanery, 2013; Palmieri et al., 2016; Platt et al.
Urban Agility and Resilience Lab
Page 3/31
2016), few studies have focused on this dimension of urban resilience, particularly among urban-
scale quantitative studies (Meerow et al., 2016). Within the smart city context, however, rapid or
real-time big data applications can mitigate damagesimpacts and enhance the capacity to recover
from extreme events quickly (Desroches & Taylor, 2018; Malik et al., 2018). For instance, early
detection of crises or emergencies and rapid responses allow cities to collect relevant information,
monitor the characteristics of events (e.g., locations, time, types), and provide timely analyses and
predictions, and thus to better coordinate relief efforts, assess damages, and restore urban system
performance (Desouza & Flanery, 2013; Khan et al., 2015; Kitchin, 2014; Kontokosta & Malik,
2018; Woetzel et al., 2018; Zhang et al. 2019).
However, most existing quantitative studies are conceptual rather than operational to enhance
resilience with smart rapidity, because it is challenging to design a specific plan for an abstract and
complex notion such as resilience (Desouza & Flanery, 2013; Hatuka et al., 2018; Wang et al.
2020). For example, Klein et al. (2017) described a vision and a conceptual framework for
monitoring and managing cities’ environmental and social dynamics without giving specific
methods or plans. Current efforts to create smart and resilient cities also suffer from a mismatch
between real-time information resources and delayed decision-making, as well as incompatible
algorithms for processing high-volume and -velocity urban streaming data (Al Nuaimi et al., 2015;
Khan et al., 2015; Yang et al., 2017). The existing prototypes of urban analytics systems (e.g.,
Huang et al., 2017; Psyllidisn et al., 2015) were designed for the analysis and visualization of a
diversity of urban topics (human movement patterns, traffic conditions, or place of interests) using
periodically updated data or a mixture of static and streaming data. These prototypes did not take
full advantage of urban streaming data for smart and rapid resilient city management.
In this research, we propose a real-time urban analytical and visual system that can detect,
track, analyze, and visualize small-scale, undefined extreme events clustered in content and space.
The system is built on the latest versions of GeoEvent Server and Online GIS for real-time data
analysis and visualization and uses geotagged streaming Twitter data. We construct several data-
mining and natural language–processing modules within the system, including online topic
modeling and sentiment analysis using Apache Spark. The Apache Spark distributed system is
especially favorable for online big-data processing with high speed and accuracy. The system is
designed for geo-textual streaming data and has the potential to be applied to various urban
management scenarios. By leveraging high-volume urban streaming data and smart technologies,
Urban Agility and Resilience Lab
Page 4/31
we hope to demonstrate the usefulness of our system to understand the dynamics of urban systems,
especially during unpredicted perturbations such as disasters. The system demonstrates the
analysis results with interactive maps to improve situational awareness and enhance community
engagement during extreme events. The system can also be integrated into a holistic, intelligent
system to play an active role in future urban planning to achieve smart and resilient cities.
2. Related Work
Recently, extracting and interpreting information from streaming data has gained increasing
prominence in the data mining domain. In addition, social media platforms, such as Twitter, have
brought valuable user-generated behavior-rich data resources in real time, offering a growing
number of opportunities to analyze the dynamics of the text streams and topics (Benhardus &
Kalita, 2013; Ghani et al., 2019). In urban contexts, these platforms allow people to share the
events they perceive, such as nearby crises or urgent needs for specific resources (Leykin et al.,
2018; Yoshinaga & Kitsuregawa, 2014). Moreover, these crowdsourced social media data exist at
the smallest possible measurement scale and represent the perceptions and emotions of citizens
and can be used to engage citizens in two-way communication (Angelidou et al., 2018; Kitchin,
2014; Neirottie et al., 2014; Woetzel et al., 2018).
2.1 Existing methods for real-time streaming text mining and topic derivation
Many methods have been proposed for topic detection and topic evolution over time. For example,
Xie et al. (2016) proposed the TopicSketch framework for detecting bursty topics on Twitter in
real time using a sketch-based topic model based on statistical data “sketches” of tweets, such as
the acceleration of the number of tweets and words. Hasan, Orgun, and Schwitter (2019) developed
the TwitterNews+ system to detect local newsworthy events from streaming tweets. This system
continually updates the most recent tweets to determine their novelty and cluster tweets into
different events.
Other studies have considered both spatial and temporal features of tweets, making them more
applicable to urban environments. For instance, Zhang et al. (2017) proposed TrioVecEvent, an
online local event-detection method that uses geotagged tweet streams. This method generates
topic clusters and selects local events by gathering information on location, time, and content to
perform online clustering using a Bayesian mixture model. Yu et al. (2017) presented a real-time
emerging-anomaly monitoring system (Ring) to detect anomalies within minutes after the events
Urban Agility and Resilience Lab
Page 5/31
happened. This system improved on a graph-stream model and was implemented on the Spark
distributed data processing system.
These streaming data-based methods were explicitly designed to achieve real-time topic
derivation and event detection through the use of low-computation solutions and quickly updating
the computation results. These methods were implemented to detect emerging topics or geo-textual
clusters without further analyzing the detected events or enabling real-time visualization of the
computation results. However, to improve the results of exploring and interpreting massive and
complex streaming data, real-time visualization is crucial. Visualization approaches couple human
intuition with computational analysis and thus help users understand the patterns of the events and
gain insights when making decisions (Chae et al., 2014; MacEachren et al., 2010; Thom et al.,
2012).
2.2 Real-time event detection and geo-visualization systems
To address the need for real-time analysis and visualization, some systems were designed to not
only process social media data streams in real time but demonstrate the results to users
interactively. Early systems such as TwitterMonitor (Mathioudakis & Koudas, 2010) could detect
trending topics in the Twitter stream and visualize their basic features (e.g., temporal line graphs,
keywords) without mapping the topics. More subsequent research was devoted to developing geo-
visualization systems or frameworks as prototypes that could be extended to automatically detect
anomalies or events and to monitor human spatiotemporal activities in tweet streams in real time.
Examples include Terpstra et al. (2012)’s Twitcident system, Huang et al. (2017)’s cloud-based
framework for disaster monitoring, and Wachowicz et al. (2016)’s workflow of querying space-
time activities (STA) via geotagged tweet streams.
Continuing these studies, many systems with similar designs could achieve real-time data
processing and geo-visualization. A large number of these systems were rooted in geospatial
visualization and enabled real-time geo-visual analytics for geo-text aggregation, spatial cluster
exploration, and crisis discovery, such as SensePlace2 (MacEachren et al., 2010), SensePlace3
(Pezanowski et al., 2018), and ScatterBlogs (Thom et al., 2012). These methods focused on making
sense of places by extracting useful information from geo-textual streams. Some studies (e.g.,
MacEachren et al., 2010; Middleton et al., 2014) parsed location information from textual data
(e.g., hashtags, name entities, or user account location) and employed reversed geocoding.
Although the geocoding approach increased the volume of data from the social media stream that
Urban Agility and Resilience Lab
Page 6/31
could be crawled, it can be problematic when mapping fine-grained local events. Some other real-
time geo-visual analysis systems were developed to recognize the importance of spatial scale and
designed to identify small-scale or local events. For example, Boettcher & Lee (2012) proposed
the Eventradar system for detecting small-scale local events using a density-based clustering
algorithm with sliding time intervals.
In addition to underlining the geographic magnitude of real-time analysis and visualization,
some systems also introduced content-relevant features, such as incorporating human perspectives
through sentiment analysis or targeting at one type of event by setting search keywords. For
example, TwitInfo (Marcus et al., 2011) can automatically identify and label spikes of tweet events
and allows users to select and track events through a visualization platform. This platform also
shows the positive and negative sentiments surrounding events and the aggregate sentiment of the
tweets within the events. Choi & Bae (2015) introduced the Social Big Board, a real-time disaster-
monitoring system that can analyze and map disaster-related tweets and their trends. This system
also analyzes people’s emotional information using pre-defined sentiment words in positive,
negative, or neutral sentiment.
Some real-time systems have been proposed to be better for specific types of events, based on
the different characteristics of extreme urban events. For instance, Avvenuti et al. (2014)
developed a real-time alert and report system specifically for earthquake disasters (EARS). Smith
et al. (2017) presented a real-time modeling framework to identify flooding areas and infer
inundation during storm weather. Șerban et al. (2019) introduced a software system SENTINEL
that classified health-related tweets to detect disease breaks and provide syndromic surveillance in
real time. We also found that a large portion of the existing studies discuss the use of streaming
data to extract crisis-relevant information for disaster management and risk control because these
events are time-sensitive and require real-time decisions.
In summary, a diverse set of data-driven methods have been designed for mining and mapping
information from streaming data. However, only a few studies have been able to pinpoint the
locations and scales of detected events and their evolution over time and space at a fine-grained
resolution. Existing real-time systems that incorporate both computational analysis and
visualization to identify emerging events only show some basic features of the events (e.g., tweet
volume change, representative tweets, and spatial clusters), but in-depth information extraction
and undefined event detection (without using predefined keywords) are lacking. Still fewer studies
Urban Agility and Resilience Lab
Page 7/31
have initiated early alerts or other actions after the systems detected emergencies or crises.
3. Developing a Real-Time Urban Analytical and Geo-Visual System
In this section, we demonstrate the design and methods of our real-time urban analytical and geo-
visual system that works for streaming geo-textual data (i.e., data with both geographical and
content features). This system is built upon Esri’s GeoEvent Server (version 10.7). The server
provides comprehensive tools and pipelines to support high-volume real-time data input,
processing, and output, making it especially suitable for geospatial streaming data. We employ
several tools provided by the server and create our customized data-processing tools with the
server’s Java SDK (Software Development Kit).
Figure 3.1 demonstrates the architecture of our proposed system, with analytical modules
embedded in the GeoEvent Server. Three analytical and visual modules make up the core of the
system: streaming data input, streaming data processing and analysis, and data output and real-
time geo-visualization. This system is supported by a large geodatabase - PostgreSQL. The input
module ingests streaming data directly from a Twitter streaming API and transforms the stream of
information into formatted raw tweets. This module also performs simple tweet screening:
collecting geotagged non-bot tweets from within defined spatial areas. The processing and analysis
module serves as the primary analytical part of the system. In this module, we develop customized
functional models (configured as processors in the server) to extract information of interest from
the data stream. The output and geo-visualization module is our platform for real-time mapping
and interactions using Online GIS maps.
Figure 3.1. The architecture and main modules of the real-time analytical and geo-visual system
3.1 Streaming data input module
Urban Agility and Resilience Lab
Page 8/31
Tweets generated during extreme events contain rich geographical and content information that is
useful for city management. Twitter offers an open API for collecting large amounts of voluntarily
reported tweets in real time. In general, about 1% of tweets can be collected by a standard account
(Wang et al. 2017; Wang & Taylor 2019). In this module, two tools are used to process the
incoming data stream into a format suitable for the GeoEvent server so that the data can be further
analyzed. The transport tool connects to the Twitter API and receives data as a raw byte stream,
and the adapter tool converts this stream into formatted tweet objects that can be processed by the
system. Each collected tweet is configured as a geo-textual object that contains both the text of the
tweet and its metadata, such as unique ID, user ID, timestamp, latitude, and longitude.
Then we set up a filter tool to remove tweets generated by robot accounts (bots), collect
geotagged tweets, and define the geographic study area using GeoFences in the server. Figure 3.2
shows an example of our filter settings. We refine a word list from a previous study that contains
terms usually used by bots, such as “temp” and “barometer(Yao & Wang, 2020a), and we use
regular expressions to match the incoming raw tweets and exclude any that contain the words from
the list. We also exclude tweets without geolocations and filter the tweets to select those from a
predefined study area using GeoFences settings. GeoFences can determine the spatial relationship
between a geographic boundary (polygon) and a geotagged tweet. We do not set keywords to filter
tweets, as we intend to design a system for detecting unexpected extreme events that meet our
design assumptions. We develop the filter tool that processes data in a computationally simple and
fast way in order to remove unnecessary data volume for the computationally expensive process
in the following steps, thus enhancing the speed and efficiency of the system.
Figure 3.2. The interface of filter settings
Urban Agility and Resilience Lab
Page 9/31
3.2 Streaming data process and analysis module
This analytical module analyzes a stream of geotagged tweets and generates the outcomes for a
dynamic map. We first apply the online topic model to identify the topic distribution for each tweet
in the stream. Then we use a clustering method to synthesize the topics and generalize geo-topics
using additional geospatial information embedded in the tweets. In this study, geo-topics are topics
clustered in space that are used to represent urban events, such as urban activities, local news, and
emergencies. The assumption is that if tweets are collected in a short period and contain similar
words and topics, they are likely to be clustered spatially and related to a specific local event
(Wang & Taylor, 2019). In the third step, we compute the sentiment score of each tweet and
average the sentiment score of each geo-topic. These scores can be used as indicators of potential
extreme events, such as emergencies or crises. We use the Java SDK provided by the GeoEvent
Server to create customized processors corresponding to these data-analysis methods.
3.2.1 Topic modeling of streaming text data based on Online LDA
One major challenge for real-time data analysis is handling the sheer volume of data rapidly. Topic
modeling is a data-mining method for discovering hidden semantic structures (topics) in large text
documents, such as tweets. The most commonly used topic models are probabilistic ones, such as
probabilistic latent semantic indexing (PLSI) (Hofmann, 1999) and latent Dirichlet allocation
(LDA) (Blei et al., 2003). These models represent each document as a mixture of topics and each
topic as a distribution over words. Most probabilistic models run offline and do not incorporate
the temporal aspect of documents (Gao et al., 2020; Yu et al., 2017). Some methods began to
exploit timestamps jointly with topic detection and topic evolution, such as Topic Over Time
(TOT) (Wang & McCallum, 2006) and Dynamic Topic Model (DTM) (Blei & Lafferty, 2006). In
social media environments, tweets arrive continuously and topics change dynamically. Topic
models that to be used online or in real time must consider this temporal aspect and be updated to
capture topic changes over time rapidly.
We employ Online LDA (Hoffman et al., 2010), an online variational Bayes algorithm for
LDA, to process streaming data and generate topics. Variational Bayes (VB) is a method of
variational inference used in a Bayesian hierarchical model. VB is based on Bayes’ theorem and
is used to approximate the true posterior by minimizing the Kullback-Leibler divergence to the
true posterior. Online LDA has advantages in handling massive collections of documents because
the method allows documents to be examined as they arrive in-stream and discarded after a single
Urban Agility and Resilience Lab
Page 10/31
look to reduce the delays. The Online LDA model is defined by some essential parameters:
(0.5, 1] is an exponential decay that controls the rate at which old topics are forgotten; its value
range can guarantee an asymptotic convergence. 0 represents a positive learning offset that
slows down the early iterations of the algorithm. Minibatch is used to consider multiple
observations per update to reduce noise. We also need to set the number of topics before running
Online LDA.
We use Apache Spark’s machine learning library (MLlib) in the Java programming language
to implement Online LDA. Apache Spark is an open-source distributed analytics system designed
for big data processing that has exceptionally high performance. We choose Java to meet GeoEvent
Server’s requirements so that we can create a processor with Java SDK. Before running the online
topic modeling, we clean and normalize the texts of tweets using the SparkNLP library and MLlib:
we remove web links, @ mentions, and stopwords, make all the words lowercase, and tokenize
the texts into single terms (Figure 3.3).
Figure 3.3. The process of data cleaning and normalization
3.2.2 Generating geo-topics with spatial features
Because topics in the data stream change over time, they cluster dynamically across space as geo-
topics, and our system is designed to monitor these spatial changes. The system uses sliding time
windows with statistical metrics to analyze the spatiotemporal dynamics of geo-topics. Sliding
time windows are often used in the online mode of topic modeling (e.g., Boettcher & Lee, 2012;
Lau, Collier, & Baldwin, 2012). Figure 3.4 shows the spatial clustering patterns of tweets for
generating geo-topics based on sliding time windows. We set the window size to one hour, and
when new tweets arrive, the window moves forward, so geotagged tweets received more than an
hour ago are discarded. All the tweets within the window are valid, and only valid tweets are
aggregated and clustered to generate geo-topics and have their corresponding features (centers and
Urban Agility and Resilience Lab
Page 11/31
ranges) calculated. As the window slides onward, the valid tweets and features of geo-topics
change. We use statistical metrics to represent the spatial centers and ranges of geo-topics. For
each geo-topic (= 0,1,2, … ), the spatial center (
,
) is represented by the arithmetic mean
of the latitudes and longitudes of the valid tweets on that geo-topic ,
,[1, ]:
=1
,
=1


where
is the latitude and
is the longitude of the center of geo-topic , and is the number
of tweets belonging to geo-topic . The geographic range of a geo-topic is the variance in the
tweets’ latitude or longitude converted to distance in kilometers:
=

 ,


where and are the coefficients for converting degrees of latitude and longitude into kilometers.
Variance can measure how geotagged tweets spread out from their geographic center. If a geo-
topic contains only one valid tweet, its range value is 0 (zero).
Figure 3.4. Demonstration of spatial clustering patterns of geo-topics in streaming data
3.2.3 Domain-specific sentiment analysis for extreme event detection
We calculate the sentiment scores of geotagged tweets and geo-topics to identify potential extreme
events that merit attention. This analytical procedure is based on the assumption that very negative
sentiments are likely to indicate crises or accidents (Caragea et al., 2014; Lu et al., 2015). We
employ a pre-trained, domain-adapted sentiment-analysis classifier to predict the sentiment score
of each tweet in the data stream (Yao & Wang, 2020b). The sentiment classifier represents a
domain-adversarial neural network (DANN) (Ganin et al., 2016) that is built on a recurrent neural
network (RNN) with an additional domain-adversarial component. RNN is specifically useful in
Urban Agility and Resilience Lab
Page 12/31
text mining and processing sequence data, such as streaming texts. The domain-adversarial
component is appended to a standard RNN learning process in the backpropagation steps, thus the
learned representations are invariant across different domains. This method can achieve high
accuracy in classification and performs robustly in distinct domains (e.g., disasters, news, and
lifestyles) when analyzing tweets. This domain-adaption feature makes the method exceptionally
suitable for streaming tweets and changing geo-topics. The DANN method calculates the
sentiment scores of individual tweets as real numbers from –2 (most negative) to 2 (most positive),
with 0 as neutral sentiment. The sentiment score of each geo-topic is the arithmetic mean of
sentiment values of the valid tweets belonging to the geo-topic.
We then use an emergency-detector tool to find abnormally negative geo-topics and provide
early alerts when potential emergencies are identified. This tool works by setting up a sentiment
threshold, usually a negative number between -2 and 0, to trigger the alerts. If the sentiment of a
geo-topic at any time is equal to or below the threshold, the system generates a potential emergency
alert and indicates it dynamically on the map. After the alert is detected, its status becomes
ongoing, and the tool tracks the total duration of the alert. If the sentiment increases to above the
threshold, the alert stops and the ongoing status ends.
3.3 Data output and real-time geo-visualization module
In the final output module, the geotagged tweets and the analytical results are stored as tables in
the geodatabase. The analytical results are also visualized via a cloud-based GIS mapping platform
- ArcGIS Online. These results include the geo-topics with their spatial, temporal, textual semantic,
and sentiment features. The online maps can display and monitor changes in such features of
detected urban events over both space and time. The maps can also show potential emergency
alerts and their locations and status changes. The system employs a large-scale geospatial database
(PostgreSQL) to support real-time geospatial data analysis and visualization. The PostgreSQL
geodatabase stores and manages a collection of geographic datasets based on a relational database
management system (RDBMS). During the real-time visualization process, the geodatabase stores
maps published by ArcGIS as a feature service. This feature service contains datasets with spatial
information that can be used to generate map layers. The datasets are updated simultaneously when
our system is running. The feature service also contains GIS map templates with symbology that
is used by Online GIS for data visualization.
4. Experiments and Results
Urban Agility and Resilience Lab
Page 13/31
4.1 Study case and system settings
We applied the proposed visual urban analytical system in a simulated real-time scenario of
Hurricane Harvey, one of the most destructive disasters to happen in the U.S. in the past decade.
The hurricane caused more than a hundred billion in damage and made landfall in a densely
populated area in south-central Texas. We use geotagged tweets collected by a Twitter streaming
API. Our study period ran from August 18 to September 12, 2017, covering the time before
(August 18 to 24), during (August 25 to 26), and after (August 27 to September 12) the disaster.
Our study area was the counties spatially overlapped by the wind swaths of Hurricane Harvey. The
area was set as the GeoFences polygon in our system. In general, our proposed system can provide
the GeoEvent Server interface for collecting real-time streaming tweets using the adapter,
transport, and filter described in Section 3.1 (Figure 4.1). For this case study, we simulated the
collected tweets at a real rate. The rate was computed from the tweets’ timestamps using the
simulator in the GeoEvent Server (Figure 4.2). This simulation can precisely capture the temporal
features of tweets and mirror the real-time collection of streaming data. In addition, we set the
topic number to 100 for online topic modeling to cover the most possible topics. We also set the
sentiment threshold to –0.15 to detect potential emergency alerts.
Figure 4.1. The interface of GeoEvent Server for managing data and analysis
Urban Agility and Resilience Lab
Page 14/31
Figure 4.2. The interface of GeoEvent Simulator
4.2 Geo-visualization of the analytical system
4.2.1 User interface
The video (uploaded as supplementary “Video Still”) and the screenshots below (Figure 4.3 and
Figure 4.4) show the user interface of our system and the analysis results for Hurricane Harvey.
The center is the online map showing the analysis results in real time and refreshing every few
seconds. The light pink polygons are county boundaries of the study area. The circles are local
urban events represented by geo-topics detected by our system. The size of each circle represents
the affected range of the geo-topic: the larger the radius, the greater the affected area. The colors
of the circles represent the average sentiment scores of different events: the darker the color, the
more negative the geo-topic is. The map also shows potential emergency alerts. The red symbol
represents ongoing emergencies and the green one represents emergencies that have ended. The
left side of the user interface shows the legend for the map, including the study area county
boundaries, spatial ranges, sentiment scores of urban events, and potential emergency alerts. The
bottom of the map is the table area, which shows detailed information on individual map layers.
For example, the urban event layer updates the event number, the location of the event center with
latitude and longitude, the number of tweets representing the events, the average sentiments of the
events, the approximate times the events appeared, and the spatial ranges of the events. The
Urban Agility and Resilience Lab
Page 15/31
emergency alert layer updates the ID, description, status, duration, and condition of each alert. Our
system also uses pop-up windows (Figure 4.3). Users can click urban events or potential
emergency alerts on the map to check all the information provided in the bottom table.
(A) Urban events pop-up window
(B) Potential emergency alerts pop-up window
Figure 4.3. The user interface with pop-up windows
4.2.2 Spatiotemporal patterns of sentiment
We examined the general spatiotemporal patterns of sentiment for detected urban events. Figure
4.4 shows screenshots of urban events (geo-topics) on different dates. Before the hurricane made
landfall (Figure 4.4 A on August 21), the overall sentiment of the geo-topics was rather positive
(see lighter colors of geo-topics). The negative sentiment was mainly distributed around the
Urban Agility and Resilience Lab
Page 16/31
metropolitan areas of Houston, Austin, and San Antonio. During the hurricane (Figure 4.4 B on
August 25), the sentiment became more negative, shown by darker colors on the map. The spatial
area of negative sentiment also expanded to surrounding areas, such as the coast. After the
hurricane (Figure 4.4 C on August 28), the sentiment was even more negative, and the spatial area
was more sprawled out because of the accumulated damage and bad weather.
(A) before the hurricane (August 21, 2017)
(B) during the hurricane (August 25, 2017)
Urban Agility and Resilience Lab
Page 17/31
(C) after the hurricane (August 28, 2017)
Figure. 4.4. Geo-visualization of sentiment on different dates
We also found that the Houston metropolitan area had the highest number of potential
emergency alerts during the hurricane. These included traffic delay (geo-topic #17), bayou flood
flow (geo-topic #62), and heavy rain that might affect citizens’ health (geo-topic #81). San Antonio
had three traffic-related emergencies on three separate days (e.g., geo-topic #23). In addition, the
detected potential emergency alerts also indicated the emergency locations reasonably well. For
example, traffic-related emergencies were found near major roads, and bayou flood emergencies
were found near Buffalo bayou. We also found that traffic delays and accidents were the most
common and long-lasting emergencies during the hurricane (geo-topic #17, 23, 44). This type of
emergency can affect social media for a period from half an hour to almost nine hours. Our system
found 12 potential emergencies at different times and locations during the hurricane. We detected
one false alarm (geo-topic #85), about fitness and training, and it lasted 25 minutes. The precision
of emergency detection in this study case is 91.67%.
4.3 System performances and data statistical features
4.3.1 Streaming data volume and temporal patterns
Our system processed about 2,707,346 tweets during the study period. The daily numbers of tweets
before and after the data filtering are shown in Figure 4.5. Before the filtering, the Twitter
streaming API received 100,000 to 140,000 tweets a day. The peak throughput (volume) of our
system is 147,589 tweets per day, and its average throughput is 117,710.
Urban Agility and Resilience Lab
Page 18/31
Figure 4.5. Daily data volume processed by the system
Focusing on the specifically hurricane-affected area, the number of clean tweets (no bots and
geotagged) had a maximum of 5,439 and an average of 3,189 per day. These tweets were
considered useful for improving situational awareness during disasters. The cleaned tweets flowed
to the following modules for analysis and visualization. After the hurricane made landfall on
August 25, Twitter activity measured by the number of clean tweets increased dramatically and
reached its peak on August 28, then became relatively stable after the disaster.
Figure 4.6 shows the hourly change in Twitter activity over the study period. The clean tweets
increased and reached a peak then decreased every day, both before and after the hurricane. During
the hurricane-affected period (after August 25 until September 1), the daily peaks were greater
than before or after the hurricane. The hourly peak throughput of cleaned data for the subsequent
analysis was 401 tweets, and the average number was 142 tweets.
Urban Agility and Resilience Lab
Page 19/31
Figure 4.6. Hourly clean data volume processed by the system
4.3.2 Feature patterns of geo-topics
Our system generated geo-topics with features such as centers, ranges, and sentiment scores over
Urban Agility and Resilience Lab
Page 20/31
time. The results showed that the system can detect a wide range of topics, including disaster-
relevant events and daily life events, and monitor their features over time. Although many disaster-
relevant events did not trigger emergency alerts, they still provided rich enough information to
improve situational awareness.
We aggregated the daily temporal patterns of several geo-topics with counts of their
represented geotagged tweets (Figure 4.7). Table 4.1 lists a few of these geo-topics and their top
keywords, which have the highest probabilities of representing those geo-topics. For example,
during the hurricane period, our system detected multiple hurricane-relevant geo-topics (e.g., geo-
topic #2, 32, and 54) with keywords such as “hurricane”, “Harvey, “floodand “accident”, and
multiple locations of hurricane-affected areas such as “Austin, “San Antonio, “downtown, or
“highway”. Geo-topic #2 indicated flash flooding and stormwater conditions during the hurricane,
geo-topic #54 was related to traffic accidents on the highways, and geo-topic #32 was about the
general situation of Hurricane Harvey. The system also found that such geo-topics reached their
volume peaks in the numbers of represented tweets during the hurricane-affected period (from
August 25 to August 29) and then gradually disappeared afterward.
Our system also detected some hurricane-irrelevant events and tracked their changes. For
example, geo-topic #31 had keywords such as “birthday”, “bar, and “restaurant”. The volumes of
such geo-topics were relatively stable over time, unlike hurricane-relevant geo-topics that reached
peak volume. We also found that some geo-topics appeared only in parts of the study period. For
example, geo-topic #88 with keywords such as “jog”, “partner”, and “produce” appeared only after
the hurricane.
Table 4.1 Selected geo-topics of Hurricane Harvey and their top keywords
Geo-topic
Top keywords
2
flood, flash, county, report, free, warning, giveaway, include, storm, public
31
birthday, drink, happy, college, great, bar, photo, food, restaurant, show
32
harvey, today, come, sanantonio, hurricaneharvey, time, hurricane, last, austin,
tonight
45
tornado, warning, lafayette, continue, lakes, city, mission, parish, weston,
brookshire
54
traffic, stop, accident, lane, delay, hwy, block, high, water, rock
71
austin, nowplaying, pool, palmillabeach, atx, downtown, livingston, near,
Urban Agility and Resilience Lab
Page 21/31
airport, long
Figure. 4.7. Temporal patterns of selected geo-topics and the count of their represented tweets
We also analyzed the sentiment changes on each geo-topic over time. To further demonstrate
the sentiment patterns and their relationships to different geo-topics, we selected several geo-topics
that represented hurricane-relevant and -irrelevant events and then compared their sentiment
patterns over time (Figure 4.8). We used the minimum sentiment scores among the tweets
belonging to the geo-topic because minimum scores revealed more apparent temporal trends. We
found that hurricane-relevant geo-topics (Figure 4.8 A) had more negative sentiment overall
throughout the study period, and the most negative sentiments during the period when the
hurricane hit the study area. By contrast, hurricane-irrelevant geo-topics (Figure 4.8 B) had more
positive and stable sentiment and were affected less by the hurricane.
Urban Agility and Resilience Lab
Page 22/31
(A) Hurricane relevant geo-topics
(B) Hurricane irrelevant geo-topics
Figure. 4.8. Temporal patterns of sentiment for selected geo-topics
5. Discussion
Urban Agility and Resilience Lab
Page 23/31
In this paper, we have presented an urban analytical and geo-visual system that automatically
collects geotagged tweets and performs urban event detection and visualization in real time.
Through the application in the simulated streaming data from a large-scale hurricane, the system
has been demonstrated to provide useful and timely information on emergencies and crises during
disasters. The system is specifically designed to address the smart rapidity aspect of resilient cities
by enabling real-time analysis and geo-visualization to decipher the dynamics of urban
environments and systems. Early detection and tracking of the urban events help provide early
alerts to the residents and assist city managers and first responders to rapidly respond to the adverse
effects.
Processing and analyzing social media streaming data in real time is exceptionally challenging
because the speed and volume of data stream require methods to update rapidly to capture the topic
changes (Goyal et al., 2019; Hasan et al. 2018; Nugroho et al. 2020). Our proposed system
improves on previous research by implementing online data-analysis methods that enable the rapid
detection and processing of streaming data (Yao & Wang, 2020a). We used a variational inference
methodOnline LDA (Hoffman et al., 2010)—to process and infer the topics of incoming data
using Bayes’ theorem. In general, topic models can be computationally complex and time-
consuming (Xie et al., 2016; Yu et al., 2017). The system reduces the redundant computation for
topic modeling by applying multiple preprocessing methods that are much faster than topic
modeling. By filtering and cleaning streaming raw tweets, and only triggering the later topic
modeling and sentiment analysis functions when necessary, we improve the accuracy and
efficiency of the overall system. We also exploit the Apache Spark distributed analytics system to
improve the scalability and speed of data processing. All these methods and settings can help our
system increase data throughput and reduce latency in data processing to achieve a real-time
analysis and visualization system.
Our developed system can continuously track multidimensional information from the data
stream, such as location, timestamp, semantics, and sentiment changes of detected events. This
property expanded on previous research that offered simple place or placetime information
reports (e.g., Bifet et al., 2011; MacEachren et al., 2010; Wachowicz et al., 2016) or spatial
clustering results (Pezanowski et al., 2018; Smith et al., 2017). Some earlier methods used bursty
keywords as prerequisites for event detection and tracking (Avvenuti et al., 2014; Li et al., 2012).
Our system does not set keywords for event detection, which gives it the potential to detect a
Urban Agility and Resilience Lab
Page 24/31
variety of unforeseeable urban events related to disasters and other unexpected perturbations.
Instead of using keywords, our system employs sentiment analysis to automatically identify
potential emergencies or crises at an early stage. Sentiment analysis in existing systems was
classified into two or three categories and demonstrated only in basic statistical features, such as
the percentage of each category of sentiment (Choi & Bae, 2015; Marcus et al., 2011). Our system
can calculate sentiments in finer-gradations and can be used to trigger early alerts about potential
emergencies to help with decision-making.
Although we have used the system to demonstrate one study case, the system is open to other
data resources and application scenarios due to its flexible and adaptable design. Supporting by
the GeoEvent Server, the system can ingest streaming API from different resources and provide
additional analysis by properly changing the adapters, filters, or processors within the system
modules. The users of our system can also customize the geo-visualization effects (colors, legends,
base maps) as needed through Online GIS maps. Thus, our system is advantageous for
transforming extracted information into a broad spectrum of applications, ranging from extreme
events such as disaster management, epidemics tracking, and crime monitoring to business-as-
usual situations such as place recommendation (Gao et al., 2020; Ghani et al., 2019; Nugroho et
al., 2020; Pezanowski et al., 2018).
Additionally, our system uses a dynamic online map to visualize the latest analysis results
update every few seconds. Managing extreme events in cities can be complicated, users of our
system may not be familiar with data analysis but still need specific information to improve
situational awareness, allocate resources, or take action. This geo-visualization feature allows
users to learn intuitively about important events happening in their areas and whether people
should be aware of those events during unpredictable extreme situations. Using crowdsourced
social media data generated by the public, our system also provides citizens opportunities to be
actively involved in the system and promotes community engagement.
However, the proposed system has several limitations that can be addressed in our future
research. Currently, the system uses Twitter as its only data source. However, people have
differencing preferences in using social networking platforms. Future systems can consider
streaming data from multiple sources to become more integrated and reduce the data bias caused
by single data source. Additionally, the proposed system performs streaming data analyses based
on pre-defined parameters, such as the number of topics and the sentiment threshold. These
Urban Agility and Resilience Lab
Page 25/31
parameters need to be set case by case due to the complexity in contents and languages online.
Future studies will focus on testing and tuning parameters across urban events with different study
areas and periods, and propose a method for system users to set parameters for best performances.
Lastly, it is challenging to compare the outcomes (e.g., small-scale local events) of the proposed
system with the ground truth because these detected events may not be reported by officials, which
makes the data unavailable. Although the system is constructed with well-developed data analysis
methods, the usage of social media data requires further credibility checks. The future system can
also be improved with additional event analysis modules when increasing the volume and
versatility of streaming data becomes available.
6. Conclusion
Building resilient cities requires smart solutions, and achieving smart rapidity is one of the most
important approaches to enhancing urban resilience when promoting smart cities. We develop a
real-time urban analytical and geo-visual system for social media streaming data to track small-
scale undefined urban extreme events and provide early emergency alerts. The system has
demonstrated the effectiveness and rapidity in processing large volumes of data with low latency.
The system has the potential to incorporate streaming data from more sources and to be involved
in cities’ emergency management tasks, such as improving situational awareness, assisting rapid
damage assessments, monitoring emergent incidents, and supporting collaborative decision-
making for multiple stakeholders. The research also contributes to developing smart city
technologies that can be integrated into holistic urban surveillance systems and achieving more
safe, resilient, and smart future cities.
Acknowledgements
This material is based upon work supported by the National Science Foundation under Grant No.
1760645, Grant No. 2028012, and the faculty start-up fund at University of Florida. Any
opinions, findings, and conclusions or recommendations expressed in this material are those of
the authors and do not necessarily reflect the views of the National Science Foundation or
University of Florida.
References:
Urban Agility and Resilience Lab
Page 26/31
Al Nuaimi, E., Al Neyadi, H., Mohamed, N., & Al-Jaroodi, J. (2015). Applications of big data to smart
cities. Journal of Internet Services and Applications, 6(1), 115.
Allam, Z., & Newman, P. (2018). Redefining the Smart City: Culture, Metabolism and Governance. Smart
Cities, 1(1), 425.
Angelidou, M., Psaltoglou, A., Komninos, N., Kakderi, C., Tsarchopoulos, P., & Panori, A. (2018).
Enhancing sustainable urban development through smart city applications. Journal of Science and
Technology Policy Management, 9(2), 146169.
Arafah, Y., & Winarso, H. (2017). Redefining smart city concept with resilience approach. IOP Conference
Series: Earth and Environmental Science, 70(1), 12065.
Avvenuti, M., Cresci, S., Marchetti, A., Meletti, C., & Tesconi, M. (2014). EARS (earthquake alert and
report system): A real time decision support system for earthquake crisis management. Proceedings
of the ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 1749
1758.
Batty, M. (2008). The size, scale, and shape of cities. Science, 319(5864), 769771.
Benhardus, J., & Kalita, J. (2013). Streaming trend detection in Twitter. International Journal of Web Based
Communities, 9(1), 122139.
Bifet, A., Holmes, G., & Pfahringer, B. (2011). MOA-TweetReader: Real-time analysis in twitter streaming
data. International Conference on Discovery Science, 46–60.
Blei, D. M., & Lafferty, J. D. (2006). Dynamic topic models. ACM International Conference Proceeding
Series, 148, 113120.
Blei, D. M., Ng, A. Y., & Jordan, M. I. (2003). Latent Dirichlet allocation. Journal of Machine Learning
Research, 3, 9931022.
Boettcher, A., & Lee, D. (2012). EventRadar: A real-time local event detection scheme using Twitter
stream. IEEE International Conference on Green Computing and Communications, 358367.
Bruneau, M., Chang, S. E., Eguchi, R. T., Lee, G. C., O’Rourke, T. D., Reinhorn, A. M., Shinozuka, M.,
Tierney, K., Wallace, W. A., & Von Winterfeldt, D. (2003). A Framework to Quantitatively Assess
and Enhance the Seismic Resilience of Communities. Earthquake Spectra, 19(4), 733752.
Caragea, C., Squicciarini, A., Stehle, S., Neppalli, K., & Tapia, A. (2014). Mapping moods: Geo-mapped
sentiment analysis during hurricane sandy. ISCRAM 2014 Conference Proceedings, 642651.
Chae, J., Thom, D., Jang, Y., Kim, S., Ertl, T., & Ebert, D. S. (2014). Public behavior response analysis in
disaster events utilizing visual analytics of microblog data. Computers and Graphics (Pergamon),
38(1), 5160.
Choi, S., & Bae, B. (2015). The real-time monitoring system of social big data for disaster management. In
Computer Science and its Applications, 809-815. Springer, Berlin, Heidelberg.
Urban Agility and Resilience Lab
Page 27/31
Desouza, K. C., & Flanery, T. H. (2013). Designing, planning, and managing resilient cities: A conceptual
framework. Cities, 35, 8999.
Desroches, R., & Taylor, J. E. (2018). The promise of smart and resilient cities. Bridge, 48(2), 1320.
Ganin, Y., Ustinova, E., Ajakan, H., Germain, P., Larochelle, H., Laviolette, F., Marchand, M., &
Lempitsky, V. (2016). Domain-Adversarial Training of Neural Networks. Journal of Machine
Learning Research, 17, 135.
Gao, W., Peng, M., Wang, H., Zhang, Y., Han, W., Hu, G., & Xie, Q. (2020). Generation of topic evolution
graphs from short text streams. Neurocomputing, 383, 282294.
Ghani, N. A., Hamid, S., Hashem, I. A. T., & Ahmed, E. (2019). Social media big data analytics: A survey.
Computers in Human Behavior, 101, 417-428.
Godschalk, D. R. (2003). Urban hazard mitigation: Creating resilient cities. Natural Hazards Review, 4(3),
136–143.
Goyal, P., Kaushik, P., Gupta, P., Vashisth, D., Agarwal, S., & Goyal, N. (2019). Multilevel Event
Detection, Storyline Generation, and Summarization for Tweet Streams. IEEE Transactions on
Computational Social Systems, 7(1), 8-23.
Hasan, M., Orgun, M. A., & Schwitter, R. (2018). A survey on real-time event detection from the Twitter
data stream. Journal of Information Science, 44(4), 443463.
Hasan, M., Orgun, M. A., & Schwitter, R. (2019). Real-time event detection from the Twitter data stream
using the TwitterNews+ Framework. Information Processing and Management, 56(3), 11461165.
Hatuka, T., Rosen-Zvi, I., Birnhack, M., Toch, E., & Zur, H. (2018). The Political Premises of
Contemporary Urban Concepts: The Global City, the Sustainable City, the Resilient City, the
Creative City, and the Smart City. Planning Theory and Practice, 19(2), 160179.
Hoffman, M. D., Blei, D. M., & Bach, F. (2010). Online learning for Latent Dirichlet Allocation. Advances
in Neural Information Processing Systems, 856-864.
Hofmann, T. (1999). Probabilistic latent semantic indexing. Proceedings of the 22nd Annual International
ACM SIGIR Conference on Research and Development in Information Retrieval, 5057.
Huang, Q., Cervone, G., & Zhang, G. (2017). A cloud-enabled automatic disaster analysis system of multi-
sourced data streams: An example synthesizing social media, remote sensing and Wikipedia data.
Computers, Environment and Urban Systems, 66, 2337.
Khan, Z., Anjum, A., Soomro, K., & Tahir, M. A. (2015). Towards cloud based big data analytics for smart
future cities. Journal of Cloud Computing, 4(1), 1-11.
Kitchin, R. (2014). The real-time city? Big data and smart urbanism. GeoJournal, 79(1), 114.
Klein, B., Koenig, R., & Schmitt, G. (2017). Managing Urban Resilience: Stream Processing Platform for
Responsive Cities. Informatik-Spektrum, 40(1), 3545.
Urban Agility and Resilience Lab
Page 28/31
Kontokosta, C. E., & Malik, A. (2018). The Resilience to Emergencies and Disasters Index: Applying big
data to benchmark and validate neighborhood resilience capacity. Sustainable Cities and Society,
36, 272285.
Lau, J. H., Collier, N., & Baldwin, T. (2012). On-line trend analysis with topic models: Twitter trends
detection topic model online. 24th International Conference on Computational Linguistics -
Proceedings of COLING 2012: Technical Papers, 1519–1534.
Leichenko, R. (2011). Climate change and urban resilience. Current Opinion in Environmental
Sustainability, 3(3), 164168.
Leykin, D., Lahad, M., & Aharonson-Daniel, L. (2018). Gauging urban resilience from social media.
International Journal of Disaster Risk Reduction, 31, 393402.
Li, C., Sun, A., & Datta, A. (2012). Twevent: Segment-based event detection from tweets. ACM
International Conference Proceeding Series, 155–164.
Lu, Y., Hu, X., Wang, F., Kumar, S., Liu, H., & Maciejewski, R. (2015). Visualizing social media sentiment
in disaster scenarios. WWW 2015 Companion - Proceedings of the 24th International Conference
on World Wide Web, 1211–1215.
MacEachren, A., Robinson, A. C., Jaiswal, A., Pezanowski, S., Savelyev, A., Blanford, J., & Mitra, P.
(2010). Geo-Twitter Analytics: Applications in Crisis Management. Proceedings of the 25th
International Cartographic Conference, 18.
Malik, K. R., Sam, Y., Hussain, M., & Abuarqoub, A. (2018). A methodology for real-time data
sustainability in smart city: Towards inferencing and analytics for big-data. Sustainable Cities and
Society, 39, 548556.
Marcus, A., Bernstein, M. S., Badar, O., Karger, D. R., Madden, S., & Miller, R. C. (2011). TwitInfo:
Aggregating and visualizing microblogs for event exploration. Conference on Human Factors in
Computing Systems - Proceedings, 227236.
Mathioudakis, M., & Koudas, N. (2010). TwitterMonitor: Trend detection over the twitter stream.
Proceedings of the ACM SIGMOD International Conference on Management of Data, 11551157.
McGowan, C.J., Biggerstaff , M., Johansson, M. et al (2019). Collaborative efforts to forecast seasonal
influenza in the United States, 20152016. Scientific Reports, 9(1).
Meerow, S., Newell, J. P., & Stults, M. (2016). Defining urban resilience: A review. Landscape and Urban
Planning, 147, 3849.
Middleton, S. E., Middleton, L., & Modafferi, S. (2014). Real-time crisis mapping of natural disasters using
social media. IEEE Intelligent Systems, 29(2), 917.
Neirotti, P., De Marco, A., Cagliano, A. C., Mangano, G., & Scorrano, F. (2014). Current trends in smart
city initiatives: Some stylised facts. Cities, 38, 2536.
Urban Agility and Resilience Lab
Page 29/31
Nugroho, R., Paris, C., Nepal, S., Yang, J., & Zhao, W. (2020). A survey of recent methods on deriving
topics from Twitter: algorithm to evaluation. Knowledge and Information Systems. 1-35.
Palmieri, F., Ficco, M., Pardi, S., & Castiglione, A. (2016). A cloud-based architecture for emergency
management and first responders localization in smart city environments. Computers and Electrical
Engineering, 56, 810830.
Pezanowski, S., MacEachren, A. M., Savelyev, A., & Robinson, A. C. (2018). SensePlace3: a geovisual
framework to analyze placetimeattribute information in social media. Cartography and
Geographic Information Science, 45(5), 420437.
Platt, S., Brown, D., & Hughes, M. (2016). Measuring resilience and recovery. International Journal of
Disaster Risk Reduction, 19, 447460.
Psyllidis, A., Bozzon, A., Bocconi, S., & Titos Bolivar, C. (2015). A platform for urban analytics and
semantic data integration in city planning. Communications in Computer and Information Science,
527, 2136.
Șerban, O., Thapen, N., Maginnis, B., Hankin, C., & Foot, V. (2019). Real-time processing of social media
with SENTINEL: A syndromic surveillance system incorporating deep learning for health
classification. Information Processing & Management, 56(3), 1166-1184.
Smith, L., Liang, Q., James, P., & Lin, W. (2017). Assessing the utility of social media as a data source for
flood risk management using a real-time modelling framework. Journal of Flood Risk
Management, 10(3), 370–380.
Soyata, T., Habibzadeh, H., Ekenna, C., Nussbaum, B., & Lozano, J. (2019). Smart city in crisis:
Technology and policy concerns. Sustainable Cities and Society, 50, 101566.
https://doi.org/10.1016/j.scs.2019.101566
Terpstra, T., Stronkman, R., De Vries, A., & Paradies, G. L. (2012). Towards a real-time Twitter analysis
during crises for operational crisis management. ISCRAM 2012 Conference Proceedings. 1-9.
Thom, D., Bosch, H., Koch, S., Worner, M., & Ertl, T. (2012). Spatiotemporal anomaly detection through
visual analysis of geolocated Twitter messages. IEEE Pacific Visualization Symposium 2012,
PacificVis 2012 - Proceedings, 4148.
UN DESA Population Division. (2018). The world’s cities in 2018. Retrieved from:
https://www.un.org/en/events/citiesday/assets/pdf/the_worlds_cities_in_2018_data_booklet.pdf
Wachowicz, M., Arteaga, M. D., Cha, S., & Bourgeois, Y. (2016). Developing a streaming data processing
workflow for querying space–time activities from geotagged tweets. Computers, Environment and
Urban Systems, 59, 256268.
Wang, X., & McCallum, A. (2006). Topics over Time: A non-markov continuous-time model of topical
trends. Proceedings of the ACM SIGKDD International Conference on Knowledge Discovery and
Urban Agility and Resilience Lab
Page 30/31
Data Mining, 2006, 424433.
Wang, Y., Wang, Q., & Taylor, J. E. (2017). Aggregated responses of human mobility to severe winter
storms: An empirical study. PloS ONE, 12(12), e0188734.
Wang, Y., & Taylor, J. E. (2019). DUET: Data-Driven Approach Based on Latent Dirichlet Allocation
Topic Modeling. Journal of Computing in Civil Engineering, 33(3), 04019023.
Wang, Y., Hulse, D., Von Meding, J., Brown, M. & Dedenbach, L., (2019) Conceiving Resilience: Lexical
Shifts and Proximal Meanings in the Human-Centered Natural and Built Environment Literature
from 1990 To 2018. Developments in the Built Environment. DOI:
https://doi.org/10.1016/j.dibe.2019.100003.
Wang, Y., Taylor, J. E., & Garvin, M. J. (2020). Measuring Resilience of HumanSpatial Systems to
Disasters: Framework Combining Spatial-Network Analysis and Fisher Information. Journal of
Management in Engineering, 36(4), 04020019.
Woetzel, J., Remes, J., Boland, B., Lv, K., Sinha, S., Strube, G., Means, J., Law, J., Cadena, A., & Tann,
V. (2018). Smart Cities: Digital Solutions for a More Livable Future. In McKinsey & Company.
Xie, W., Zhu, F., Jiang, J., Lim, E. P., & Wang, K. (2016). TopicSketch: Real-time bursty topic detection
from twitter. IEEE Transactions on Knowledge and Data Engineering, 28(8), 22162229.
Yang, C., Su, G., & Chen, J. (2017). Using big data to enhance crisis response and disaster resilience for a
smart city. IEEE 2nd International Conference on Big Data Analysis, ICBDA 2017, 504507.
Yao, F., & Wang, Y. (2020a). Tracking urban geo-topics based on dynamic topic model. Computers,
Environment and Urban Systems, 79,101419. DOI:
https://doi.org/10.1016/j.compenvurbsys.2019.101419.
Yao, F., & Wang, Y. (2020b). Domain-specific sentiment analysis for tweets during hurricanes (DSSA-
H): A domain-adversarial neural-network-based approach. Computers, Environment and Urban
Systems, 83, 101522. DOI: https://doi.org/10.1016/j.compenvurbsys.2020.101522.
Yoshinaga, N., & Kitsuregawa, M. (2014). A self-adaptive classifier for efficient text-stream processing.
Proceedings of the 25th International Conference on Computational Linguistics, 2014: Technical
Papers, 10911102.
Yu, W., Li, J., Bhuiyan, M. Z. A., Zhang, R., & Huai, J. (2017). Ring: Real-Time Emerging Anomaly
Monitoring System Over Text Streams. IEEE Transactions on Big Data, 5(4), 506519.
Zhang, C., Liu, L., Lei, D., Yuan, Q., Zhuang, H., Hanratty, T., & Han, J. (2017). TrioVecEvent:
Embedding-based online local event detection in geo-tagged tweet streams. Proceedings of the
ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, 595604.
Zhang, X., & Li, H. (2018). Urban resilience and urban sustainability: What we know and what do not
know? Cities, 72, 141148.
Urban Agility and Resilience Lab
Page 31/31
Zhang, F., Li, Z., Li, N., & Fang, D. (2019). Assessment of urban human mobility perturbation under
extreme weather events: A case study in Nanjing, China. Sustainable Cities and Society, 50,
101671.
Zhu, S., Li, D., & Feng, H. (2019). Is smart city resilient? Evidence from China. Sustainable Cities and
Society, 50. https://doi.org/10.1016/j.scs.2019.101636
Zobel, C. W. (2011). Representing perceived tradeoffs in defining disaster resilience. Decision Support
Systems, 50(2), 394403.
... How to scientifically simulate this complex, dynamic interaction process is the key procedure of urban-resilience research. Moreover, in the context of smart cities, the response speed of risk is crucial to urban resilience [2,17,18]. Real-time big-data technology can reduce the impact of disaster losses and improve the ability of urban rapid recovery [19,20]. Increasingly more studies have been concerned with the quantitative assessment, spatial visualization, and dynamic simulation of urban resilience. ...
... The research on the multi-element dynamic-evolution process of urban resilience is still limited [6]. Therefore, future work should emphasize urban-resilience computation simulation leveraging an advanced "smart" approach (e.g., big data, urban computing, and artificial intelligence) to realize the transition from static mode to dynamic process and strengthen the ability of intelligent decision-making in cities [6,18,19]. ...
Article
Full-text available
Simulating the dynamic process of urban resilience and analyzing the mechanism of resilience-influencing factors are of great significance to improve the intelligent decision-making ability of resilient urban planning. The purpose of this article is to implement a comprehensive literature review on the quantitative computation and simulation studies of urban resilience, investigating the characteristics of current research, including the most commonly applied methods, the most frequently space–time scales, the most popular research topics, and the most commonly involved risk types. Then, the study provides recommendations for future research: (1) research on multiple risk disturbance scenarios, (2) the computation of urban resilience from the public perspective, and (3) a computation-simulation framework with the goal of revealing the mechanism. Finally, this study constructs a resilience-computation simulation framework for resilient urban planning, which lays a foundation for the further development of urban-resilience dynamic-simulation computing and planning-scenario applications in the future.
... In fact, it is impossible to collect such data over a large geographic space. The recent emergence of location-aware big data, such as the smartphone call records and social media data, has opened a new gate for people-oriented rainstorm perception studies (Roy et al., 2021;Yao & Wang, 2020). For example, Wang et al. (2020) used social media check-in data to evaluate the spatiotemporal characteristics of public response behaviors in response to a heavy downpour in Nanjing, China in July 2016. ...
Article
Extreme weather events become more frequent in the context of global climate change. Understanding how the public sense and perceive extreme weather events such as rainstorms is crucial for rainstorm-induced hazard mitigation. However, it is not clear how public's rainstorm perception and sensitivity vary across a large geographic scale. In this study, we examined over 210 million microblogs and studied rainstorm perception and perception sensitivity across China in 2017. Our results show that, on average, when the rainstorm-derived rainfall increase 1 mm, city dwellers would post 0.178 more rainstorm-related microblogs. There are also significant variations in rainstorm perception sensitivity across our study area. Dwellers living in the cities in the southeastern coast are not as sensitive to rainstorms as those in the north. Such spatial variations could be explained by annual rainfall and terrain relief variables. Dwellers who live in cities with annual rainfall less than 839 mm and terrain relief greater than 264 m show more attention to rainstorms by posting more microblogs.
... Answering these questions will inform practical applications of mining social media data to improve response and recovery. This approach can be applied in future disasters to extract relevant information in near real-time for emergency response and management (Ferner et al. 2020;Yao and Wang 2020). ...
Article
Full-text available
Twitter can supply useful information on infrastructure impacts to the emergency managers during major disasters, but it is time consuming to filter through many irrelevant tweets. Previous studies have identified the types of messages that can be found on social media during disasters, but few solutions have been proposed to efficiently extract useful ones. We present a framework that can be applied in a timely manner to provide disaster impact information sourced from social media. The framework is tested on a well-studied and data-rich case of Hurricane Harvey. The procedures consist of filtering the raw Twitter data based on keywords, location, and tweet attributes, and then applying the latent Dirichlet allocation (LDA) to separate the tweets from the disaster affected area into categories (topics) useful to emergency managers. The LDA revealed that out of 24 topics found in the data, nine were directly related to disaster impacts—for example, outages, closures, flooded roads, and damaged infrastructure. Features such as frequent hashtags, mentions, URLs, and useful images were then extracted and analyzed. The relevant tweets, along with useful images, were correlated at the county level with flood depth, distributed disaster aid (damage), and population density. Significant correlations were found between the nine relevant topics and population density but not flood depth and damage, suggesting that more research into the suitability of social media data for disaster impacts modeling is needed. The results from this study provide baseline information for such efforts in the future.
... Thành phố thông minh, hay đô thị thông minh (từ thuật ngữ "smart city") là một khái niệm mới đề cập đến mức độ phát triển mới của các đô thị hiện đại, nơi công nghệ thông tintruyền thông (ICT -Information communication technology) đóng góp vào việc quản lý và vận hành thành phố, đem lại nhiều lợi ích cho cả cư dân và nhà quản trị, tạo nên một môi trường sống tốt hơn. Khái niệm đô thị thông minh đã được đề cập và làm rõ trong nhiều văn liệu khoa học [1]- [6], trong đó nhấn mạnh tới vai trò của công nghệ thông tin -truyền thông [6]- [10], vai trò của internet vạn vật (IoT -Internet of things) [6], [8], [11]- [13] gắn với việc vận hành các hoạt động chức năng của đô thị, quản lý, lập kế hoạch và phát triển đô thị theo hướng bền vững [2], [4], [6], [8], [10], [14]. ...
Conference Paper
Full-text available
Đô thị thông minh có thể được hiểu là nơi công nghệ thông tin và truyền thông (ICT) được hợp nhất với cơ sở hạ tầng truyền thống hiện có, sau đó được điều phối và quản lý bằng công nghệ số. Hướng tới phát triển đô thị thông minh đang là một xu hướng trên toàn cầu, trong đó có Việt Nam. Việc phát triển đô thị thông minh, bên cạnh tập trung các công nghệ cao trong điều hành và quản lý còn cần chú ý tới yêu cầu phát triển bền vững, trong đó có vai trò rất lớn của quy hoạch. GIS, với tư cách là một công nghệ cao, cần có vai trò trong cả quá trình quy hoạch và quản lý đô thị thông minh. Thông qua việc phân tích vai trò của GIS trong các bài học phát triển đô thị thông minh trên Thế giới, hiện trạng và tiềm năng đô thị thông minh tại Việt Nam, nghiên cứu này xem xét các thách thức trên ba khía cạnh: công nghệ, kinh tế xã hội và môi trường với các chi tiết cụ thể của từng khía cạnh. Kết quả của nghiên cứu là chỉ rõ được các thách thức của bản thân hệ thống GIS hiện nay cũng như các thách thức thuộc về môi trường phát triển tại Việt Nam. Từ đó, nghiên cứu cũng đề xuất một số giải pháp cho đô thị thông minh tại Việt Nam, nhấn mạnh vào vai trò của GIS.
Article
Full-text available
Why some places are popular but others are not? Many studies have captured the drivers of a place's popularity, but few have explained how they work. The contribution of this study is that it proposes a methodology composed of “calculation” and “explanation” so as to explore the key factors and mechanisms that influence cultural venues to become popular places. 41 cultural and creative industry parks in Beijing were chosen and points-of-interest (POIs) and Dazhong Dianping (DZDP) social media data was used in this study. Park popularity was judged by the DZDP star rating and the average number of comments per year (ANCPY). The “calculation” of the main influence factors was based on fuzzy set qualitative comparative analysis (fsQCA), and the “explanation” of their rules was based on Word2vec. It is found that “coordination between old and new”, “density of cultural, scientific and educational facilities”, “building height diversity”, “parking lot density”, “number of bus stops” and the “number of universities and scientific research institutions” influence park popularity by affecting the aesthetics of framing scenes, the cultural functions, parking system, and accessibility. The conclusions can be referred to by park planners and operators to improve cultural ecosystem service and passenger flow volume.
Article
Urban flooding is one of the most widespread natural hazards in modern cities. Risk mapping provides critical information for flood risk management to reduce life and economic loss. As a widely used approach to support flood risk mapping, physical-based modelling suffers from model accuracy and computation complexity. Empirical methods rely on availability of rich disaster data and are not normally transferable for fast flood prediction to different cities. Both methods require high-quality hazard data and disaster information for reliable prediction, which are not always available. This paper presents an alternative near real-time flood risk mapping method for data scarce environments developed using social sensing and region-stable deep neural network (RS-DNN). By extracting disaster information in near real-time using social sensing techniques and considering risk distribution factors rather than flood influencing factors, this new method enables flood risk mapping and analysis cross a large domain in minutes, with all input data openly available. The proposed method can be adapted to different disaster process and different case study cities through timely social sensing.
Chapter
With the advent of information communications technology (ICT), online exchange of information and opinions is common on the Internet. The sentiment analysis on social media networks like Facebook, Twitter, etc., has turned into a reliable method for ascertaining the users’ opinions encompassing different scenarios. Sentiment analysis is extremely useful to calculate and find out public sentiments and, overall, assist in better decision-making. Sentiment analysis comes under artificial intelligence. The paper presents its applications in the context of safety and event detection in smart cities. It describes data acquisition and data analysis in short. It also discusses the future possibilities and applications of sentiment analysis as a whole. KeywordsArtificial intelligence (AI)Natural language processing (NLP)Sentiment analysisSecuritySafetySmart cities
Chapter
Delay-tolerant networks (DTNs) have potential of working in disconnected environment and tolerate high delays. Due to its promising service behaviour, researchers are encouraged to work in the area of DTNs. Over the last few years a verity of work has been carried out in the area of DTNs. To enlighten the researchers about current trends and to understand the scope of further research, 134 research papers have been referred and formulated a graphical and organized perspective. We have performed a bibliometric analysis by designating referred papers using research objectives, citations, publishers, article efficiency and article type. This analysis would cater a perception for researchers, students, experts and publishers to investigate modern research trends in the area of delay-tolerant networks.KeywordsDelay-tolerant networksGraphical interpretationTrends analysis
Article
Climate change and rapid urban development have intensified the impact of hurricanes, especially on the Southeastern Coasts of the United States. Localized and timely risk assessments can facilitate coastal communities’ preparedness and response to imminent hurricanes. Existing assessment methods focused on hurricane risks at large spatial scales, which were not specific or could not provide actionable knowledge for residents or property owners. Fragility functions and other widely utilized assessment methods cannot model the complex relationships between building features and hurricane risk levels effectively. Therefore, we develop and test a building‐level hurricane risk assessment with deep feedforward neural network (DFNN) models. The input features of DFNN models cover the meta building characteristics, fine‐grained meteorological, and hydrological environmental parameters. The assessment outcomes, that is, risk levels, include the probability and intensity of building/property damages induced by wind and surge hazards. We interpret the DFNN models with local interpretable model‐agnostic explanations (LIME). We apply the DFNN models to a case building in Cameron County, Louisiana in response to a hypothetical imminent hurricane to illustrate how the building's risk levels can be timely assessed with the updating weather forecast. This research shows the potential of deep‐learning models in integrating multi‐sourced features and accurately predicting buildings’ risks of weather extremes for property owners and households. The AI‐powered risk assessment model can help coastal populations form appropriate and updating perceptions of imminent hurricanes and inform actionable knowledge for proactive risk mitigation and long‐term climate adaptation.
Article
Full-text available
Hurricanes are one of the most frequent and destructive disasters in the United States. The events are large scale and have relatively long-term impacts. Social networking platforms such as Twitter can provide real-time information for disaster managers and affected populations during large-scale disasters (e.g., hurricanes), but extracting useful information and interpreting data accurately for disaster management is still challenging. Sentiment analysis of social media data helps detect the concerns of affected people and understand individuals' responses on the ground at unprecedented scales, but the method is known to be domain-dependent. The same words or expressions can indicate opposite sentiments in different domains. This paper proposes a domain-specific sentiment analysis approach specifically for tweets posted during hurricanes (DSSA-H). DSSA-H can retrieve hurricane-relevant tweets with a trained supervised-learning classifier, Random Forest (RF), and classify the sentiment of hurricane-relevant tweets based on a domain-adversarial neural network (DANN). We built a dataset of tweets posted during six recent hurricanes and applied the DSSA-H approach for sentiment analysis. After evaluation, we found that each classifier (i.e., RF and DANN) outperforms baseline classifiers and that DSSA-H outperforms two high-performing general sentiment classification approaches when classifying sentiments of tweets posted during hurricanes. We also applied DSSA-H in examining sentiment patterns across six recent hurricanes in the U.S. This domain-specific sentiment analysis approach can be used by the first responders and affected communities to more accurately and rapidly detect crises and emergent events, allocate resources, and assess disaster's impact during hurricanes. DSSA-H contributes to an intelligent and adaptive disaster information system for the data-rich human and the built environment system.
Preprint
Full-text available
Improving urban resilience to disasters becomes well-recognized in both industry and academia, but resilience remains challenging to be operationalized, especially in the complex urban contexts. Currently, longitudinal empirical studies on measuring resilience at fine-grains of space and time are lacking. Few methods can quantify resilience at an urban scale based on collective responses of individuals in a real disaster and can be adopted in distinct disaster contexts with crowdsourced data. We explored the potential advantages of network analysis to describe a complex human-spatial system (HSS). We integrated insights from the research field of socio-environmental systems, finding Fisher information (FI) to be an effective tool to quantify the dynamics of resilience. Consequently, we propose a quantitative framework, combining network analysis and FI, to measure resilience of HSS to disasters. We generated spatial networks with aggregated geolocations from a Twitter Streaming API, and computed and compared network-wide metrics pre, during and after a disaster. FI was employed to detect mobility perturbations and to reveal the dynamic process of resilience over time. We applied our spatial-network analysis and FI framework to examine Hurricane Harvey and the subsequent flood in Greater Houston, Texas in 2017. The analysis uncovers changed statuses and durations in the spatial network and suggests an intrinsic resilience of the HSS. The data-driven analytical framework contributes to an enhanced spatiotemporal understanding of urban resilience through a human-mobility perspective and to improved management of integrated cyber, human and infrastructure systems.
Article
Full-text available
In recent years, studies related to topic derivation in Twitter have gained a lot of interest from businesses and academics. The interconnection between users and information has made social media, especially Twitter, an ultimate platform for propagation of information about events in real time. Many applications require topic derivation from this social media platform. These include, for example, disaster management, outbreak detection, situation awareness, surveillance, and market analysis. Deriving topics from Twitter is challenging due to the short content of the individual posts. The environment itself is also highly dynamic. This paper presents a review of recent methods proposed to derive topics from social media platform from algorithms to evaluations. With regard to algorithms, we classify them based on the features they exploit, such as content, social interactions, and temporal aspects. In terms of evaluations, we discuss the datasets and metrics generally used to evaluate the methods. Finally, we highlight the gaps in the research this far and the problems that remain to be addressed.
Article
Full-text available
The complex dynamics of the human-centered natural and built environment (HNBE) have been characterized by emerging and diverging conceptions of resilience (e.g., climate resilience, disaster resilience, social-ecological resilience). Each resilience modifier has produced rich bodies of literature, drawing on distinctive meanings of the term “resilience.” As resilience modifiers have continued to multiply and evolve over the past three decades, the relationships among them have become less clear. This impedes effective management of resilience in a more integrated HNBE. To improve understanding of the evolution and commonalities of resilience modifiers, we surveyed 3,181 articles concerning seven resilience modifiers published from 1990 to 2018. Using bibliometric analysis tools, we clarify lexical meanings of these resilience concepts and find a convergence among them based on their shared concerns and intellectual genealogy. This converging conception of resilience may motivate and catalyze more integrated knowledge, methods, and expertise in understanding and managing resilience across disciplines.
Article
Users acting as real-time sensors post information about current events on various social media sites like Twitter, Facebook, Instagram, and so on. This generates a huge amount of data requiring significant effort to process and filter it to detect events/topics. It becomes more challenging when data are generated as a tweet stream because of its speed, presence of noise, slangs, phrases, abbreviations, and so on. In recent years, many approaches have been proposed either for detecting small- or large-scale events, individually. There is a lack of a complete solution that provides analysis from different perspectives. We propose a novel approach Mythos that detects events, subevents within an event, and generates abstract summary and storyline to provide different perspectives for an event. There are three modules in Mythos. Online incremental clustering algorithm identifies small-scale events in the form of small clusters, the event hierarchy generator generates bigger events in the form of hierarchies, and the summarization module produces summary of events/subevents. The summarization module uses a long short-term memory (LSTM)-based learning model to generate summaries at different levels--from the most abstracted to the most detailed. The summaries at different levels are used to generate a storyline for the event. Our experimental analysis on a variety of twitter data sets from different domains compares Mythos against the known existing approaches for event detection and summarization. It outperforms baseline approaches for both. The generated summaries are evaluated against summaries provided by external reference sources like Guardian and Wikipedia.
Article
Topic evolution mining on short texts is an important research topic in natural language processing. Existing methods have been focused either on the topic evolution of normal documents or on the evolution of topics along a timeline. In this paper, we aim to generate topic evolutionary graphs from short texts, which not only capture the main topic timeline, but also reveal the correlations between related subtopics. Firstly, we propose an Encoder-only Transformer Language Model (ETLM) to quantify the relationship between words. Then we propose a novel topic model, referred as weighted Conditional random field regularized Correlated Topic Model (CCTM), which leverages semantic correlations to discover meaningful topics and topic correlations. Finally, topic evolutionary graphs are generated by an Online version of CCTM (OCCTM) to capture the evolutionary patterns of main topics and related subtopics. Experimental results on real-world datasets demonstrate our method outperforms baselines on quality of topics and presents motivated patterns for topic evolution mining.
Article
Modern cities are facing critical environmental and social problems that are difficult to solve using conventional planning approaches due to the cities' magnitude and complexity. Recent developments in sensing technologies and urban computing, however, integrate new data resources and technologies to tackle these challenges. Popular social networking platforms such as Twitter provide new data sources on important events (e.g., cultural activities, political campaigns, accidents, crises) providing rich knowledge about urban systems and human dynamics. This research is intended to develop a method for effectively monitoring important information during such events and helping with planning and policymaking. We use semantically similar and geographically close geo-topics to represent important local events. This research proposes a data-driven system for detecting and tracking the semantic, spatial, and temporal dynamics of these geo-topics, specifically designed for geo-tagged tweets. The system consists of data preprocessing, geo-topic generation, and geo-topic tracking modules. The preprocessing module can remove robotic and semantically trivial texts. In the geo-topic generation module, we use spatial factors to measure the spatial impacts of geo-tagged tweets by applying an exponential decay function to the pairwise distances between tweets. We then improve the dynamic topic model (DTM) by embedding the spatial factors to enable the generation of geo-topics in semantic, spatial, and temporal dimensions simultaneously. The geo-topic tracking module monitors semantic change by detecting changes in certain keywords' probabilities and the volumes of tweets belonging to different geo-topics. This module also uses radius of gyration and trajectory-pattern mining to track and analyze the movement patterns of geo-topics. We employed the tracking system in three disaster cases in different U.S. cities to track small-scale emergencies and crises. These implementations demonstrated the effectiveness of the system for identifying and tracking geo-topics at fine temporal and geographic scales. The system also has strong potential in creating planning-related analyses for policy makers, improving the situational awareness of the general public, and serving as a basis for urban information systems that contribute to smart, agile, and resilient city developments.
Article
Extreme weather events (EWEs), due to their high uncertainty, massive scale, irreversibility and destructiveness, may significantly impact cities, including causing notable perturbation to urban human mobility. Recent research has substantially advanced the knowledge on general human mobility patterns in cities, primarily about the spatiotemporal characteristics of trajectories of urban population, but has rarely examined the perturbation of these mobility patterns during EWEs. To quantitatively assess human mobility perturbation, this study proposes to measure both the instantaneous perturbation at any given moment during an EWE, and the accumulated perturbation over the entire timespan of the EWE. Using two metrics that are developed for the above purpose, a case study is conducted in Nanjing, a major city in China, which recently experienced record-breaking rainstorm and snowstorm events. Based on trajectories of all taxies and buses in Nanjing during these events, the case study quantitatively assesses the perturbation of human mobility in the city, compares it between two EWEs and between two modes of transport, and analyzes the geographical distribution of the perturbation within the city boundary. Based on the results, further insights into the impacts of EWEs on urban human mobility are discussed in the paper.
Article
Smart city is originally aimed at dealing with various urban problems due to rapid urbanization, like energy shortage, congestion, and environmental pollution. The Chinese government has been devoting to the promotion of smart cities for many years. However, it is unconfirmed whether the city is more resilient with all the modern technologies provided when unexpected predicaments like climate changes or disasters occur. Therefore, it is urgent to consider resilience in the smart city. This paper provides a MCDM approach to assess and rank the resilience of 187 smart cities in China. The results demonstrate that the overall resilience of smart cities is at a relatively low level. There is also a significant unbalance of resilience between smart cities due to different infrastructural, economic, social, institutional, and environmental conditions. The potential links between urban smartness and resilience were also explored, and the results showed significant positive relationship between the smartness of a city and its resilience. Evidence also proved that developing smartness is more or less useful for improving urban resilience. Suggestions such as strengthening the development of infrastructure and economy, and enhancing the multi-stakeholders’ cooperation are proposed to further promote the smart and resilient development in China.