Supporting Decision Making during Emergencies through Information

Visualization of Crowdsourcing Emergency Data

Paulo Sim

oes Jr.

, Pedro O. Raimundo

, Renato Novais

2,3

, Vaninha Vieira

1,3

and Manoel Mendonc¸a

1,3

Department of Computer Science, Federal University of Bahia, Av. Adhemar de Barros, Salvador, Brazil

Department of Computer Science, Federal Institute of Bahia, Rua Em

ıdio dos Santos, Salvador, Brazil

Fraunhofer Project Center for Software and Systems Engineering at UFBA, Av. Paralela, Salvador, Brazil

Keywords:

Emergency Management, Information Visualization, Crowdsourcing, Decision Support.

Abstract:

Decision making during an emergency response requires having the right information provided in the right way

to the right people. Relevant information about an emergency can be provided by several sources, including

the crowd at the place where the emergency is happening. A big challenge is how to avoid overwhelming the

decision makers with unnecessary or redundant information provided by the crowd. Our hypothesis is that

appropriate information visualization techniques improve the understanding of information sent by a crowd

during an emergency. This work presents an approach for emergency information visualization, gathered

through crowdsourcing, which improves context-aware decision making by keeping a real-time emergency

state board. This approach was implemented in ERTK, as a proof of concept, and evaluated with 15 emergency

management experts in Brazil. The yielded results show that our approach has the potential to assist a context-

aware decision making during an emergency response.

1 INTRODUCTION

According to the United Nations Department of Hu-

manitarian Affairs (DHA, 1992), an emergency is “a

sudden and usually unforeseen event that calls for

immediate measures to minimize its adverse conse-

quences”. These can be natural events (e.g. earth-

quakes, tsunamis) or men-made (e.g. terrorist attacks,

stampedes, mass shootings) and their consequences

may be a disaster for life, environment and facilities.

The emergency management cycle has four dif-

ferent stages, aligned with the timeline of the emer-

gency: preparedness, response, recovery and mitiga-

tion(Horita et al., 2013; Poser and Dransch, 2010).

The response phase is the most critical one, and in-

volves several activities related to decision making,

rescuing and combat, intended to save lives and deal

with the emergency consequences in the fastest pos-

sible way (Waugh and Streib, 2006; Lanfranchi et al.,

2013). During the response phase, emergency man-

agers need accurate and reliable information to sup-

port their decisions. A late or inaccurate decision may

result in damage to human life and hinder the over-

all emergency management process (Thompson et al.,

2006).

Crowdsourcing aims to use a crowd to do a set of

tasks that would otherwise be performed by a smaller

number of people (Howe, 2006). This technique has

been used to support emergency management and

their utility has been ratiﬁed in real emergency scenar-

ios, as is the case of the Ushahidi platform (Heinzel-

man and Waters, 2010; Duc et al., 2014). People in

the place of an incident often have the most updated

information regarding the emergency, and are able to

provide what the emergency managers need to sup-

port their decision-making (Gao et al., 2011).

We developed a crowdsourcing-based emergency

management platform called RESCUER which aims

to support command and control centers on acquir-

ing and processing context-aware multimedia infor-

mation about an emergency, using the crowd as the

main information source. However, a challenge is

how to avoid overwhelming the decision makers with

inappropriate, inaccurate, outdated, or redundant in-

formation coming from the crowd (Gao et al., 2011),

which can impair their decisions. Several informa-

tion visualization (Mazza, 2009; Ware, 2012) (Info-

Vis) techniques have been used to improve the pre-

sentation of emergency information to decision mak-

ers (Dusse et al., 2016; Wu et al., 2011; Besaleva and

Weaver, 2013).

A recent systematic mapping study (Dusse et al.,

178

Jr., P., Raimundo, P., Novais, R., Vieira, V. and Mendonça, M.

Supporting Decision Making during Emergencies through Information Visualization of Crowdsourcing Emergency Data.

DOI: 10.5220/0006370701780185

In Proceedings of the 19th International Conference on Enterprise Information Systems (ICEIS 2017) - Volume 3, pages 178-185

ISBN: 978-989-758-249-3

2016) shows that social networks (e.g. Twitter, Face-

book) are the main tool for acquiring crowdsourcing

information in emergency management. This study

also shows that most tools using InfoVis apply similar

techniques, which consists of devising a set of icons

or shapes to represent the information on a map.

In our research, we argue that decision making

in emergency response may be improved by focus-

ing on presenting the most relevant information about

the emergency to the emergency managers. In our ap-

proach, we employed several heuristics to summarize

gathered information, using icons and text-based vi-

sual metaphors. We also incorporated different charts

detailing the distribution of crowdsourcing informa-

tion, which the emergency manager can use to reﬁne

his/her judgment about the emergency.

To evaluate our InfoVis approach, we imple-

mented a proof of concept tool called Emergency Re-

sponse Toolkit (ERTK), a component of RESCUER

platform aimed to support command and control cen-

ters on analyzing information from the crowd. We

also conducted a case study with 15 experts in emer-

gency management to gather their impressions and

feedback on the proposed approach. The results in-

dicate that our approach has the potential to improve

decision making in emergency management.

The rest of this paper is organized as follows. Sec-

tion 2 discusses related works about the use of InfoVis

in the emergency management. Section 3 presents our

visualization approach. Section 4 describes the de-

sign and and execution of our evaluation. Section 5

presents our experimental ﬁndings. Finally, we draw

our conclusions and indicate future directions of re-

search in Section 6.

2 RELATED WORK

The visualization of crowd data faces several unique

challenges due to the nature, the amount and the for-

mat of data involved (Gao et al., 2011). The visualiza-

tion tool has to allow the user to gather an overall view

of the data, and also be able to drill down and look

into speciﬁc data points (Dusse et al., 2016). Crowd

data must also usually be categorized and transformed

into useful information for decision making. In this

section, we will discuss existing tools and approaches

to visualize crowd data in emergency management,

their achievements and shortcomings.

Ushahidi is an open-source platform that has been

used in several high-proﬁle disasters such as the Haiti

earthquake. This platform provides tools for users

to load, ﬁlter and visualize crowd data from multiple

sources (email, twitter, SMS et cetera) (Heinzelman

and Waters, 2010; Duc et al., 2014). Data that ar-

rives from the available channels are gathered if they

contain one or several of predeﬁned keywords. Ad-

ditionally, Ushahidi provides a web-based form and a

mobile app that users can utilize to input structured

data in the system.

The main visualization component of the Ushahidi

platform is an annotated map. This map allows users

to obtain a generalized view of the emergency and

clusters reports based on the level of zooming of the

map and their proximity. There is also a timeline of

the emergency and widgets that allow the user to ﬁlter

the reports by category, area and time. The clustering

strategy adopted by Ushahidi provides an adequate

view of which areas need more attention (i.e., have

more people). But sometimes results in undesirable

clutter and occlusion of informational elements.

Sahana Eden (Duc et al., 2014) is another promi-

nent open-source emergency management tool that

has been successfully deployed in the Haiti earth-

quake of 2010. Sahana Eden focuses on the pre-

paredness and mitigation phases of emergency man-

agement and has a plethora of features to organize and

manage emergency management assets (human and

otherwise) whose investigation are beyond the scope

of this paper.

Despite being used to display static information,

the map provided by Sahana Eden has ﬁltering and

visualization categories. Nevertheless, both Sahana

Eden and Ushahidi lack the ability to represent what

is happening as anything more than points, as high-

lighted by Duc et al. (2014) .

Another platform to deal with crowdsourcing data

is presented by Wu et al. (2011), with focus on par-

ticipatory multimedia generation. This platform con-

sists of a mobile application that sends geo-tagged

pictures to a server, a web-based desktop application

(called CIVIL) that allows the emergency manage-

ment professionals to view the received information,

and server infrastructure (Wu et al., 2011).

CIVIL displays the pictures received from the

users in real-time as thumbnails on a map. The

thumbnails can be interacted with, allowing the user

to contact the sender of the picture. The desktop

interface of the CIVIL application provides: “a pri-

vate map, a public map, and a set of analysis tools

to support collaboration and decision making” (Wu

et al., 2011). Each user can make annotations and

sketches on his private map or the public map, which

is shared across all users. The analysis tools provided

by CIVIL include “a chatting panel, a table sorting

all annotations, a chart showing accumulated annota-

tions, and a timeline visualizing individual annotation

activities” (Wu et al., 2011). CIVIL also tracks the

Supporting Decision Making during Emergencies through Information Visualization of Crowdsourcing Emergency Data

179

location of the users sending pictures.

The map utilized in CIVIL has more features than

the ones presented this far, as it supports annotations

and different visualization modes. However, it dis-

plays the individual pictures to the user, and thus will

generate more clutter and noise as more people use

the mobile application.

CrowdHelp (Besaleva and Weaver, 2013) is an-

other existing tool to deal with crowdsourcing infor-

mation in emergency management. Using this tool,

mobile users are able to provide information about

the event, their physical condition and symptoms.

The system uses this information to facilitate medi-

cal triage and provides visualizations for emergency

management professionals through a web interface,

in real-time.

CrowdHelp provides several ways to cluster the

information. Giving users full control over the ma-

chine learning algorithms utilized for clusterization.

In terms of visualization, however, these clusters are

displayed as data points with different colors. Color

is a sub-par visual attribute to represent the clustering

of information because it does not remove any of the

clutter and clashes with the geographical view, which

induces a proximity-based clustering metaphor.

As shown so far, most InfoVis solutions applied

to emergency management focus on map-based vi-

sualizations of the collected information which dif-

fer mainly in the speciﬁc format, clusterization ap-

proach and associated data of the points displayed on

the map. Some of the tools presented here also pro-

vide non map-based visualizations; however, these

are just plain tabular displays of the gathered data.

Map-based visualizations of georeferenced data are

staple in emergency management systems; therefore,

we implemented our approach to this kind of visual-

ization attempting to minimize noise and occlusion.

However, the main appeal of this work is that we

proposed, implemented and evaluated (Section 4) a

novel way to provide a summary of crowdsourced

data to the user, utilizing icons, text, and several cus-

tom widgets. This approach differs slightly from ex-

isting work, but shows promise, as we highlight in

Section 5. In the next section, we elaborate on how

ERTK deals with the visualization of crowdsourcing

information and the aspects of our approach.

3 AN APPROACH FOR INFOVIS

IN EMERGENCY RESPONSE

Our research set out to identify proper ways to present

crowdsourcing emergency data so that emergency

managers can quickly identify the most relevant set

Figure 1: Example of the new incident alert for when a ﬁre

incident is detected by the system.

of information during the emergency response. Based

on workshops conducted with emergency manage-

ment specialists from Brazil and Europe, and feed-

back received during during the evaluation of our ﬁrst

efforts (Sim

oes J

unior et al., 2015), we focused on

three speciﬁc needs of emergency managers: to iden-

tify the type of incident and where it is happening, to

know the available information for the incident at all

times; and to know how reliable a given information

is.

To allow users to identify where the emergency is

happening we use a map view together with an icono-

graphic representation of the incident type. This com-

bination (map and icons) allows the emergency man-

agers to quickly establish the place of incident, and

what is happening. Additionally, we display a visual

alert with a message describing what is the type of

the incident, to counteract the possible misinterpre-

tation of the icons by some users. Figure 1 shows a

situation where a emergency with ﬁre incident is hap-

pening, presenting the ﬁre icon (middle-right) and the

incident alert (top, red frame).

The Incident Detail View is the main view of our

approach. Figure 3 presents this view, which lets the

emergency managers quickly identify what is happen-

ing at the incident site, based on the information sent

by the crowd. This view can be subdivided in 5 areas:

(1) an interactive timeline that shows and can be used

to alter the incident status; (2)the aggregation results

of the crowdsourcing data; (3) a multimedia carousel

that shows the images and videos sent by the crowd,

with an area that highlights the current selected me-

dia; (4) a map that shows the most relevant reports

within the incident area (e.g. civilian reports within

the hot zone, workforce reports); (5) and a header

containing the incident ID and its type.

The Aggregated Incident Data is the main con-

tribution of the incident detail view. It presents the

most relevant data for each information collected

from the crowd. By applying a set of heuristics

to each report received, we identify the most rele-

vant data for a given information at an speciﬁc time

(e.g. the current number of injured, the environmen-

ICEIS 2017 - 19th International Conference on Enterprise Information Systems

180

tal damages). These results are then expressed using

icons and text, as shown in Figure 2. The labels of

each information (e.g. Number of Injured People, En-

vironmental Damage) are given a color that indicates

the most reliable proﬁle that conﬁrms the data being

shown. We use the red color to represent civilians, or-

ange for supporting forces and green for workforces.

Figure 2: Aggregated information from civilians (top left),

supporting forces (bottom) and workforces (top right).

To make the emergency managers aware that

the value of an information changed, we highlight

the representation of that information with every

change. We consider an information as “changed”

when a more reliable proﬁle conﬁrms that informa-

tion, and/or when the most relevant data for the in-

formation changes (e.g an increase in the number of

injured, a change in the color of the smoke). Figure 2

shows the highlighted state for the information ”En-

vironmental Damages”. To ensure that the changes

were acknowledged, the highlight only disappears af-

ter the emergency manager interacts with it.

We expected that these set of features could help

emergency managers to quickly identify the most rel-

evant information provided by the crowd during the

emergency response phase. To evaluate if emergency

managers could understand and use these features,

and ultimately if they would help them in the deci-

sion making process, we planned and executed a case

study described in the next section.

3.1 Implementing the InfoVis approach

in ERTK

We implemented the described visualizations in the

ERTK module of the RESCUER project to evaluate

our approach. Emergency managers will be able to

use it during the emergency response phase in large-

scale events or industrial parks.

ERTK is a web application with many features de-

signed to help in emergency management tasks (e.g.

map annotation, alerts, messages broadcast). Among

these, we implemented the proposed visualizations

to allow quick identiﬁcation of the type of incident,

where it is happening, which emergency information

is available and the reliability of the information.

From our planning, the users would see the inci-

dent alert, apprehend the type of incident that hap-

pened, click on the incident alert link to go to the

incident detail view, obtain the knowledge about the

current available information and accompany the evo-

lution of the incident though the update feature. The

displayed information and its reliability level would

aid the decision making of the emergency managers.

4 EXPERIMENTAL EVALUATION

This section describes the experimental evaluation we

conducted in order to evaluate ERTK.

4.1 Evaluation Tasks

We created an emergency scenario where the evalua-

tion participants would perform a set of tasks. In this

scenario, an explosion happened during a large-scale

event and the people in the crowd started to send re-

ports through the RESCUER mobile app. The partic-

ipants would play the role of a C&CC operator and

would try to complete the tasks present at Table 1.

For each task, we expected the participants to use a

speciﬁc visualization feature.

The proposed tasks are the result of interviews

with emergency management experts, and reﬂect

what they would try to do, in the proposed scenario,

using ERTK. The feature column shows which visu-

alization feature we expected the participants to use

to complete each task.

In tasks T1 and T2 we wanted to know if the new

incident alert, and the iconographic incident repre-

sentation in the map, provided the participants with a

clear view of what type of incident was detected and

where it started. We also evaluated if the participant

could easily understand how to access the detailed in-

cident information.

With T3 we sought to see if the participants were

able to identify all available information of the in-

cident, and outline what aspects of the visualization

they would notice and those they would not (e.g. if

they would perceive the colors in the title, the inci-

dent status, the highlights, etc.).

Through T4, we aimed to evaluate if the partici-

pants would perceive the incident status timeline not

only as an information source but as an interactive

widget as well.

From T5, we wanted to conﬁrm if the informa-

tion highlight would clearly indicate a change in the

available data. This is directly related to T7, were we

provoked a change in the title colors of some informa-

tion, and also changed the pictures being shown. We

wanted to see if the participants noticed these changes

and, in case they had passed the mouse over updated

Supporting Decision Making during Emergencies through Information Visualization of Crowdsourcing Emergency Data

181

Figure 3: Incident Detail View showing: the Incident Status Timeline (1); the Aggregated Incident Data (2); the Incident

Media (3); the Incident Map (4); the Incident Id and Type (5).

information, we wanted to see if they would indicate

that the information changed again.

On T6 the system would be in a state in which the

aggregated information would be indicating a smoke

color (e.g. dark grey) and a image with a different

smoke color (e.g. white grey) would be in the images

carousel . We wanted to see if the participants would

give more importance to the image or to the aggre-

gated information.

Finally, we expected the participants to notice

the different colors in the information titles, but we

did not predict that they would associate it to differ-

ent sender proﬁles (e.g. civilian, workforces) with-

out training. With T8, we wanted to conﬁrm if the

color/proﬁle relationships we used, namely: red for

civilians; orange for supporting forces; and green for

workforces matched the participants’ view.

4.2 Evaluation Design

A researcher accompanied the participants during the

tasks, telling them when to start a task, taking notes

about the participants’ remarks and helping then in

case they expressed any doubts.

We did not set time limits for any of the tasks and

did not train any participants either. We wanted to

observe if they could perceive and use the features

we described in section 3 without previous knowledge

about the system and, if they could not, which other

features they would use. To compensate the lack of

training we informed the participants that they could

skip tasks. We decided to execute the evaluation pro-

cedure with one participant at time, and to record the

computer screen and participants remarks during the

whole process.

We prepared three sets of reports that would alter

the system state during the duration of one or more

tasks. Two researchers were responsible for feed-

ing the data to the system at speciﬁc times, thus, we

would be able to compare if the information provided

by the participants while answering the tasks was

complete (e.g. if they were able to list all available

information or the changes that occurred). These sets

could be sent manually, using the RESCUER mobile

app, or through a program we created. The two auxil-

iary researchers would be in contact with the evalua-

tor researcher during the whole experiment. The eval-

uator would tell the auxiliaries the moment to send the

report sets. Figure 4 displays the general workﬂow of

our evaluation.

ICEIS 2017 - 19th International Conference on Enterprise Information Systems

182

Table 1: Evaluation tasks and expected visualization features to solve them.

ID TASK VISUALIZATION FEATURE

Identify the type of incident that happened

and where it happened

Incident alert, incident icon on Map

T2 Find the screen with information about the incident Incident alert

T3 List all available information about the incident

Incident status timeline, aggregated incident data,

incident media carousel

T4 Change the incident status to conﬁrmed Incident status timeline

In case you identify any, list the changes

in the information regarding the incident

Highlighted information

T6 Identify the color of the smoke Incident media carousel, aggregated incident data

Again, in case you identify any, list the changes

in the information regarding the incident

Incident media carousel, aggregated incident data

Colors indicate the proﬁle of the user who sent

an information (i.e. civilian, supporting force, workforce).

According to your feelings, associate a color to

each user proﬁle

Color from the title of aggregated incident data

Figure 4: Evaluation Procedure Dynamic.

After all tasks were completed, we asked the par-

ticipants to answer a feedback questionnaire and ﬁll

in a characterization form.

4.3 Evaluation Execution

We performed the evaluation at the International

Congress of Mass Disaster (CIDEM 2016) and at the

Command Center of Porto Alegre (CEIC

In both places, all participants were randomly se-

lected. A criteria to participate in the evaluation was

to be a member of an organization that would per-

form emergency response activities (regardless of the

emergency type). To encourage the candidates partic-

ipation we offered them to be part of a 32 GB thumb

drive rafﬂe.

We gave the chosen participants an overview of

RESCUER purpose and explained that we wanted

them to use ERTK to execute some tasks and give us

feedback at the end. We told each of them that there

were no right or wrong answers, and that they could

skip any task if they wanted to. They were free to

http://www.portoalegre.rs.gov.br/ceic

use the system as they wanted. When they accepted

the conditions, we applied the evaluation as we had

planned.

5 RESULTS AND DISCUSSION

We had a total of ﬁfteen participants, six on CIDEM

and nine at CEIC. Their time of experience with emer-

gency situations ranged from a few months to 40

years. Their roles and the organization they belonged

to greatly varied: police, ﬁreﬁghters and civil defense

are some examples. Also, all CIDEM participants

declared not having previous experience with emer-

gency management systems, which was the opposite

for the CEIC participants.

The participants took an average of one minute to

complete each task, counting the time they spent giv-

ing us feedback about their feelings while doing each

task (e.g. they explained us what they were think-

ing, why they made some decisions). The feedback

gathered from the evaluation in the form of the ap-

plied questionnaire, the participants feedback during

the evaluation and by observing the way, which they

interacted with the visualizations, is very encourag-

ing.

Only one participant considered the system as use-

less to help in emergency management tasks. He

could not identify the meaning of most icons and the

system responses to his actions were not as he ex-

pected. His feedback was that the system should be

clearer in the messages it was passing. We found this

most useful, while most of the other participants did

not had the same difﬁculties, we could identify inter-

esting points in his interaction with the system, and

Supporting Decision Making during Emergencies through Information Visualization of Crowdsourcing Emergency Data

183

Figure 5: Results of the feedback form given by the evaluation participants. One participant did not answer Q3.

his expectations, that can help us to improve the visu-

alization mechanisms applied.

Overall, the participants identiﬁed the type of in-

cident and the place where it was happening using

either the map or the new incident alert, as we

were expecting. In general, most systems related

to emergency management support work with maps

and icons, which allow for quickly identiﬁcation that

”something is happening at this place”. We observed

that the participant that considered this task very dif-

ﬁcult, was able to complete it in 10s. While the

one who considered it difﬁcult is the same one that

deemed the system useless. While we are certain that

there is room for improvement, based on the overall

feedback received, we consider the provided features

for helping in identifying the place and type of inci-

dent show great promise.

”Finding the screen with information about the in-

cident” (Task T2) was the task that the participants

took more time to complete as the majority of them

did not notice, or did not understood, the link in the

incident alert. Those who did not use the link, either

skipped the task (two of them) or followed almost the

same steps to complete it (using tabs within ERTK

interface). This shows that there needs to be a better

indication that there is a link within the alert or that

other interaction mechanisms should be used to take

the users to the incident detail view.

All participants correctly identiﬁed the informa-

tion we were expecting with tasks T3, T5 and T7. The

results of these tasks were very important for us, be-

cause they showed that the participants perceived the

information we were expecting using the features we

expected them to use. The participants were able to

tell which information was available. They perceived

when those information changed and which were the

changes. This result is aligned with our efforts to shift

the focus from the traditional map and iconographic

visualizations present in most emergency manage-

ment system to an approach focused on the aggrega-

tion of information gathered through crowdsourcing.

The answers of the feedback questionnaire

were

mostly positive (Figure 5), and supported the in-

sights we have while observing the participants using

ERTK. The majority of them considered to be fast in

ﬁnding the information about the incident. They were

also able to understand the icons we used in the vi-

sualizations and found to be easy to perceive when an

information changed. Most important, 12 participants

deemed the system as very useful regarding the possi-

bility of being of help on supporting decision making

during an emergency.

6 CONCLUSION

The people present in the site where an emergency is

happening are an invaluable source of information for

emergency managers. However, they need support to

help them make sense of all information available.

Current emergency management systems that

make use of crowdsourcing data, often use maps and

icons to present the collected data. While this have

been proven helpful, we believe that a visualization

focused on the results of an aggregation of the crowd-

sourced data that can improve the way emergency

accessible at https://goo.gl/Ym1Gaa

One of the participants did not answer Q3, hence the

sum for this question is less than the number of participants

ICEIS 2017 - 19th International Conference on Enterprise Information Systems

184

managers make sense of it.

In this paper we presented an InfoVis approach

to visualize the emergency data gathered through

crowdsourcing. We implemented this approach in

ERTK, whose goal is to allow emergency managers

to quickly identify the relevant information about

an emergency during the response phase, supporting

their decision making process.

We designed and ran an evaluation with ﬁfteen

emergency specialists to understand if and how they

perceived perceive and interacted with the proposed

visualizations. The results of this evaluation were

ultimately positive, highlighting the strengths of the

approach and highlighting several points of improve-

ment within the proposed visualizations. Most of

the participants said that the approach proposed here

would be helpful in supporting decision making dur-

ing the emergency response.

Future directions for this work include: improving

our approach based on the received feedback, devis-

ing new interaction mechanisms, proposing visualiza-

tions that better support the visualization of contradic-

tory information that arrives from a crowd.

ACKNOWLEDGEMENTS

This work was supported by the RESCUER project,

funded by the CNPq/MCTI (Grant: 490084/2013-3)

and by European Commission (Grant: 614154). We

thank all researchers that participate in the RESCUER

project for their valuable work. We also thank the

commander of CEIC - POA and the organizers of CI-

DEM for allowing us to perform our evaluations.

REFERENCES

Besaleva, L. I. and Weaver, A. C. (2013). Crowdhelp:

A crowdsourcing application for improving disas-

ter management. In Global Humanitarian Technol-

ogy Conference (GHTC), 2013 IEEE, pages 185–190.

IEEE.

DHA, U. (1992). Internationally agreed glossary of ba-

sic terms related to disaster management. UN DHA

(United Nations Department of Humanitarian Af-

fairs), Geneva.

Duc, K. N., Vu, T.-T., and Ban, Y. (2014). Ushahidi and

sahana eden open-source platforms to assist disaster

relief: Geospatial components and capabilities. In

Geoinformation for Informed Decisions, pages 163–

174. Springer.

Dusse, F., J

unior, P. S., Alves, A. T., Novais, R., Vieira, V.,

and Mendonc¸a, M. (2016). Information visualization

for emergency management: A systematic mapping

study. Expert Systems with Applications, pages 424–

437.

Gao, H., Barbier, G., Goolsby, R., and Zeng, D. (2011).

Harnessing the crowdsourcing power of social media

for disaster relief. Technical report, DTIC Document.

Heinzelman, J. and Waters, C. (2010). Crowdsourcing crisis

information in disaster-affected Haiti. US Institute of

Peace.

Horita, F. E. A., Degrossi, L. C., de Assis, L. F. G., Zipf, A.,

and de Albuquerque, J. P. (2013). The use of volun-

teered geographic information (vgi) and crowdsourc-

ing in disaster management: a systematic literature re-

view.

Howe, J. (2006). The rise of crowdsourcing. Wired maga-

zine, 14(6):1–4.

Lanfranchi, V., Mazumdar, S., and Ciravegna, F. (2013).

Evaluating the real usability of a c2 system–short and

controlled vs long and real. Proceedings of ISCRAM

2013, pages 62–66.

Mazza, R. (2009). Introduction to information visualiza-

tion. Springer Science & Business Media.

Poser, K. and Dransch, D. (2010). Volunteered geographic

information for disaster management with applica-

tion to rapid ﬂood damage estimation. Geomatica,

64(1):89–98.

Sim

oes J

unior, P., Novais, R., Santos, V. V., Pedraza, L. G.,

Mendona Neto, M. G., and Lima, K. V. C. (2015). Vi-

sualization mechanisms for crowdsourcing informa-

tion in emergency coordination. In XIV Brazilian

Symposium on Human Factors in Computer Systems,

2015, Salvador.

Thompson, S., Altay, N., Green III, W. G., and Lapetina, J.

(2006). Improving disaster response efforts with deci-

sion support systems. International Journal of Emer-

gency Management, 3(4):250–263.

Ware, C. (2012). Information visualization: perception for

design. Elsevier.

Waugh, W. L. and Streib, G. (2006). Collaboration and lead-

ership for effective emergency management. Public

administration review, 66(s1):131–140.

Wu, A., Yan, X., and Zhang, X. L. (2011). Geo-tagged mo-

bile photo sharing in collaborative emergency man-

agement. In Proceedings of the 2011 Visual In-

formation Communication-International Symposium,

page 7. ACM.

Supporting Decision Making during Emergencies through Information Visualization of Crowdsourcing Emergency Data

185