Karmegam et al.

Geoenviron Disasters (2021) 8:25 Geoenvironmental Disasters

https://doi.org/10.1186/s40677-021-00195-x

RESEARCH Open Access

Near real time flood inundation mapping

using social media data as an information
source: a case study of 2015 Chennai flood
Dhivya Karmegam1* , Sivakumar Ramamoorthy2 and Bagavandas Mappillairaju3

Abstract
During and just after flash flood, data regarding water extent and inundation will not be available as the traditional
data collection methods fail during disasters. Rapid water extent map is vital for disaster responders to identify the
areas of immediate need. Real time data available in social networking sites like Twitter and Facebook is a valuable
source of information for response and recovery, if handled in an efficient way. This study proposes a method for
mining social media content for generating water inundation mapping at the time of flood. The case of 2015 Chennai
flood was considered as the disaster event and 95 water height points with geographical coordinates were derived
from social media content posted during the flood. 72 points were within Chennai and based on these points water
extent map was generated for the Chennai city by interpolation. The water depth map generated from social media
information was validated using the field data. The root mean square error between the actual water height data
and extracted social media data was ± 0.3 m. The challenge in using social media data is to filter the messages that
have water depth related information from the ample amount of messages posted in social media during disasters.
Keyword based query was developed and framed in MySQL to filter messages that have location and water height
mentions. The query was validated with tweets collected during the floods that hit Mumbai city in July 2019. The
validation results confirm that the query reduces the volume of tweets for manual evaluation and in future will aid in
mapping the water extent in near real time at the time of floods.
Keywords: Flood, Inundation map, Social media, Geographical information system

Introduction et al. 2016). Unanticipated heavy precipitation within a

Change in climate, urbanization and other human activi- short time span followed by flash floods in urban areas
ties across the globe disturbs the hydrological cycle and causes a greater loss in terms of lives, infrastructure and
cause various water related issues like water pollution, properties (Duan et al. 2014). In the past decade, occur-
floods, droughts, etc., (Lyu et al. 2019a; Luo et al. 2019, rence of urban floods increased drastically across India
2020). Especially cities face the problem of uneven dis- (Rafiq et al. 2016). Flood extent or water depth details
tribution of rainfall very often which leads to subsequent are required immediately after the disaster to identify
urban floods (Lyu et al. 2019b; Zou et al. 2020). Flash inundated areas that need quick attention(Blyth 1997).
flood disasters leave a massive social, environmental and Emergency managers need appropriate and rapid infor-
psychological impact on the affected community (Duan mation about severity of flooding for planning rescue
and response operations. Information on variations in
flood water depth with respect to time and space were
*Correspondence: [email protected] required for effective flood risk management (Luo et al.
1
School of Public Health, SRM Institute of Science and Technology, Tamil 2018; Mu et al. 2020). The unexpected, quick nature
Nadu, Chennai 603203, India
Full list of author information is available at the end of the article
of flood in urban localities due to very intense rainfall

© The Author(s) 2021. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which
permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the
original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or
other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line
to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory
regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this
licence, visit http://​creat​iveco​mmons.​org/​licen​ses/​by/4.​0/.
Karmegam et al. Geoenviron Disasters (2021) 8:25 Page 2 of 11

restricts getting water depth information during floods. time inundation mapping during floods is explored least
In general inundation map is prepared based on the field in Indian context. One of the biggest challenges of uti-
data, remote sensing data and hydraulic models(Grimaldi lizing social media data during disasters is huge volume
et al. 2016). Field data will be collected by sending field of posts shared by people on different aspects. Extract-
workers to the flooded areas, inspect the highest water ing useful and required information from the noisy,
mark after floods and based on this inundation maps volumous text becomes a barrier for emergency manag-
will be generated. Collecting field data have practical dif- ers (Hiltz and Kushma 2014). Earlier studies filtered the
ficulties and fail to provide timely data regarding flood flood related content based on hash-tags or geographic
extent. For instance, during Chennai flood 2015, author- locations (Lu et al. 2015; Woo et al. 2015; Murzintcev
ized official report on inundation map was released by and Cheng 2017), but again segregating inundation infor-
Disaster Management Support (DMS) Division, National mation containing messages from the flood related post
Remote Sensing Centre (NRSC/ISRO), India on March becomes a laborious task.
2016 after a field survey that was carried out on Decem- In this article, results of a feasibility study to filter and
ber 24 to 26, whereas flood disaster occurred on Decem- utilize social media content for flood mapping in Indian
ber 2 2015 (National Remote Sensing Centre 2015). context was provided. Flooding in the Chennai city
Utilizing remote sensing data for rapid water extent (Tamil Nadu), India in December 2015 was one of the
mapping have some limitations that includes restricted worst devastating, unexpected flooding events in India.
availability(Mason et al. 2012), limited spatial and tem- At the time of Chennai floods, affected people used social
poral resolutions(McDougall and Temple-Watts 2012). networking sites as a communication platform and that
Apart from these traditional data content, user generated helped in identification of people in need. Social media
crowd sourced content called volunteered geographical platforms played a major role after floods in rescue and
information(VGI) were also widely used for water extent relief operations (Prakash and Anand 2016). Disaster
mapping and validation(McDougall 2011; Hirata et al. management stakeholders and volunteers utilized social
2018; Rollason et al. 2018). Geo referenced data from networking sites to connect people in need and peo-
social media like Facebook, Twitter, etc., are also consid- ple who came forward to offer help (Yadav and Rahman
ered as VGI. The role of social media in management of 2016). Hence Flooding event in Chennai 2015 was con-
disaster situations was widely researched in past decade sidered as the case scenario in this study. Flood extent
(Lindsay 2011; Verma et al. 2011; Cameron et al. 2012; map was generated using social media information on
Middleton et al. 2014; Takahashi et al. 2015; Anson et al. water depths and location posted during Chennai floods
2017). Social media data ascending from the affected 2015 and validated against field data collected after
population has the potential to aid in creating situation floods. A simple keyword based query was developed to
awareness and planning response and rescue operations filter social media data that contains water depth infor-
(Huang and Xiao 2015; Lin et al. 2016; Mart et al. 2017). mation in case of urban floods in India. The developed
As the data from social media are posted real time with query was also validated with another disaster scenario.
no time delay, the same can be mined for rapid water
inundation mapping. Methods
Although social media was widely used as a tool for Study setting
information dissemination, early warnings and situational Greater Chennai Corporation (GCC), which is located
awareness during floods (David et al. 2016; Kaewkitipong at the state of Tamil Nadu in India, was considered
et al. 2016; Lin et al. 2016; Yadav and Rahman 2016; Alias as our study area. GCC is divided in to fifteen zones,
et al. 2020), exploring its utilization in water depth map- which is further subdivided into 200 wards. Chennai
ping was at infancy. Few recent studies investigated the city receives almost 60% of the annual rainfall during
potential of using social media information in inunda- north east monsoon period (from October to Decem-
tion mapping (Eilander et al. 2016; Brouwer et al. 2017; ber). Due to flat topography, some localities in Chen-
Li et al. 2018). Information got from social media about nai deal with the problem of poor drainage during
water depths were used in studies to validate the flood monsoons. Chennai experienced severe flooding due
extent from other bases and frameworks (Cervone et al. to heavy rainfall and the normal life of the population
2016; Smith et al. 2017). Previous researches also exam- across the city was troubled in 1976, 1985, 1996, 2005
ined the possibility of using water logging information and 2015. That is approximately once in every decade
in social media along with other sources of information (National Remote Sensing Centre 2015). During 2015,
for inundation mapping and risk assessment (Zhang et al. as per meteorological reports, Chennai received very
2016; Rosser et al. 2017; Wu et al. 2018). To the best of heavy rainfall of 1471.6 mm, far excessive than that it
our knowledge, the usage of social media content for real receives usually (915.6 mm—Normal rainfall) during
Karmegam et al. Geoenviron Disasters (2021) 8:25 Page 3 of 11

the monsoon between October and December (Indian and in some localities even up to second floor. The
Meteorological Department (IMD) 2016). City expe- population unaware of this sudden rise in water level
rienced one episode of substantial rainfall in the end were stranded at the terrace without any basic needs
of November 2015 about 1049 mm, which filled up all like food, water, etc. (National Disaster Management
the water bodies and water logging in some low lying Authority (NDMA) Government of India 2017). As per
areas. Again extremely high intensity of rainfall was government reports, around 1.8 million people from
recorded on December 1 and 2, 2015 at Nungambak- various localities were sent to relief camps at the time
kam and Chembarambakkam rain gauge stations, that of flood. As estimated by media, approximately 500
flooded the entire city. There was sudden increase in people lost their lives and there was around 200 million
water levels about 6–8 m in many areas across the city Rupees economic losses due to flood (Mujumdar et al.
on December 1st 2015. In few residential areas water 2016). Figure 1 shows the ward map of GCC (Chennai
entered in to the houses and reached till first floor Corporation 2011) with its zone name shown.

Fig. 1 Ward map of Greater Chennai Corporation

Karmegam et al. Geoenviron Disasters (2021) 8:25 Page 4 of 11

Methodology Maps. Following location the extraction of location,

Flood depth mapping water depths in these locations needs to be derived. In
Data from Twitter and Facebook public pages were uti- some messages, the data on water depth was specified
lized in this study for getting water depth information. directly in feet and meter. An example message where
The process followed for generating water extent maps the height of water was specified straight in a particular
from social media content is provided in Fig. 2. location is provided below:
As the objective is to generate water extent map
“Kannan Avenue (2nd Floor), Near annai arul hos-
rapidly in real time, messages posted on Twitter and
pital, Mudichur Road, Old Perungulathur. Stuck
Facebook public pages between 1 December 2015 and
up in second floor with a kid. No power and food.
3 December 2015 were considered. In Facebook, mes-
Area with surrounded with 11ft water”
sages shared in public pages only were considered
because messages posted by individuals are restricted There were also messages where the water height in
to public access based on their privacy settings. But particular area was mentioned with reference to some
the messages shared in the public pages can be used other aspects like up to first-floor, or hip-level, etc. One
without restriction. The public pages in Facebook cre- such example message is given below:
ated before December 3rd 2015 to share the details
“Managed to get out of west mambalam. Water
about flood situation and rescue activities in Chen-
levels we at knee/calf level at arya gowda road till
nai were identified. In Twitter, the hash-tags related to
panigraha hall.”
floods 2015 were identified by general look up of the
tweets. Twitter messages in English and Tamil with Figure 3 shows an example message posted in Face-
the identified hash-tags and tweets that originated book, from which location and water height were
from Chennai (25 miles around Chennai geographical derived from the text. In the message shown in Fig. 3,
coordinates—search option in Twitter) in the above the geo-coordinates of the location were extracted
mentioned time frame were collected. The messages based on the address and land mark provided using
in Twitter and Facebook pages were screened manu- Google Maps. The water height at that location was
ally to identify the messages that have both location mentioned as “up to 1st floor”, that which means the
and water depth information regarding the flood situ- water height will be approximately 3 m.
ation. The location information from the geo-coded Figure 4 shows two images (a and b) posted on social
messages was obtained directly from the coordinates media, from which water height and location were
specified. In other messages, the location was derived extracted. In the first image (Fig. 4a), location was
either from the address provided in the text or from the derived from the text and water height from the image.
image shared. Then the geographical coordinates of the In the second image (Fig. 4b), both location and water
specified location was derived utilizing using Google depth were derived from the photo shared. Messages

Twitter and Facebook community pages (Dec 1-3) Data collection and screening

Geo-tagged, location reference in text or image Derive locations

From text and image (from ground level) Water height extraction

Quantum GIS Mapping

IDW interpolation -QGIS Interpolation

Water height from field survey – Reports Validation

Fig. 2 Framework of water extent mapping from social media data
Karmegam et al. Geoenviron Disasters (2021) 8:25 Page 5 of 11

Fig. 3 Example message with location and water height mentions

Fig. 4 Sample images in social media with reference to location and water height

that reveal, a particular locality that had no water log- Chennai Corporation using Quantum geographical
ging was also considered as water depth point. Example information system (QGIS). The water height points
messages that reveal no inundation are given: that were outside the geographical extent of the city
were excluded. The water inundation map to the extent
“Our flat in #ValmikiNagar, # Thiruvanmiyur is dry
of Chennai city was generated by interpolating the
with Internet and electricity. Please get in touch if
water height points using Inverse Distance Weight-
you need help #ChennaiFloods”
ing (IDW) interpolation in QGIS. IDW is one of the
“Loyola college, Nungambakkam has accommoda-
commonly used deterministic, spatial interpolation
tion for rain victims. They have electricity”
methods in hydrological modelling. IDW interpolation
assumes that nearer values are more connected than
Once the water height and locations were derived, farther values and this method works best with dense
they were mapped over the base map of Greater point values in flat zones (Ly et al. 2013).
Karmegam et al. Geoenviron Disasters (2021) 8:25 Page 6 of 11

The generated water extent map after interpolation was level. Figure 5 shows the distribution of water height
validated with water height points reported after the field points over the base map.
assessment by the Disaster Management Support (DMS) The derived water height and location (geographical
Division, National Remote Sensing Centre (NRSC), India coordinate) were given in Additional file 2. Among these,
(National Remote Sensing Centre 2015). A field survey 72 points were within the geographical extent of Chen-
was done by DMS, NRSC, along with the Indian Institute nai. The generated water extent map for the entire city by
of Technology (IIT) Madras on 24th and 25th Decem- IDW interpolation was given in Fig. 6.
ber 2015 to collect data on water depth marks in the The actual water height from field survey report and
flood-affected areas in and around Chennai. From this water height from interpolated map got from social
report, field information on water depths in twelve loca- media at the 12 locations were tabulated in Table 1. As
tions were used to validate the water height derived from per the results of the t-test, there was no significant
social media data. The water height in those these twelve difference between mean of the actual data and the
locations (where actual information on water depth is extracted data of the water height from social media. The
available) was extracted from the interpolated map gen- root mean square error between the actual data and the
erated using social media content. In order to examine extracted social media data (water height) was 0.3.
the significance in the difference between the averages of The query in MySQL filters the messages, if it contains
two water heights (actual water depth and depth result- both water depth and location keywords. The combina-
ing from social media data), t-test was performed. Root tion of inundation keywords was handled using regular
mean square error (RMSE) was calculated to understand expressions in MySQL. The query was validated using
the error between the actual heights and water heights the tweets collected during Mumbai flood and that was
generated from social media. 17,846 messages excluding duplicates and re-tweets.
When these messages were filtered using the query, it
Query development returned 156 tweets. On checking the tweets manually
A simple keyword-based query particularly attuned we found that, 102 messages had water depth and loca-
to Indian urban setting was proposed to filter the mes- tion information that will be used for rapid water extent
sages that have both location and inundation informa- mapping. The query written to filter messages regarding
tion. Based on the experience on manual screening of Mumbai flood data was given in Additional file 3. The
messages posted during Chennai flood scenario, the key- screen shot of the executed query in MySQL workbench
words which the affected population used to mention the with results was given in Fig. 7. Sample tweets that had
water depths in their messages were identified. Water location and water height mentions, filtered by the query
depth keywords are usually a combination, such as a were given below.
number followed by metre or feet (5 feet, 10 cm), number “my college basement was floodedwalked in neck
followed by floor (for example—2nd floor), or indicative deep water with my colleagues up to lbs marg and
levels like ankle-high and neck-deep. Location informa- then on in waist deep water walked home for nearly
tion keywords include area and road names in the city. an hour and half this is from bhandup to mulund
The query was framed in such a way that the messages mumbairains”
will be filtered if it contains both the location and inun- “water raises above 3 feets at cm high school bad-
dation keywords. The query was framed in MySQL, an lapur water has started entering kitchen godown
open-source database management system. As the water and classrooms destroying uniforms and food. mum-
depth keywords were identified based on the experi- bairains badlapur mahalaxmiexpress ndrf mum-
ence with manual screening of Chennai flood data, we bairainsliveupdates”
validated the filtering query with tweets collected during
the floods that hit Mumbai city in July 2019. The tweets
related to Mumbai floods in 2019 were collected using
search API based on the hash-tags related to Mumbai Discussion
floods. Areas and road names of Mumbai city was used This study examined the feasibility of using social media
as the location keywords in that query. data, for water depth mapping in Indian disaster scenar-
ios. This article is the proof of concept that confirms the
Results potential of utilizing social media information for rapid
The Facebook pages related to Chennai flood, 2015 and flood mapping in Indian context for immediate response
the Twitter search query was given in Additional file 1. and recovery.
On manual screening, we derived 95 points with geo- The water depth map was generated from informa-
graphical coordinates and water height from ground tion in Twitter and Facebook messages, for the case of
Karmegam et al. Geoenviron Disasters (2021) 8:25 Page 7 of 11

Fig. 5 Distribution of water height points

Chennai flood, 2015. Based on derived water height water in to Adyar River(Mujumdar et al. 2016). As Adyar
points, we found that localities in Saidapet, Jafferkhanpet River flows through the city, the areas lying in close prox-
and Ashok nagar had the highest water depth more than imity to the river were highly inundated. That might be
4 m. During flood, Chembarambakam, one of the tanks the reason for very high water depth in above mentioned
that supplied water to the city, breached due to unex- areas, as they are closely located to the Adyar River. The
pected heavy downpour, releasing thousands of cusecs of interpolated water extent map also confirms that the
Karmegam et al. Geoenviron Disasters (2021) 8:25 Page 8 of 11

Fig. 6 Water inundation map after interpolation

areas close to Adyar River had higher water levels. As affected areas during floods (National Remote Sensing
per the results, the wards in Kodambakkam (ward 142) Centre 2015). We found that the error between the field
and Adyar (ward 171) zones had water height greater data and social media data was about ± 30 cm. This can
than 6 m. Both these ward are located in the bank of the be acceptable as this map was generated rapidly in real
river. These areas were also reported among the worst time at the time of disaster, when no other source of data
Karmegam et al. Geoenviron Disasters (2021) 8:25 Page 9 of 11

Table 1 Comparison of field data and derived data from social We believe that if this query is applied to the stream-
media (Water heights) ing online data in future during floods in India, this will
Id Latitude Longitude Water height in meter provide flood extent or inundation information without
delay for emergency management. This will aid in plan-
Field data Social
(actual) media
ning the rescue operations in accordance with the need
data of affected population.
As the water depth and locations are extracted manu-
1 13.01913 80.178095 3.51 3.35
ally with approximations, there is a possibility of inaccu-
2 13.024089 80.240774 3.20 2.97
racies in data. But at emergencies, when the information
3 12.973916 80.228116 1.37 1.22
on flood is very scarce, this information will definitely
4 13.027433 80.208179 3.66 3.58
give some understandings towards the flood situation. In
5 12.875667 80.226843 1.83 1.59
case of Chennai floods, some of the localities had power
6 12.980011 80.234378 0.76 1.19
cuts and network issues. This prevented the affected pop-
7 13.037385 80.244238 1.68 1.37
ulation from updating the flood status in social media.
8 12.904028 80.229995 0.61 1.35 So there is also a possibility that updates from highly
9 13.047948 80.177572 0.30 0.44 inundated and affected areas may be limited in social
10 13.028559 80.225438 3.05 3.10 media. Instead using social media information as stand-
11 13.056767 80.251894 0.30 0.29 alone information source, it can be used along with other
12 13.02217 80.206351 1.67 2.20 sources like remote sensing data and already available
Digital Elevation model to fill the information gap during
crisis.
was available. The main advantage of using social media Previous researches also mention that there is a possi-
data in emergency situations is that they are available at bility of uncertainties regarding location information in
near real time and also from the affected population, who social media posts (Brouwer et al. 2017; Ogie and Fore-
are the eye witnesses of the disaster situation (Fohringer head 2018). Development of national scale integrated
et al. 2015). The query that filters messages based on key- framework to fuse multiple data sources (social media,
words reduces the volume of messages considerably for remote sensing, topographic and environmental data) in
manual screening. This handles the problem of informa- real time by duly taking into account uncertainties in data
tion overload and aids in creating the flood map in time. sources for the purpose of generating precise real-time

Fig. 7 MySQL Workbench—Keyword based query and results (screen shot)

Karmegam et al. Geoenviron Disasters (2021) 8:25 Page 10 of 11

water depth maps at the time of flood could be a focus Declarations

area for future research.
Competing interests
The authors declare that they have no competing interests.
Conclusion
This study presented a method and keyword based query Author details
1
School of Public Health, SRM Institute of Science and Technology, Tamil Nadu,
to filter messages from social media that support extrac- Chennai 603203, India. 2 Department of Civil Engineering, SRM Institute of Sci-
tion of water height information for near real time flood ence and Technology, Chennai, Tamil Nadu 603203, India. 3 Centre for Statistics,
inundation mapping during urban floods in India. The SRM Institute of Science and Technology, Chennai, Tamil Nadu 603203, India.

results of the application circumstance Chennai flood, Received: 12 August 2020 Accepted: 2 September 2021
2015 was positive. The advantage of proposed methodol-
ogy to use social media information for mapping is the
rapid availability of data when compared to other tradi-
tional sources of information like remote sensing, satellite References
data, etc., particularly in urban setting. In future, during Alias NE, Salim NA, Taib SM et al (2020) Community responses on effective
flood dissemination warnings—a case study of the December 2014
floods this rapid flood map in real time will improve situ- Kelantan Flood, Malaysia. J Flood Risk Manag 13:e12552. https://​doi.​org/​
ational awareness and aid in efficient flood management. 10.​1111/​jfr3.​12552
Social media information on water heights will close the Anson S, Watson H, Wadhwa K, Metz K (2017) Analysing social media data for
disaster preparedness: understanding the opportunities and barriers
information openings in traditional information sources. faced by humanitarian actors. Int J Disaster Risk Reduct 21:131–139.
In future mapping framework and tool can be developed https://​doi.​org/​10.​1016/j.​ijdrr.​2016.​11.​014
that automatically derive information from social media Blyth K (1997) Floodnet: a telenetwork for acquisition, processing and dis-
semination of earth observation data for monitoring and emergency
by text and image analysis and integrating with other management of floods. Hydrol Process 11:1359–1375. https://​doi.​org/​10.​
sources of information available to acquire a more accu- 1002/​(sici)​1099-​1085(199708)​11:​10%​3c135​9::​aid-​hyp529%​3e3.0.​co;2-6
rate inundation maps in real time. Brouwer T, Eilander D, Van Loenen A et al (2017) Probabilistic flood extent
estimates from social media flood observations. Nat Hazards Earth Syst
Sci 17:735–747. https://​doi.​org/​10.​5194/​nhess-​17-​735-​2017
Cameron MA, Power R, Robinson B, Yin J (2012) Emergency situation aware-
Abbreviations ness from twitter for crisis management. In: Proceedings of the 21st
DMS: Disaster Management Support; NRSC: National Remote Sensing Centre; international conference companion on World Wide Web—WWW ’12
ISRO: Indian Space Research Organization; VGI: Volunteered geographical Companion. ACM Press, New York, New York, USA, p 695
information; GCC​: Greater Chennai Corporation; QGIS: Quantum geographi- Cervone G, Sava E, Huang Q et al (2016) Using Twitter for tasking remote-
cal information system; IDW: Inverse Distance Weighting; RMSE: Root mean sensing data collection and damage assessment: 2013 Boulder flood
square error. case study. Int J Remote Sens 37:100–124. https://​doi.​org/​10.​1080/​01431​
161.​2015.​11176​84
Supplementary Information Chennai Corporation (2011) Welcome to Corporation of Chennai. https://​
chenn​aicor​porat​ion.​gov.​in/​Land_​Estate/​city_​expan​sion.​html. Accessed
The online version contains supplementary material available at https://​doi.​ 29 Jul 2020
org/​10.​1186/​s40677-​021-​00195-x. David CC, Ong JC, Legara EFT (2016) Tweeting supertyphoon Haiyan: evolving
functions of twitter during and after a disaster event. PLoS ONE 11:1–19.
Additional file 1. List of Facebook community pages and Twitter search https://​doi.​org/​10.​1371/​journ​al.​pone.​01501​90
query for data collection. Duan W, He B, Takara K et al (2014) Anomalous atmospheric events leading
to Kyushu’s flash floods, July 11–14, 2012. Nat Hazards 73:1255–1267.
Additional file 2. Location and water height derived from social media https://​doi.​org/​10.​1007/​s11069-​014-​1134-3
data. Duan W, He B, Nover D et al (2016) Floods and associated socioeconomic
Additional file 3. MySQL Query. damages in China over the last century. Nat Hazards 82:401–413. https://​
doi.​org/​10.​1007/​s11069-​016-​2207-2
Eilander D, Trambauer P, Wagemaker J, Van Loenen A (2016) Harvesting
Acknowledgements social media for generation of near real-time flood maps. Procedia Eng
Not applicable. 154:176–183
Fohringer J, Dransch D, Kreibich H, Schröter K (2015) Social media as an infor-
Authors’ contributions mation source for rapid flood inundation mapping. Nat Hazards Earth
Conceptualization and methods: BM, SR and DK. Data collection and analysis: Syst Sci 15:2725–2738. https://​doi.​org/​10.​5194/​nhess-​15-​2725-​2015
DK. Supervision: BM and SR. Writing Draft: DK. Review and editing draft manu- Grimaldi S, Li Y, Pauwels VRN, Walker JP (2016) Remote sensing-derived water
script: BM and SR. All authors read and approved the final manuscript. extent and level to constrain hydraulic flood forecasting models: oppor-
tunities and challenges. Surv Geophys 37:977–1034
Funding Hiltz SR, Kushma J (2014) Use of SOCIAL MEDIa by U.S. public sector emer-
None. gency managers : barriers and wish lists. In: Proc 11th Int ISCRAM Conf,
pp 602–611. https://​doi.​org/​10.​13140/2.​1.​3122.​4005
Availability of data and materials Hirata E, Giannotti MA, Larocca APC, Quintanilha JA (2018) Flooding and inun-
The datasets used and analyzed during the current study are available from dation collaborative mapping—use of the Crowdmap/Ushahidi platform
the corresponding author on reasonable request. in the city of Sao Paulo, Brazil. J Flood Risk Manag 11:S98–S109. https://​
doi.​org/​10.​1111/​jfr3.​12181
Karmegam et al. Geoenviron Disasters (2021) 8:25 Page 11 of 11

Huang Q, Xiao Y (2015) Geographic situational awareness: mining tweets for Mujumdar P, Narasimhan B, Bhallamudi M et al (2016) Chennai floods 2015,
disaster preparedness, emergency response, impact, and recovery. ISPRS a rapid assessment. Interdisciplinary Centre for Water Research Indian
Int J Geo-Inf 4:1549–1568. https://​doi.​org/​10.​3390/​ijgi4​031549 Institute of Science, Bangalore
Indian Meteorological Department (IMD) (2016) District—wise rainfall Tamil Murzintcev N, Cheng C (2017) Disaster Hashtags in Social Media. ISPRS Int J
nadu2015 Geo-Inf 6:204. https://​doi.​org/​10.​3390/​ijgi6​070204
Kaewkitipong L, Chen CC, Ractham P (2016) A community-based approach National Remote Sensing Centre (2015) Chennai floods, 2015 [A satellite and
to sharing knowledge before, during, and after crisis events: a case study field based assessment study]
from Thailand. Comput Human Behav 54:653–666. https://​doi.​org/​10.​ National Disaster Management Authority (NDMA) Government of India (2017)
1016/j.​chb.​2015.​07.​063 Best practices adopted by government of Tamil Nadu post 2015 floods—
Li Z, Wang C, Emrich CT, Guo D (2018) A novel approach to leveraging social a study report by NDMA
media for rapid flood mapping: a case study of the 2015 South Carolina Ogie RI, Forehead H (2018) Investigating the accuracy of georeferenced social
floods. Cartogr Geogr Inf Sci 45:97–110. https://​doi.​org/​10.​1080/​15230​ media data for flood mapping: the PetaJakarta.org case study. In: Proc
406.​2016.​12713​56 2017 4th Int Conf Inf Commun Technol Disaster Manag ICT-DM 2017
Lin X, Spence PR, Sellnow TL, Lachlan KA (2016) Crisis communication, learning 2018-January, pp 1–6. https://​doi.​org/​10.​1109/​ICT-​DM.​2017.​82756​72
and responding: best practices in social media. Comput Human Behav Prakash G, Anand E (2016) Indian News Media and natural calamities: case of
65:601–605. https://​doi.​org/​10.​1016/j.​chb.​2016.​05.​080 Chennai floods. Int J Multidiscip Approach Stud 03:166–177
Lindsay BR (2011) Social media and disasters: current uses, future options and Rafiq F, Ahmed S, Ahmad S, Khan AA (2016) Urban floods in India. Int J Sci Eng
policy considerations. Congr Res Serv Reports 13 Res 7:721–734
Lu Y, Hu X, Wang F et al (2015) Visualizing social media sentiment in disaster Rollason E, Bracken LJ, Hardy RJ, Large ARG (2018) The importance of volun-
scenarios. In: Proceedings of the 24th international conference on World teered geographic information for the validation of flood inundation
Wide Web—WWW ’15 Companion. pp 1211–1215 models. J Hydrol 562:267–280. https://​doi.​org/​10.​1016/j.​jhydr​ol.​2018.​05.​
Luo P, Mu D, Xue H et al (2018) Flood inundation assessment for the Hanoi 002
Central Area, Vietnam under historical and extreme rainfall conditions. Sci Rosser JF, Leibovici DG, Jackson MJ (2017) Rapid flood inundation mapping
Rep 8:1–11. https://​doi.​org/​10.​1038/​s41598-​018-​30024-5 using social media, remote sensing and topographic data. Nat Hazards
Luo P, Kang S, Apip et al (2019) Water quality trend assessment in Jakarta: a 87:103–120. https://​doi.​org/​10.​1007/​s11069-​017-​2755-0
rapidly growing Asian megacity. PLoS ONE. https://​doi.​org/​10.​1371/​journ​ Smith L, Liang Q, James P, Lin W (2017) Assessing the utility of social media
al.​pone.​02190​09 as a data source for flood risk management using a real-time modelling
Luo P, Sun Y, Wang S et al (2020) Historical assessment and future sustainabil- framework. J Flood Risk Manag 10:370–380. https://​doi.​org/​10.​1111/​jfr3.​
ity challenges of Egyptian water resources management. J Clean Prod 12154
263:121154. https://​doi.​org/​10.​1016/j.​jclep​ro.​2020.​121154 Takahashi B, Tandoc EC, Carmichael C (2015) Communicating on Twitter
Ly S, Charles C, Degré A (2013) Different methods for spatial interpolation during a disaster: an analysis of tweets during Typhoon Haiyan in the
of rainfall data for operational hydrology and hydrological modeling at Philippines. Comput Hum Behav 50:392–398. https://​doi.​org/​10.​1016/j.​
watershed scale A review. Biotechnol Agron Soc Env 17:392–406 chb.​2015.​04.​020
Lyu J, Luo P, Mo S et al (2019a) Towards sustainable water regulation based on Verma S, Vieweg S, Corvey WJ et al (2011) Natural language processing to
a distributed hydrological model for a heavily polluted urban river, north- the rescue? Extracting “Situational Awareness’’’ tweets during mass
west China. Hydrol Res 50:961–973. https://​doi.​org/​10.​2166/​nh.​2019.​005 emergency.” In: Proceedings of the fifth international AAAI Conference on
Lyu J, Mo S, Luo P et al (2019b) A quantitative assessment of hydrological Weblogs and Social Media, pp 385–392
responses to climate change and human activities at spatiotempo- Woo H, Cho Y, Shim E et al (2015) Public trauma after the Sewol Ferry disaster:
ral within a typical catchment on the Loess Plateau, China. Quat Int the role of social media in understanding the public mood. Int J Environ
527:1–11. https://​doi.​org/​10.​1016/j.​quaint.​2019.​03.​027 Res Public Health. https://​doi.​org/​10.​3390/​ijerp​h1209​10974
Mart Y, Carolina S, Carolina S (2017) Leveraging Twitter to gauge evacuation Wu X, Cao Y, Xiao Y, Guo J (2018) Finding of urban rainstorm and waterlogging
compliance: spatiotemporal analysis of Hurricane Matthew. PLoS ONE. disasters based on microblogging data and the location-routing problem
https://​doi.​org/​10.​1371/​journ​al.​pone.​01817​01 model of urban emergency logistics. Ann Oper Res. https://​doi.​org/​10.​
Mason DC, Davenport IJ, Neal JC et al (2012) Near real-time flood detection 1007/​s10479-​018-​2904-1
in urban and rural areas using high-resolution synthetic aperture radar Yadav M, Rahman Z (2016) The social role of social media: the case of Chen-
images. IEEE Trans Geosci Remote Sens 50:3041–3052. https://​doi.​org/​10.​ nai rains-2015. Soc Netw Anal Min 6:101. https://​doi.​org/​10.​1007/​
1109/​TGRS.​2011.​21780​30 s13278-​016-​0410-5
McDougall K, Temple-Watts P (2012) The use of lidar and volunteered Zhang N, Chen H, Chen J, Chen X (2016) Social media meets big urban data:
geographic information to map flood extents and inundation. In: ISPRS a case study of urban waterlogging analysis. Comput Intell Neurosci.
annals of the photogrammetry, remote sensing and spatial information https://​doi.​org/​10.​1155/​2016/​32645​87
sciences. Copernicus GmbH, pp 251–256 Zou S, Abuduwaili J, Ding J et al (2020) Description and attribution analysis
McDougall K (2011) Using volunteered information to map the Queensland of the 2017 spring anomalous high temperature causing floods in
floods. In: Proceedings of the 2011 Surveying and Spatial Sciences confer- Kazakhstan. J Meteorol Soc Jpn Ser II 98:1353–1368. https://​doi.​org/​10.​
ence. Surveying and Spatial Sciences Institute, pp 13–23 2151/​jmsj.​2020-​070
Middleton SE, Middleton L, Modafferi S (2014) Real-time crisis mapping of
natural disasters using social media. IEEE Intell Syst 29:9–17. https://​doi.​
org/​10.​1109/​MIS.​2013.​126 Publisher’s Note
Mu D, Luo P, Lyu J et al (2020) Impact of temporal rainfall patterns on flash Springer Nature remains neutral with regard to jurisdictional claims in pub-
floods in Hue City, Vietnam. J Flood Risk Manag 14:e12668. https://​doi.​ lished maps and institutional affiliations.
org/​10.​1111/​jfr3.​12668