the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Exploring the utility of social media data for urban flood impact assessment in data scarce cities
Abstract. The growing amount of social media data is a valuable and rapidly available information source to inform flood response and recovery. In this study, a workflow framework is developed to assess urban flood impacts by extracting and analysing social media data, as well as identifying the intensive public response areas, using the case of 2020 China Chengdu rainstorm-induced flooding. A crawler-algorithm is applied to extract and filter the social media data from the commonly used social platforms, namely Weibo (static data) and Tiktok (dynamic data). Based on the spatiotemporal analysis and the identified 232 flood sites with geological locations, the study shows that, social media activities and precipitation have a significant positive correlation temporally. The temporal evolution analysis of social media topics reveals the process of flooding enabling quickly to determine the severely affected areas. Spatially, social media data can give spatial flood information and social media activities are generally associated with the demographical distribution of users. Based on a flood simulation, the framework can generate reliable data source of urban flooding from social media, which can enhance flood risk modelling with the aid of hydrodynamic model. This study demonstrates the utility of social media data for urban flood assessment.
- Preprint
(1649 KB) - Metadata XML
-
Supplement
(31951 KB) - BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on nhess-2022-109', Anonymous Referee #1, 02 Jun 2022
The manuscript deals with a very relevant topic and is generally well written. It is the result of an ambitious and impressive effort to explore the utility of social media data for urban flood impact assessment in data scarce cities. The method presented here promises in itself to be very valuable, and applicable in future flood impact studies. However, I have only major concern on the flood modeling and validating. Can you provide more informations on the methodology used (modeling process) and some disscusions.
Citation: https://doi.org/10.5194/nhess-2022-109-RC1 -
AC1: 'Reply on RC1', Kaihua Guo, 13 Jul 2022
Response: We appreciate the reviewer for the positive comments. For the flood modeling and validation part, we will present more detailed description about the input data, the model, and the setup of model parameters in section 4.1of revised manuscript. The model we are using is the commonly used model based on Shallow Water Equations that has been well studied in past decade. The supportiveness of social media to the simulation results will be further discussed in this section as well, to explore the applicability of social media data in flood impact assessments in data-scare cities. Please noted: as indicated in the manuscript, the main aim of the manuscript is to develop a replicable and optimized social media data processing method in order to better filter and identify valuable flood data from a large amount of ordinary unstructured public social information full of intensive unknown interventions. Therefore, the flood modelling part here is mainly for the purpose of demonstrating the utility of the gathered data.
Citation: https://doi.org/10.5194/nhess-2022-109-AC1
-
AC1: 'Reply on RC1', Kaihua Guo, 13 Jul 2022
-
RC2: 'Comment on nhess-2022-109', Zhang Hongping, 12 Jun 2022
Urban flooding is a hot issue in the world. There are still many challenges in urban flood emergency response, one of which is collecting the timely flooding information. This research demonstrates a technological workflow to get flooding information from social media and deeply mine valuable information from them. It’s doubtlessly a good try, though there is still a long way to put it in practical use. Here are some specific comments and suggestions as following.
- About the introduction section:
The introduction section seems a little logically weak and needs a better organization. For example:
(1)The content in the first paragraph seems a little far away from the research topic. I would like to introduce the topic directly from urban floods in the world, and then explain why we need to get data from social media for urban flood emergency management and what the difficulties are.
(2)At the beginning of the second paragraph, it would be better to state the urban flood situation in a worldwide view, not only from China, because urban flooding is a worldwide hot issue. China can be used as one of examples.
(3)The traditional observation methods are mentioned as “costly, delay and time-consuming” methods in Line 57. But discussion about the traditional observation methods seems insufficient. It would be better to give more discussion on what kinds of traditional observation methods there are and how costly, delay and time-consuming they are before Line 57.
(4) Line 45 seems to indicate that the existing researches use the unified format data. If there are unified format data, why do we need to use the unformatted data from Weibo and Tiktok? Please give more explanations about it. It would be better to give more details about the existing researches on utilizing the social media data for flood emergency management. What have they done and what are their limits?
(5)From Line 57, two research questions are listed. There might be one more question: how are the data from social media validated before they are used? Or what are the uncertainties when they are used? Please give a little more statements about it.
- About Table 2:
(1)Table 2 seems a little confusing. Usually the first row is the header which defines the content of each column, and each column has contents matching with the header. In this table, row 5 seems like a new header, and row 7 doesn't seem to match either headers. It would be better to reorganize this table.
(2) More explanation about time flag in Table 2 would be better. Does it mean: the Pearson correlation coefficient between the hourly precipitation and the hourly tweet number with a certain time lag? If it does, please specifically clarify the hours of precipitation and hours of tweets analyzed during calculating their Pearson correlation coefficient.
(3) Line 212 describes Storm 1 filtering as by excluding relevant discussions. Please give more explanations on how relevant discussions are identified in a huge number of tweets.
- About Table 4.
It might be more figurative to transpose this table by placing the check-in data and flood-point data for one district in one row in order to give a better comparison.
- About Section 4
Some contents in Section 4 seem a week relation with the results of this research. For example:
(1)In Section 4.1 the author seems to discuss the value or the role of social media for various stakeholders. It’s a fact, but not a result of this research. It might be better to use this content to develop the necessity of this research in the introduction section.
(2)Section 4.3 is titled as “comparison with other platforms and countries”. I would like to make comparison just between researches, weakening the idea of countries. Furthermore, the content in this section seems like the progress of other similar research. Is it possible to put it in the introduction section as existing research progress?
(3)Section 4.2 gives a good example of deeper mining of the value of social media data. It might be better to focus on this example and give more details about it in Section 4.
- Others
Some other minor questions or suggestions are seen in the manuscript.
- AC2: 'Reply on RC2', Kaihua Guo, 13 Jul 2022
Status: closed
-
RC1: 'Comment on nhess-2022-109', Anonymous Referee #1, 02 Jun 2022
The manuscript deals with a very relevant topic and is generally well written. It is the result of an ambitious and impressive effort to explore the utility of social media data for urban flood impact assessment in data scarce cities. The method presented here promises in itself to be very valuable, and applicable in future flood impact studies. However, I have only major concern on the flood modeling and validating. Can you provide more informations on the methodology used (modeling process) and some disscusions.
Citation: https://doi.org/10.5194/nhess-2022-109-RC1 -
AC1: 'Reply on RC1', Kaihua Guo, 13 Jul 2022
Response: We appreciate the reviewer for the positive comments. For the flood modeling and validation part, we will present more detailed description about the input data, the model, and the setup of model parameters in section 4.1of revised manuscript. The model we are using is the commonly used model based on Shallow Water Equations that has been well studied in past decade. The supportiveness of social media to the simulation results will be further discussed in this section as well, to explore the applicability of social media data in flood impact assessments in data-scare cities. Please noted: as indicated in the manuscript, the main aim of the manuscript is to develop a replicable and optimized social media data processing method in order to better filter and identify valuable flood data from a large amount of ordinary unstructured public social information full of intensive unknown interventions. Therefore, the flood modelling part here is mainly for the purpose of demonstrating the utility of the gathered data.
Citation: https://doi.org/10.5194/nhess-2022-109-AC1
-
AC1: 'Reply on RC1', Kaihua Guo, 13 Jul 2022
-
RC2: 'Comment on nhess-2022-109', Zhang Hongping, 12 Jun 2022
Urban flooding is a hot issue in the world. There are still many challenges in urban flood emergency response, one of which is collecting the timely flooding information. This research demonstrates a technological workflow to get flooding information from social media and deeply mine valuable information from them. It’s doubtlessly a good try, though there is still a long way to put it in practical use. Here are some specific comments and suggestions as following.
- About the introduction section:
The introduction section seems a little logically weak and needs a better organization. For example:
(1)The content in the first paragraph seems a little far away from the research topic. I would like to introduce the topic directly from urban floods in the world, and then explain why we need to get data from social media for urban flood emergency management and what the difficulties are.
(2)At the beginning of the second paragraph, it would be better to state the urban flood situation in a worldwide view, not only from China, because urban flooding is a worldwide hot issue. China can be used as one of examples.
(3)The traditional observation methods are mentioned as “costly, delay and time-consuming” methods in Line 57. But discussion about the traditional observation methods seems insufficient. It would be better to give more discussion on what kinds of traditional observation methods there are and how costly, delay and time-consuming they are before Line 57.
(4) Line 45 seems to indicate that the existing researches use the unified format data. If there are unified format data, why do we need to use the unformatted data from Weibo and Tiktok? Please give more explanations about it. It would be better to give more details about the existing researches on utilizing the social media data for flood emergency management. What have they done and what are their limits?
(5)From Line 57, two research questions are listed. There might be one more question: how are the data from social media validated before they are used? Or what are the uncertainties when they are used? Please give a little more statements about it.
- About Table 2:
(1)Table 2 seems a little confusing. Usually the first row is the header which defines the content of each column, and each column has contents matching with the header. In this table, row 5 seems like a new header, and row 7 doesn't seem to match either headers. It would be better to reorganize this table.
(2) More explanation about time flag in Table 2 would be better. Does it mean: the Pearson correlation coefficient between the hourly precipitation and the hourly tweet number with a certain time lag? If it does, please specifically clarify the hours of precipitation and hours of tweets analyzed during calculating their Pearson correlation coefficient.
(3) Line 212 describes Storm 1 filtering as by excluding relevant discussions. Please give more explanations on how relevant discussions are identified in a huge number of tweets.
- About Table 4.
It might be more figurative to transpose this table by placing the check-in data and flood-point data for one district in one row in order to give a better comparison.
- About Section 4
Some contents in Section 4 seem a week relation with the results of this research. For example:
(1)In Section 4.1 the author seems to discuss the value or the role of social media for various stakeholders. It’s a fact, but not a result of this research. It might be better to use this content to develop the necessity of this research in the introduction section.
(2)Section 4.3 is titled as “comparison with other platforms and countries”. I would like to make comparison just between researches, weakening the idea of countries. Furthermore, the content in this section seems like the progress of other similar research. Is it possible to put it in the introduction section as existing research progress?
(3)Section 4.2 gives a good example of deeper mining of the value of social media data. It might be better to focus on this example and give more details about it in Section 4.
- Others
Some other minor questions or suggestions are seen in the manuscript.
- AC2: 'Reply on RC2', Kaihua Guo, 13 Jul 2022
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
753 | 528 | 56 | 1,337 | 73 | 52 | 52 |
- HTML: 753
- PDF: 528
- XML: 56
- Total: 1,337
- Supplement: 73
- BibTeX: 52
- EndNote: 52
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1