What We Have Learned about the Characteristics and Differences of Disaster Information Behavior in Social Media—A Case Study of the 7.20 Henan Heavy Rain Flood Disaster

Abstract

As a significant human behavior, disaster information behavior may operate as a catalyst for affecting the evolution of disaster occurrences in social-ecological systems and the sustainability of social systems. Yet little research has been carried out on this subject, particularly on the information behavior of major natural disasters. Based on the case of the 7.20 Henan heavy rain flood disaster, this study constructs an information behavior composite index from the four dimensions of temporal, spatial, content, and behavioral agents and statistically identifies and quantifies the characteristics and differences of disaster information behavior in social media. The results are as follows. (1) Disaster information behaviors have an obvious life cycle with three phases, essentially following the “formation-development-extinction” process; disaster areas, near-disaster areas, and economically and technologically developed areas exhibit higher levels of information behavior. (2) A total of 47% of the content is related to the case, while 53% is unrelated; the most related microblogs (43.88%) were about “Disaster response/relief”. (3) Females (54.19%) engage in more information behavior than males (45.81%) and they also exhibit more positive behavior; the 20–29-year-old age group is dominated by positive and neutral comments with the highest level of information behavior, whereas the lowest level of information behavior occurs in the 50+ age group; neutral and irrelevant comments in the 30–49-year-old age group dominated. This case study enables a scientific understanding of the necessity of information dissemination for disaster prevention and mitigation and further demonstrates the hazard, psychological distance, societal, and individual factors that all affect how disaster information behaves and performs differently.

1. Introduction

Information behavior is a very broad notion that encompasses all human behavior toward information resources and channels, including both positive and negative information-seeking and information-use behavior [1]. It is a stage that humans invariably go through in the process of exploring knowledge. The process by which the general population searches for and then uses various sorts of information linked to disaster occurrences and disaster aid can be described as disaster information behavior in the context of natural disaster events. The Web 3.0 era, which is signaling the next stage or iteration of internet development, is founded on the fundamental ideas of decentralization, openness, and better user utility, enabling quicker and easier information distribution via social media. Sina Weibo, the most widely used social media platform by the Chinese public, has changed the way information is disseminated, enabling the instant sharing of information. Through Sina Weibo, information about the disaster spread rapidly, and everything users did, including topic searches, posting, forwarding, and commenting, was finally reflected in the Sina Weibo records.

Analyzing the processes and mechanisms of disaster information dissemination is an essential part of integrated disaster risk governance and disaster management [2]. However, integrated disaster risk governance is one of the most critical tasks of social management and an essential component of achieving sustainable socioeconomic development [3]. Natural disasters caused by extreme climate events are having a devastating impact on the socioecological system against the backdrop of global warming (SES). From the beginning of the International Decade for Natural Disaster Reduction (IDNDR, 1990–1999) to the 7th Session of the Global Platform for Disaster Risk Reduction in 2022, how to effectively adapt to climate change, reduce disaster risks, reduce disaster vulnerability, and strengthen social-ecosystem resilience, achieving the Sustainable Development Goals (SDGs) at the national and global levels is a topic that has been closely followed and deeply discussed [4,5,6,7].

In recent decades, the research on disaster information dissemination has mostly focused on crisis communication and the reporting of emergencies, primarily on social security emergencies and public health emergencies, but less on major natural disasters and other emergencies relatively [8,9,10]. It is also uncommon to discuss research that integrates the dissemination of information about major natural disasters with all-encompassing disaster risk governance. The existing case studies have been undertaken on the topic of integrated disaster risk governance under the influence of climate change from various angles, such as risk identification, risk assessment, risk prevention, and control measures (engineering technology measures, infrastructure development, government organization construction, etc.) [11,12,13,14]. Additionally, several studies have produced unique viewpoints on how to approach integrated disaster-risk prevention and management by integrating human behavior in social-ecological systems [15,16]. The analysis of governmental organizational behavior and governmental/social disaster-response behavior, for instance, is used to analytically illustrate the processes and mechanisms of disaster-risk prevention [17,18,19,20]. Yet few studies on information behavior regarding disasters have been studied, despite it being a significant human behavior in integrated disaster risk governance.

With the advent of the Web 3.0 era, disaster or crisis information spreads rapidly on social media platforms, further amplifying the impact of related events on the social-ecological system. The public, however, exhibits an extremely high level of information that is required in the face of disasters or crises, with the public more likely to seek out pertinent event information via searching, posting, forwarding, and commenting behaviors [21]. The information behavior in social media breaks the restriction of time and space and is more participatory and interactive [22,23]. The information behavior that carries various sentiments becomes the main trigger to accelerate the dissemination of disaster information, and through the feedback function of social media, it collects ideas and opinions from community groups and the public who may be affected by the disaster or crisis events, which affects the subsequent information behavior, and connections to the public’s thoughts, sentiments, and feelings to then influence the subsequent related events response [24,25,26,27,28]. The question, thus, becomes, what characteristics will the behaviors in social media for specific disaster information show?

As an essential element that influences the evolution of disaster events in social systems, disaster information dissemination pervades the whole process and all aspects of integrated disaster risk governance and is particularly vital in the prevention, preparation, and response phases of the disaster emergency management cycle [2]. Previous studies have shown that disaster information behavior can play an important role in regulating and controlling the evolution of disaster events, including communication, decision-making assistance, and stimulating, amplifying, or inhibiting the transition of social operating states [29,30]. Natural disaster events, such as extreme precipitation, can, for example, superimpose the changes of natural hazards and eventually evolve into a composite event in which sudden natural disasters interact with elements of the social system via the interaction and interweaving of the system’s cascade coupling utility. China is a sensitive and significant area of global climate change, and as global warming intensifies, the extreme climate events in the region tend to intensify, and the level of climate risk is on the rise [31]. In July 2021, Henan Province experienced a historically rare rainstorm, resulting in severe flooding and significant losses to social and economic development. By 16:00 on July 21, Henan’s heavy rain had generated 30.3233 million reports/discussions on the Weibo platform, garnering widespread social attention [32]. The public used the Sina Weibo platform to share emergency and rescue information during the “7.20 Henan heavy rain flood disaster”. Tens of thousands of pieces of information were commented on and retweeted, bridging the gap between emergency seekers and rescue teams within 24 h of the disaster and allowing for timely, effective, and organized disaster relief; disaster information behavior may be a catalyst for influencing the evolutionary process of disaster events in social-ecological systems.

Therefore, in this study, we specifically focus on the public’s disaster information behavior related to the “7.20 Henan heavy rain flood disaster” by asking the following research questions:

• RQ1 What characteristics and differences will the disaster information behavior of social media show in the context of severe and sudden natural disasters?
• RQ2 What is the reference significance of this particular performance in terms of actively promoting integrated disaster risk governance?

We gathered original data on precipitation and information behavior during the 7.20 Henan heavy rain flood disaster in order to respond to the questions above. After processing and analyzing the original data, we identified the characteristics of disaster information behavior in social media in four dimensions: temporal, spatial, content, and behavioral agents, and we discussed the reasons for the differences in disaster information behavior. Finally, we put forward some implications for integrated disaster risk governance in the context of network information dissemination. The results of this research provide evidence for understanding the public’s information behavior in scenarios involving abrupt, significant natural disasters.

2. Materials and Methods

2.1. Study Case Selection

China’s Henan Province was hit by a historically rare heavy rain event from 17 July to 23 July 2021. Especially on 20 July, the capital of Henan Province, Zhengzhou, suffered from severe urban waterlog, river flooding, landslides, and other multidisaster complications due to heavy rainfall, which significantly damaged both the people’s lives and property. The number of deaths and disappearances in Zhengzhou City accounted for 95.5% of the province.

The abnormal distribution of mid-latitude and low-latitude multi-scale pressure and associated vortex systems caused continuous water vapor transportation to Henan province. The abnormal weather system, coupled with local topographical effects, resulted in extreme precipitation. The recorded first precipitation process from 17 to 18, 2021, mostly affected northern Henan (Jiaozuo, Xinxiang, Hebi, and Anyang); the recorded second precipitation process from 19 to 20, the rain center moved south to Zhengzhou; the recorded third precipitation process from 21 to 22, the rain center again moved north, the precipitation process gradually weakened to end on 23 [33].

A State Council Disaster Investigation Team of China verified and determined that this heavy rain has the characteristics of long duration, wide range, large total, and extreme precipitation process. The precipitation process has produced surface rain totals of up to 589 mm, process point totals of up to 1122.6 mm, and the strongest hourly point totals of up to 201.9 mm, all of which are much above the extreme values of the past. The north-central part of Henan Province suffered severe flooding as a result of this extremely heavy rain, with 12 main rivers cresting the warning level. This disaster affected 150 counties (cities and districts) in Henan Province, resulting in 302 fatalities and a total of 14.786 million people exposed to the disaster, with direct economic damages of RMB 120.06 billion (Figure 1) [33]. According to the disaster grade calculation standard, the case of a heavy rain flood disaster in Henan Province has a disaster grade of up to 16.7, which is much greater than the average disaster grade of 9 [34]. Consequently, the “7.20 Henan heavy rain flood disaster” is a typical case of natural disaster that has a serious impact on the economy and society.

2.2. Data Sources

We extended the time range of the required data, specifically from 12 July to 5 September 2021, which facilitates the complete reproduction of the dynamic changes in the information behavior of this case during the development of the disaster. This is because the precipitation process of the 7.20 heavy rain flood disaster in Henan Province started on 17 July and ended on 23 July 2021.

(1)

Precipitation data are obtained from the China Meteorological Data Service Centre Daily Value Dataset of China Surface Climate Information (V3.0) (data source: http://cdc.cma.gov.cn (accessed on 12 September 2022)), which has undergone strict quality control and data checking to present the precipitation process of the case with high accuracy and resolution;

(2)

Information behavior data were obtained from the text and number of microblogs that posted, forwarded, and commented on via the public about the case on Sina Weibo (Weibo). Sina Weibo is a Chinese microblogging website. Launched by Sina Corporation on 14 August 2009, it is one of the biggest social media platforms in China, with over 582 million monthly active users as of Q1 2022. It is used by people looking for informational and trending content, including topic searches, posting, forwarding, and commenting, and was finally reflected in the Sina Weibo records.

2.3. Research Methods

2.3.1. Information Behavior Data Acquisition

The study filtered five Hot Topics using the well-known Weibo topics filter, including “#Henan rainstorm#”, “#Henan rainstorm rescue#”, “#Henan rainstorm 24 h later#”, and “#Henan rainstorm mutual help#”, among others. Social media data is multifarious and disorderly; thus, to make the data more representative for this study, the hot Weibo topics filtering function was used. We used Python to extract 25,590 popular comments and 16,922 microblogs from the case of “7.20 Henan heavy rain flood disaster” instance between 12 July and 5 September 2021. Examples of the original data discovered there include the user ID, user gender, user age, post content, post time, post location, and the number of comments and forwards.

2.3.2. Data Preparation

(1)

Identification, classification, and statistics of original data content

This section was mainly carried out using grounded theory, natural language processing, machine learning, and literature information interpretation methods.

① According to

Table 1. Attribute/status judgment words/phrases of case in hot microblogs.

Category		Attributes/Status Words/Phrases	Attributes/Status Extensions
Hazard factors		Rain	Coudburst, Rainstorm, Downpour, Torrential rain, Heavy rain, Severe rain, Heavy precipitation, etc.
Disaster loss and exposure	Disaster loss	Disaster, Hazard, Death	Damage, Missing/disappearing, Injured, Trapped, etc.
		Direct economic Losses	Reduce the tax sources, Reduce non-tax revenue, etc.
		House collapse	Damaged house, Flooded house, etc.
		Blocked/disrupted Traffic	Road closures, Stranded passengers, Delayed/Canceled flights, Late trains, etc.
		Knocked-out Electricity grids	Power outage, Paralysis of power grid facilities, Communication interruption, etc.
	Disaster exposure	Roads	Highways, Civil aviation, Stations, Airports, etc.
		Infrastructure	Electric facilities, Natural gas pipelines, Water pipelines, Houses, Communication base stations, Power plants, etc.
		Materials	Vegetables, Compressed cookies, Bread, Porridge, Eggs, Flashlights, Kayaks, Raincoats/boots, etc.
		Agriculture	Forestry, Planting, Farming, Crops, Wheat, Soybeans, Sweet potatoes, etc.
Disaster response/relief		Forecasting/warning Release	Activation of early warning emergency response, Issuance of weather alerts, Development of emergency protection plans, etc.
		Rescue	Eliminate danger and do rush repairs, Drainage measures, Evacuation, Laying antislip mats, Setting generators, Diversion, Guidance, Search and relief, etc.
		Donation	Donations, Donations of materials, Grants, Solicitations, etc.
		Supervision	Price regulation, Regulate and control, etc.
		Resettlement	Centralized resettlement, Relocation, Setting up tents, Setting up transition houses, Consolation, Peace of mind, etc.
		Reconstruction	Restoration, Emergency construction, etc.

, the 16,922 hot microblogs related to the case were crawled and judged by the attributes/status of the text fields one by one, which were classified into three categories based on the judgment of the results-hazard factors, disaster loss and exposure, and disaster response/relief, where the disaster relief response was further classified into seven categories, as shown in

Table 2. Explanation of the classification of disaster response/relief contents in hot microblogs.

Category	Category Description
#1 Science popularization	The government and various departments explain such things as the causes of heavy rain in Henan, the methods and measures to prevent and avoid disasters, and production and selfhelp.
#2 Authority announcement	The government and various functional departments issue specific measures for disaster relief work and deploy various stages of relief and security work.
#3 Prewarning & forecasting	All service departments continue to monitor the meteorology, traffic, and other aspects of the disaster area and make early warning and forecast tips.
#4 Appeal	Appeal to social organizations or individuals outside the affected areas to donate money or goods; appeal to people in the affected areas to produce and help themselves.
#5 Donation	Donations and contributions from the government, social organizations, and individuals.
#6 Humanistic care	Government visits disaster areas in depth, disaster relief performances, video clips, and poems celebrating rescue teams, etc.
#7 Remind	The public is reminded of similar disasters in the past, such as the heavy rain in Zhumadian, Henan Province, in 1975 and the heavy rain in Beijing in 2012.

. We classified the hot microblogs as “unrelated” if their attributes/status were not in the judgment words/phrases.

② According to

Table 3. Attribute/status judgment words/phrases of cases in hot comments.

Category	Attributes/Status Words/Phrases
*1 Cheering	Come on, Hang in there, Stand together through storm and stress, Insist, etc.
*2 Praying for safety	Be safe, Protect yourself, Take care, All the best, Bless, Pray, Good people live in peace, etc.
*3 Worried & afraid	Distress, Scary, So scared, So dark, I can’t hold on, Horror, Fear, I can’t live, etc.
*4 SOS & rescue	Rescue, Relief, Save lives, Ask for diffusion, Ask for help, Rescue team, Help each other, etc.
*5 Rational appeal	Do not joke, Do not brush the screen, Public screen left to help people, Proliferation of rescue information, Please host to focus, Pay attention to Zhengzhou, etc.
*6 Forwarding science	Rainstorm causes, Typhoon causes, Drowning self-help guide, Psychological tips, etc.
*7 Forwarding disaster situations	Trapped, flooded, Fever, Diarrhea, Death due to disaster, Lack of supplies, Power outage, Loss of communication, Lost contact, Shipwrecked, Water level, etc.
*8 Overwhelmed	Cannot see, Do not know what to do, Do not know what the situation, Who will help me, etc.
*9 Complained	Hate natural disasters, Raining all night, So sad, Do not rain, etc.
*10 Surprised & puzzled	My God, What happened recently, Hasn’t it been repaired, etc.
*11 Irrelevant	Zheng Shuang surrogate, Wang Yibo concert, Xiao Zhan scandal, Marriage under the COVID-19 epidemic, Takeaway wages, etc.
*12 Anxious & restless	Nerves are dead, I can’t sleep, I can’t eat, Leave it unsaid, I’m bored, etc.

, the 25,590 hot comments related to the case were classified into 12 categories, including “*1 Cheering”, “*2 Praying for safety”, “*3 Worried & afraid”, “*4 SOS & rescue”, “*5 Rational appeal”, “*6 Forwarding science”, “*7 Forwarding disaster situations”, “*8 Overwhelmed”, “*9 Complained”, “*10 Surprised & puzzled”, “*11 Irrelevant”, and “*12 Anxious & restless”.

③ The naïve Bayesian algorithm was chosen to perform sentiment analysis on 13 comment types, and the text library used for sentiment analysis was the open-source text library that comes with the Python class library-SowNLP. Further, the comment types were classified into four categories: positive, negative, neutral, and irrelevant. The details are shown in

Table 4. Classification results of comment type sentiment analysis.

Sentiment	Comment Types
Positive	“Cheering”, “Praying for safety”, “Rational appeal”, “SOS & rescue”, “Forwarding science”, “Forwarding disaster situations”
Negative	“Complained”, “Worried & afraid”, “Anxious & restless”
Neutral	“Surprised & puzzled”, “Overwhelmed”
Irrelevant	“Irrelevant”

.

(2)

Time series construction

This section mainly classified and counted the original data by time.

① Precipitation series: These series include the daily precipitation series of Henan Province and the daily precipitation series of Zhengzhou, which indicates changes in the case’s hazard factor, which were constructed by calculating the daily mean of precipitation at all Henan stations and single-point in Zhengzhou.

② Posting series, forwarding series, and commenting series: These three sequences were constructed by counting the daily microblog posts, forwards, and comments of the case, respectively.

③ Information behavior composite index series.

Weibo posting, retweeting, and commenting are all expressions of information behavior, but heavy rain and flood disasters stand out due to their extremeness, suddenness, complexity, devastation, urgency, and unpredictability. The instantaneous nature of social media information dissemination has resulted in disaster-related information being multifarious and disorderly at various stages of the evolution of disaster events. To more accurately quantify disaster information behavior, an index-information behavior composite index was constructed by selecting the number of Weibo posts, forwards, and comments within three days. First, the total number of information behaviors for the day is the number of Weibo posts, forwards, and comments for the day plus the previous two days. Second, the data are standardized via the function relation in formula (1):

$$X^{\ast }=\frac{X-min}{max-min}$$

(1)

where X is the total number of information behaviors, max and min represent the maximum and minimum values of this index in a certain time period, and X* is the standardized index, which has a value between 0 and 1.

2.3.3. Statistical Analysis Methods

The VAR model-impulse response analysis is used to observe the temporal response of endogenous variables induced by the imposed impulse function and to analyze the impact of shocks on endogenous variables. In this paper, the method is selected for considering how certain information from a disaster event drives opinion orientation and has a dynamic impact on information behavior during the evolutionary development of a social system.

Granger causality analysis (GCA) is an effective method for establishing a cause-and-effect relationship and can quantitatively validate closed relationships between variables. Notably, it is worth noting that before GCA, it is necessary to subject each series, including the posting series, forwarding series, commenting series, and information behavior composite index series, to an Augmented Dickey-Fuller (ADF) test to determine the smoothness of the time series data. The most extensively used and generally accepted Schwarz information criteria (SIC) are selected as the statistical criteria for time delay indices.

3. Results

3.1. Spatiotemporal Characteristics Recognition of Disaster Information Behavior

The spatial distribution results (Figure 2b) show the following. The level of information behavior in disaster areas (Henan Province) is higher than that in nondisaster areas. The areas with a high level of information behavior are primarily distributed in Guangdong, Shandong, Jiangsu, Zhejiang, Beijing, Hebei, Sichuan, Shanghai, Hubei, and other provinces, and the areas with a low level of information behavior are primarily distributed in the Gansu Qinghai, Hainan, Ningxia, Xinjiang, and Xizang provinces.

The temporal series analysis results (Figure 2a) show that the trend for the information behavior composite index is noticeably to the left when compared to the normal distribution curve, meaning that the peak of information behavior occurs before the normal distribution. This trend is consistent with the research conclusions of the big data information dissemination law in the era of network communication, meaning that the interactive communication characteristics of instant interaction exhibited by social media make people participate more deeply, quickly, and conveniently in the dissemination of disaster information. The information behavior in disaster occurrences has an obvious life cycle that essentially follows the “formation-development-extinction” change process.

The impulse response of the disaster information behavior results (Figure 2c) shows the following. When the endogenous variables, including microblog posts, forwards, and comments, are affected by external variables shocks, the results are immediately transmitted to disaster information behavior. Disaster information behavior is significantly affected by conformity mentality, but this effect will gradually diminish as the disaster event evolves in the social system until the adverse effects of a disaster event “Exit” the social system, and the adverse effects of a disaster event “Entry-Exit Transition” the social system is completely completed.

3.2. Content Characteristics Recognition of Disaster Information Behavior

By analyzing the content of the hot microblogs text item by item, in terms of quantity, it was noticed that 47% of the microblog content was related to the disaster case, and 53% of the content was unrelated to it. From 12 to 18 July 2021, the percentage of information related to cases (37%) was lower than the percentage of irrelevant information (63%); from 19 to 25 July, the percentage of relevant information (94%) was higher than the percentage of irrelevant information (6%), and by 26 July, the irrelevant events had vanished, and the percentage of relevant information had skyrocketed to 100% (Figure 3a).

In terms of specific content, most microblogs were about “Disaster response/relief” (43.08%), followed by “Disaster loss and exposure” (41.54%) and “Hazard factors” (15.38%), respectively (Figure 3b). From highest to lowest, the highest percentage of “disaster response/relief” categories were classified by message content: #3 Prewarning & forecasting (21.19%), #5 Donation (20.21%), #4 Appeal (17.39%), #2 Authority announcement (15.84%), #1 Science popularization (14.48%), #6 Humanistic care (9.49%), and #7 Remind (1.40%) (Figure 3c).

3.3. Behavioral Agents’ Gender and Age Characteristics Recognition of Disaster Information Behavior

An analysis of the gender of the hot commenters shows that more females than males engage in information behavior, with 45.81% and 54.19% of males and females, respectively (Figure 4b).

In terms of the content of the comments by gender, the proportion of male comments on “*9 Complained” (54%) and “*11 Irrelevant” (52%) was slightly higher than that of female comments, while females showed high-level information behavior in “*4 SOS & rescue”, “*2 Praying for safety” and “*7 Forwarding disaster situations” (Figure 4a).

In terms of the content of the comments by age, it can be seen that the 20–29 age group has the highest level of information behavior (45%), while the 50+ age group has the lowest (1.96%) (Figure 4c). Positive comments dominate (67.67%) when the sentiment component of the comment content of different age subjects is refined, and most information behaviors convey positive messages. Positive comments predominate under the age of 19, neutral comments predominate between the ages of 20 and 29, negative comments predominate between the ages of 30–39, and irrelevant comments predominate between the ages of 40 and 49 (

Table 5. Percentage of hot comment sentiment types by age group.

Sentiment	All Ages	≤19	20~29	30~39	40~49	≥50
Positive	67.67%	15.38%	59.62%	21.79%	0.64%	2.56%
Negative	11.17%	12.50%	45.83%	37.50%	0.00%	4.17%
Neutral	16.00%	5.00%	77.50%	15.00%	2.50%	0.00%
Irrelevant	5.17%	0.00%	66.67%	16.67%	16.67%	0.00%

).

4. Discussion

4.1. The Reasons for Differences in Disaster Information Behavior Are Numerous and Complex

In this paper, we discuss the characteristics and differences of disaster information behavior in terms of temporal, spatial, content, and behavioral agents, which are influenced by multiple variables, including hazard factors, psychological distance factors, social factors, and individual factors.

(1) As shown in Figure 2a, the information behavior composite index has small variations and maintains a stable trend during the latent phase of information behavior (12–19 July). When compared to the precipitation series, this phase is in the first precipitation process, with precipitation occurring in some areas with relatively small rain and intensity, which has not evolved into a disaster in the social system, so there is not a great deal of social concern.

Beginning on 20 July, the information behavior composite index increased and peaked on 21 July, indicating that society paid the most attention to the disaster event during this phase, although there was a one-day lag. The reason for this is that, on the one hand, the evolution of disaster events in the social system becomes more intense as the intensity of the hazard factors increases gradually, and this stage occurs in the second and third precipitation processes. The more intense the hazard factors, the more serious the disaster situation and the greater the potential for widespread social attention and information behavior. The 24 h precipitation in Zhengzhou on 20 July reached 942.7 mm, indicating the severity of the hazard factor. On the other hand, the disaster damaged communication facilities, forcing the disaster area to become an “information island”, the victims were too busy with self- and mutual medical aid to express their emotions and sentiments through social media, but as the State Flood Control and Drought Relief Headquarters of China launched the Level III emergency response on 20 July, the disaster area gradually restored communication, with some of the victims and nondisasters in the latent period of information behavior as the disaster intensified such that, following the heat of the event, the disaster impacted on the social system.

As the precipitation processes gradually weakened and ended on the 23rd, the information behavior composite index exhibits a falling trend during the receding phase of information behavior (23 July–5 September), which is characterized by a decline in both social attention to the case and information behavior. In conclusion, the number of information behaviors will increase as the disaster event becomes increasingly ingrained in the social system, and the degree of public awareness of the disaster event will also have an impact. Disaster information behavior partially reflects the level of social concern about natural disasters.

(2) Psychological distance includes spatial distance and social distance, where spatial distance refers to the distance of the public from the disaster area, i.e., the closer the spatial distance is to the disaster area, and vice versa; social distance refers to whether the public has previously experienced similar Henan heavy rain flood disasters, and if so, it indicates the existence of a closer social distance, and vice versa. The public will exhibit different information behaviors in the event of a disaster depending on the difference in psychological distance. Related research has found that the level of interpretation of events is a function of psychological distance [35,36], and that psychological distance influences the public’s perception and decision-making behavior, i.e., people emphasize the goal of the event and view the event’s development more abstractly for distant events, and more specifically for what is going to happen soon by emphasizing the means to reach the goal [37,38]. According to the analysis of the spatial distribution characteristics of disaster information behavior in this paper, the majority of the low-level information behavior areas are located in China’s northwest region. Because these regions are at different spatial distances from the disaster areas, there are differences in the temporal distance and the ability to receive disaster information, such as the quantity and speed of information dissemination, which follows the law of disaster information diffusion, where it shows distance decay.

Furthermore, the public’s awareness and risk perception of disaster events will be strongly affected by the frequency of previous heavy rain flood disasters. If the public rarely sees it, they will not have the same emotional experience and resonance as those in the disaster area and thus will not be able to promote subsequent behaviors, such as posting, forwarding, and commenting on helpful information in the disaster area. This is unquestionably the result of growing public social distance.

(3) Depending on gender, age, and level of disaster awareness, the public’s psychological and behavioral reactions to natural disaster events differ.

As shown in Figure 4a, we discovered that females are more likely to exhibit high levels of information behavior in “*4 SOS & rescue”, “*2 Praying for safety”, and “*7 Forwarding disaster situations”. Positive and neutral comments predominate the 20–29 age group, and neutral and irrelevant comments predominate the 30–49 age group.

In a study of 62,903 Chinese respondents’ willingness to provide disaster relief, it was found that females were twice as likely to share early warning and disaster situations information compared to males, and females were more willing to donate than men [39]. This conclusion is consistent with the results obtained in this paper that females are more likely to present high levels of information behavior in “*7 Forwarding disaster situations”.

Possible explanations for the occurrence of this phenomenon are, on the one hand, from a psychological standpoint, “mood”—a typical sentiment condition brought on by outside shocks and capable of continually influencing a person’s entire mental activity—is crucial to regulating the various information behaviors of the general public [40]. It has been demonstrated that females are more likely than males to experience “moods”, and that female “mood” persists for a longer period and has a bigger influence on general behavioral performance [41]. Meanwhile, positive mood had a facilitative effect on response compared to negative mood, but only in females, who were more susceptible to mood changes than males [42]. Thus, more females are posting, forwarding, and commenting on positive information such as “*4 SOS & rescue” and “*2 Praying for safety”. On the other, females are relatively less able to resist disasters due to their physical fitness and physiological structure [43] and are more likely to generate information behavior for “*4 SOS & rescue” based on empathy. The “mood” also differs among age groups [44]. The key characteristics of the 19–29 age group include obtaining a mature level of physical and psychological development and taking on societal responsibilities. The moral development of this age group is positive and normal, and they believe that upholding social order is a civic duty and that social norms and regulations should be observed, according to Kohlberg’s “Heinz’s drug theft” dilemma test [45], resulting in more positive and neutral information behavior. When tackling diverse, complicated challenges, the 30–49 age group eventually demonstrates a relativistic, flexible, and pragmatic way of thinking. Adults sometimes exhibit limitations and staleness when faced with complex real-world challenges and, as a result, exhibit negative and irrelevant information behavior.

People of different ages have varying levels of social experience, disaster cognition, and perception ability. People aged 20–29 years old have diversified access to information and actively search for information driven by curiosity and are more sensitive to social hotspots, and previous studies show that this stage is also the age group with the strongest disaster perception ability [27] so this age group exhibits a higher level of information behavior than others. One issue to take into account is specifically the utilization of the internet by various age groups. For instance, the solitary access to information and the typically low usage of social media among people over 50 may be responsible for the generally low level of information behavior shown in the internet medium. Furthermore, we discovered that as disaster events evolve in the social system, the proportion of information related to the case increases while the proportion of information unrelated to the case decreases. The regularity with which such information behavior emerges may be affected by the process of public cognition of disaster events.

(4) Subject to economic and technological advancement, developed regions have higher levels of public education, citizenship, information literacy, and news and information awareness than less developed regions [46]; as a result, developed regions’ information behavior is more than that of less developed regions. For instance, despite their geographic distance from the disaster area, the provinces of Sichuan and Guangdong exhibited high-level information behavior due to their economic development.

Additionally, different social information requirements arise at various stages of the evolution of a disaster event, causing the public to pay attention to the hazard factors, disaster loss and exposure, and disaster response/relief differently, and, as a result, to behave differently when it comes to information.

4.2. Differences in the Influence of Disaster Information Sentiment on Information Behavior

In terms of sentiment attributes, identifying disaster information content can be divided into two categories: positive and negative. It has been proven that sentiment information on the internet inspires the generation of information behaviors (posting, forwarding, commenting, etc.) and speeds up the dissemination of information [47,48]. According to the emotional infection theory, images, videos, texts, and other information relating to events all trigger and infect the audience’s sentiment, which has a mediating influence on the transmission of other sorts of information. When severe, sudden natural disasters happen, in particular, their distinct and unique emergent characteristics stimulate the dissemination of numerous sentimental information. Upon receiving this various sentimental information, the public awakens its intrinsic physiological mechanisms and unconsciously produces the corresponding sentiment, which in turn, stimulates them to produce behaviors that are in line with the emotions [25,49].

The rapid spread of disaster information through social media platforms further amplifies the impact of disaster events, making disaster information dissemination more susceptible to emotions. In terms of information-process theory, when faced with different sentiment information, audiences preferentially access the emotional meaning of sentimental words [48]. Positive sentiment information can motivate audiences to have a stronger willingness to further process the information, which, in turn, enhances the infection of sentiment information [50]. For instance, positive comments remarks like “*1 Cheering”, “*2 Praying for safety”, “*5 Rational appeal”, and “*6 Forwarding science” express more public prayers and blessings for the disaster area, and such positive sentiment can subsequently stimulate information behaviors such as posting, forwarding, and commenting and may inspire a series of positive disaster response behaviors among “Tweet Viewer”, whereas negative comments, such as “*3 Worried & afraid”, “*9 Complained”, and “*12 Anxious & restless” generate uncertainty and negative expectations to regulate public risk perceptions and inhibit information behavior.

4.3. Disaster Information Behavior Is Driven by Conformity and Interest Mentality

As shown in Figure 3c, when compared to “#7 Remind” information that recalls similar disasters in the past, the number of “#3 Prewarning & forecasting”, “#5 Donation”, and “#4 Appeal” information in “Disaster response/relief” related to the public interest is much more numerous, and the information behavior exhibits a conformity phenomenon in the time series with impulse-response analysis.

Behavioral theory claims that the public seems to have a propensity for benefit tending, harm avoidance, and conformity when making decisions [51]. On the one hand, the public typically uses a heuristic information-processing approach, which involves processing the information received using straightforward decision criteria, skipping some perceptions, and making decisions or acting in certain ways immediately [52]. While obtaining information, individuals frequently choose the information that is pertinent to their interests [53]. This pattern of behavior is particularly noticeable when an unpredictable natural disaster occurs. Any behavior is the result of individual information processing [54]. The public, on the other hand, will be influenced by the behavioral patterns of the majority and will act in accordance with public opinion in their judgment and cognitive decision-making. Additionally, due to differences in public perceptions and sensitivities to disaster events, the outcomes of the public’s behavioral choices under conformity and interest-driven mentalities are also indirectly impacted [55]. Hence, in the disaster information public opinion field, communicators’ conformity manifests itself in spreading information preferred by the public, while recipients focus only on information that is practically related to their interests and generate follow-up information behaviors based on their perceptions of disaster risks, which highlights the role of social media in guiding public opinion on disaster events in the age of online cohesion. This can also be used to explain why the level of information behavior in disaster areas is higher than the level of information behavior in nondisaster areas.

Further to this, we discovered that 53% of the microblogs were unrelated to the case, demonstrating that some of the public’s information behavior was motivated by interests and conformity but that this behavior was limited to “Tweet Viewers” and did not include offline disaster relief. The majority of “Tweet Viewers” is “Interest Viewers” and “Entertainment Viewers”, which is different from other social emergencies that tend to gather “Moral Viewers” and “Outraged Viewers”, based on information behavior (publishing, forwarding, commenting, etc.), they generally do not extend to actual disaster responders.

4.4. Implications for Integrated Disaster Risk Governance

The emergence of interactive new media, such as Weibo and WeChat, has broken the monopoly of traditional media on the right to speak and has provided the possibility for people to participate in science. Here, we analyze the information behavior of social media based on the case of the “7.20 Henan heavy rain flood disaster”. Accordingly, the following inspirations were obtained:

(1)

With the advent of the Web 3.0 era, the evolution of disaster events in the social system has been stimulated, amplified, and inhibited by internet-based public sentiment. Therefore, the government should release authoritative information in a timely and objective manner, grasp the information requirements, focus on public opinion at various stages of disaster event evolution, and avoid “information islands”;

(2)

Public opinion monitoring organizations should always pay close attention to various sentimental information in social media. Meanwhile, official media should continuously monitor and control negative information, as well as focus on the virtuous guidance of the public information behaviors of various age groups and genders in order to avoid the social instability caused by the continuous fermentation of rumors and negative public opinions;

(3)

The scientific propaganda of disaster prevention and mitigation knowledge for the whole society should be strengthened, cultivate people’s awareness of safety and danger, and improve disaster-response ability.

5. Conclusions

In this study, based on the case of the “7.20 Henan heavy rain flood disaster”, the characteristics and differences of disaster information behavior via social media use were identified and quantified. The main conclusions derived from this study are summarized as follows:

(1)

Public awareness and risk perception of the disaster events, etc., are influenced by multiple factors, including the intensity of hazard factors. Disaster information behaviors in the temporal series have an obvious life cycle with three phases: the latent period, outbreak period, and the receding period, essentially following a “formation-development-extinction” process, and the climax point of information behaviors is ahead of the normal distribution; information behavior increases gradually with the intensity of the hazard factors and the depth of public concern, although there was a one-day lag; the more serious the disaster and the higher the social concern, the greater the disaster information behavior; disaster information behavior shows the conformity phenomenon in a temporal series;

(2)

Disaster information behavior is influenced by multiple factors, including spatial distance, socioeconomics, education level, etc. Disaster information behaviors show obvious differences in their spatial distribution, meaning the disaster areas are higher in activity than the nondisaster areas; areas that are near to a disaster area are higher in activity than areas that are distant to disaster areas, and economically and technologically developed areas are higher in activity than less developed areas;

(3)

A total of 53% of the content was unrelated to disaster cases, and 47% of the content was related to disaster cases in the microblogs; the most related microblogs (43.08%) were about “disaster response/relief” followed by “disaster loss and exposure” (41.54%), and “hazard factors” (15.38%), respectively. In the microblogs related to “disaster response/relief”, the top three category proportions were “#3 Prewarning & forecasting” (21.19%), “#5 Donation” (20.21%), “#4 Appeal” (17.39%), and the lowest category proportion was “#7 Remind” (1.40%);

(4)

Depending on gender and age, the public’s psychological and behavioral reactions to natural disaster events differ. Females (54.19%) engage in more information behavior than males (45.81%), and females showed more positive information behavior than males, with a high-level information behavior level in “*4 SOS & rescue”, “*2 Praying for safety”, and “*7 Forwarding disaster situations”. The 20–29 age group had the highest level of information behavior (45%), while the 50+ age group had the lowest (1.96%). Positive comments predominate under the age of 19, neutral comments predominate between the ages of 20 and 29, negative comments predominate between the ages of 30–39, and irrelevant comments predominate between the ages of 40 and 49;

(5)

The implications for integrated disaster risk governance in the context of network information dissemination are as follows: the government and official media should release authoritative information in a timely and objective manner in the current climate of rapid information dissemination through social media. They should also understand the public’s information requirements and focus on public opinion at various stages of disaster event evolution and avoid “information islands”. They ought to concentrate on various sentimental information in social media. Additionally, the government needs to increase the general public’s understanding of disaster prevention and mitigation, cultivate people’s awareness of safety and danger, and improve disaster-response ability.

We have to point out that we did not distinguish between analyzing the characteristics of information behavior in disaster and nondisaster areas. In the meantime, the impact of information behavior on natural disaster events is also not discussed. These two issues will be the focus of follow-up studies. Additionally, the mechanisms and pathways by which information dissemination stimulates and regulates the action of natural disaster events on social systems also require further investigation.