Unveiling the Nexus Between Farmer Households’ Subjective Flood Risk Cognition and Disaster Preparedness in Southwest China

Abstract

Understanding Farmer households’ subjective flood risk cognition is important for effectively mitigating the impacts of flood, and adequate disaster preparedness reduces the impact of floods on the sustainability of farmers’ livelihoods. The existing literature focuses on objective flood risk assessment and subjective–objective risk consistency and less systematically explores the correlation between Farmer households’ subjective flood risk cognition and disaster preparedness. Therefore, this study aims to explores the correlation between Farmer households’ subjective flood risk cognition and disaster preparedness. This study employed a random sampling method to conduct a survey among 540 households in Gaoxian County, Jiajiang County, and Yuechi County, which are flood-prone areas in Southwest China. Based on the survey results, this research framework can be used to evaluate systems of subjective flood risk cognition and farmers’ disaster preparedness. We chose the Tobit Regression Model to empirically explore the correlation between subjective flood risk cognition and farmers’ disaster preparedness. The results showed that among the 540 surveyed farmers, their overall subjective flood risk cognition was at a medium-high level (3.58), with self-efficacy more than response efficacy, more than threat, and more than probability. Further, the overall disaster preparedness of farmers was at a medium level (0.5), with physical disaster preparedness more than emergency disaster preparedness and more than knowledge and skills preparedness. The regression analysis showed that the probability of flooding and the threat in Farmer households’ subjective flood risk cognition were positively related to disaster preparedness, whereas self-efficacy, response efficacy, and overall risk cognition in Farmer households’ subjective flood risk cognition were negatively related to disaster preparedness. This study is representative of or may serve as a reference for building governance systems and disaster prevention in other flood risk areas in Southwest China.

1. Introduction

Floods are one of the most destructive and common natural disasters worldwide, causing the greatest economic losses [1]. Floods and their associated disasters, such as typhoons, landslides, and mudslides, have become more frequent over the past few years. These disasters are attributed to changes in precipitation patterns and accelerated melting of glaciers, which have caused catastrophic damage globally [2,3]. Farmers are an important part of modern society and agriculture plays an extremely important role in economic development [4,5]. Sichuan is one of the worst flood-affected provinces in China [6]. The Sichuan Emergency Management Bureau reported that floods and geological disasters in 2022 affected 2,225,000 people in 180 counties in 21 cities (states), with direct economic losses amounting to CNY 4.9 billion. Therefore, assessing farmers’ flood risk cognition in flood-threatened environments can help mitigate their flood losses, increase their resilience to flood risks, and develop appropriate disaster prevention measures.

Generally, the literature describes risk cognition as having two dimensions—cognitive and affective. For example, Paek and Hove [7] describe cognitive risk cognition as the extent to which people understand risk, while the affective dimension covers how people feel about risk. Lechowska [8,9] provides a simplified description of these two dimensions of risk cognition, which are referred to as awareness and worry, respectively. The research in this study focuses on the cognitive dimension of risk cognition among farmers. Mañez et al. they argue that risk cognition is critical in determining the impact of risk management and vulnerability reduction initiatives. Farmers in flood threatening environmental jobs are generally categorized into subjective and objective risk cognition [1,10]. Existing relevant research focuses on the objective cognition dimension, which mainly refers to the government’s interpretation of floods, and farmers’ cognition of floods are usually passive, as in the case of flood warnings, etc. Bradford et al. emphasize in their study that prioritizing public cognition of risk is crucial, as the authorities’ lack of understanding of societal perspectives has been identified as a key factor contributing to inadequate flood risk management [11]. Moreover, in flood risk areas, farmers are the mainstay of flood defense. Understanding farmers’ subjective cognition of flood risk is a key strategy for risk reduction [12]. Paek and Hove describe risk cognition as the extent to which people understand risk, and assessing farmers’ subjective risk cognition in flood threatened environments can help to mitigate their losses and increase their resilience to flood risk [7]. Farmers’ subjective risk cognition of floods are their overall knowledge and attitudes towards floods from the farmers themselves [13,14]. According to Anderson et al. this information may help to enhance risk communication and predict risk reduction behaviors such as preparedness [15]. For example, farmers, based on their past experiences, know the probability of the next arrival of a flood, how much damage will be caused to their personal property (farmland, house), whether they can cope with it and whether it is necessary to relocate. Therefore, this study referred to the existing studies by Xue et al. [16] and other scholars to explore the four dimensions of probability, threat, self-efficacy, and coping efficacy of flood occurrence from the subjective perspective of farmers and assigned values from 1 to 5 to measure their subjective cognition of flood risk [3,16,17,18,19].

Subjective cognition of flood probability by farmers is the probability of making appropriate adjustments and adaptations when they realize that flooding is imminent [20]. Farmers’ cognition of the extent of flood damage is subjective cognition of the threat of flooding, which also influences the conditions under which they make avoidance choices [21]. In addition, self-efficacy and coping efficacy in terms of flood risk are gaining public attention [18]. These are key psychological dimensions that national factors for natural disaster adaptation behavior [22]. Strong risk awareness influences people’s disaster preparedness and response. Understanding the role of self-efficacy and coping efficacy in disaster risk management is essential for developing effective intervention strategies [23]. Self-efficacy is one of the strongest and most stable drivers of individual flood prevention behavior and one of the most closely related motives [24], and self-efficacy is believed to contribute to public preparedness for disasters. Coping efficacy [25], which corresponds to self-efficacy, refers to the ability to develop coping strategies to deal with disasters when faced with disaster risks [26,27].

In a recent study by Billman et al. [28], it was noted that farmers’ risk cognition influence sustainable agricultural practices and will prioritize risk reduction efforts, i.e., adopting disaster preparedness measures. Disaster preparedness behaviors are activities or measures taken by individuals prior to a disaster event to reduce the severity of the impact of the disaster. Another study by Alam reported a strong relationship and causality between risk cognition and disaster preparedness [29], with direct disaster experience having a mediating role. In real life, disaster preparedness measures are needed to cope with risks and shocks when faced with flood threats [30], and adequate disaster preparedness measures can significantly mitigate losses caused by disasters [31]. However, in flood risk zones, farm households are affected by different types of risks interacting with each other in a complex network of relationships, leading to increased vulnerability of their livelihoods [32,33]. Farm households do not have sufficient livelihood capital to always support their best preparedness for disaster avoidance. In general, farmers’ disaster preparedness is influenced by their subjective cognition of flood risk, and disaster preparedness usually varies with the external environment and livelihood capital [34]. A review of the existing literature reveals that prior emergency preparedness to cope with disasters has been suggested by Lam et al. [35]. Emergency preparedness is not effective in reducing flood risk, and the use of farmers’ initiative can be effective in reducing flood risk. Monteil et al. encourage people at risk of floods to learn about floods and to learn disaster avoidance skills [36]. In addition, they also pointed out that reducing flood-related disaster risk requires multi-scale physical measures to cope with flood risk. Therefore, this study mainly deals with the coping strategies and the dimensions of disaster preparedness proposed by Hoffmann and Muttark [37], Wu et al. [20], and He [38]. Here, we categorize preparedness into three dimensions, emergency response, knowledge and skills, and physical preparedness, and measure farmers’ preparedness for disaster avoidance through dichotomous variables [39].

Currently in academia, scholars have focused their attention mainly on the concept of flood risk cognition. For example, Netzel and others have already explored the need to categorize flood risk cognition into individual and global risk cognition, highlighting the differences between the two [40]. Wang et al. explored what correlation exists between farmers’ subjective and objective flood risk characteristics, subjective and objective flood risk congruence, and willingness to purchase natural disaster insurance [1]. Faruk also introduced the Protection Motivation Theory (PMT) to measure farmers’ perceived flood risk and adaptation assessment [41]. In addition to this, academics have also studied areas related to disaster preparedness, such as Naz et al. [42], who in their study identified the factors influencing gender on disaster preparedness by analyzing the gender of the respondents. Hossain et al. explored the impact of floods on the standard of living of the farmers of Chal village and discussed their disaster avoidance strategies [43]. Faruk and Maharjan identified the impact of farmers’ collective participation on their flood adaptation [44]. However, our study differs from the above literature in that it endeavors to identify the determinants that influence farmers’ subjective cognition of flood risk and their preparedness behavior for disaster avoidance. Based on the above considerations, this study aims to answer two crucial questions: (1) What are the characteristics of farmers’ subjective cognition of flood risk and their selection of avoidance behaviors in flood risk zones? (2) What is the correlation between the subjective cognition of flood risk and the selection of disaster avoidance behaviors of farmers in flood risk zones? The remainder of this paper is organized as follows. Section 2 presents the study hypotheses; Section 3 presents the study area, data sources, variable settings, and study methods; Section 4 presents the results and analysis; Section 5 presents the discussion; and Section 6 provides a summary and recommendations.

2. Research Theory and Research Hypothesis

2.1. Protection Motivation Theory

Protection Motivation Theory (PMT) was developed by Rogers. The basic concept of PMT is that when a person has the intention to initiate or maintain an action, the value of his or her willingness to do so is related to the threat assessment and coping assessment as the influencing factors. Threat assessment includes the likelihood of the threat occurring, the perceived severity of the threat, and the fear of the threat occurring; coping assessment includes three elements: self-efficacy, response efficacy, and response cost [45]. The basic assumption of PMT is that exposure to risk-related information enables individuals to assess the severity of the risk, vulnerability to the risk, and ability to mitigate the risk, which in turn motivates the motivator [46,47]. This theory is widely used in academic research on disaster prevention and mitigation actions and focuses on the idea that a certain level of risk-related information can create the motivation necessary to determine the severity of risk, vulnerability, and ability to reduce people’s risk. Yu found that if people perceive their ability to cope with relevant disaster situations (self-efficacy) to be low, they are less likely to take protective measures [48,49]. However, if local officials regularly communicate disaster mitigation knowledge to villagers, this can increase individuals’ level of protection against disasters, and this effect is stronger for individuals with higher levels of self-efficacy. Therefore, protection motivation theory is commonly used in research areas related to disaster perception and preparedness for disaster avoidance.

2.2. Research Hypothesis

This study mainly explores the impact of subjective cognition of flood risk among farmers in southwestern China on disaster preparedness. Three counties, nine towns, and twenty-seven villages were selected as specific and significant representatives of flood risk areas in southwestern China, in order to provide a theoretical basis for flood control and disaster relief in the entire southwestern flood risk area and reduce flood risk. Figure 1 shows the theoretical framework of this study, based on which the following three hypotheses are proposed.

Farmers’ subjective flood risk cognition is closely related to disaster preparedness, and different subjective risk cognition of farmers influence them to prepare for disaster avoidance accordingly [41]. Farmers’ threatening cognition of flood risk are expressed in terms of individuals’ assessments and cognition of the magnitude and severity of floods that will occur in the future. Zabini [12] found that there is a correlation between the level of perceived threat of flood risk and the preparedness behaviors adopted by individuals. Flooding, as a high-risk natural disaster, can cause varying degrees of damage to houses and other infrastructure such as farmland [50]. However, when floods are more threatening than a certain level, farmers’ physical disaster preparedness may not be effective in mitigating losses. The higher the subjective flood risk score of rural residents, the lower the corresponding physical disaster preparedness. Therefore, hypothesis H1 is proposed:

H1: 

There is a negative relationship between farmers’ subjective cognition of flood risk as threatening and physical disaster preparedness.

Farmers’ perceived flood risk self-efficacy refers to their confidence in their ability to withstand floods. Subjective assessments of competence cover both self-efficacy and response efficacy [18,19], with higher levels of self-efficacy indicating greater confidence in one’s ability to protect against floods [17]. Disaster management research has shown that individuals with higher self-efficacy are more likely to engage in risk mitigation and emergency preparedness activities. This finding is consistent with Wurjatmiko’s finding that there is a positive correlation between self-efficacy and flood risk preparedness (p < 0.001, r = 0.63). Therefore, research hypothesis H2 is proposed:

H2: 

There is a positive relationship between farmers’ self-efficacy on subjective cognition of flood risk and their overall disaster preparedness and all its three dimensions (emergency disaster preparedness behavior, knowledge and skills preparedness and physical disaster preparedness).

Farmers’ probability cognition of flood risk, i.e., their cognitive assessment of the probability of actual flood occurrence, which influences their decision making to undertake disaster preparedness [20]. In fact, the higher the probability of flood occurrence as perceived by farmers, the more they are motivated to prepare for disaster avoidance, creating a positive push. Similarly, farmers’ response efficacy scores for flood risk indicate their perceived level of risk mitigation. Higher response efficacy scores indicate that farmers perceive that flood risk is more likely to be mitigated by taking measures to cope with flood risk, and therefore they are more likely to prevent and mitigate disasters [26]. Hoffmann and Muttarak [37] found that an individual’s subjective cognition has a significant impact on preparedness, and that people with experience in flood resilience can acquire knowledge and skills that motivate and facilitate flood mitigation actions. Consequently, those who perceive themselves to be able to respond effectively to disasters are more inclined to take preparedness measures.

Farmers’ subjective cognition of overall flood risk includes their cognition of flood probability, threat, self-efficacy, and coping efficacy [16]. The higher the farmers’ subjective cognition of overall flood risk indicates that in their subjective cognition, the higher the probability of floods coming and the higher the threat of floods, and accordingly, they believe and are confident that they can cope with floods by some means. Therefore, farmers will take some precautionary measures to reduce losses. Therefore, research hypothesis H3 is proposed:

H3: 

There is a positive relationship between farmers’ probability of subjective risk cognition of floods, coping efficacy, and overall flood risk cognition with their overall disaster preparedness and its three dimensions. (emergency disaster preparedness, knowledge and skills preparedness, and physical disaster preparedness).

3. Material and Methods

3.1. Overview of the Study Area and Data Sources

Floods have become a serious disaster worldwide, mainly due to geographical location and climatic conditions. Southwest China, characterized by high mountains and a subtropical monsoon climate, is a highly flood-prone region. Jiajiang in Sichuan Province is located in the hinterland of the Chengdu Plain Economic Zone, west of Emei Mountain, south of Leshan Giant Buddha, north of Meishan Sanshu’s hometown, is a subtropical humid climate with abundant rainfall. Jiajiang territory is densely populated with rivers, and in addition to the Qingyi River, there are also the Macun River and the Jinniu River. Jiajiang County is low-lying terrain; “rainfall + low-lying terrain,” a combination that indicates that flood season drainage is poor, and with the Minjiang River tributaries to top off the impact, it is prone to flooding. Gaoxian County is located in the southern edge of the Sichuan Basin; the terrain is high in the south and low in the north, belongs to the mid-subtropical humid monsoon climate, there is precipitation throughout the year, and a total of fourteen rivers in the county. This combination of “high precipitation + medium slope” can easily lead to short-term flash floods. Yuechi is located in the northeastern part of the Sichuan Basin, with high terrain in the northwest and low terrain in the southeast and is a triangular section at the confluence of the Qujiang and Jialing Rivers. It belongs to the central subtropical monsoon climate zone, and Yuechi County is rich in transit water resources. Yuechi County has “medium rainfall + high river network density”, a combination that makes Yuechi prone to flooding. The three locations include most of the risk characteristics of the flood risk zone in Southwest China and are representative. Figure 2 shows that these three places have a common subtropical humid monsoon climate. Although some seasons are dry, heavy rainfall occurs during the rainy season from June to August each year, making these areas prone to severe flooding and posing a serious threat to the safety of local lives and assets. In 2022, floods and geological disasters affected 21 cities (states), 180 counties, and 2,225,000 people in the province. Furthermore, 51 people were reported missing, and the emergency relocation of 59,000 people was required. The area of affected crops was 55,000 hectares. Additionally, 13,000 houses were damaged to different degrees, and the immediate economic damage totaled CNY 4.9 billion. As mentioned previously, Jiajiang, Yuechi, and Gaoxian County, which are located in flood risk areas, were selected as the focus of this study.

The data used in this study mainly included those on farmers’ subjective risk cognition of floods and the selection of disaster preparedness, both of which came from field research conducted by the team from July to September 2021 in nine towns in three counties: Jiajiang (Ganjiang, Huangtu, and Mucheng), Gaoxian (Jiale, Qingling, and Shengtian), and Yuechi (Fulong, Luodu, and Zhonghe) in the flood risk area of Sichuan Province. The research method was random sampling. According to the amount of rainfall, flood risk areas were divided into three groups. The topography and economic development differences for each group were considered, and sample districts, towns, and villages were randomly selected. Random selection was also used to identify the same proportion of farmers. The methodology employed was individual, in-person interviews. The essence of this research predominantly involved Farmer households’ subjective flood risk cognition, disaster preparedness, and basic personal information. Each questionnaire took about 1.5 h to answer. To maintain the authenticity and comprehensiveness of the questionnaire, the team conducted a pre-survey in Jiajiang and made systematic revisions. In addition, the team conducted systematic training for the researchers and unified their understanding of the questionnaire before conducting official research. Based on this process, the team obtained 540 valid questionnaires from individuals in three counties and nine towns.

3.2. Variable Settings

3.2.1. Explanatory Variables

Farmer households’ subjective flood risk cognition was used as the explanatory variable in this study, according to Qing et al. [51], Xue et al. [16], Xu et al. [52], and others. We designed words for subjective flood risk measurement mainly from the four dimensions of farmers’ perceptions regarding the probability of flooding, the threat, self-efficacy, and response efficacy for the occurrence of disasters. The entropy method is suitable for multi-indicator evaluation scenarios, calculating the weight of each independent variable, which is beneficial for distinguishing indicators with different sample differences and contributes more to comprehensive evaluation. The probability rating in Figure 3 is an assessment by farmers of the probability or threat of flood arrival, both of which are evaluations of flood events. Self-efficacy and response efficacy are evaluations of the extent to which farmers receive relief from floods. Among them, the alpha coefficients of all dimensions are higher than 0.8, far exceeding the acceptable level of 0.7, indicating that the variables in each dimension have good internal consistency, the scale reliability meets the research requirements, the data reliability is high, and it is suitable for subsequent analysis.

3.2.2. Explained Variables

In this study, farmers’ disaster preparedness was regarded as an explanatory variable, and the variable settings refer to the studies of Hoffmann and Muttarak [37] and Ma et al. [39]. As shown in Figure 4 and as mentioned previously, disaster preparedness is classified as emergency, physical, and knowledge and skills preparedness. These three dimensions are aggregated to generate overall disaster preparedness, with the aim of discussing the effect of farmers’ subjective risk cognition of floods on disaster preparedness. During the research process, farmers were asked whether they purchased natural disaster insurance, prepared emergency items, secured valuables properly, learned about disaster mitigation and prevention, reinforced their houses, and engaged in other behaviors that could reduce the impact of disasters.

3.2.3. Control Variables

To enhance the accuracy of the model, we referred to Xue et al. [16] and Gammoh et al. [18] and set factors affecting residents’ preparedness for disasters as control variables. The control variables displayed in

Table 1. Control variables.

Variables	Meaning and Assignment	Mean	SD
Gender	Sex of respondent (Male = 0, Female = 1)	0.40	0.49
Age	Age of respondents (Years)	58.48	11.84
Education	Educational attainment of respondents (Years)	6.55	3.44
Marriage	Whether the respondent is married (0 = No, 1 = Yes)	0.91	0.29
Health	Health status (1-very bad-5-very good)	3.67	1.14
Length of residence	Length of time living in the home (Years)	50.32	17.31
House structure	Whether the house is the concrete structure (0 = No, 1 = Yes)	0.41	0.49
Number of rivers	Number of rivers in your neighborhood (Number)	1.07	0.47
Nearest river	Distance of your home from the nearest river (Meters)	1060.39	1542.71
experiences	Number of flood experiences (Times)	0.93	0.26
asset value	Total present value of physical assets (RMB ten thousand)	6.39	11.63

included farmers’ basic personal information characteristics (e.g., marriage and age) and social and economic characteristics (e.g., household economic income, years of residence, etc.).

3.3. Research Methodology

3.3.1. Entropy Value Method

The entropy value method constitutes is an objective weighting method. It assigns weights to indicators according to the significance of the information provided by the observed values of each indicator, and the subjective weights derived have higher accuracy and credibility than the subjective assignment method [53]. Hence, in the present study, we used this method to ascertain indicators pertaining to farmers’ subjective perceptions of flood risk. Xue et al. [16] described the specific steps required. The specific steps are as follows:

Firstly, the indicators were standardized, and the evaluation index system constructed had inconsistent units with significant differences. The article standardized the raw data of positive and negative indicators before calculating the weight of flood risk indicators for farmers. The specific formula is as follows:

Positive indicator calculation formula: $x_{ij}={\displaystyle \frac{X_{ij}-X_j^{min}}{X_j^{max}-X_j^{min}}}$

Calculation method for negative indicators: $x_{ij}={\displaystyle \frac{X_j^{max}-X_{ij}}{X_j^{max}-X_j^{min}}}$

In the formula i represents the county, town, and village in the research area; j represents various indicators; $X_{ij}$ is the j-th indicator value of the i-th research site before standardization; and $x_{ij}$ is the j-th indicator value of the i-th research site after standardization.

Secondly, calculate the proportion of the i-th sample value under the j-th indicator to the sum of all sample values for that indicator $P_{ij}$:

$$P_{ij}={\displaystyle \frac{x_{ij}}{\sum _{i=1}^n{x}_{ij}}}$$

Thirdly, calculate the entropy value $e_j$ of the j-th indicator:

$$e_j=-{\displaystyle \frac{1}{ln\left(n\right)}}\sum _{i=1}^n{P}_{ij}ln\left(P_{ij}\right)$$

In the formula, ${\displaystyle \frac{1}{ln\left(n\right)}}$ > 0; $e_j$ ≥ 0.

Among them, n is the total number of counties, towns, and villages in the research area, which is 9 in this case.

Fourth, calculate the information entropy redundancy $d_j$:

$$d_j=1-e_j$$

Fifth, calculate the evaluation index coefficient $W_j$:

$$w_j={\displaystyle \frac{d_j}{\sum _{j=1}^m{d}_j}}$$

And, m is the total number of evaluation indicators, which is 16 in this case.

Finally, calculate the comprehensive score $S_j$:

$$S_j=W_j{X}_{ij}$$

3.3.2. Tobit Regression Model

The goal of this study was to systematically estimate the impact of farm households’ subjective understanding of flood risk on their level of disaster preparedness using quantitative methods. The dependent variable in this investigation was farmers’ disaster preparedness, a continuous variable between 0 and 1. Depending on the attributes of the variable, the study aim was to investigate the relationship between subjective flood risk and farmers’ disaster preparedness using the Tobit regression model. The model is briefly summarized as follows.

Assumption $y_i^{*}={\partial }_i^{\prime }+{ℇ}_i$ ($y_i^{*}$ unobservable, perturbation terms $\left({\mathsf{\epsilon }}_{\mathrm{i}}|{\mathrm{x}}_{\mathrm{i}}\right)~\mathrm{N}(0,{\mathsf{\delta }}^2)$). Given that the stage point is set to c = 0, the value of Y can be determined by referencing Equation (1), and the expected value given certain conditions $\mathrm{E}\left(y|x\right)$ for the entire sample can be expressed by Equation (2):

$$y_i=\left\{\begin{array}{c}{y}_i^{*}, if y_i^{*}> 0\hfill \\ 0, if{ y}_i^{*}\le 0\hfill \end{array}\right.$$

(1)

$$\mathrm{E}\left(\mathrm{y}|\mathrm{x}\right)=0\mathrm{P}\left({\mathrm{y}}_{\mathrm{i}}=0|{\mathrm{x}}_{\mathrm{i}}\right)+\left({\mathrm{y}}_{\mathrm{i}}|{\mathrm{x}}_{\mathrm{i}};{\mathrm{y}}_{\mathrm{i}}>0\right)\mathrm{P}\left({\mathrm{y}}_{\mathrm{i}}>0|\mathrm{x}\right)=\mathrm{E}\left({\mathrm{y}}_{\mathrm{i}}|{\mathrm{x}}_{\mathrm{i}};{\mathrm{y}}_{\mathrm{i}}>0\right)\mathrm{P}\left({\mathrm{y}}_{\mathrm{i}}>0|\mathrm{x}\right)$$

(2)

Based on the introduction in this paper, the precise formulation of the Tobit model established herein is outlined in Equation (3):

$$\mathrm{T}\mathrm{o}\mathrm{b}\mathrm{i}\mathrm{t}\left({\mathrm{y}}_{\mathrm{i}}\right)={\partial }_0+{\partial }_{1\mathrm{i}}\times {\mathrm{D}\mathrm{P}}_{\mathrm{i}}+{\partial }_{2\mathrm{i}}\times {\mathrm{R}\mathrm{C}}_{\mathrm{i}}+{\partial }_{3\mathrm{i}}\times {\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}}_{\mathrm{i}}+{\mathsf{\epsilon }}_{\mathrm{i}}$$

(3)

where ${\mathrm{D}\mathrm{P}}_{\mathrm{i}}$ denotes farmers’ disaster preparedness, ${\mathrm{R}\mathrm{C}}_{\mathrm{i}}$ denotes the subjective cognition of farmers’ flood risk, ${\partial }_0$ denotes the constant term, ${\partial }_{1\mathrm{i}}$_i, ${\partial }_{2\mathrm{i}}$_i, ${\partial }_{3\mathrm{i}}$_i represent the model parameters that must be estimated, ${\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}}_{\mathrm{i}}$ is the control variable, and ${\epsilon }_i$ is the random disturbance term. All data analyses were performed using Stata 16.0.

4. Results and Analysis

4.1. Descriptive Statistics

Study findings revealed various degrees of differences and gaps in the subjective flood risk of farm households based on family labor force size, income, education level, and region. Figure 5a shows that when the household labor force population was 0, the overall subjective flood risk cognition of the household was the highest, reaching 0.368. With the increasing household labor force population, the average subjective flood risk of farm households showed an overall decreasing trend, and the resilience of households with larger labor force populations was significantly higher than that of other small-scale households. The subjective flood risk cognition of farm households was also related to the level of education. Due to shortcomings in rural economic and education levels, the overall education level of farm households was low, with the largest number of households having primary and middle school education levels. Figure 5b shows that as the education level of farmers increased, their subjective flood risk cognition showed a clear downward trend, and the subjective flood risk cognition of farmers with university and higher education levels (0.290) was much lower than that of illiterate farmers (0.376). We also found that, among the three counties displayed in Figure 5c, the overall flood risk awareness of farm households in Yuechi County (0.331) was the highest, followed by Gaoxian County (0.316), and the overall flood risk cognition of farm households in Jiajiang County (0.280) was the lowest. Figure 5d shows that Gaoxian County had the lowest flood risk response effectiveness, indicating that its overall disaster preparedness was poor. Thus, it was not able to effectively resist risks in the face of floods. The probability and threat of flood risk in Jiajiang County were lower than in the other two counties, and the probability and threat of flood risk in Yuechi County were notably higher than those in the other two counties, indicating that its rural areas are more susceptible to flooding than those of the other two counties. Jiajiang County was the least prone to experience flooding.

4.2. Interpretation of Regression Results

In Stata 16.0, we conducted multicollinearity tests on the independent variables to prevent overfitting of the core explanatory variables to the model and avoid multicollinearity issues that could affect regression analysis. The results showed that the maximum VIF value was 3.97, far less than 10, showing that there is no problem of multicollinearity between the variables, and the assumption that variables are independent is satisfied. Figure 6 shows the heat map of the correlation coefficient matrix for the core variables of the model. This study used the entropy method to calculate the correlation coefficients between the Farmer households’ subjective flood risk cognition and disaster preparedness in flood risk areas.

Table 2. Regression results related to farmers’ subjective cognition of flood risk and emergency disaster preparedness and knowledge and skills preparedness.

Variables	Emergency Disaster Preparedness				Knowledge and Skills Preparedness
	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6	Model 7	Model 8
Probability	0.632 **		0.594 **		0.003		0.252
	(0.288)		(0.290)		(0.304)		(0.299)
Threat	0.103		0.305		−0.112		0.056
	(0.323)		(0.317)		(0.340)		(0.327)
Self-efficacy	−0.440 ***		−0.342 **		−0.434 **		−0.369 **
	(0.163)		(0.162)		(0.172)		(0.167)
Response efficacy	−0.366		−0.161		−0.626 ***		−0.3
	(0.227)		(0.228)		(0.24)		(0.235)
Overall risk		−0.194 **		−0.070		−0.386 ***		−0.203 **
		(0.453)		(0.088)		(0.484)		(0.090)
Gender			0.006	0.007			0.015	0.015
			(0.028)	(0.028)			(0.029)	(0.029)
Age			0.001	0.001			−0.001	−0.001
			(0.002)	(0.002)			(0.002)	(0.002)
Education			0.013 ***	0.013 **			0.022 ***	0.022* **
			(0.004)	(0.004)			(0.004)	(0.004)
Marriage			−0.009	0.001			−0.079 *	−0.072 *
			(0.039)	(0.039)			(0.040)	(0.040)
Health			0.020 *	0.016			0.006	0.003
			(0.011)	(0.011)			(0.011)	(0.011)
Length of residence			−0.001	−0.001			0.001	0.001
			(0.001)	(0.001)			(0.001)	(0.001)
House structure			0.026	0.025			0.050 **	0.050 **
			(0.023)	(0.023)			(0.024)	(0.024)
Number of rivers			0.012	0.015			−0.009	−0.007
			(0.024)	(0.025)			(0.025)	(0.025)
Nearest river			−5.000	−4.760			−5.64	−5.480
			(7.280)	(7.390)			(7.51)	(7.550)
Experiences			0.150 **	0.182 ***			−0.007	0.016
			(0.045)	(0.044)			(0.465)	(0.045)
Asset value			0.001	0.001			0.002 *	0.002 *
			(0.001)	(0.001)			(0.001)	(0.001)
Constant	0.397 ***	0.453 ***	0.040	0.088	0.454 ***	0.484 ***	0.268 **	0.301 **
	(0.031)	(0.029)	(0.114)	(0.115)	(0.033)	(0.030)	(0.118)	(0.118)
Observation	540	540	540	540	540	540	540	540
LR chi2 (×2)	26.12	5.07	62.75	44.80	24.37	18.31	85.25	77.35
Prob > chi2 (×2)	0.000 ***	0.024 **	0.000 ***	0.000 ***	0.000 ***	0.000 ***	0.000 ***	0.000 ***
Pseudo R2	0.234	0.045	0.562	0.401	0.1455	0.109	0.509	0.462

Notes: *** p < 0.01, ** p < 0.05, * p < 0.1.

and

Table 3. Regression results related to farmers’ subjective cognition of flood risk and physical disaster preparedness and overall disaster preparedness.

Variables	Physical Disaster Preparedness				Overall Disaster Preparedness
	Model 9	Model 10	Model 11	Model 12	Model 13	Model 14	Model 15	Model 16
Probability	0.921 ***		1.056 ***		0.518 ***		0.634 **
	(0.269)		(0.275)		(0.196)		(0.193)
Threat	−0.469		−0.518 *		−0.159		−0.052
	(0.301)		(0.301)		(0.219)		(0.212)
Self-efficacy	−0.547 ***		−0.548 ***		−0.474 ***		−0.420 ***
	(0.152)		(0.153)		(0.111)		(0.108)
Response efficacy	0.249		0.241		−0.248		−0.073
	(0.212)		(0.216)		(0.154)		(0.152)
Overall risk		−0.093		−0.087		−0.224 ***		−0.120 **
		(0.081)		(0.085)		(0.059)		(0.060)
Gender			0.022	0.025			0.014	0.016
			(0.027)	(0.027)			(0.019)	(0.019)
Age			0.000	−0.001			0.0002	0.000
			(0.001)	(0.002)			(0.001)	(0.001)
Education			0.006 *	0.006			0.014 ***	0.014 ***
			(0.004)	(0.004)			(0.003)	(0.003)
Marriage			−0.012	−0.002			−0.033	−0.024
			(0.037)	(0.038)			(0.026)	(0.027)
Health			−0.139	−0.019 *			0.004	0.0002
			(0.010)	(0.010)			(0.007)	(0.007)
Length of residence			0.001	0.001			0.000	0.000
			(0.001)	(0.001)			(0.001)	(0.001)
house structure			−0.006	−0.015			0.023	0.020
			(0.022)	(0.022)			(0.015)	(0.016)
Number of rivers			0.001	0.006			0.001	0.005
			(0.023)	(0.235)			(0.016)	(0.017)
Nearest river			0.000 **	0.000			−8.760 *	−8.170
			(6.910)	(7.080)			(4.860)	(5.000)
Experiences			−0.080 *	−0.028			0.021	0.056 *
			(0.043)	(0.043)			(0.030)	(0.030)
Asset value			0.000	0.000			0.001	0.001
			(0.001)	(0.001)			(0.001)	(0.001)
Constant	0.531 ***	0.573 ***	0.589 ***	0.612 ***	0.461 ***	0.503 ***	0.299 ***	0.334 ***
	(0.029)	(0.027)	(0.109)	(0.110)	(0.021)	(0.020)	(0.076)	(0.078)
Observation	540	540	540	540	540	540	540	540
LR chi2 (×2)	26.49	1.33	45.54	17.31	43.14	14.35	98.61	66.02
Prob > chi2 (×2)	0.000 ***	0.249	0.000 ***	0.138	0.000 ***	0.000 ***	0.000 ***	0.000 ***
Pseudo R2	0.694	0.035	1.193	0.453	−0.149	−0.050	−0.341	−0.228

Notes: *** p < 0.01, ** p < 0.05, * p < 0.1.

show the results of the regressions correlating the subjective flood risk cognition of farmers with disaster preparedness, with the dependent variable being the farmers’ disaster preparedness. Model 1 shows the correlation between the four subjective risk cognition and emergency disaster preparedness, Model 2 shows the correlation between overall flood risk cognition and emergency disaster preparedness, Model 3 shows the correlation between the four subjective risk cognition and emergency disaster preparedness with the addition of control variables, and Model 4 shows the correlation between overall flood risk cognition and emergency disaster preparedness with the addition of control variables. A similar treatment was performed for the results of Models 5–16, where the dependent variables for Models 5–8 were knowledge and skills, physical, and overall disaster preparedness. In addition, the chi-square value for Model 10 was 0.249 and the chi-square value for Model 12 was 0.138, which was not significant, while all other models were significant. The following applies only to Models 3, 4, 7, 8, 11, 12 and 15, 16 after the inclusion of control variables.

Model 3 shows that the probability of subjective cognition of flood risk among farmers is significantly positively related to emergency preparedness and self-efficacy is significantly negatively related to emergency preparedness. In Model 4, there was no correlation between farmers’ subjective flood risk cognition and emergency disaster preparedness. For Model 7, self-efficacy of subjective flood risk cognition of farmers was significantly negatively correlated with knowledge and skill preparedness. For Model 8, subjective flood risk cognition of farm households was significantly negatively related to knowledge and skill preparedness.

Model 11 showed that the probability of the subjective flood risk cognition of farmers was significantly positively related to physical disaster preparedness and threat, and self-efficacy was significantly negatively related to physical disaster preparedness. In Model 12, farmers’ subjective flood risk cognition was not correlated with physical disaster preparedness. For Model 15, the probability of the subjective flood risk cognition of farmers was significantly positively related to overall disaster preparedness, and the self-efficacy of subjective flood risk cognition of farmers was significantly negatively related to overall disaster preparedness. In Model 16, the subjective flood risk cognition of farmers was significantly negatively related to overall disaster preparedness.

5. Discussion

This study aims to examine the correlation between farmers’ subjective cognition of flood risk and disaster preparedness in flood risk zones, exploring the collective cognition of flood threats faced by farmer groups [45], and how this influences disaster avoidance measures. This study correlates farmers’ subjective cognition with disaster preparedness, which summarizes and upgrades past research in this area and explores the mechanisms by which the two influences work. In this study, we used survey information from 540 farmers in flood-prone areas of southwestern China and investigated the link between farmers’ subjective flood risk cognition and disaster preparedness using the Tobit model. The results of the study showed that farmers’ overall subjective flood risk cognition were at a moderate level (3.58), and their response efficacy (3.89) and self-efficacy scores (4.22) for flood risk were high. However, their perceived probability (3.06) and threat (3.14) of flood risk occurrence scores were low. Overall, farmers’ flood disaster preparedness (0.5) was moderate, characterized by low knowledge and skills preparedness (0.37) and low emergency disaster preparedness score (0.47), while the physical disaster preparedness score was high (0.78).

There are several similarities and differences between our findings and those in the existing literature. Specifically, we find that farmers’ subjective cognition of flood risk as threatening is negatively associated with physical disaster preparedness, consistent with Hypothesis H1. This is in line with the findings of Zabini et al. [12], where respondents generally perceived that floods cause damage and that the threat of floods is positively related to vulnerability. Any physical disaster avoidance methods such as reinforcing the house adopted by farmers in the affected area will be ineffective if the threatening nature of the flood exceeds a certain limit [52,54]. Mızrak et al. suggested that farmers in the affected area need to look for more effective ways of avoiding disaster to mitigate their losses such as short-term evacuation or relocation [20,22]. Gammon et al. [18] and Seebauer [24] et al. argued that people with higher self-efficacy believe they can cope effectively with disasters and will be more active in preparing for disaster avoidance. However, our findings are inconsistent with H2, where we found that the higher farmers perceived self-efficacy for flood risk, the less prepared they are for disaster avoidance overall. This is related to the local economic and geographical characteristics of the risk zone, where the livelihood structure of farmers in a country is more likely to be influenced by the identity and occupation of the study participants [55,56], the proportion of migrant laborers in the risk zone continues to increase, and this group of farmers is no longer engaged in monoculture farming, with an increase in the proportion of wages in the farmers’ income, and a decrease in the share of agricultural income in the total household income [43]. For them, the threat of floods to their crops is not enough to put their families in a difficult situation. Moreover, Akpan’s study found that the income of farmers engaged in agriculture alone is not even enough to cover all the basic household expenses, and the farmers cannot afford to take extra money for disaster preparedness behaviors such as agricultural insurance, believing that they can cope with floods with their own hard work instead of adopting scientific preparedness to deal with floods. This is actually a negative effect of self-efficacy and they are reluctant to invest further in agricultural products. This phenomenon is in line with PMT, which believes that agriculture no longer requires high investment and uses the money for more important purposes. Thus, resulting in high self-efficacy and low disaster preparedness.

Our results are inconsistent with H3, in which farmers’ probability of flood risk cognition is positively associated with overall disaster avoidance, but overall risk cognition of floods is negatively associated with disaster preparedness., while response efficacy is not associated with disaster preparedness. This is partially similar to the findings of Ao et al. [21], suggesting that the finding that farmers’ probability of and concern about disaster risk cognition are positively and significantly related to disaster preparedness is appropriate for the risk zone. Our results are consistent with Hossain’s [43] findings in Bangladesh, which has a similar natural environment to the flood risk zone in Southwest China, where Hossain et al. concluded that local farmers who perceive a higher probability of flood risk occurrence are more likely to purchase crop insurance to cope with floods [43]. Meanwhile, the study by Gumasing et al. [26] differs from the present study in that they conclude that higher response efficacy ratings indicate that farmers perceive flood risk as more likely to require measures to mitigate the response. With high response efficacy, people are more likely to take precautionary measures to minimize losses [27]. There are also differences between the results of Frauk’s study in Riverside Island, Bangladesh, and the conclusions of this paper [41]. Frauk used the Protection Motivation Theory (PMT) to measure farmer’s perceived flood risk and adaptation assessment. The results of the study in Riverside Island showed that the higher the risk cognition of local farmers indicates the higher the sense of self-efficacy, coping efficacy of the farmers and the better the preparedness for disaster. This study is contrary to previous studies, possibly because farmers in the risk area do not have a proper understanding of flood response effects related to disaster preparedness [57]. It may also be that the situation in the flood risk area in Southwest China is different from the situation in the risk area of previous studies, the limitation of smallholder economy, poverty traps discouraging investment, etc. Therefore, under the effect of different factors, the results of the study show that there is no correlation between the two. The reason for the negative correlation between overall flood risk cognition and disaster preparedness, specifically, is that households living in flood risk areas have experienced at least one flood more than once; fortunately, after the last flood, farmers strengthened their houses, built flood protection infrastructure, and the risk resilience of their houses and other infrastructures was greatly improved [20]. Thus, reducing the level of farmer’s cognition of flood risk and the next flood of overall preparedness [50].

This study has some limitations. First, we selected a sample of farmers in Gaoxian County, Jiajiang, and Yuechi in Southwest China, and the applicability of these results to places with different natural environmental characteristics needs to be investigated. Second, this study constructed three types of preparedness for disaster avoidance, but the preparedness for disaster avoidance is affected by farmers’ subjective cognition of flooding, which are dynamic. Firstly, future research could consider collecting longitudinal data to observe the dynamics of farmers’ subjective cognition to improve the robustness of the findings. Second, after ensuring the robustness of the findings, we can consider adding the variable of farmers’ objective risk cognition of flooding to study the impact of farmers’ subjective and objective flood risk cognition on disaster preparedness, which will help to establish a disaster prevention and mitigation system in other flood risk areas of southwestern China.

6. Conclusions and Policy Implications

Understanding farmers’ subjective cognition of flood risk is crucial for their protection motivations and avoidance preparedness behaviors in response to floods. However, research in this area, especially in Southwest China, is limited and it is difficult to find research results to build upon. In this study, we analyzed farmers’ subjective cognition of flood risk and disaster preparedness through survey responses from 540 farmers in a flood risk area in Southwest China and constructed a Tobit model to explore the link between these two factors. Our conclusions are as follows:

Firstly, there are significant differences in farmers’ subjective flood risk cognition in different areas, and the level of subjective flood risk cognition of farmers in Yuechi is higher than that of farmers in Jiajiang and Gaoxian County.

Second, the overall subjective risk cognition of farmers in flood risk areas was at a medium-high level (3.58), with the highest self-efficacy score for flood risk (4.22), indicating that farmers were optimistic about flood risk and able to cope with the impacts of flood risk even though they were in flood risk areas.

Finally, there is a significant relationship between farmers’ subjective flood risk cognition and preparedness for disaster avoidance. Farmers’ probability of flood risk cognition was significantly and positively related to disaster preparedness; farmers’ threat of flood risk cognition was significantly and negatively related to physical disaster preparedness; farmers’ self-efficacy of flood risk cognition was significantly and negatively related to disaster preparedness; and farmers’ flood risk cognition was significantly and negatively related to knowledge and skill preparedness and overall disaster preparedness.

The results of this study not only fill a gap in this area for academics, but also provide valuable insights into flood prevention and mitigation efforts in flood risk areas and the development of adaptation policies The following recommendations are made in response to the findings of this study:

Firstly, the subjective flood risk cognition of farmers should be improved. This is mainly due to the fact that farmers scored the lowest in terms of their awareness of the probability of flooding, and their lack of awareness of flooding may lead to serious consequences. Therefore, at the individual level, the government should conduct flood learning and education, and farmers should conduct in-depth learning about floods to improve their subjective risk cognition.

Second, farmers’ types of disaster avoidance should be enriched. In the study, it was found that farmers have different ways of avoiding disasters between different regions, and farmers between different regions can communicate with each other to learn excellent disaster avoidance behaviors and improve the level of disaster avoidance.