The Determinants of Farmers’ Perceived Flood Risk and Their Flood Adaptation Assessments: A Study in a Char-Land Area of Bangladesh

Abstract

Floods are the most frequent and devastating disasters in Bangladesh. The riverine islands, known as char-lands, are particularly vulnerable to flooding. As flooding poses a significant threat to the lives and livelihoods of residents, especially farmers, it is crucial to understand how they perceive flood risk and assess their adaptation strategies in this geographically susceptible context. However, the existing literature has not adequately addressed these issues. Therefore, this study aims to analyze the factors influencing farmers’ perceptions of flood risk and their assessments of flood adaptation. In a survey of 359 farmers in Bangladesh’s char-land region, located in the Chauhali sub-district (Upazila) of Sirajganj district, we used the protection motivation theory (PMT) to measure farmers’ perceived flood risk and adaptation assessments. Multiple regression analysis was employed to identify factors influencing them. Farmers prioritized the risk to livelihoods (production and income) over psychological aspects (health and diseases). Larger farms, more flood experience, and greater risk awareness are associated with higher overall flood risk perception and better flood adaptation, indicating higher self-efficacy, response efficacy, and response cost among farmers. Farmers perceived lower flood risk in exchange for greater house distance from the river and more trust in government actions. Hence, strengthening campaigns and programs is crucial to understanding flood risk in char-lands for improved adaptation to floods. The study highlights the application of PMT to assess farmers’ perceptions of flood risk and their attitudes towards adaptation, suggesting further research opportunities.

1. Introduction

Natural hazards pose significant threats worldwide, and it is projected that a substantial portion of the global population will be exposed to catastrophic events by 2050. Among various natural hazards, flood disasters stand out as a major cause of death and devastation, with significant economic and social consequences [1]. In fact, over 30% of all natural disasters in the last century have been attributed to floods [2]. Certain regions are more vulnerable to specific hazards, and Bangladesh, a densely populated nation in South Asia with low-lying terrain, is highly susceptible to flooding due to its geographical characteristics. Approximately 80% of the country consists of floodplains formed by rivers like the Ganges, Brahmaputra, and Meghna [3]. Moreover, the unique natural environment of Bangladesh, coupled with the characteristics of the tropical monsoon climate, significantly contributes to the country’s elevated flood risk [4]. Among the flood-prone regions in Bangladesh, char-lands, which are riverine islands formed by the dynamics of erosion and accretion, are particularly susceptible to flooding. Almost 10% of the population lives on these islands (chars) of the world [5,6]. In addition, there is an estimation that approximately 4–5% of the population in Bangladesh lives in the char lands, which cover almost 7200 km² [7,8], and Sirajganj district covers 44,000 hectares, or 440 km² [9]. There are 56 big and 226 small chars in Bangladesh [10]. These char-lands periodically protrude from the riverbed, offering opportunities for settlement and agricultural activities. However, during the monsoon season, widespread flooding wreaks havoc on char settlements, agricultural crops, infrastructure, and communication networks. Most char families’ livelihoods, reliant on sharecropping, agricultural labor, and livestock farming, suffer substantial setbacks due to floods and riverbank erosion. This underscores the pressing need to safeguard livelihoods in flood-prone char communities. Moreover, char-lands’ agricultural fertility contributes notably to national production, necessitating measures to mitigate flood-induced agricultural losses.

In recent years, flood adaptation in char-lands has garnered significant attention in scholarly research. Naz et al. [11] conducted a gender-based analysis to determine the factors affecting their flood adaptation strategies separately. Hossain et al. [12] explored the impacts of floods on the livelihoods of people in char villages, particularly on income and occupation, and also explored their coping strategies. Faruk and Maharjan [13] identified the impact of farmers’ participation in community-based organizations on their flood adaptation adoption. However, our study distinguishes itself from the aforementioned literature by aiming to assess farmers’ perceived flood risk (threat appraisal) and flood adaptation measures (coping appraisals: self-efficacy, response efficacy, and response cost). Furthermore, our research endeavors to pinpoint the determinants that shape farmers’ perceptions of flood risk and their evaluations of flood adaptation. Many studies abroad have also extensively explored the key drivers of flood adaptation. For example, Tasantab et al. [14] explored the factors influencing flood risk adaptation intentions in Ghana. Grothmann and Reusswig [15] identified the factors to answer the question of why some people take precautionary actions while others do not in Germany. Poussin et al. [16] conducted a survey of 885 households in three flood-prone regions in France and recognized the influential factors of flood mitigation behavior. These studies focused on analyzing the factors that influence flood adaptation behavior using the Protection Motivation Theory (PMT). Originally proposed by Rogers [17] to explain how people defend themselves against health risks, PMT has been adapted for studying flood risk by researchers. PMT has gained popularity to explain how locals take precautions against natural disasters [15,18,19]. PMT involves two cognitive processes: threat appraisal and coping appraisal. Threat appraisal encompasses risk perception or perceived risk, including perceived risk probability and consequences associated with a hazard [15,20]. As outlined by Bubeck et al. [21], most of the studies [22,23,24,25,26] examined in their review focus primarily on evaluating risk perception by considering solely the flood risk probability. Some studies identified risk perception as having three main components: awareness, worry, and preparedness [27,28]. Recognizing the way in which the community perceives the flood risk (threat assessment) holds significant importance in identifying an appropriate approach for disseminating flood-related information. This approach aims to enhance the public’s confidence in their government, ultimately resulting in improved preparedness for floods and bolstered societal resilience [29]. As emphasized by Bradford et al. [29], prioritizing the public’s perception of risk is paramount, as a lack of comprehension by authorities regarding societal perspectives has been identified as a key factor leading to inadequacies in flood risk management. Indeed, flood risk as perceived by the community often diverges from the assessment of flood risk determined by experts [30,31,32]. Interestingly, individuals tend to have a more accurate understanding of flood risk compared to the broader societal perception [33]. An important challenge in flood management is the significant number of people who underestimate the actual flood risk. This issue highlights the need for a comprehensive understanding of the factors that shape subjective perceptions of flood risk. When threat assessment or perceived risk reaches a certain level, coping appraisal, or consideration of adaptation assessments, is initiated to reduce or prevent the threat. According to PMT, individuals will adopt adaptive measures against a specific risk if they perceive it to be high (threat assessment). Poussin et al. [16] revealed that high coping appraisals or adaptation assessments occur when individuals rate their ability to adapt as high (high self-efficacy), effective (high response efficacy), and cost-effective (low response costs). High perceived risks can result in the adoption of a protective response if they are accompanied by high adaptation assessments, but they can also result in maladaptive responses like fatalism, denial, or wishful thinking if they are accompanied by poor adaptation assessments [34]. Therefore, it is crucial to understand the interplay between residents’ flood risk perception and flood adaptation assessment for better flood adaptation.

Given the crucial role of perceived risk and adaptation assessments in influencing protection motivation and subsequent behavior, it becomes essential for policies aiming to enhance individual risk preparedness to understand the factors influencing these perceptions. Numerous studies have explored various factors related to risk perception. For instance, some researchers [35,36] discovered an inverse relationship between respondents’ level of education and their risk perception due to better access to disaster-related information among highly educated individuals. Wang et al. [37] conducted a survey in Jingdezhen, China, revealing a strong connection between public risk perception and influencing factors. Ardaya et al. [38] highlighted flood experiences and demographic characteristics as the primary determinants of risk perception. Risk perception can also be strengthened or weakened by indirect experience. In this respect, communication networks in the form of media and personal interactions with people play a major role [39]. Lechowska [40] reviewed multiple sources and categorized factors affecting flood risk perception into cognition, behavior, socioeconomic and demographic aspects, geography, information, and cultural background. This collective analysis identified measurement methods and explored influential factors in risk perception. However, despite the long history of floods in Bangladesh, there is a scarcity of research concerning flood risk perception, particularly when it comes to the farming community in the highly flood-prone char-lands. Therefore, this study aims to conduct quantitative measurements of how farmers perceive flood risks concerning different aspects of their lives, such as physical health and diseases, household assets, crops and livestock, and income, and identify the influential factors. Although previous studies [15,16,20,41] have analyzed all flood adaptation assessments (self-efficacy, response efficacy, and response cost) to assess their impact on flood adaptation behavior, little attention has been given to analyzing the factors influencing adaptation assessments (self-efficacy, response efficacy, and response cost). Babcicky and Seebauer [42] explicitly investigate factors that influence only one component of adaptation assessment (self-efficacy) of flood-affected households in Austria, while Lindell et al. [43] identified a relationship between socioeconomic variables and perceived attributes of earthquake preparedness measures related to response efficacy and response cost. Bubeck et al. [44] considered all three components (self-efficacy, response efficacy, and response cost) to show the relationship of different factors with them based on structural and non-structural measures. As farming is the main livelihood option in the char-lands, this study separated the flood adaptation measures into two broad categories: farming and non-farming flood adaptation measures, and assessed the flood adaptation based on those measures. While Dang et al. [45,46] identified significant factors influencing farmers’ perceived risk and appraisal of private adaptive measures to climate change in the Mekong Delta, Vietnam, there is a notable absence of research addressing the distinct influential factors simultaneously affecting the two major components of PMT-threat appraisal (perceived risk) and coping appraisal (adaptation assessments) in the context of flood adaptation within geographically vulnerable areas such as char-lands in Bangladesh. Considering the frequency of flooding in char-land and the predominant reliance on agriculture as the main livelihood in this region, it becomes crucial to comprehend how flood risk and adaptive measures are evaluated by char-land farmers. Identifying the various factors involved in this assessment is also of paramount importance. Therefore, this study aims to:

• To quantitatively measure farmers’ perceived risks of flooding and to appraise farmers’ flood adaptation measures.
• To explore different factors influencing farmers’ perceived flood risk and flood adaptation assessments.

We conducted an empirical investigation to examine various socio-demographic factors such as farmers’ age, year of schooling, family size, children under 10 years, farm size, livestock ownership, annual income, distance from the river, and cognitive factors (flood experience, flood risk awareness, and trust in government actions) that may influence farmers’ perceived flood risk and adaptation assessments. We analyzed data from 359 flood-affected char-land farmers using multiple linear regression analysis to gain a better understanding of these factors. This study’s risk assessment methodology is noteworthy, constructing an aggregate perceived flood risk score by considering both risk probability and consequences across four risk dimensions. Furthermore, this research is unique in simultaneously analyzing factors influencing both perceived flood risk (threat appraisal) and flood adaptation assessments (coping appraisals). This study also contributes significantly by providing systematic insights into the key components of the Protection Motivation Theory (PMT), namely threat appraisals (perceived flood risk) and coping appraisals (flood adaptation assessments), within the context of the char-land region, known for its vulnerability to recurrent flooding. The remainder of this article is structured as follows: Section 2 outlines the materials and methods, emphasizing the data collection process and the analytical methods employed. Section 3 presents the results or main findings of the study, while Section 4 focuses on discussing the factors influencing perceived flood risk and flood adaptation assessments. Finally, Section 5 encompasses the conclusions and recommendations. This research provides valuable insights that can inform flood risk management strategies in vulnerable regions like char-lands.

2. Materials and Methods

2.1. Description of the Study Area

Sirajganj, located in northern Bangladesh, is a flood-prone district predominantly consisting of char regions. The district is intersected by the Brahmaputra River, commonly known as the Jamuna River, which regularly experiences overflowing during the rainy season, resulting in the inundation of various parts of Sirajganj. The Chauhali sub-district (Upazila) within Sirajganj district is particularly susceptible to severe flooding due to its location on the banks of the Jamuna River (Figure 1). Considering the significant impact of floods in the region, the Chauhali sub-district was chosen as the focus area for this investigation.

The Chauhali sub-district is geographically divided into two halves by the Jamuna River. In this region, the most prevalent and severe disasters are frequent floods and riverbank erosion. The land in this sub-district is regularly lost to the river due to riverbank erosion occurring at various intervals. The study was conducted in six villages in Ghorjan and Sthal unions in Chauhali sub-district. These unions, situated in the womb of the Jamuna River, are flooded frequently every year. The map above represents the river system in the study area and indicates the flood risk union under the Chauhali sub-district.

2.2. Sampling and Data Collection

A standardized questionnaire was used to conduct face-to-face farmer interviews in August 2021. The questionnaire was initially intended to address three primary topics, such as flood risk perception, flood adaptation assessments, and several associated factors (e.g., socio-demographic and cognitive factors) based on the flood occurred in 2020. The questionnaire was improved and finalized using qualitative information from an earlier preliminary survey and different key informants (sub-assistant agricultural officer interviews, NGO workers, local leaders, etc.) in the study area. After that, it was pre-tested and modified to verify its relevancy and comprehensibility. The study ethics committee at the Graduate School of Humanities and Social Sciences, Hiroshima University, authorized the questionnaire for conformity with ethical concerns such as basic human rights, the protection of personal information, and data security prior to performing the final survey.

A multistage sampling procedure was used to determine the sample size. First, two unions, Ghorjan and Sthal, in Chowhali subdistrict were purposely designed based on the severity of the floods. Moreover, three villages from each union—Harghorjan, Boroghorjan, and Muradpur from the Ghorjan union—and three villages from the Sthal union—South Nouhata, North Nouhata, and Chaluhara—were selected at random. A village-wise list of farmers was obtained from the sub-district agricultural office. From each village, a random sample was chosen. Using these techniques, 359 farmers were chosen as the sample size from a limited population.

2.3. Variable Description and Analysis

Table 1. Description of variables.

Variables	Description	Measuring Unit
Explanatory Variables
Age	Age of respondents	Age in years
Gender	Gender of farmers	1 if male, 0 if otherwise
Years of schooling	Years of formal schooling received by respondents	No. of years
Family size	Members of the family	No. of family members
Children under 10 years	No. of children under 10 years old	No. of children
Farm size	Agricultural land under cultivation in decimals	Land size in decimals
Annual income	Total annual income in BDT (after cost)	Income in thousand BDT
Livestock ownership	How many livestock farmers owned?	No. of livestock
Distance from river	House distance from the river in kilometers	Distance in kilometers
Flood experience	Flood severity experienced in the last 10 years	No. of flood severities
Trust in government actions	Respondents were asked to rate the frequency of showing trust in govt. flood protection measures four items)	1—very untrustworthy, 2—untrustworthy, 3—neutral, 4—trustworthy, 5—very trustworthy
Flood risk awareness	Respondents were asked their opinions on five flood-related statements	1—strongly disagree, 2—disagree, 3—neutral, 4—agree, 5—strongly agree
Dependent variables
Perceived flood risk	There are a total of eight components in this personal evaluation of the probability of a future occurrence (a) and the consequent damage (b)	(a) 1—very unlikely, 2—rather unlikely, 3—neutral, 4—rather likely, 5—very likely (b) 1—not bad at all; 2—rather not bad; 3—neutral; 4—rather bad; 5—very bad;
Self-efficacy	The respondent thinks that he/she is capable of taking the described 21 measures.	1—very unable, 2—rather unable, 3—neutral, 4—rather able, 5—very able
Response efficacy	Effectiveness of the described 21 flood adaptation strategies	1—very ineffective, 2—rather ineffective, 3—neutral, 4—rather effective, 5—very effective
Response cost	To what extent the adaptation measures are costly (21 items)	1—very costly, 2—rather costly, 3—neutral, 4—rather not costly, 5—not very costly

shows how variables were measured. The respondents were asked to rate on a five-point Likert scale how likely and severe they believed floods would impact different aspects of their lives, including valuable assets, crops and livestock, physical health, and income, if they did not implement any flood adaptation measures. The scale ranged from 1 (very unlikely) to 5 (very likely) for perceived probability and from 1 (not bad at all) to 5 (very bad) for perceived severity [20].

To calculate the perceived risks associated with flood hazards for each dimension and the total perceived risk, the formulae suggested by Dowling [49] were employed. These formulas are based on attitude models widely used in marketing and psychology, providing an information-processing view of decision-making [49]. The application of these formulas aligns with the concept of “threat assessment” discussed in the introduction part.

$$\begin{gathered} \mathrm{Perceived} \mathrm{Risk} = \mathrm{perceived} \mathrm{risk} \mathrm{probability}\times \mathrm{perceived} \mathrm{risk} \mathrm{consequences} \\ \mathrm{Overall} \mathrm{Perceived} \mathrm{Risk}={{\displaystyle \sum }}_{\mathrm{i}=1}^{\mathrm{n}}\mathrm{perceived} \mathrm{risk} \mathrm{probability}\times \mathrm{perceived} \mathrm{risk} \mathrm{consequences} \end{gathered}$$

where n is the number of dimensions (i.e., household assets, crops and livestock, physical health, and income).

For each risk dimension that is believed to be impacted by flooding, the perceived risk was calculated by multiplying the estimated risk probability by the perceived risk consequences. All four risk dimensions were added up to get the overall perceived flood risk.

Each risk dimension and the overall perceived flood risk were determined for 359 char-land farmers. Overall perceived flood risk was used as the dependent variable for multiple linear regressions in this study.

Initially, a list of 21 individual flood adaptation measures was developed using the literature [16,20,46]. To check the validity of these flood adaptation measures, they were brought up for debate in FGDs during the preliminary survey. The adaptive measures were pretested before the final survey. Farmers were asked to assess their own ability to execute each adaptive measure (self-efficacy), to rate the effectiveness (response efficacy), and to rate the perceived cost of implementing each adaptation strategy (response cost). The average of adaptation assessments (perceived self-efficacy, perceived response efficacy, and perceived response cost) has been used as three dependent variables in separate regression models.

In this study, multiple regression was used to examine the relationship between several independent variables (socio-economic and cognitive factors) and a dependent variable (perceived flood risk and flood adaptation assessments). The multiple regression equation can be written as:

$$y_i ═ {\beta }_1{x}_1+{\beta }_2{x}_2+{\beta }_3{x}_3+\dots +{\beta }_i{x}_i+{\mathsf{\epsilon }}_i$$

Here $y_i$ denotes dependent variables (perceived flood risk and flood adaptation assessment-perceived self-efficacy, response efficacy, and response cost), x_i denotes independent variables (shown in Table 1), and ${\mathsf{\epsilon }}_i$ is the error term.

Multiple linear regression captures complex relationships between multiple independent variables and a dependent variable, aiding in variable importance quantification and prediction; however, it is sensitive to multicollinearity, overfitting, outliers, and violations of assumptions, potentially leading to misleading results [50].

3. Results

3.1. Descriptive Statistics of the Study

Before descriptive analysis of the independent variables, histogram plots for continuous and discrete variables were made to present the data distribution as shown in Appendix A (Figure A1). The study’s descriptive statistics are presented in

Table 2. Descriptive statistics.

Variables	Mean	SD	Minimum	Maximum
Age	48.68	12.78	25	85
Gender	0.70	0.46	0	1
Years of schooling	2.82	3.05	0	12
Family size	5.61	1.77	0	12
Children under 10 years	1.33	0.87	0	4
Farm size (dm)	131.79	59.43	33	363
Annual income (000′ BDT)	44.14	22.10	6	150
Livestock ownership	0.86	0.37	0	1
House distance from river (km)	1.68	0.79	0.25	4
Flood experience	2.50	0.74	1	4
Trust in government actions	2.78	0.49	1.6	4.6
Flood risk awareness	3.87	0.55	2.6	5

. The results indicate that the average age of the farmers is 48.68, suggesting that older individuals are more likely to be engaged in agriculture. Males constituted over 70% of the survey respondents in the char-lands. In the char-lands, the average number of years of schooling was relatively low (2.82), reflecting primary-level education, which is typical in the context of Bangladesh. This is consistent with similar findings in the Padma floodplain, where the average schooling years were 1.9 years [51]. Additionally, another study [52] found that 45% of the floodplain population had only primary education. On average, there are around six people per household, with approximately one child under the age of 10. The average land under cultivation is 131.79 decimals, indicating that agriculture is the primary livelihood option for char-dwellers. However, despite their reliance on agriculture, the farmers’ average annual income was relatively low at 44,140 BDT.

In the char-lands, most farmers (86%) owned livestock. Due to their location on riverine islands, farmers lived near the river, with an average distance of 1.68 km. On average, farmers in the char-lands experienced 2.50 flood-severity events in the last decades. Trust in government action was relatively low, with an average rating of 2.78, emphasizing the significance of private flood adaptation measures. The average flood risk awareness among farmers in the char-lands was 3.87.

3.2. Farmers’ Adoption of Flood Adaptation Measures

Table 3. Farmers’ adoption of flood adaptation measures.

Variables	Frequency and Percentage of Adoption
	Frequency	Percentage
Farming flood adaptation measures
Growing seedlings in a pot or sandbag	137	38.16
Mixed cropping	202	56.27
Changing crop varieties	145	40.39
Adjustment of planting and harvesting times	190	52.92
Fodder arrangement	294	81.89
Raising livestock shed	278	77.44
Relocating livestock	166	46.24
Money savings	180	50.14
Informal credit	222	61.84
Formal credit	187	52.09
Alternative occupations during floods	160	44.57
Non-farming flood adaptation measures
Construction or raising the plinth of the house	150	41.78
Fencing house	137	38.16
Raising tube wells	164	45.68
Flood-proof sanitation	162	45.13
Portable stoves	303	84.40
Arrangement of the boat	135	37.60
Macha preparation	260	72.42
Dry food collection	207	57.60
Shifting family	209	58.22
Shifting valuable goods	212	59.05

presents the percentage of adoption for twenty-one flood adaptation strategies utilized by farmers in the char-lands. According to the preliminary survey, farmers tended to implement most of these measures, both related to farming and non-farming practices. Each adaptation measure was scored 1 if adopted and 0 if not. Given that agriculture is the primary livelihood in the char-lands, local farmers have traditionally employed various agricultural and livelihood adaptation measures. For instance, 81.89% of farmers arranged fodder for their livestock, and about three-fourths (77.44%) raised their livestock to a safer place before the onset of floods, reflecting their concern for livestock safety. To minimize crop losses, more than half of the respondents (56.27%) adopted mixed cropping practices, and 52.92% adjusted planting and harvesting times. Regarding financial measures, around 61.84% of farmers utilized informal sources for financing, while 52.09% used official sources to continue their farming livelihood.

Additionally, more than half of the respondents saved money for emergency requirements in anticipation of the impending flood disaster, which may hamper their agricultural production. These findings highlight the proactive approach of char-land farmers in implementing various adaptation strategies to cope with flood risks.

Among the non-farming flood adaptation measures, a significant proportion of farmers (84.4%) installed portable cooking stoves before floods in the char-lands, as most homes with permanent cooking stoves have yards that become submerged during floods. Another crucial non-farming flood adaptation strategy is the construction of “macha,” with 72.42% of farmers adopting this measure. A macha is a high stage or bed made primarily of low-cost material such as bamboo. During flooding, farmers take refuge on top of the macha and safeguard their essential belongings.

However, other structural improvements had lower adoption rates, including building or elevating the house plinth (41.78%), raising tube wells (45.68%), and implementing flood-proof sanitation measures (45.13%). These findings suggest that while some non-farming strategies are widely adopted, there is scope for further improvements in other structural measures to enhance flood resilience in the char-lands.

3.3. Farmers’ Perceived Flood Risk

The overall perceived risk associated with floods is derived by summing up the four distinct perceived risks, calculated by multiplying perceived risk probability by perceived risk consequences for each risk dimension. The minimum and maximum values for each specific perceived risk are assumed to be 1 and 25, respectively, resulting in a range of 4 to 100 for the total perceived risk. In our sample, the overall perceived flood risk is regularly distributed, ranging from 28 to 100.

The average overall perceived flood risk is 62.52, which is above half of the maximum level of perceived flood risk. This suggests that among the surveyed farmers, there is a relatively high level of perceived flood risk. Farmers perceive flood hazards differently in various aspects of their lives. They tend to place higher importance on the probability of floods affecting their crops, livestock, and income, with average perceived risks of 17.75 and 17.04, respectively, compared to household assets and physical health, which have average perceived risks of 15.09 and 12.64, respectively (

Table 4. Means and standard deviations (SD) of perceived risk probability, perceived risk severity, and specific perceived risks.

Risk Dimensions	Perceived Risk Probability		Perceived Risk Consequences		Specific Perceived Risk
	Mean	SD	Mean	SD	Mean	SD
Household assets	3.91	0.74	3.76	0.72	15.09	4.99
Crops and livestock	4.23	0.60	4.16	0.57	17.75	4.19
Income	4.13	0.72	4.05	0.69	17.04	5.19
Physical health and diseases	3.52	0.98	3.39	0.91	12.64	6.10

).

In the context of perceived risk probability and perceived consequences, crops and livestock are considered to have the highest mean values (4.23 and 4.16, respectively) compared to the other dimensions when assessing the potential impact of floods. On the other hand, floods are perceived to have a lesser impact on farmers’ physical health and diseases, as evidenced by the lowest mean values of perceived risk probability (3.52) and perceived risk consequences (3.39) in Table 4. These findings suggest that farmers prioritize and perceive greater vulnerability for their crops and livestock compared to other aspects of their lives when it comes to the risks posed by floods.

Furthermore, we assessed the contributions of each perceived risk dimension to the overall perceived risk and compared their percentages. As shown in Figure 2, among the risk dimensions, perceived risk for crops and livestock had the highest contribution, accounting for 28.39% of the overall perceived risk. It was followed by farmers’ income at 27.25%, household assets at 24.14%, and physical health and diseases at 20.22%. These results indicate that farmers attribute the highest importance to the risks posed to their crops and livestock, followed closely by their income and household assets, while considering physical health and diseases to have a relatively lower impact on the overall perceived flood risk. To comprehend how risk is perceived, additional Pearson correlation and simple regression analysis showing the relationship of different risk dimensions with overall perceived flood risk have been shown in Appendix A (

Table A1. Relationship of different risk dimensions with overall perceived flood risk.

Different Dimension of Flood Risk	Overall Perceived Flood Risk
	Pearson Correlation Coefficient	Regression Coefficient
Perceived risk probability on
Physical health	0.772 ***	3.320 ***
Household assets	0.723 ***	3.678 ***
Crop and livestock production	0.591 ***	3.864 ***
Income	0.651 ***	3.696 ***
Perceived risk consequences on
Physical health	0.744 ***	3.453 ***
Household assets	0.693 ***	3.930 ***
Crop and livestock production	0.549 ***	4.222 ***
Income	0.654 ***	4.215 ***

Note: *** denotes a significance level of 1%.

).

3.4. Farmers’ Assessment of Flood Adaptation

Farmers’ flood adaptation assessments encompass perceived self-efficacy, response efficacy, and response cost.

Table 5. Farmers’ assessments of flood adaptation measures.

Flood Adaptation Measures	Self-Efficacy		Response Efficacy		Response Cost
	Mean	SD	Mean	SD	Mean	SD
Farming flood adaptation measures
Growing seedlings in a pot or sandbag	3.13	0.99	2.89	0.87	3.60	0.69
Mixed cropping	3.29	1.09	3.09	0.93	3.68	0.71
Changing crop varieties	2.70	0.95	2.74	0.86	2.97	0.84
Adjustment of planting and harvesting times	3.03	0.95	3.06	0.84	3.37	0.73
Fodder arrangement	3.78	0.78	4.07	0.64	2.82	0.98
Raising livestock	3.48	0.98	3.83	0.67	2.96	1.05
Relocating livestock	3.18	0.87	3.30	0.79	3.15	0.75
Money savings	2.77	0.97	4.00	0.75	3.22	0.52
Informal credit	3.21	0.90	2.99	0.82	2.67	0.86
Formal credit	3.26	0.88	3.45	0.77	3.32	0.80
Alternative occupations during floods	2.95	1.02	3.74	0.73	3.23	0.72
Non-farming flood adaptation measures
Construction or raising the plinth of the house	2.78	1.16	4.21	0.78	1.45	0.56
Fencing house	3.20	1.05	3.06	0.87	2.77	1.01
Raising tube wells	2.92	1.03	3.90	0.59	2.20	0.73
Flood-proof sanitation	2.92	1.06	4.13	0.61	1.97	0.57
Portable stoves	4.42	0.68	4.11	0.61	4.34	0.80
Arrangement of the boat	2.81	1.05	3.99	0.70	2.18	0.74
Macha preparation	3.95	0.74	3.48	0.76	3.86	0.69
Dry food collection	3.70	0.90	3.40	0.80	3.60	1.00
Shifting family	3.35	0.90	3.54	0.86	3.48	0.77
Shifting valuable goods	3.37	0.82	3.54	0.66	3.25	0.80

presents the means and standard deviations of these three variables. Regarding farming-related flood adaptation measures, farmers exhibited higher self-efficacy in activities like arranging fodder (3.78), raising livestock (3.48), and practicing mixed cropping (3.29). Notably, farmers demonstrated greater self-efficacy in obtaining credit from formal sources (3.26) compared to informal sources (3.21).

For non-farming flood adaptation measures, higher levels of self-efficacy were observed among farmers for using portable stoves (4.42), preparing macha (3.95), and collecting dry food (3.70). In contrast, self-efficacy was relatively lower for structural measures such as house construction or elevation (2.78), raising tube wells (2.92), implementing flood-proof sanitation (2.92), and preparing boats (2.81). These findings highlight farmers’ varying levels of confidence in different flood adaptation strategies, reflecting their perceptions of efficacy and potential outcomes associated with these measures.

In terms of response efficacy, farmers considered activities such as arranging fodder (4.07), saving money (4.00), and raising level of livestock shed (3.83) as more effective in the context of farming-related flood adaptation measures. For non-farming adaptation measures, higher levels of response efficacy were reported by farmers for activities like house construction or elevation (4.21), flood-proof sanitation (4.13), using portable stoves (4.11), and preparing boats (3.99). These responses indicate that farmers perceive certain measures as more efficacious in mitigating flood-related challenges, highlighting their beliefs in the effectiveness of these strategies to adapt to flood risks.

Farmers perceived mixed cropping (3.68), growing seedlings in pots or sandbags (3.60), and obtaining formal credit (3.32) as more cost-effective in comparison to other farming-related adaptation measures. Among non-farming flood adaptation measures, portable stoves (4.34), macha preparation (3.86), and dry food collection (3.60) were considered more financially feasible by farmers, in contrast to other non-farming measures such as house construction or elevation (1.45), flood-proof sanitation (1.97), and boat arrangement (2.18). This indicates that farmers consider the affordability of different adaptation measures when making decisions to address flood risks, considering the associated costs and benefits. Additionally, the Pearson correlation coefficient value for each adaptation measure has been shown to demonstrate the relationship with flood adaptation assessments such as self-efficacy, response efficacy, and response cost (

Table A2. Relationship of different flood adaptation assessments with overall perceived self-efficacy, response efficacy, and response cost.

Flood Adaptation Measures	Flood Adaptation Assessment
	Perceived Self-Efficacy	Perceived Response Efficacy	Perceived Response Cost
Self-efficacy, response efficacy, and response cost of farming flood adaptation measures
Growing seedlings in a pot or sandbag	0.476 ***	0.437 ***	0.474 ***
Mixed cropping	0.450 ***	0.476 ***	0.545 ***
Changing crop varieties	0.404 ***	0.446 ***	0.392 ***
Adjustment of planting and harvesting times	0.514 ***	0.429 ***	0.462 ***
Fodder arrangement	0.470 ***	0.396 ***	0.455 ***
Raising livestock	0.562 ***	0.401 ***	0.606 ***
Relocating livestock	0.216 ***	0.159 ***	0.483 ***
Money savings	0.572 ***	0.522 ***	0.226 ***
Informal credit	0.288 ***	0.236 ***	0.311 ***
Formal credit	0.568 ***	0.504 ***	0.581 ***
Alternative occupations during floods	0.308 ***	0.243 ***	0.390 ***
Self-efficacy, response efficacy, and response cost of non-farming flood adaptation measures
Construction or raising the plinth of the house	0.654 ***	0.498 ***	0.518 ***
Fencing house	0.592 ***	0.326 ***	0.638 ***
Raising tube wells	0.642 ***	0.456 ***	0.532 ***
Flood-proof sanitation	0.556 ***	0.527 ***	0.295 ***
Portable stoves	0.467 ***	0.409 ***	0.473 ***
Arrangement of the boat	0.350 ***	0.397 ***	0.135 **
Macha preparation	0.367 ***	0.407 ***	0.422 ***
Dry food collection	0.546 ***	0.432 ***	0.566 ***
Shifting family	0.259 ***	0.273 ***	0.478 ***
Shifting valuable goods	0.295 ***	0.303 ***	0.545 ***

Note: *** and ** denote significance levels of 1% and 5%, respectively.

in Appendix A).

3.5. Factors Influencing Farmers’ Perceived Flood Risk and Flood Adaptation Assessments

This section presents the results of multiple regression models that analyze the effects of socio-demographic and cognitive characteristics on farmers’ perceived flood risk across four dimensions and their overall perceived flood risk. Additionally, multiple regression analysis was also conducted to identify factors influencing farmers’ flood adaptation assessments in the categories of self-efficacy, response efficacy, and response cost.

Prior to conducting the regression analysis, a multicollinearity test was performed to assess variance inflation factors (VIFs). Results showed that most of the explanatory variables exhibited VIFs of less than 2, with a maximum VIF of 3.7 (

Table 6. Multicollinearity test: variance inflation factor.

Variables	VIF	1/VIF
Age	1.21	0.83
Gender	1.09	0.92
Years of schooling	1.19	0.84
Family size	1.47	0.68
Children under 10 years	1.27	0.79
Farm size (dm)	2.12	0.47
Annual income	1.65	0.61
Livestock ownership	1.03	0.97
House distance from river	1.59	0.63
Flood experience	1.92	0.52
Flood risk awareness	2.24	0.45
Trust in government actions	1.25	0.80

). This suggests the absence of significant multicollinearity. A standard VIF cutoff value of 10 is often used to identify substantial multicollinearity [53]. Kehinde and Ogundeji [54] further suggest that variables with VIF exceeding 10 should be regarded as highly collinear and can potentially be excluded from the model.

Even when VIF is far below the cut-off criterion [53] or more than 5, the degree of multicollinearity should be carefully evaluated [55]. It is nevertheless justified, to consider the multicollinearity problem beyond the cut-off point. In this situation, removing or combining variables might result in more severe issues.

3.5.1. Factors Influencing Farmers’ Perceived Flood Risk

Table 7. Factors influencing farmers’ perceived flood risk.

Variables	Physical Health and Diseases	Household Assets	Crops and Livestock	Income	Overall Perceived Flood Risk
Age	−0.111 ** (0.022)	−0.081 * (0.018)	0.065 (0.015)	−0.009 (0.018)	−0.056 (0.042)
Gender	−0.144 *** (0.597)	−0.117 ** (0.476)	0.079 * (0.410)	0.028 (0.489)	−0.065 * (1.133)
Years of schooling	0.032 (0.093)	0.060 (0.074)	−0.006 (0.064)	0.034 (0.077)	0.043 (0.177)
Family size	−0.028 (0.178)	−0.026 (0.142)	0.023 (0.123)	0.067 (0.146)	0.009 (0.339)
Children under 10 years	−0.005 (0.339)	−0.008 (0.270)	0.015 (0.233)	−0.062 (0.278)	−0.021 (0.643)
Farm size (dm)	0.206 *** (0.006)	0.244 *** (0.005)	0.049 (0.004)	0.064 (0.005)	0.197 *** (0.012)
Annual income	0.119 ** (0.015)	0.052 (0.012)	0.096 * (0.010)	−0.095 * (0.012)	0.059 (0.029)
Livestock ownership	−0.069 (0.763)	−0.003 (0.608)	0.030 (0.525)	0.031 (0.626)	−0.010 (1.450)
House distance from river	−0.057 (0.416)	−0.021 (0.331)	−0.164 *** (0.286)	−0.174 *** (0.341)	−0.134 *** (0.790)
Flood experience	0.174 *** (0.486)	0.196 *** (0.387)	0.204 *** (0.334)	0.175 *** (0.398)	0.249 *** (0.923)
Flood risk awareness	0.190 *** (0.712)	0.232 *** (0.568)	0.129 ** (0.490)	0.336 *** (584)	0.301 *** (1.353)
Trust in government actions	−0.069 (0.593)	−0.057 (0.472)	−0.118 ** (0.407)	−0.070 (0.486)	−0.102 *** (1.126)
R squared	0.368	0.399	0.366	0.411	0.636
Adjusted R squared	0.346	0.378	0.344	0.390	0.626
F test (p value)	0.000	0.000	0.000	0.000	0.000

Note: Significance at * p < 0.1, ** p < 0.05, and *** p < 0.001. Standard error in parenthesis. Standardized coefficients are used in regression results.

displays the outcomes of the multiple linear regression analyses. The dependent variable is perceived flood risk, divided into four dimensions (perceived risk in physical health and diseases, household assets, crops and livestock, and income), as well as the overall perceived risk. The explanatory factors remain consistent across all five models (as presented in Table 1). The regression coefficients for the five regression models are presented in Table 7, which indicates how much the perceived risk for each dimension and the overall perceived flood risk are expected to vary because of a change of one unit in each significant explanatory variable.

The R squared, adjusted R squared, and F-test are also present in Table 7. The explanatory factors in our models may account for between 36.6 and 63.6% of the variation in perceived flood risks, according to the R squares for the five models. All models are highly significant according to the F-test (p-value < 0.01).

Among the socio-economic factors, age is negatively associated with perceived risk in all models except crops and livestock. The perceived risk in physical health and diseases, household assets, income, and overall risk perception increased with the decrease in farmers’ age, which indicates that younger farmers were likely to have a higher perceived flood risk compared to older ones. Female farmers had significantly higher risk concerns for their health and valuable assets, whereas male farmers were found to have significantly higher perceived risks for crops and livestock than female farmers.

However, the overall perceived risk was significantly higher for women compared to men. Farmers with higher amounts of farmland were found to have significantly higher overall perceived flood risk compared to farmers with a low amount of farmland, with significantly higher concerns about health, diseases, and valuable assets. Farmers who had higher annual income possessed more risk in their health and agriculture (crops and livestock) and exhibited lower risk in income compared to poor farmers, while the overall perceived risk was insignificant. Flood experience and flood risk awareness were found to be vital drivers affecting all specific flood risk dimensions and overall risk perception, indicating that farmers with higher flood experience and risk awareness were likely to perceive more risk compared to farmers with less experience and awareness. Farmers’ overall perceived risk decreased with the increase in house distance from the river and trust in government actions.

3.5.2. Factors Influencing Farmers’ Flood Adaptation Assessments

This section shows how important elements have an influence on how farmers evaluate their adaptation behavior. In the multiple regression models, the three dependent variables, namely, perceived self-efficacy, perceived response efficacy, and perceived response cost, were considered with the explanatory variables listed in Table 1. The estimated coefficients (as shown in

Table 8. Factors influencing farmers’ flood adaptation assessments.

Variables	Farmers’ Flood Adaptation Assessments
	Self-Efficacy	Response Efficacy	Response Cost
Age	−0.017 (0.001)	0.019 (0.001)	0.043 (0.001)
Gender	0.013 (0.032)	0.000 (0.027)	0.012 (0.029)
Years of schooling	0.028 (0.005)	−0.033 (0.004)	0.096 ** (0.005)
Family size	0.069 (0.010)	−0.005 (0.008)	0.070 (0.009)
Children under 10 years	0.019 (0.018)	0.017 (0.015)	0.018 (0.017)
Farm size (dm)	0.259 *** (0.000)	0.255 *** (0.000)	0.098* (0.000)
Annual income	0.118 ** (0.001)	0.037 (0.001)	−0.045 (0.001)
Livestock ownership	0.047 (0.041)	0.050 (0.035)	0.046 (0.037)
House distance from river	0.022 (0.022)	−0.079 (0.019)	0.085 * (0.020)
Flood experience	0.112 ** (0.026)	0.099 * (0.022)	0.128 ** (0.024)
Flood risk awareness	0.414 *** (0.038)	0.331 *** (0.032)	0.571 *** (0.035)
Trust in government actions	−0.017 (0.031)	−0.027 (0.027)	−0.005 (0.029)

Note: Significance at * p < 0.1, ** p < 0.05, and *** p < 0.001. Standard error in parenthesis. Standardized coefficients are used in regression results.

) for farm size, flood experience, and flood risk awareness of char-land farmers were statistically significant for all three models. Farmers who owned or cultivated more land were more likely to have a higher ability to adopt flood adaptation, consider the adaptation measures more effective, and perceive less adaptation cost compared to farmers with fewer farmlands.

Similarly, farmers who experienced more flood severity and had a higher level of risk awareness tended to possess higher perceived self-efficacy, response efficacy, and response cost. The year of schooling was significant for perceived response cost, which indicates that farmers who had a higher education level would have perceived less response cost. Farmers with higher annual income were found to have higher self-efficacy, indicating that they have more ability to implement the adaptation measures. House distance from the river was positively correlated with the perceived response cost, explaining that farmers who were far from the river perceived a higher response cost for adopting flood adaptation measures.

4. Discussion

In this study, the evaluation of farmers’ perceived flood risk was conducted by measuring both the perceived risk probability and the consequences of the risk. The research aimed to delve into how farmers perceived the various threats posed by floods across multiple facets of their lives, encompassing areas such as their physical well-being, household assets, crops, and livestock, as well as their income in response to 2020 flood. Given the distinct significance attached to different factors, farmers were prompted to assess flood risks differently within each dimension. The findings from our sample of farmers indicate a prevailing emphasis on perceiving heightened flood risks in relation to crop and livestock production as well as income. This observation underscores the farmers’ overarching priority for the psychological dimension, like health and diseases, which are of relatively marginal importance. Examining the effects of various factors on perceived flood risk, socio-demographic variables, particularly age, exhibited a negative relationship with farmers’ perceived flood risk in the majority of models. This suggests that younger farmers tended to perceive a higher flood risk compared to their older counterparts, which contrasts with the findings of Kellens et al. [56] that associate higher risk perception with older individuals. This discrepancy might be attributed to the greater flood experience of older farmers in the char-land region. Nevertheless, other studies have presented evidence indicating that age does not exert a significant influence on flood risk perception [57,58]. Gender is recognized as a pivotal factor that shapes farmers’ perceived risk [36]. Notably, overall perceived risk and risks associated with specific dimensions, such as health and assets, were found to be significantly higher among women compared to men, which is aligned with the previous studies [26,38,56]. However, men tended to perceive higher risk in relation to crops and livestock, given that agriculture constitutes the primary livelihood strategy in the char-land region and male farmers are extensively involved in farming activities. This study did not establish a relationship between risk perception and years of schooling, which aligns with the findings of Kellens et al. [59] and Armas et al. [57]. Farmers’ annual income demonstrated positive significance in some models, even though the overall perceived risk remained insignificant, a pattern supported by earlier studies [58,60]. Notably, Dang et al. [45] identified agricultural land as a pivotal factor influencing perceived climate change risk. Given our focus on understanding farmers’ risk perception, we included the size of the farming land as a critical factor impacting risk perception. The overall perceived flood risk increased notably with the addition of farming land. This could be attributed to the assumption that farmers with larger land holdings in the char-land possess greater household assets, agricultural production, and income, thereby intensifying their perceived flood risk.

In terms of cognitive factors, farmers’ flood experience and flood risk awareness emerged as significant predictors influencing perceived flood risk across all five models. Similar findings have been observed in earlier studies [59,61,62], where previous flood experience was identified as a significant determinant of flood risk perception. Given the varying degrees of flood severity experienced by individuals in the char-land, their perceptions may differ accordingly. Consequently, farmers exposed to a higher number of severe floods were more likely to possess elevated flood risk perceptions in the char-land. In char-land regions, flood awareness levels among farmers are not uniform, with those who exhibit a heightened awareness of flood vulnerability tending to have a higher perceived flood risk, a finding that aligns with the study of Papagianakki et al. [63]. Farmers’ trust in government actions exhibited a negative correlation with their overall perceived risk level. Farmers who perceive that government adaptive measures are well-coordinated tend to have a reduced overall perceived risk as well as lower perceived risks in crops and livestock production, which is consistent with the findings of Dang et al. [45].

The significant factors influencing farmers’ adaptation assessments include farm size, previous flood experience, and flood risk awareness. Given that agriculture is a primary livelihood strategy in the char-land, farmers are primarily concerned with mitigating flood damage. Those with larger agricultural land holdings tend to have a greater ability to adapt, perceive higher effectiveness of adaptation measures, and consider adaptation strategies to be less costly compared to farmers with smaller land holdings. This could be attributed to the fact that larger land holdings generally entail higher crop production, thereby increasing the susceptibility to flood-related risks.

Flood experience and flood risk awareness also emerged as crucial factors affecting self-efficacy, response efficacy, and response cost. As floods are recurrent events in the char-land, farmers are exposed to them frequently. However, the severity of floods experienced by individuals varies. Farmers who have encountered more severe floods in the past tend to possess higher levels of self-efficacy, response efficacy, and response cost. Similarly, those with a higher level of risk awareness also exhibit increased self-efficacy, response efficacy, and response cost. Since flood risk experience and risk awareness were both significant predictors of flood adaptation assessments, it is assumed to have a positive correlation between risk experience and risk awareness, which is supported by the study of Lindell and Hwang [26], where empirical evidence showed that individuals who had experienced a disaster were more aware of it compared to those who were unfamiliar with it. Consequently, awareness tends to increase alongside flood experience, a relationship affirmed by the research findings of Bradford et al. [29].

5. Conclusions

Understanding how farmers perceive flood risk and evaluate their private adaptive measures, as well as identifying the factors that shape their flood risk perception and adaptation assessments, is essential for comprehending their protective motivation and adaptive behavior in response to flooding in char-lands. However, research in this area, particularly in Southeast Asia, has been limited and largely absent.

Research conducted in the char-lands of Bangladesh has revealed that a significant number of farmers perceive flood risk as primarily concerning agricultural production and income. In the subsequent phase, this study delves into how farmers evaluate their flood adaptation strategies in the char-land. This examination of the char-land context has revealed that most farmers employ adaptive measures predominantly related to their farming practices, including actions like raising livestock shelters, arranging fodder, practicing mixed cropping, and adjusting planting and harvesting schedules. Among non-farming activities, the adoption of portable stoves and macha preparation emerge as common flood adaptation strategies among farmers. Several socio-economic and cognitive factors, such as farm size, flood experience, and risk awareness, have been identified as influential in shaping both farmers’ perceived flood risk and their assessments of flood adaptation strategies.

Given the significant influence of previous flood experience and risk awareness on both farmers’ perceived flood risks and their flood adaptation assessments, it is recommended that authorities prioritize efforts to enhance awareness about the probabilities and severe consequences of floods. This could involve disseminating timely information and organizing various awareness-building programs. Implementing disaster-related education in schools is also highly advisable, as it establishes a direct and impactful channel of information from children to their parents.

The study’s findings also provide valuable insights for shaping adaptation policies. Recognizing the potential impact of risk awareness on farmers’ perceptions and assessments of their adaptive capabilities is crucial for enhancing flood risk awareness building programs. Moreover, it is important to acknowledge that successful adaptation strategies in char-lands hinge on the accessibility and utility of local services, particularly those related to agricultural extension and rural development.

Lastly, it is important to note that farmers’ perceptions of flood threats and their evaluations of flood adaptation strategies are context-dependent and relative. The intention here was not to categorize these perceptions and assessments as universally high or low. Rather, the aim was to discern the relative importance of different facets of farmers’ lives in relation to flood threats and how they gauge their adaptive capacity. Equally important was understanding the extent to which external factors influence farmers’ perceived risks and assessments of adaptation. These insights can be instrumental in the development of targeted adaptation policies.