Determinants of households' flood risk coping strategies in a high exposure system of the Manafwa catchment and Lake Kyoga Basin

Globally, floods represent the most frequent natural hazard (UNISDR, 2015) and over the period 1995–2015, they have affected more than 2 billion people (UNISDR 2015). The intensity of the detrimental effects of floods is projected to further increase over the 21st century under the combined effect of climate and demographic evolutions, unprecedented urbanization as well as economic development (Ward et al., 2020). Floods have put mounting pressure on society and economies through life and economic losses (Gao et al., 2019) as well as the destruction of public infrastructure (Khalil, 2018).

Africa, especially sub-Saharan Africa, has experienced a large number of floods, and although there is considerable uncertainty, predictive modelling studies suggest that flooding will in the future become more frequent because of anthropogenic climate change (IPCC, 2014). Risks due to flood events disrupt the livelihoods of the majority of people in developing countries and since the majority of the population in these countries depends on climate-sensitive sectors like agriculture for their living, the occurrence of such events adversely affects their livelihoods (Padhan & Madheswaran, 2022). In terms of sensitivity and adaptive capacity, the impacts of natural hazards like floods are disproportionately large, especially in Africa, most especially in rural Africa and over the last two decades, the region has lost over US$300 billion due to flood-associated damage (Mensah & Ahadzie, 2020). Increased flood frequency has led to losses across different sectors and has affected assets, entitlements and livelihood security, especially for the poor in hazard-affected Africa (Mbereko et al., 2018; Balgah et al., 2019). The direct effects include forceful displacement, destruction of houses and fixed assets, increased disease prevalence and loss of human lives (Balgah et al., 2019).

Within the East African region, the frequency, duration, severity and areas impacted by droughts and floods have increased significantly over the last two decades (Gebremeskel et al., 2019). Extreme flood events in Uganda leading to disasters have increased over the last 30 years (World Bank, 2020). Increased intensity of heavy rainfall has led to a greater impact of floods and is causing more damage due to expanded infrastructure, human settlement and general development of the country (World Bank, 2020). In 2007, devastating floods occurred in northern and northeastern Uganda (World Bank, 2020). Hundreds of thousands of people were displaced. Nationally, half a million people were affected, 21 were reported dead and 170 schools were flooded (World Bank, 2020). Since 2011, about 1,000 flood events have been reported with 480 deaths, 50,000 hectares of cropland were destroyed which indirectly affected over 4 million people (World Bank, 2020). Flood devastation intensity is driven largely by proximity to the flood path. For example, households located near rivers are always prone to floods (Mondal et al., 2021). Furthermore, the level of loss and damage is often influenced by poor disaster management systems and rampant state and market failures to contain flood aftermaths (Balgah et al., 2019).

It has been argued that reducing loss and damage from adverse flood impacts is made possible by adopting different adaptation measures at the household and community levels (Mavhura et al., 2017). A combination of structural and non-structural adaptation measures is an effective way to combat flood risks (Ran & Budic, 2016). Adaptation strategies can be classified as private and public (on the basis of ownership), simple and complex (on the basis of investment), ex-ante and ex-post or precautionary and reactive (on the basis of timing and purpose), hard and soft, and autonomous and planned adaptations (Baylie & Fogarassy, 2022). Adaptations can also be classified on the basis of timing, reasons for implementation and spontaneity (Baylie & Fogarassy, 2022). On the basis of timing, short-term coping strategies and long-term adaptation strategies can be differentiated (Mabuku et al., 2019). Autonomous adaptations are taken by individuals or households, while planned adaptations, on the other hand, are taken by governments, but they may sometimes hinder autonomous adaptation when individuals wait for government provisions (Boakye et al., 2018). Ex-ante strategies include risk reduction and risk mitigation strategies taken by the households in pre-hazard periods, while ex-post strategies refer to risk coping strategies to recover from disasters in post-disaster periods (Siegel & Alwang, 1999). Coping is made up of immediate and short-term measures in response to an event, culminating in the ‘here and now’ capacity of a system and/or community to mitigate and respond to the event (Birkmann, 2011). In a post-disaster period, households adopt different coping strategies, including loan arrangements; sale of assets, livestock, or labour; temporary migration; clearing savings; living on charity; receiving emergency support from external actors and starvation (Sultana et al., 2019). Coping strategies do not lessen vulnerability; however, understanding the rationale behind coping behaviours might help with the effective targeting of those who are at their greatest risk (Adams et al., 1998). Successful coping may enable households to recover from the impact of a disaster. On the other hand, when coping strategies turn ineffective, households face difficulties in recovering from a disaster. However, the severity of the impact may vary across households and most often, poor people, who have limited coping capacities, bear the greatest risks (Patnaik & Narayanan, 2015). Due to the involvement of several stakeholders in technical, institutional, social, economic and psychological issues, decision-making about flood dangers is complex (Twerefou et al., 2019).

Public adaptation measures in developing countries are rare because investment is costly and as such, private household adaptation measures must be encouraged (Baylie & Fogarassy, 2022). This, however, necessitates a better understanding of the socioeconomic, demographic and psychological factors that influence households' adaptation choices (Mashi et al., 2020). The decision of households to take flood-protection measures prevents 80% of economic damage (Baylie & Fogarassy, 2022). Despite this, not all households are willing to take precautionary measures against extreme flood damage. Exploring the reasons for ‘prophylactic actions’ in response to ‘noxious’ natural events such as flooding is critical for both private and public decision-makers to effectively offset the negative effects of flooding (Baylie & Fogarassy, 2022). Identifying the various mechanisms of flood coping in the study area has two advantages. First, it assists in distinguishing between simple mechanisms that can be handled privately by households and mechanisms that require a large sum of investment, and in assigning responsibilities accordingly ((Baylie & Fogarassy, 2022). Second, it serves as a reference tool for engineering method-specific policy and strategy to address flooding challenges at the lowest possible cost, because indigenous and conventional adaptation methods are both cost-effective and environmentally friendly (Pathak, 2021). Furthermore, understanding the specific nature of adaptation mechanisms as simple and complex, private and public, ex-ante and ex-post is required to take appropriate action based on their nature (Pathak, 2021).

Within the households and communities, differential impacts are experienced because the manner in which flood risk management is undertaken is largely influenced by the factors affecting the choice of mitigation measures and constraints in mitigating and adapting to flood risks (Birkholz et al., 2014). Several studies have found that socioeconomic characteristics such as gender, education, age, family size and monthly income are significantly correlated with various risk perception variables such as controllability, knowledge of mitigation options and perceived likelihood of disaster (Mavhura et al., 2017; Balgah et al., 2019). These characteristics influence people to take appropriate measures for reducing disaster losses (Ahmad & Afzal, 2020). There is, however, a lack of studies on households' risk perceptions in many African nations, including Uganda. Most of the flood-related research has focused on the economic effects of floods on local livelihoods or agricultural productivity and little work has been done on the post-flood effects and mitigation strategies adopted at the local level (Abid et al., 2016). There is a need for place- and context-specific assessments of households' responses to flooding, since responses may differ with respect to local characteristics of flooding (Mondal et al., 2021). Keeping in view the current research gap, this study aims to examine post-disaster coping strategies adopted by riverine households and identify the determinants to adopt a particular coping measure to respond and recover from the impact of a flood disaster. There are two research questions to be answered: (i) What post-disaster coping strategies did a household employee use to respond immediately after the floods? (ii) Which factors influenced households' choices to adopt these coping strategies? To answer these questions, we surveyed 210 households, supplemented with focused group discussions (FGDs) and key informant interviews and used a multivariate probit model (MVP) to analyse the determinants of individual coping strategies among the studied households in Butaleja, Eastern Uganda, representing the high exposure areas within the Manafwa catchment. Descriptive statistics represented in the form of pie charts were also used to show the percentage of locals using each coping strategy. It is imperative to examine how people in these communities cope with flooding, so that the findings can be factored into national policies and local strategies geared towards more reliable flood management. An understanding of why a household chooses coping strategies and whether these strategies help them recover from the disaster can guide policymakers in promoting effective flood risk management by identifying target variables (Mondal et al., 2021).

The Cabinet is the body that produces government policy and provides advice to the President on disaster management. The standing committee of Cabinet is the Ministerial Policy Committee (MPC), which ‘handles cross-sectoral matters relating to disaster preparedness and management [… and ensures] that disaster preparedness and management is mainstreamed in the governance of Uganda’ (Republic of Uganda, 2010). According to Uganda's Disaster Management Policy, an Inter-Agency Technical Committee was formed, which is ‘comprised of focal point technical officers from line ministries, UN agencies, NGOs and relevant stakeholders chaired by the Permanent Secretary of the Office of the Prime Minister’ (Republic of Uganda, 2010).

Uganda also has a National Emergency Coordination and Operations Centre (NECOC), which deals with emergencies that have a sudden onset (e.g. landslides, floods and collapsed buildings). The NECOC was established by the Office of the Prime Minister's Department for Disaster Preparedness and Management in 2014. The NECOC ‘is responsible for the effective coordination and networking of the various emergency response institutions of government such as the fire brigade, Police Rapid Response Units, UPDF Emergency Support Units, Uganda Red Cross Society, hospital's emergency units and the private emergency firms’ (Republic of Uganda, 2010). NECOC serves to provide and disseminate early warning information, and to establish mechanisms for the effective coordination and networking of emergency response and recovery assets and resources. It also helps in assigning responsibilities and establishing procedures to safeguard the lives and properties of the population in case of emergencies or disasters through organizational, planning and training activities designed to enhance the country's preparedness and response capabilities (GOU, 2015).

At district levels, the Uganda Police set up a District Emergency Coordination and Operations Centre, which operates from the district police station and reports to the NECOC. There is also a District Disaster Policy Committee (DDPC). In every city, there is a City Disaster Policy Committee (CDPC) and a City Disaster Preparedness and Management Technical Committee (CDP&TC). The CDPC is run by the City Mayor, while the CDP&TC is run by the City Town Clerk (Republic of Uganda, 2010). Furthermore, in every municipality of the country, there is a ‘Municipal Disaster Policy Committee’ and in every Town Council in the country, there is a ‘Town Disaster Policy Committee’ (Republic of Uganda, 2010). At the village level, there is a Village Disaster Management Committee to reach the local communities. The Village Disaster Management Committee is chaired by the chairperson of the village's Local Council and the other members of this committee encompass all of the adult members of the village.

Uganda, however, still lacks a national law governing DRR and management and a law that guides the alignment of local structures with international and regional commitments (GOU, 2019). Other challenges facing Uganda's DRR management include poor coping and relief mechanisms, which according to the Office of the Prime Minister are still not enough. Even with a national disaster management institutional framework in place, little is known about the effectiveness or financing of some DRR components such as climate change adaptation activities and local government structures for DRR, for example, the district disaster management committees that are not directly funded by the government (Irish Aid 2015). The Government of Uganda still spends the bulk of its DRR investment on managing and responding to disasters, as opposed to managing and reducing disaster risk. While this is partially attributed to the frequent occurrences of disasters like floods, landslides and droughts that affect particular regions, the lack of direct funding to local government structures that are at the forefront of dealing with disasters could equally be a contributing factor (GOU, 2019).

The issues identified through the household surveys were validated during the FGDs. Both secondary and primary qualitative data related to flood disasters occurrence (period of occurrence), severity, affected places, number of dead/affected persons and current and potential socioeconomic impacts, and possible factors of the damage levels were collected. The respondents were divided into two groups. The residents of the flood-prone areas formed one group that analysed perceptions and responses from the perspective of local citizens. The second group was of institutional representatives (environmental officers, LCIII chairpersons) who analysed perceptions and responses from the perspective of decision-makers and their respective institutions.

The observed household's decision on a coping strategy to hedge against damage caused by an anticipated or experienced risk factor (e.g. floods) is discrete in nature, thus necessitating the use of qualitative choice models. Univariate logit and probit methodologies for each type of coping strategy are not appropriate, because these models assume the independence of error terms of the different coping strategies, whereas a household may adopt a mix of coping strategies and the decision to adopt one strategy could be influenced by adoption decisions for other practices (Greene, 2003). Failing to account for such issues leads to biased and inefficient estimates (Wooldridge, 2012). The use of multinomial response models is equally not appropriate. Multinomial response models have a strict assumption of a household choosing only one of the available alternatives and independence of the irrelevant alternatives (Piya et al., 2013) which implies that removing one or more alternative coping strategies should not affect the probabilities of selecting any of the remaining strategies. Multinomial response models are ruled out on the grounds of possible interdependence between coping strategies and the observations of most households in the data choosing more than one coping strategy. Households' decisions are inherently multivariate in that they often use the information on several strategies while making adoption decisions of one strategy, thus, the decision to adopt one strategy may influence the decision to adopt another. In such a case, using univariate techniques could exclude crucial information about interdependent and simultaneous adoption decisions (Greene, 2003). The MVP model helps us to determine possible complementarities (positive correlation) and substitutability (negative correlation) between various adaptation strategies.

Therefore, we applied an MVP model, which allows for the interrelationships among the coping practices, i.e. the potential correlation among the unobserved disturbances in the adoption equation. This study used an MVP model to analyse the determinants of individual coping strategies among the studied households in the Manafwa Catchment. MVP enabled modelling of the probability of choosing more than one coping strategy simultaneously since more than one coping strategy can be applied concurrently; it is likely that the decision to adopt one strategy may influence the choice of other strategies. Using an MVP model, this study recognized the possible interdependence between choices and the possible association among unobserved random error terms across these strategies, thus generating unbiased and efficient estimates (Greene, 2003).

In this study, the dependent variables were the coping strategies identified, which were binary to justify the use of probit. The coping strategies were classified into short-term and long-term. The short-term strategies adopted by locals in coping with flood risks were eight, namely, preparation of house, preparation of agricultural facilities, adjusting crop calendar, elevating assets, evacuating children and migrating to other areas. The eighth category worth considering was the category of locals who did not adopt any coping strategy, usually referred to as non-adopters (NoA). Similarly, six long-term coping strategies were identified, namely, afforestation, avoiding marginal lands, relocation, proper siting of buildings, adhering to proper building codes and no adaptation. In the first decision stage – coping status – we used the dummy variable (Yes/No), which indicated whether households had taken any coping measures. Adaptation intensity was calculated using the simple unweighted sum of coping measures households had adopted. We assumed that each coping measure was equally important depending on the location where it was implemented. Descriptive statistics represented in the form of bar charts were used to show the percentage of locals using each strategy. We developed two different models to explore short-term and long-term strategies, respectively. For each dependent variable, we assigned value one to the ith household, which adopted a specific measure and zero otherwise.

The choice of explanatory variables was based on the literature review and results from participatory group discussions (Table 1).

The independent variables in this study represented some of the many factors that affect the use of coping options to reduce flood risk at a household level. Although there might be many factors affecting household use of coping options, this study identified the nine variables listed in Table 1 to be most appropriate in explaining the use of different coping strategies by households. In the empirical mode, each explanatory variable was included in all the equations to help test if the impacts of the variables differed from one coping strategy to the other.

The socio-demographic characteristics of respondents used in the MVP model are presented in Table 2. The findings of this study from the returned questionnaires (n = 197), as shown in Table 2, showed a higher number of female respondents (59%) compared with male respondents. Most (72%) of these respondents had attained a primary level of education with the ability to read and write. Meanwhile, 86% were engaged in agriculture as their main source of livelihood and had no other employment. Most of the households are headed by males (55%) and the average family size is seven people, which is above the national average of five people (UBOS, 2017). The average monthly income is 380,000 Uganda shillings and the smallest income is 5,000 shillings per month (Table 2). Over 85% (majority) of the household heads were married and most of them had lived in the area for more than 11 years (75%) as shown in Table 2. The majority of the respondents lived in semi-permanent houses (60%), which are made up of mud and wattle under iron roofs and located mainly close to the river (35%) or in the flooded area (25%), as shown in Table 2.

The study first probed to determine whether households' coping decisions were complementary or substitutes for one another (Table 3). Results showed 30 pair correlations among the coping strategies were positive, while 34 were negative at the 0.05 significant level (two-tailed).

The determinants of the choice of short-term coping strategies among households were determined using the MVP analysis (Table 4). Results showed that type of house, location of house, time of residence, family size and number of males in the household are statistically significant determinants of preparing the house as a coping strategy. Out of these, only family size had a positive sign. Table 4 shows that family size and the number of males are statistically significant determinants for the other coping strategies as well. The positive sign for the family size variable confirms the fact that the probability of adopting short-term coping strategies increases with household size. Similarly, the determinants of the choice of long-term coping strategies among households were determined using the MVP analysis (Table 5). Results showed that type of house, location of house, time of residence, family size and a number of males in the household are statistically significant determinants of avoiding marginal lands and relocation as long-term coping strategies.

Gender has a negative but insignificant relationship with most of the short-term adaptation measures (Table 4) and a positive but insignificant relationship with all long-term coping strategies except proper siting of buildings (Table 5). The gender of the household head has a mixed influence on the choice of adaptation strategy. It has a negative and insignificant relationship with the preparation of the house, adjusting crop calendar, evacuating children and selling animals (Table 4), afforestation, avoiding marginal lands and relocation (Table 5) and, on the other hand, a positive but insignificant relationship with the other adaptation strategies.

In the current study, age has a mixed effect on the coping strategies, i.e. positive but insignificant relationship with all short-term coping strategies except adjusting crop calendar (Table 4). Age also has negative but insignificant relationships with most long-term strategies except for proper siting of buildings (Table 5).

In this study, education also had a mixed effect on the adaptation of flood coping strategies. Higher educational status encourages the adoption of preparation of the house, adjusting crop calendar, preparing agricultural facility relocation, ensuring proper siting of buildings, following proper building codes and discouraging the elevation of assets and selling of animals as flood coping tools in the study area (Tables 4 and 5).

The study revealed that households secured themselves against flood events by preparing the house (erecting temporary barriers around their houses and making gravel (murrum) buffer fences), elevating important assets to prevent them from being submerged, selling animals and agricultural products, evacuating to higher places and migrating to other relatives in non-flood-prone regions as coping strategies for the pending floods. The inspiration behind selling animals and other agricultural products was minimizing the anticipated loss so that they could be replaced after the flood event. In addition, the income generated from the sales could be able to finance other basic household needs. This analogy was also supported by Weldegebriel & Amphune (2017). Such basic pre-flood preparations have been reported to be effective in reducing the impact of floods at the household level in Bangladesh as indicated by Dewan (2015). Thus, the majority of flood victims used more reactive coping strategies than preventive ones, but such reactive measures are costly and sometimes ineffective. More efforts should be concentrated on devising preventive approaches to solving the flooding menace.

Much as floods have happened repeatedly in the area, they have still happened as a surprise to the people, according to the study's findings. This is an indication that the existing early warning system is either inadequate or the information provided is not understood by the community members. Raising awareness of natural hazards and preparing the community by providing holistic information is very important for preparedness at both the household and community levels, a notion supported by Frigerio & De Amicis (2016). For any community-based early warning system to be effective, it needs to integrate locally based and indigenous practices, as stated by Chowdhooree et al. (2018) that integrating the local knowledge and understanding of risk into the decision-making process can lead to better-targeted approaches on resolving the flood question and therefore attract better interventions from the non-state actors. The indigenous practices include watching the positioning of the clouds in the sky, the degree of rainfall in the upstream catchments, the agility of ants, abnormal crying/voices of animals and birds, the intensity of thunderstorms and wind, the position of the stars and the moon, and magnitude of hotness, odd noises from the rivers, a muddy scent in the water and increasing levels of water flow (Dewan, 2015).

It is noted that pre-flood preparations by the households are wholly focused on structural strategies of securing their lives and property against the effects of the pending floods but with little or no attention placed on the after-effects of floods. For instance, there was no mention of how households prepare themselves against diseases that may emerge, fuel (both cooking and lighting) and food shortages. Yet, research has proven that such effects always follow after any floods. Ghatak et al. (2012) assert that the after-effects of floods normally include water-borne diseases, hiking of commodity prices and social insecurity. The study results showed that over 300 hectares of crop area were lost to floods in just one event. It is also true that the affected crops were basic food and cash crops, upon which the livelihood of the majority of households is anchored. For instance, 95.3% of Butaleja depend on crop farming (district specific profile – UBOS, 2017). Such an impact would definitely spark a food shortage and other nutritional-related deficiencies. The failure by households to prepare for the after-effects of the floods could be attributed to weaknesses within the coordination systems that do not provide a comprehensive pre-flood preparation information package to the victims.

Respondents' perceptions of the long-term interventions focused on what they could receive from the government and/or other agencies to bolster their individual and/or community capacities. For example, there was a strongly held belief that government support in the provision of seedlings for afforestation, awareness campaigns on the dangers and causes of floods, the enforcement of building standards and relocation of settlements to safer grounds could help reduce the net loss and damages in the area. The Government of Uganda through the Northern Uganda Social Action Fund (NUSAF) has made a number of interventions in the area, whose impact is yet to be evaluated. Perhaps, more importantly, there is a need to build trust between the community and the government. When people trust the government, they tend to respond positively to the measures it takes (Shao et al., 2017). No policy measure is likely to succeed in the absence of trust. The government should prioritize the development of cost-effective systems for early flood warning, flood prevention strategies and programmes to educate rural communities on how to adapt to flooding, which includes technical agronomic advice on flood resilient crop management. Rural communities that are heavily reliant on farming and lack diversified sources of income would benefit most from targeted resilience-building measures (Ashar et al., 2021).

Effective awareness is created if community members are fully involved in the entire process and without disregarding their indigenous knowledge. By allowing them to exercise their local capacity and only be given technical support by the experts, they become a critical component of the process with a responsibility to manage their own community flood combating programmes and not just be information and technology recipients. This increases the level of adaptability to the strategies that reduce the flood impacts. Effective involvement happens when the participants are fully aware, empowered and skilled (World Meteorological Organization, 2017).

Low education levels in the study area greatly impacted people's perceptions of disaster risk management practices and the level of adaptability of such practices aimed at lessening the flood-related impacts. The study results showed that only about 29% of the respondents had gone beyond primary level education. This result was consistent with other research findings, for instance, Mavhura et al. (2017) stated that low levels of education negatively affect someone's knowledge and awareness of disaster risk management practices that may be probable even in the prone areas. Therefore, awareness programmes in the study area should be cognizant of the low education levels and ensure that messages delivered are simple enough and in the local languages for them to be comprehended by the lowest person. Also, integrating Functional Adult Literacy (FAL) programmes in the awareness campaigns can be of great help in enhancing the abilities of community members to comprehend the information on the causal effects of floods. It is therefore important to employ an integrated approach that will raise awareness on both the early warning system and other flood impact mitigation strategies such as afforestation programmes, better farming practices without cultivating marginal land, embracing settlements relocation programmes, adopting proper siting of buildings and adherence to proper building codes. This will enhance the community's capacity to technically apply the acquired knowledge and information in reducing the impact of floods on their own lives and property.

Gender has a positive but insignificant relationship with most of the adaptation strategies, indicating a positive relationship between gender and the decision to adopt the most flood risk management tools. One possible explanation for these findings is that the study region is mainly male-dominated where men have more liberty to implement certain measures than women due to local customs and traditions. Similarly, various measures such as house construction need more physical work and construction knowledge, which are limited in women household heads. Therefore, male household heads tend to adopt more measures to safeguard their property and household from such catastrophes. Our results are corroborated by those from Murphy et al. (2005) who also found that men are dominant in both indoor and outdoor activities and are responsible for most kinds of risk reduction strategies. Gender analysis by Tu & Nitivattananon (2011) has also confirmed that men and women play different roles, with women being less likely to adapt to floods. The reason for this is that women are more prone to health risks, sexual harassment and increased responsibilities.

The housing type owned is considered to be an important factor in determining the household's adaptive capacity and choice of certain mitigation strategies. The house type is negatively and significantly associated with the preparation of the house, evacuation of children, afforestation and avoiding marginal lands and this implies that households living in semi-permanent houses are less inclined to adopt those measures compared with households living in permanent houses.

The coefficient of inhabitants in or near the vicinity of the river has a positive sign for some of the adaptation measures. This is true in the sense that households living near to the river need more precautionary measures than others living away from the river. Our results are consistent with other studies (Bantilan et al., 2015; Gioli et al., 2014; Mondal, 2014), which found that location was an important factor in determining the choice of mitigation measures in developing countries.

Age is an important social indicator of vulnerability, particularly in rural areas where people's capacities or potentials need to be improved (Buckle et al., 2000). It is an important fact that people's ability and capacity to respond and recover from natural hazards like floods in hazard-prone areas depends on age (Cannon, 2000). In our study, age has a mixed effect on the adaptation measures. The results were consistent with the findings of Tadesse & Dereje (2021) who noted that age influences the choice of adaptation strategies both positively and negatively. The positive age coefficient for preparation of agricultural facilities, evacuation of children and avoiding marginal lands implies that more aged households would prefer to implement preparation of agricultural facilities, evacuation of children and avoiding marginal lands compared to young heads. Similar positive results were found in other studies (Sultana & Rayhan, 2012; Berman et al., 2014; Ullah, 2014; Dasmani et al., 2020). It can be concluded from the negative coefficient that as people grow older, the household head is reluctant and shows no sensitivity to modification to take some actions of adaptation to flood risk. To summarize, growing older is connected with decreased physical capacity, a worse likelihood of obtaining credit, limited mobility to new locations and less demand in the work market (Mondal et al., 2021).

The time spent in the area has a negative and significant relationship with most of the adaptation strategies. This is best explained by the fact that the more time people reside in flood-prone areas, the more they seem to get accustomed to fate and may not get involved in any adaptation strategies. However, the time of residence has a positive and significant relationship with relocation and ensuring proper building codes as long-term adaptation strategies. This means that people who have spent longer times in flood risk areas are more willing to relocate to other areas.

Education is an important determinant of a flood coping strategy to enhance one's resilience and quality of life in response to natural hazards (Tong et al., 2012). Households with high levels of education tend to be more informed and therefore more capable of choosing a better coping strategy in any emergency. The uneducated households, on the other hand, tend to wait on government and other social capital, including seeking assistance from relatives and friends. Education level is also very important in generating awareness of flood forecasting. In our study, education also had a mixed effect on the adaptation of flood coping strategies. Higher educational status encourages the adoption of preparation of the house, adjusting crop calendar, migration, relocation, ensuring proper siting of buildings, following proper building codes and discouraging the adoption of the preparation of agricultural facilities, elevation of assets and selling of animals as flood coping tools in the study area. These findings are supported by those of Ullah et al. (2015b) who found a mixed effect of education on risk coping tools among agricultural producers in KP Province, Pakistan. The protective motivation theory can explain the negative coefficient of educational status for literate household heads (Baylie & Fogarassy, 2022). Individuals (households), according to this idea, tend to respond to any threat (even a flood) if two conditions are met: the hazard is perceived as a higher risk, and their capacity to handle the hazard is greater (Twerefou et al., 2019). As a result, even if the individual is well aware of the potential threat from the hazard as a result of being educated and able to use information, the household may ignore the threat (decide not to adopt any strategy) if one of the conditions is missed, particularly due to barriers such as money and the cost of adaptation ((Baylie & Fogarassy, 2022).

The effects of floods are multi-sectoral and they require a multi-sectoral approach to designing coping mechanisms to be adopted by the victims. Our findings demonstrate that the vulnerabilities, capacities and risk perceptions of households in relation to disasters are diverse. Households living in this flood-prone area have developed some level of disaster resilience by employing coping strategies that enable them to deal with frequent flooding every year. The differing frequencies of performing coping strategies seem to indicate that households employ a set of coping strategies according to their perception of flood risk. However, some of the short-term coping strategies undertaken by households were damaging and irreversible, for instance, the sale of livestock and other household property traded off for survival may never be replaced. The coping strategies should be comprehensive, including a variety of risk-lessening measures that are not only related to flooding but also encompass issues such as food security, sanitation and water supply.

Weaknesses within the flood response coordination system were exhibited by the lack of preparedness of the households due to inadequate pre-flood preparation information packages given to the victims. Obtaining the opinion and perception of the affected people before choosing a coping strategy is quite important in optimizing the already scarce resources, especially if it involves moving people from their ancestral areas. The limited adaptive capacity, resilience and high sensitivity of households and communities in the study area increase their vulnerability to hazard exposure, necessitating urgent external support.

As mentioned earlier, the country has a well-documented national policy for disaster preparedness and management. However, gaps in the implementation of this policy have contributed greatly to the annual incidence of flooding in the communities. This gap emanates from the politics of giving a ‘human face’ when residents in risky areas need to be simply evacuated. The friction between the government and people living in flood-prone areas as well as the ‘weak’ implementation of the policy contributes to the flooding problem in the study area. To curb the problem, this study recommends that the Office of the Prime Minister should intensify education to residents in the catchment about the dangers associated with their activities along watercourses. The sensitization should be followed by a law enforcement system that punishes perpetrators.

Future studies can look into the details of the barriers preventing the implementation of the existing policies on flooding. There should also be policies geared towards establishing early warning systems to alert communities about natural hazards. One way of establishing early warning systems is by directing the Uganda National Meteorological Authority (UNMA) to collaborate with radio stations in the district to broadcast in local dialect the possible natural hazards (floods and droughts) that may occur. When this initiative is taken into consideration, households would be informed early enough, and that will enable them to prepare for possible disaster, thus increasing their coping capacity.

The coping capacity of local households needs to be enhanced by providing more access to financial means and diversified sources of income to safeguard livelihood sources in case of floods. Common mitigation strategies within communities also need to be developed and implemented at the local level to reduce the mitigation cost. This could be done by the government, private sector and community by developing strong linkages and partnerships among different stakeholders.

Furthermore, research needs to be done on low cost and advanced mitigation options for households and communities living close to river areas in order to make them less vulnerable and more resilient. Future studies could also look into quantitative measures of the direct impact and benefit of the coping strategies employed and the length of time needed by the families to recover from the disaster. Quantifying the losses and disruptions incurred by families across varying flood risks would provide valuable data in determining the actual economic impact of disasters.

G.E., A.E. and Y.B. conceptualized the study; G.E., A.E. and R.A. prepared methodology; R.A. used the software; G.E., A.E. and R.A. validated the data; G.E., A.E. and R.A. did formal analysis; G.E. investigated the study; Y.B. managed the resources; G.E. did data curation; G.E., A.E. and Y.B. wrote and prepared – original draft; G.E., A.E., Y.B., R.A., A.G. and I.K. wrote and edited – review; G.E., A.E. and R.A. did visualization; A.E., Y.B., A.G. and I.K. supervised the study; Y.B. did project administration and acquired funds. All authors have read and agreed to the published version of the manuscript.

This research was funded by the Swedish International Development Agency (SIDA), grant number 331.

The research of this article was supported by SIDA under the Building Resilient Ecosystems and Livelihoods to Climate Change and Disaster Risks (BREAD) project (Project, 331). The authors and publisher are fully responsible for the content. The authors are extremely thankful to the respondents and local stakeholders for their support during the entire period of the field survey. Furthermore, we would like to thank our survey team members for conducting interviews during March 2019.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.