The study area is situated in southeastern Senegal within the administrative boundaries of the Tambacounda region, covering approximately 9546.9 km². It lies along the Gambia River, between 13.517˚ North latitude and −12.995˚ West longitude, and includes the municipalities of Missirah, Nétéboulou, Dialacoto, and Tambacounda, with extensions toward the Vélingara department in the Kolda region ([2]). Its geographical location is illustrated in Figure 1.

Figure 1. Location of the study area (Source: Administrative boundaries from the National Agency of Statistics and Demography (ANSD, Senegal). Map produced by the author using GIS data. Coordinate Reference System: WGS 84/UTM Zone 28N).

The region is characterized by a North Sudanian climate, with annual rainfall ranging from 568.8 to 1218 mm and a mean annual temperature of around 28.4˚C, providing highly favorable bioclimatic conditions for banana cultivation when supplemented by irrigation systems ([6]; [10]; [22]). The physical landscape consists of low-lying alluvial plains at altitudes between 10 and 20 m along the river margins, as well as slightly elevated plateaus ranging from 30 to 60 m, which significantly influence surface runoff patterns and flood dynamics ([6]). The soils are predominantly hydromorphic, clay-rich, and high in organic matter, offering strong water retention capacity but presenting a high risk of saturation during intense rainfall episodes ([12]). The Gambia River constitutes the primary water source for irrigation, which is generally carried out using motor pumps and PVC piping, although the system remains unevenly developed and technologically limited ([6]).

More than 1000 hectares are currently devoted to banana cultivation, representing nearly 80% of national production and organized into Economic Interest Groups coordinated by the Association des Producteurs de la Vallée de la Gambie (APROVAG) ([2]; [6]). The original vegetation cover, formerly dominated by natural savanna ecosystems, has been substantially transformed by the expansion of Musa spp. plantations, while regional biodiversity remains partially preserved but increasingly threatened, particularly given interactions with the adjacent Niokolo-Koba National Park. Banana cultivation was initially introduced in the 1970s by the NGO OFADEC with support from international partners including USAID and Secours Catholique, before being progressively taken over and reorganized by local producer structures ([2]).

The methodological approach adopted in this study combines qualitative and spatial analytical tools in order to assess both the historical dynamics of flooding and the perceptions of risk among agricultural communities. It is structured around three complementary components: the reconstruction of past flood events and the calculation of their recurrence, the collection and processing of empirical data through field surveys and documentary analysis, and the use of statistical and geospatial tools to analyze hydrological variability and map the spatial distribution of vulnerable zones. This integrated approach provides a solid framework for understanding the interactions between environmental processes, agricultural practices, and local perceptions of risk.

2.2.1. History and Calculation of Flood Frequency

1)Collection of local historical data

The reconstruction of the flood history in the Gambia River basin was carried out using a qualitative approach combining the collection of local and institutional data. Semi-structured interviews were conducted with key stakeholders in the area, including the presidents of Economic Interest Groups, private agricultural producers, the Secretary General of CORPROBAT, and officials from the hydraulic brigade in Tambacounda. These interviews made it possible to identify and precisely date major flood events, particularly those affecting agricultural production zones such as banana plantations. At the same time, a review of local archives was undertaken with the technical services and the municipal administration of Tambacounda, including technical reports, meeting minutes, and documentation related to disaster management. For the purpose of calculating flood recurrence, a “significant flood event” was defined according to three criteria: 1) inundation of at least 10 hectares of irrigated agricultural land, 2) reported crop losses exceeding 30% of the affected plots, or 3) disruption of transportation or irrigation infrastructure for more than 48 hours. Only events meeting at least one of these criteria were included in the final frequency calculation to ensure methodological consistency.

2)Collection of local historical data

Based on the documented history covering the period from 2003 to 2024, each year in which a significant flood occurred was counted as a distinct event. The total number of flood years was then divided by the total duration of the study period, corresponding to 22 years, in order to calculate the mean annual flood frequency using Equation (1):

where N denotes the number of years affected by flooding and P represents the duration of the study period in years. This frequency provides a preliminary quantitative estimation of flood risk in the region derived from the synthesis of qualitative and documentary evidence.

3)Data processing and analysis tools

Statistical analyses related to hydrological series and parameter estimation using extreme value theory were carried out with the R software environment, which enabled trend testing, flood modeling, and the production of graphical outputs for interpretation. Spatial analyses, including watershed delineation, slope extraction, and hydrographic network mapping, were conducted within a GIS environment using QGIS for geospatial processing and cartographic representation of monitoring stations and flood-prone areas.

2.2.2. Perceptions of Agricultural Producers

The analysis of producers’ perceptions was based on field surveys conducted within the study area. A participatory and integrated methodological approach was adopted, combining quantitative and qualitative techniques. This approach aims to examine how producers perceive the causes and risks associated with flooding and to assess the impacts of these events on local agricultural systems.

1)Sampling strategy

The survey was conducted in the Gouloumbou valley, an area highly exposed to flooding from the Gambia River and characterized by the vulnerability of irrigated banana plantations and the organizational presence of CORPROBAT. The target population consisted of 3836 producers, including 2881 men and 955 women, representing both members of Economic Interest Groups and private producers. The sampling frame was developed using the CORPROBAT database, which provided essential information such as producer identity, gender, cultivated area, geographic location, and contact details.

To ensure balanced representation across villages and organizational categories, a stratified sampling approach was applied. The population was distributed across 23 strata, including 12 Economic Interest Groups and 11 private farms. The initial sample aimed to cover 75 percent of the total entities, equivalent to 17 Economic Interest Groups and 18 private producers, in order to maximize representativeness and minimize sampling error. However, due to field constraints related to accessibility, producer availability, and logistical limitations, the number of surveyed private farms was adjusted to 11. Within each Economic Interest Group, five producers were randomly selected to ensure diversity of perspectives while maintaining operational feasibility. This selection process incorporated spatial variability by considering production plots located near the river, intermediate zones, and more distant areas, thereby capturing heterogeneous levels of exposure to flooding.

To preserve proportionality between groups and reduce sampling bias, a stratified sampling method with proportional allocation was used. The sample size for each stratum was calculated according to the following Equation (2):

where n_i represents the sample size for stratum i, N_i is the population size of stratum i, and N corresponds to the total target population. This approach, widely used in statistical surveys, ensures that the sample composition reflects the structure of the broader population while improving the reliability of the estimates ([5]).

The sampling strategy implemented in this study aimed to ensure balanced representation of producers across the different villages and organizational structures within the Gouloumbou valley. Table 1 shows the distribution of interviewed producers across villages and Economic Interest Groups, illustrating the spatial organization and diversity of production settings. Table 2 complements this by presenting the sample of private producers and their level of participation in the data collection.

Table 1. Sampling for economic interest groups.

Table 2. Sampling for PRIVATE individuals.

2)Data collection and analysis tools

Several digital tools were employed for data collection, processing, and analysis. The mWater platform was used to design questionnaires and interview guides. Survey data were analyzed using the R software, while summary tables, graphs, and pivot tables were generated with Microsoft Excel. Spatial data were processed using QGIS, which was employed for mapping and GIS analyses.

The chronological analysis of floods in Tambacounda highlights a progressive intensification of flood events since 2003, a year marked by significant agricultural and material losses. The floods of 2005, 2007, and 2009 illustrate the lack of effective drainage systems and the resulting extensive damage to agricultural production, compounded by health risks associated with stagnant water. Although some years of relative calm occurred, flood recurrence resumed in 2013, with further intensification observed in 2015, 2017, and 2020, reflecting increased climatic variability and the growing isolation of rural areas. The floods of 2022 and 2023 further reveal the negative consequences of unplanned urbanization, which exacerbated impacts on infrastructure and rural populations. Table 3 presents the chronological record of flood events reported in the Tambacounda region over the study period. The rows highlighted in red correspond to the years in which flooding events were effectively observed by producers and confirmed through local data sources. This visual distinction facilitates a quick understanding of the frequency, recurrence, and irregularity of floods over time, enabling a clearer interpretation of critical periods and their potential impacts on banana production systems.

This context underscores the urgent need for integrated risk management, combining the mapping of vulnerable zones, the adaptation of agricultural practices, and the strengthening of flood control infrastructure.

Table 3. Chronology of floods in the tambacounda region.

The following table presents the chronology of floods in the Tambacounda region, highlighting for each event the main characteristics and associated impacts. The highlighted columns provide detailed descriptions of each recorded flood. Regarding flood frequency, Tambacounda experienced 11 years with significant flooding between 2003 and 2024, corresponding to a mean annual frequency of approximately one year in every two. The intervals between successive floods ranged from 1 to 4 years, with an average interval of 2.1 years and a standard deviation of 0.83, indicating a relatively regular recurrence of flooding over this period.

Table 3 provides a detailed chronology of flood events in the Tambacounda region, highlighting the timing, main characteristics, and associated impacts of each episode. This overview allows for a comprehensive understanding of the temporal patterns of flooding, the severity of their effects on agricultural production and rural infrastructure, and the evolving vulnerability of local communities over the period under study.

The study focused on a group of producers that was predominantly male (75% men versus 25% women), a distribution reflecting national trends in irrigated agriculture ([1]) and providing an important context for analyzing their risk perceptions and adaptive practices in response to flooding. Although exposure levels are similar for both men and women, some women reported constraints related to access to land, financial resources, and mobility, which may limit their ability to implement certain adaptation strategies. These dimensions deserve further investigation in future research.

3.2.1. Perceived Causes of Flood Risk

The analysis of perceived causes of flooding in Tambacounda, as illustrated in Figure 2, highlights a predominance of natural factors in local representations, with 30% of respondents citing river overflow and flooding, and 25% referring to heavy rainfall, together accounting for 55% of the responses. This perception aligns with observations in the Kou basin in Burkina Faso, where the majority of rural respondents identified intense precipitation as the primary cause of floods ([15]), confirming that climate-related hazards, exacerbated by climate change, are widely regarded as the main triggers of flooding in agricultural areas. This interpretation is often based on direct experience of past events, a factor recognized as central to heightened risk perception ([24]).

Figure 2. Perceived causes of flooding in Tambacounda.

At the same time, anthropogenic causes including insufficient infrastructure (15%), unplanned land use (12%), and deforestation (10%) account for 37% of responses, reflecting a growing acknowledgment of the role of human activities in flood vulnerability. These results corroborate findings from Léogâne, Haiti, where demographic pressure, weak planning policies, and environmental degradation were identified as aggravating factors ([16]).

The mention of “divine will” as a perceived cause (8%) reflects a form of cultural resignation, often interpreted within the framework of protection motivation theory: when a threat is perceived as high but the capacity to respond is considered low, individuals tend to adopt passive or fatalistic attitudes ([14]). This reference to “divine will” must be understood within the broader cultural and religious context of eastern Senegal, where Islamic beliefs and traditional worldviews coexist and strongly shape perceptions of environmental phenomena. In many rural communities, extreme events are interpreted as manifestations of divine testing or fate, which can reduce the perceived ability to anticipate or prevent hazards. Such fatalistic interpretations have been documented in several Sahelian contexts and help explain why some producers adopt passive attitudes toward flood preparedness ([17]; [21]).

It is important to note that risk perception is uneven among producers. According to survey data, around 38% of respondents report strong awareness of flood hazards, whereas 35% declare limited or no knowledge. This disparity is consistent with structural education gaps in the region. The Situation Économique et Sociale de la Région de Tambacounda (2020-2021) reports that 77.9% of unemployed individuals have no formal education, and literacy remains low among women aged 15 - 49, with 75.1% unable to read a complete sentence ([1]). Such indicators suggest that both limited education and access to information may affect how populations including agricultural producers perceive, understand, and interpret flood risks.

These heterogeneous perceptions reveal a risk interpretation that is both grounded in local experience and shaped by cultural beliefs and structural factors, highlighting the importance of an integrated approach that combines awareness-raising, territorial planning, and community empowerment to enhance resilience to flooding.

3.2.2. Perception of Flood Risk

The analysis of flood risk perception in the banana plantations of Tambacounda, as shown in Figure 3, reveals a shared but uneven awareness among producers. Approximately 38% of respondents consider themselves fully aware of the danger, 27% report moderate awareness, while 35% either underestimate or are unaware of the risk. This distribution reflects a common situation in areas exposed to climatic hazards. Individuals who have previously experienced flooding often develop a heightened perception of risk, as demonstrated in studies conducted in Burkina Faso and Haiti, where farmers affected by losses are more sensitive to the threat ([16]).

However, the fact that one-third of respondents remain poorly or not at all aware of the risk indicates that prevention messages do not always reach the target population. This lack of awareness may be attributed to the isolation of certain areas, limited access to information, or low levels of education.

Figure 3. Flood risk perception in the banana plantations of Tambacounda.

Finally, those who are “moderately aware” often hesitate between acknowledging the risk and taking action. This may result from a lack of infrastructure, technical support, or safe alternatives. In this context, it is essential to strengthen awareness-raising efforts in the banana plantations of Tambacounda. Messages should be simple, concrete, and adapted to local conditions, while also providing visible local solutions. Only under these conditions will communities fully recognize the risk and be able to adapt more effectively to flooding.

3.2.3. Distribution of Months Favorable to Flooding According to Producers

As illustrated in Figure 4, banana producers in Tambacounda primarily identify September and October as the months most prone to flooding, corresponding to regional rainfall peaks ([22]). During this period, soil saturation promotes overflows, as observed during the 2009 flood that inundated more than 113 hectares in Gouloumbou ([8]). Rainfall variability, with intense episodes concentrated over short periods, increases crop vulnerability, particularly when water infrastructure is insufficient ([3]; [25]).

However, some producers are unable to precisely identify the high-risk period, reflecting the uncertainty associated with the spatial and temporal variability of rainfall in this semi-arid climate ([22]). This unpredictability reinforces a sense of vulnerability, as shown by [14] and [20]. Exceptional floods in 2009 and 2013 also heightened farmers’ caution and anticipation of flood risks.

To better manage this uncertainty, it is essential to develop anticipatory tools such as risk zone mapping, flood modeling, and farmer training. Effective risk management relies on the dissemination of accurate and contextually relevant information, supported by awareness-raising initiatives ([18]). Experiences in Léogâne demonstrate that participatory mapping and local awareness campaigns enhance prevention efforts ([16]). Therefore, integrating local perceptions with scientific data is a key lever for guiding risk reduction policies in the banana plantations of Tambacounda.

Figure 4. Months most prone to flooding according to producers.

In the face of this uncertainty, anticipatory measures such as risk zone mapping, flood modeling, and farmer training are indispensable for improving flood risk management. As emphasized by [18], effective risk management depends on the provision of accurate, comprehensible, and locally contextualized information, accompanied by training programs adapted to the profiles of exposed populations. In this regard, the work of [16] in Léogâne shows that participatory mapping and local awareness initiatives foster better appropriation of early warning tools and preventive behaviors. Thus, a better understanding of perceived critical periods, combined with scientific data, constitutes an essential lever for guiding local risk reduction policies in the banana plantations of Tambacounda.

3.2.4. Impacts of Floods and Low Water Levels on Banana Production

Banana plantations located in the Gambia River valley, particularly in Gouloumbou, are strongly affected by fluctuations in river levels, amplified by the return of heavy rains since the 2000s. Floods regularly inundate plantations near the riverbanks, rendering the land temporarily unproductive. In 2003, an exceptional flood with a flow rate of 1844 m³/s affected 737.5 ha over 57 km, particularly impacting plots situated below 13 m in elevation ([6]). Such floods, like the one in 2015 in Gouloumbou, caused major production losses.

The isolation of plantation areas and river level fluctuations also complicate irrigation. During the dry season, the distance to water sources makes irrigation difficult and costly. Producers are forced to relocate pumps or invest in more efficient systems, which limits the expansion of cultivated areas. Floods also affect producers’ mobility and the transport of harvests. Inundated roads, impassable access routes, and longer journeys to the fields result in additional logistical costs and deterioration in banana quality, impacting their marketability. These seasonal constraints require constant adaptation. The implementation of sustainable solutions, such as developing plots outside flood-prone areas, using high-pressure pumps, or constructing windbreaks, is essential to enhance the resilience of banana plantations to the hydrological hazards of the river.

3.2.5. Flood Management in Banana-Producing Areas

Floods, whether caused by river overflow or heavy rainfall, pose complex challenges for banana producers. They affect the sustainability of plantation areas, imposing constraints on plot layout and agricultural practices. To cope with these challenges, various adaptation strategies are implemented, ranging from collective village-level interventions to initiatives specific to Economic Interest Groups (GIEs).

Individual strategies for flood management

Analysis of individual strategies in response to flooding in Tambacounda, as illustrated in Figure 5, reveals a wide variety of behaviors. The most common strategy is abandoning flooded plots (30%), followed by replanting after stabilization and continuing cultivation despite the risk (25% each), and relocating to non-flooded areas (20%).

Land abandonment is often a default choice, made when farmers lack the technical means or support necessary to continue production. This type of response has also been observed in Léogâne, Haiti, where many producers refrain from cultivating their land in the face of flood risks ([16]).

Figure 5. Individual flood management strategies.

In contrast, those who replant or continue cultivation despite the risk demonstrate an active adaptation strategy, often based on experience and economic necessity. These producers have learned to live with the risk, as described by [24]. Sometimes they have no other choice, particularly due to limited land availability or their dependence on production for survival. This also aligns with [4] coping theory, which explains that risk responses are often dictated by what is feasible rather than what is ideal. Relocation, although less common (20%), indicates that some producers attempt to adapt by moving to other areas. However, this option is rarely easy to implement due to land scarcity, relocation costs, or food insecurity. Similar situations have been observed in the Kou basin in Burkina Faso, where populations sometimes temporarily move their activities when state interventions are ineffective ([15]).

Overall, these behaviors reflect an individual and uncoordinated response to a collective risk, with each person acting according to their means, perception of danger, and available resources. In a context where floods are becoming more frequent and severe due to climate change, these individual strategies are no longer sufficient. There is an urgent need to establish structured institutional support to shift from reactive management to organized prevention, adapted to the realities of local producers.

Collective Flood Management Strategies

As shown in Figure 6, the analysis of collective strategies for flood management in the banana plantations of Tambacounda reveals a strong preference for simple, locally implemented technical solutions. The majority of actors use gabions and embankments (90%), drainage channels (60%), or continuous water pumping (50%). These responses demonstrate a willingness to directly address the physical effects of flooding, attempting to limit damage on-site. This local mobilization is largely explained by the lived experience of risk. The more communities have suffered from floods, the more they organize to respond. This is consistent with the psychometric theory of risk, which posits that experience, proximity to danger, and repeated flooding events strengthen collective response ([23]; [24]).

Figure 6. Collective flood management strategies.

In contrast, more substantial measures are much less common. Only 40% mention the collective removal of plots, and 30% refer to relocation to other areas. This indicates that economic, land, and social constraints limit long-term adaptation options.

Thus, a two-speed dynamic is observed in Tambacounda: on one hand, effective but limited local actions; on the other, structural adaptations that are difficult to implement. This highlights the need to strengthen local governance, provide technical and institutional support, and promote coordination among stakeholders.

The results obtained in the banana plantations of Tambacounda reveal a mixed perception of flood risk, marked by a predominance of natural causes, but also an increasing recognition of anthropogenic factors. While a significant portion of the population is fully aware of their vulnerability, a notable proportion continues to underestimate the risk, thus hindering the widespread adoption of preventive behaviors. Individual responses range from abandonment and pragmatic adaptation to resigned exposure, while collective strategies prioritize immediate technical measures such as embankments, drainage channels, or pumping rather than long-term solutions like relocation.

These dynamics illustrate the limitations of a fragmented response, largely dependent on perceived capacity to act, land constraints, and lack of institutional support. Consistent with previous studies (notably Terpstra, Grothmann, Hyppolite, and Guelbeogo), these findings emphasize the urgent need to strengthen awareness, collective organization, and the integration of local knowledge into adaptation policies to address flood risks in a context of climate change.

In practice, several constraints made it difficult to apply the initial methodology. First, the database was outdated: many recorded producers had ceased their activities, either due to death, abandonment of banana cultivation, or a shift to other economic sectors. Second, the number of active producers had significantly decreased, reducing the accessible population and challenging the representativeness of the samples. Additionally, factors such as migration caused by flooding or economic difficulties within zonal unions had not been anticipated, further increasing discrepancies with initial projections. Finally, logistical issues, including inaccessibility of certain areas and absences during surveys, limited geographic coverage and the number of responses obtained.

To overcome these difficulties, several adjustments were made. The number of surveyed producers was reduced, particularly among private producers, where the number of villages was adjusted from 12 to 7, and the number of private respondents decreased from 18 to 11. Moreover, for hard-to-reach areas, online questionnaires were used to complement the field surveys in these villages.