The rural municipality of Kourthèye is situated between 13˚45' and 14˚27' North latitude and 1˚30' and 1˚52' East longitude, in the southwest section of the Tillabéri Department. Geographically, the urban municipality of Tillabéri borders this municipality to the north; the rural municipality of Karma borders it to the south; Namaro borders it to the southwest; the rural and urban municipalities of Simiri and Ouallam border it to the east; and the rural municipality of Gothèye borders it to the west ( Figure 1 ).

A variety of socio-economic infrastructures, including educational, health, hydraulic, agricultural, pastoral, and road infrastructures; a diverse human capital that includes the population, municipal services, decentralized state services, and development partners; and social capital composed of community organizations primarily composed of farmers, herders, and fishermen are the foundations of the economic potential. Other socio-economic infrastructures include agricultural production (through rain-fed, irrigated, and recession systems), pastoral activities, fishing/aquaculture, commerce, and crafts. The natural aspects of relief, soils, climate, vegetation, fauna, and water resources are all part of the physical environment. Understanding these essential components will help place the municipality in its proper context. The Sahelian climate has a precipitation gradient that is negative from south to north. In fact, from north to south, the average annual rainfall totals vary between 250 and 400 mm. Even at their lowest, temperatures are still rather high, averaging between 18˚C and 45˚C all year round. There are two different kinds of watercourses in the municipality of Kourthèye: permanent and semi-permanent ponds, as well as the Niger River and its seasonal tributaries on the left bank. In the municipality’s far west, the Niger River runs 55 kilometers across it [18] .

To achieve its objectives, three survey devices (SD) were implemented. The first involves “observing the landscape” (I), which means conducting a field mission initially to observe the various agricultural operations. This phase aims to discern how family farms (EA) are organized (geographical situation, potential climate risks that could affect the farms, observing the landscape, and household behaviors regarding the farms), but mainly to prepare the ground for subsequent activities: focus groups and household surveys. This first phase is purely qualitative. The second survey device allows for discussions with farmers in “focus groups” (II) to gain an overview of their operations and the constraints they face. In this sense, it seeks to understand their perceptions of climate change, its impacts on the farms, and on household well-being, particularly the health of farmers. The selection of participants for the focus group was facilitated by the leaders of the various villages in the commune, who proposed participants based on their experience in agriculture. However, this does not prevent the participation of other farmers who may not have significant experience. Finally, the last device involves a questionnaire that enabled the collection of individual data from farmers (III) regarding production factors, the effects of climate change, and their health status in relation to climate.

The following Table 1 provides the main components and indicators considered during the individual survey, along with their descriptions and measurements.

The sample size for the survey is set at 250 agricultural households. The following mathematical formulas were used in three steps to determine the final sample size. The first step (1) involves determining the initial sample size (ISS). Its calculation formula is as follows:

where:

$ISS=\frac{z^2\cdot P\left(1-P\right)}{e^2}$

The second step (2), called adjustment of the sample size (ASS), takes into account the number of farmers. Its calculation formula is:

$ASS=\frac{ISS\cdot N}{N+ISS}$

where:

The third step (3) coincides with the adjustment for the response rate to determine the final sample size (FSS) to ensure a response rate (RR) of 90%. Its formula is written as:

$FSS=\frac{ASS}{RR}$

This is a purely qualitative method. It involves working with farmers and mobilizes multiple disciplines to understand the functioning and organization of agricultural operations. The study focused on aspects related to climate change, the health of farmers, and their agricultural households. This approach allowed for establishing a cause-and-effect relationship between these dimensions through an analytical framework.

According to the literature, several methods are used to create a typology of agricultural operations. The methods employed to realize these typologies depend on the objectives sought and the selected discriminating indicators. However, structural typologies and functional typologies can be distinguished based on the nature of the variables used [19] . The former uses multiple discriminating criteria simultaneously. Principal component analysis (PCA), correspondence factor analysis (CFA), and hierarchical ascending classification (HAC) are distinguished in this context. PCA and CFA are used to characterize operations concerning the selected variables, while HAC is used to group operations based on the weight of the considered variables [6] . PCA is a method from the family of data analysis, more generally from multivariate statistics, which involves transforming correlated variables into new uncorrelated variables. These new variables are called “principal components” or main axes. It allows practitioners to reduce the number of variables and make information less redundant.

These new variables are referred to as “principal components” or main axes. This method allows practitioners to reduce the number of variables and make the information less redundant. It is important to specify that the choice of one of these methods depends on the nature of the researcher’s data. For quantitative data, PCA is preferred, while for qualitative data, CFA is more suitable. In hierarchical ascending classification (HAC), the goal is for individuals grouped within the same class (intra-class homogeneity) to be as similar as possible, while the classes should be as dissimilar as possible (inter-class heterogeneity). Two identical observations will have a distance of zero. The more dissimilar the two observations are, the greater the distance will be. Based on a review of the literature, it appears that PCA is more commonly used in executing typologies of agricultural operations, as seen in studies by [20] - [22] . Given the quantitative nature of the data, PCA combined with HAC has been maintained in this study. For PCA, a total of 17 variables were used to construct the typologies of agricultural operations: age, household size, agricultural assets, wage labor, mil production in kg, sorghum production in kg, rice area in hectares, rice production in kg, area of dry crops, access to climate information, availability of modern agricultural materials, access to agricultural support, membership in a farmers’ organization, access to agricultural credit, availability of non-agricultural income sources, income (both agricultural and non-agricultural), and capital factors (which include equipment and intermediate consumption). Access to agricultural credit is essential for allowing farming operations to invest in technological improvements, quality seeds, equipment, and sustainable farming practices. This has a direct impact on the productivity and profitability of the farm. This will help distinguish agricultural operations. HAC will then iteratively group individuals to produce a dendrogram or classification tree. The classification is ascending because it starts from individual observations; it is hierarchical because it produces increasingly larger classes or groups that include sub-groups within them.

Descriptive statistics, as well as analysis of variance (ANOVA) for structural and functional typologies is used. ANOVA tests the null hypothesis that all group means are equal against the alternative hypothesis that at least one group mean is different. The Chi-Square test for organizational and technical production typologies of agricultural operations, is used to analyze whether significant differences exist between categories.

Linear Models and Quadratic Models are fundamental concepts in statistics and data analysis, used to describe relationships between variables. The two models have been used to analyze the relationships between major diseases and the number of people who fell ill (chronic diseases and malaria).

Table 2 presents the description of the economic and financial indicators considered in this study. This allowed us to analyze the economic performance of agricultural holdings.

The systemic approach taken has established a link between various production factors, production systems, and the health impacts of climate change. Indeed, the agricultural operations of farming households in the municipality are facing alarming climatic conditions that prevent them from ensuring sufficient and sustainable agricultural production. The effects of climate change (drought, flooding, soil degradation, locust invasions, temperature rise, etc.) have had consequences on household health, particularly affecting the workforce such as family labor and, at times, hired labor. In this context, certain agricultural expenditure items, such as intermediate consumption and equipment (very rarely), are used to cover healthcare costs for household members. Additionally, households have other sources of income outside agriculture. These activities include trade, daily labor, fishing, crafts, civil service jobs, migration, and others. These income sources significantly contribute to expenditure items related to agricultural operations (notably, capital factors represented here by equipment and consumption). Furthermore, this operation still retains an extensive character, primarily manual labor employing rudimentary tools (hoes, plows, etc.). The largest portion of agricultural production obtained is intended for self-consumption ( Figure 2 ).

After analysis (PCA), three dimensions were retained. These dimensions explained a total of 65.97% of the variability. The three dimensions can group three categories of variables: one category related to services, a second related to technics of production, and a third concerning the socioeconomic factors of farming households. The results below indicate that most variables in category one are positively correlated with dimension 1. However, some variables from dimensions 2 and 3 are also linked to this factor. For example, access to agricultural credit (0.5) from axis 2 and income (0.3) from axis 3 are included. This gives this axis the power to hold most of the information and explain the interrelationships that exist among them in the agricultural production process. Axis 2, primarily composed of variables from category 2, has an explanatory power of 15.848% (of variability). Regarding the third factor, it includes age (0.7), household size (0.8), agricultural assets (0.8), income, and capital factors (0.7) like shows the figure below ( Figure 3 ).

The variance decomposition indicates that the intra-group variance is 73.28%, while the inter-group variance (between groups) is 26.72%. The high proportion suggests that the groups created by the hierarchical method contain relatively heterogeneous operations. This could reflect significant differences among operations within the same group in terms of structure, resources, or resilience to climate change. Although the inter-group variance is lower than the intra-group variance, it is still significant. This indicates that the groups exhibit distinct characteristics regarding their response or vulnerability to climate change. The high intra-group variability (73.28%) may also reflect challenges in grouping operations into homogeneous sets. This can be explained by multiple factors, such as the diversity of agricultural practices within the same municipality, varying capacities for adapting to climate change among operations in a single group, and the influence of unaccounted variables in the model, such as cultural or religious factors ( Table 3 ).

The results of the hierarchical ascending classification have defined three types or groups of agricultural operations (G_A, G_B, and G_C), as represented in the figure below. There is an irregular progression between the groups, reflecting a diverse nature of operations according to the context ( Figure 4 ). Based on the analysis

from the PCA, agricultural operations can be categorized into three types or groups:

This classification suggests that the identified groups correspond to distinct typologies of operations based on their vulnerabilities to climate shocks and their health impacts.

The analysis of the variance results recorded in Table 3 shows that there is no significant difference in the age of operators across different agricultural operations (P > 0). The average age is 57.65 years in G_A agricultural operations, 56.02 years in G_B, and 52.68 years in G_C. In terms of labor force, the difference is statistically significant (P < 0.05). G_A operations have an average of 4 agricultural workers, compared to 5 in G_B and 8 in G_C. The cropping systems are dominated by cereal cultivation (millet, sorghum, cowpeas) across all categories, with production results differing significantly at the 5% level. These productions are primarily intended for household self-consumption, especially for small and medium-sized farmers ( Table 4 ). The practice of rice cultivation is more significant among farmers in the G_B and G_C groups.

Significant at the level of: *** = 1%, ** = 5%, and * = 10%.

In an environment marked by climate change, rice farms face various challenges, including yield losses, rising costs, and economic vulnerabilities. The data reveals significant disparities among the three types of farms characterized by their sizes, performances, and expenses. High production losses for all three types (with respective relative losses of 48%, 48%, and 43%) indicate potential impacts from climatic hazards (droughts, floods) and possible inefficiencies in adapting agricultural practices. Variable costs represent a substantial portion of total expenses, ranging from 37% to 38% depending on the type. Large farms (G_C) incur high fixed costs due to greater investments; however, these costs are proportionately well-managed compared to small (G_A) and medium-sized farms (G_B). G_C operations are more profitable, with a profitability index of 1.80 (very satisfactory profitability), attributed to better management of fixed costs and higher yields. In contrast, G_A farms remain less profitable with a marginal profitability index of 0.64 due to lower yields and significant losses. Health expenditures, often linked to climate-sensitive diseases such as malaria or malnutrition, diminish households’ financial capacities to invest in agricultural activities. This directly impacts productivity and resilience against climate shocks. Health expenses constitute a notable share of total costs: 8.4% for G_A, 12% for G_B, and 11% for G_C ( Table 5 ).

Agricultural operations develop various types of relationships with certain public or private structures to increase their productivity. However, disparities are observed among them. The unavailability of modern agricultural equipment, or its complete absence in production units, is noted in small (G_A) and medium-sized (G_B) farms. Only large farms (G_C) have access to such equipment, albeit in a relatively low proportion (30%). The majority of these farmers lack access to agricultural support due to a shortage of advisors. Nonetheless, a significant proportion across all categories belong to a peasant organization and have access to climate-related information through platforms such as community radio and WhatsApp. Additionally, farming households have other sources of income outside agriculture. The practice of these activities depends on the type of agricultural operation to which the household belongs. Activities such as crafts, fishing, daily labor, and rural exodus are more commonly practiced by those in G_A. For medium-sized farmers (G_B), daily labor, rural exodus, trade, and other activities are prevalent. In contrast, large agricultural operations (G_C) tend to engage more in trade, livestock farming, and fishing ( Table 6 ).

Significant at the level of: *** = 1%, ** = 5%, and * = 10%.

Drought remains the primary major climatic risk across all categories of rain-fed agricultural operations. No statistically significant difference was observed (P = 0.361) between the groups. Regarding irrigated crops, it is noted that flooding and the invasion of caterpillars or bird attacks remain the main major risks affecting all operations. However, this inter-category difference is significant at the 1% level ( Table 7 ).

Significant at the level of: *** = 1%, ** = 5%, and * = 10%.

Figure 5(A) and Figure 5(B) illustrate the level of production (in irrigated and rain-fed/dry crops) according to the increase in the number of active individuals suffering from malaria. It is observed that rice production among households gradually declines as the number of sick individuals increases. In contrast, for rain-fed crops, the decline in production is minimal as the number of sick individuals rises ( Figure 5 ).

The significance value of the F statistic is less than 0.05 for both models (logarithmic and linear). The results from both models are therefore statistically significant (P < 0.05), indicating that the variation explained by each model is not due to chance. The curve fitting graph provides a quick visual assessment of how well each model fits the data. The estimated parameters show negative results for the constant and b2 in the logarithmic model, while they are positive for the CS ( Table 8 ). Consequently, the analysis indicates that the marginal effect of parameter b2 could reach—64.7 kg of production when there is one sick individual (especially among the heads of operations).

Regarding the impact of chronic diseases on production, the variation in the production level of households is not as significant, although it represents the second most prevalent illness among farming households ( Figure 6 ). All estimated parameters are positive except for b1.