Integrating Fish Farming into Runoff Water Harvesting Ponds (RWHP) for Sustainable Agriculture and Food Security: Farmers’ Perceptions and Opportunities in Burkina Faso

Abstract

Integrated aquaculture–agriculture systems are recognized as sustainable solutions to optimize resources, support livelihoods, and enhance food security in climate-sensitive Sahelian regions. In contexts like Burkina Faso, runoff water harvesting ponds (RWHPs) improve agricultural yields during the rainy season but remain underutilized for the rest of the year. This study assesses the feasibility of integrating fish farming into these ponds. Using the Waso-2 tool, structured perception interviews were conducted with 51 farmers across 17 localities. Welch ANOVA and Games–Howell tests revealed, on a scale of 20, that water insufficiency scored 16.01 among experienced farmers without additional water access as a key obstacle, while pond degradation scored 17.69 for those with water access. For motivations, income generation scored 16.24 among inexperienced farmers, whereas training opportunities scored 17.65 for experienced ones, highlighting varying priorities across strata. Farmers preferred fish farming effluents over NPK for vegetables, scoring 15.99. Some favored raw effluents for immediate use, scoring 13.91, while others preferred decanted water with dried sludge for gradual nutrient release, scoring 12.39. This study demonstrates strong farmer interest in integrated RWHP systems. Enhancing pond retention, supplementing groundwater, and providing tailored training in aquaculture practices, pond maintenance, and water management are recommended to encourage adoption.

1. Introduction

In the current global context marked by rapid population growth and an increasing demand for food resources, the pressure on natural resources [1,2,3], particularly water, is a major concern [4,5,6]. The rising demand for freshwater in agriculture, which currently accounts for approximately 70% of global freshwater consumption [7,8], underscores the urgent need for sustainable and efficient water management practices [9]. Additionally, recurrent droughts during the rainy season and ongoing soil degradation, have further exacerbated declining agricultural yields, especially in Sahelian regions.

To address these challenges, strategies, such as runoff water harvesting through RWHPs, have emerged as an innovative solution to store water and support agricultural activities during dry periods [10,11,12,13]. At the same time, integrated systems combining aquaculture and agriculture, particularly the integration of aquaculture with irrigation (IIA), have gained recognition as effective approaches to optimize water use and diversify family farms incomes [14,15]. Fish farming effluents, rich in nitrogen and phosphorus, have demonstrated the potential to reduce chemical fertilizer use by up to 40% [16,17,18], contributing to more sustainable agriculture.

This integrated agricultural approach has been tested in diverse agroecological con-texts, in Asia [19,20,21,22] and Africa [23,24,25], where they have shown their potential to reduce the environmental impacts of fish farming effluents, provide animal protein for family farms, and supply biofertilizers for soils and crops [26,27,28]. In West Africa, efforts to integrate aquaculture and irrigated agriculture date back to the mid-20th century, and significant case studies were presented and discuss at the FAO-ADRAO workshop in 2003 [29].

In Burkina Faso, the first integration experiments focused on rice-fish farming. They started in 1980s, notably in the Kou Valley and the irrigated perimeter of Bagré [30]. Research on rice-fish farming, has highlighted multiple benefits including financial savings on household rice purchases and supple-mental income from fish sales. Despite these promising outcomes, challenges such as inadequate extension services, limited credit access, lack of information, and insufficient fingerling supply have hindered widespread adoption of such systems [30]. Between 2015 and 2017, a collaboration between FAO and the National General Direction of Fisheries Resources allowed implementation of three pilot sites to promote integrated systems. Six types of integration including rice-fish farming and association with vegetables crops production were then validated and disseminated. Promising results in terms of yields and incomes led some private operators to adopt and replicate these systems [31].

Despite the government’s clear willingness to support fish farming and such systems, as well as the growing interest and investment of private operators [32], local fish production remains low(about 17% of the total demand in 2018) [33]. The use of RWHPs as platforms in integrated systems could contribute to yields improvement on both sides, as shown by case studies in Asia [34] and Africa [1,35,36]. Indeed, these studies demonstrated the feasibility of such system to diversify agricultural activities [37,38] while providing local protein sources [39] to rural communities in Sahelian regions [12,40,41]. Increase in agricultural yields and sustainable improvements in rural livelihoods [35,36] were also reported.

However, annual rainfall ranges between 600- and 900-mm in Burkina Faso, concentrated over four months, and agriculture remains predominantly rainfed [42,43]. As a result, farmers face significant challenges during the long dry season. RWHPs were therefore introduced by an initiative of the Ministry of Agriculture to secure cereal crops through supplemental irrigation, resulting in yield increases of nearly 40% [44]. A recent study by [12] revealed that 93.48% of the total production carried out using these ponds, involves vegetable crops, reflecting a diversification of agricultural practices. Conversely, the use of these ponds met challenges related to sealing and the labor required for their implementation. Scientific research has thus mainly focused on developing effective and affordable linings [45], implementation techniques, as well as their socio-economic [44,46] and climatic impacts [13].

During the eight dry months, RWHPs remain largely underutilized in Burkina Faso. Although some RWHP owners integrate fish farming into their use, this practice is not well adopted. There is also a significant gap in empirical studies documenting the subject and farmers’ perceptions of it. Similarly, farmers’ opinions on the reuse of fish farming effluents in integrated aquaculture-agriculture systems remain underexplored and documented through scientific research.

This study is a contribution to filling these gaps by evaluating the potential for integrating fish farming into RWHPs in Burkina Faso, in analyzing farmers’ perceptions of this approach. A survey using the innovative Waso-2 [47]; a tool allowing the prioritization of illiterate respondents ‘opinions, was conducted among RWHP owners. The study identifies the main obstacles, opportunities and motivating factors associated with this practice. It also assesses farmers’ knowledge of reusing fish farming effluents as agricultural fertilizers.

The findings recommend targeted interventions for policymakers and other de-velopment stakeholders to promote sustainable resource management through RWHPs in family farms within the Sahelian context.

2. Materials and Methods

2.1. Study Area

The study area, shown in Figure 1, is located in Burkina Faso, a Sub-Saharan Africa country, within the Sudano-Sahelian climatic zone, characterized by an average annual rainfall of 600 to 900 mm. It specifically focused on six (06) municipalities in Bazega Province (South-Center region) and two (02) municipalities in Kadiogo Province (Center region). Farmers in theses area face significant challenges with water access and management for agriculture, often struggling with water scarcity. To address these challenges, RWHPs (runoff water harvesting ponds) have been widely adopted, with previous studies demonstrating their positive impact on agricultural yields [12].

Across the eight (08) municipalities, seventeen (17) localities were selected for the study as shown in Figure 1. In Kadiogo province, the study included the municipalities of Koubri and Komsilga, covering areas such as Gomtoaga, Wemtenga, Koubri, and Rawelgué. In Bazega province, the municipalities of Kombissiri, Saponé, Doulougou, Gaongo, Toécé, and Ipelcé were included, encompassing localities such as Goghin, Kamsando, Kond-Koakin, Kierma, Ouidi-Ouafé, Damkièta, Saponé, Belegré, Kombougo, Timboin, Koulpele, Sandeba, and Ipelcé. These localities were chosen due to the high prevalence of RWHPs, reflecting strong community interest in this technology.

2.2. Data Collection Process

The data collection process was carefully designed and executed to ensure clarity, reliability, and inclusivity. By combining structured interviews, local language facilitation, and a visual scoring method, the process ensured that the collected data accurately reflected participants’ perceptions and preferences.

2.2.1. Sampling

To ensure that the data collected accurately reflect respondents’ real needs and can be generalized to the entire population, the sample was selected, ensuring respondent independence and randomization in questionnaire administration. In the Waso method, since each respondent evaluates every anticipated response (AR), it is possible to compute an average score for each AR based on individual ratings. The minimum sample size $n_i$ was determined by Equation (1) [45,47,48].

$$n_i\ge (1+\sqrt{g+1})\times {\displaystyle \frac{{(z_{1-\alpha /2}+z_{1-\beta })}^2}{{\delta }^2}}+{\displaystyle \frac{({z^2}_{1-\alpha /2}\times \sqrt{g+1})}{2(\sqrt{g+1})}}$$

(1)

with $\alpha =$ the probability of making a false positive; $g=$ the probability of making a false negative; $g=$ the maximum number of possible responses that is equal to 6 for Waso-2; $\delta =$ the actual difference between factors levels; $n_i=$ the sample size. The total population is estimated by ${\displaystyle {\sum }_1^g{n}_i}$.

The usual values for Waso-2 are: $\alpha =0.04;\beta =0.20;g=6;\delta =1$ [49]. This gives Equation (2):

$$Z_{0.975}=1.960;Z_{1-\beta }=0.8416\leftrightarrow n_i\ge 27$$

(2)

The minimum sample size was therefore 27. Then, the list of RWHP owners of the two provinces investigated provided by the Bazega Provincial Department of Agriculture, Animal Resources and Fisheries allowed the final stratum selection.

An examination of the total population revealed distinct strata based on two key characteristics influencing perceptions of fish farming adoption in RWHPs: prior experience in fishing or fish farming and access to an additional water source during the dry season. These stratification criteria are essential, as prior experience informs familiarity with aquaculture practices, while water access determines the feasibility of year-round fish farming. For instance, respondents with limited water access are more likely to prioritize infrastructure improvements, whereas those with additional water sources may focus on technical training. By stratifying the population based on these factors, the study ensures representation of diverse perspectives and challenges. This approach enables a comprehensive understanding of the variables influencing adoption.

Thus, the two primary strata are defined as RWHP owners with prior experience in fishing or fish farming (E) and those without such experience (IE). Each stratum was further divided based on water source availability, resulting in four groups: RWHP owners with fish farming experience but no additional water source (E_NW); those with fish farming experience and an additional water source (E_W); those without fish farming experience or an additional water source (IE_NW); and those without fish farming experience but with an additional water source (IE_W).

A proportional stratified random sample was then adopted to ensure each stratum’s accurate representation in the final sample relative to its proportion in the initial population [45,47,49]. This approach required first defining the total population size (N), identified here as the official RWHP owners in Burkina Faso. Due to national security considerations, authorities have classified areas into three security levels unsafe (red), moderately safe (orange), and safe (green) [50,51]. For safety issues, only RWHP owners in the moderately safe and safe zones, totaling N_AA = 288 [45], were considered for this study’s baseline population.

Given a minimum sample size previously calculated at 27, a larger sample size was set at n_s (E) = 36 for the largest stratum (owners without prior experience in fishing or fish farming, N(IE) = 195), and a smaller sample size was set at ns (IE) = 15 for the smallest stratum (owners with experience in fishing or fish farming, N(E) = 93).

In order to ensure that each stratum is adequately represented in the sample, maintaining the balance between strata sizes in the population and in the sample, the proportionality coefficients fs(E) and fs (IE) were then computed according to Equation (3):

$$f_s={\displaystyle \frac{n_s}{N_s}}$$

(3)

where $f_s$ is the proportionality coefficient for the related stratum; $n_s$ the size of the related stratum in the sample; and $N_s$ the total stratum size in the population.

Once the coefficient $f_s$ is known, the substratum sample size is computed according to Equation (4):

$$n_{ss}=f_S\times N_{ss}$$

(4)

where $n_{ss}$ is the size of the related substratum in the sample and $N_{ss}$ the total substratum size in the related stratum.

The final sample size is then computed according to Equation (5):

$$n_i={\displaystyle \sum n_{ss}}$$

(5)

where $n_i$ is the final sample size; $n_{ss}$ the size of substrata in the final sample.

The complete process is detailed in

Table 1. Sample determination process.

Population	Total Number of RWHP Owners in Accessible Areas in Burkina Faso NAA= 288
Stratum	Number of RWHP with fishing or fish farming experience Ns(E) = 93		Number of RWHP owners without fishing or fish farming experience Ns(IE) = 195
Proportionality coefficient	fs(E) = 15/Ns(E) fs(E) = 16.13%		fs(IE) = 36/Ns(E) fs(IE) = 18.46%
Substratum size	Number of experienced RWHP owners with access to a supplemental water source Nss (EN_W) = 50	Number of experienced RWHP owners without access to a supplemental water source Nss (E_W) = 43	Number of unexperienced RWHP owners without access to a supplemental water source Nss (IE_NW) = 108	Number of unexperienced RWHP owners with access to a supplemental water source Nss (IE_W) = 87
Substratum sample size	Number of experienced RWHP owners with access to a supplemental water source nss (EN_W) = 8	Number of experienced RWHP owners without access to a supplemental water source nss (E_W) = 7	Number of unexperienced RWHP owners with access to a supplemental water source nss (IE_NW) = 20	Number of unexperienced RWHP owners without access to a supplemental water source nss (IE_W) = 16
Total Sample size	$\mathbf{ni}\mathbf{=}{\displaystyle \sum n_{ss}}$ = 51

.

The final sample size was $n_i$ = 51 functional RWHP owners, including 7 in E_NW, 8 in E_W, 16 in IE_NW, and 20 in IE_W.

2.2.2. Survey Tool

The Waso tool, created by [47] is an innovative device that combines two traditional tools: Awalé, a well-known African board game [52] with regional variations, and the Soroban, a Japanese abacus [53] known for enhancing cognitive and arithmetic skills. Awalé, a variation of the Mancala game [54], is widely played, not only in African communities but also in parts of Asia and the Caribbean.

Crafted from wood, Waso serves multiple functions: it can be used as a game, a calculator, and a survey instrument [55]. When used for a survey, the Waso method is applied in individual interviews to minimize the influence of group dynamics on respondent’s answers [45]. Researchers have shown that in focus group settings, individual opinions are often overshadowed by collective bias, potentially leading to decisions that do not accurately reflect true needs [56,57]. In contrast, the Waso method prioritizes individual perspectives, from which group consensus is later derived, ensuring a more inclusive and representative decision-making process [47]. Furthermore, due to its origin and familiarity, its easy adoption by rural, often illiterate populations is a major advantage. In fact, illiteracy can introduce biases in transcribing the respondents’ opinions, a challenge faced by researchers conducting rural surveys [58,59]. Statistics indicate that approximately 73% of the farming population in Burkina Faso is not literate [60]; hence, the selection of the Waso-2 tool is particularly appropriate for this study. Its tactile and visual design ensures that respondents can engage intuitively, minimizing the risk of bias or misunderstanding. This unique approach enhances the reliability and depth of the data collected. Applied across various previous case studies, this tool has proven to be very effective in identifying converging opinions among independent farmers and highlighting key solution pathways [45,47,61].

The current version of this tool, Waso-2, features six columns and four rows of rectangular slots, with wooden sticks as counters (Figure 2), all handcrafted by local artisans.

2.2.3. Questionnaire Design and Administration

The survey questions were designed with up to six responses to score. To enhance precision, anticipated responses were carefully designed to reflect the diversity of possible opinions. For this study, a closed, coded, and scored questionnaire was employed [47,62,63]. This approach facilitates analysis, enhances response comparability, minimizes biases, and improves respondent comprehension, particularly among low-literacy populations. Designing such a questionnaire, however, requires an in-depth understanding of the subject to ensure that questions and anticipated responses accurately reflect the realities of the target population.

The authors ‘extensive experience in rural contexts was instrumental in achieving this.

The questionnaire design followed three key steps [47]. First, the research question was structured as a process, identifying its inputs and outputs. Next, factors potentially influencing the output were identified, with each factor linked to a Survey Theme Question (STQ) and grouped into broader Survey Themes (STs). Finally, anticipated responses (ARs) or factor levels were defined [64] for each STQ. A fishbone diagram (Figure 3) was developed to visualize these factors.

2.2.4. Questionnaire Administration

The questionnaire was administered by 12 teams, each consisting of four interviewers, with at least one team member fluent in the local language, Mooré (Figure 4). This arrangement ensures effective communication and minimized the risk of misinterpretation. The interviews began with Part 1, which covered general information, starting with obtaining respondents’ informed consent (Appendix A) to participate in the study and authorize data usage, followed by questions on resources and activities.

Part 2 focused on seven (07) Waso questions (Appendix A). After a detailed explanation of the tool’s usage, respondents interacted with the Waso-2 board to allocate scores. The board features up to six (06) anticipated responses, where each assigns a column of four slots for scoring (Figure 2). Respondents are then given twenty sticks, which they distribute to assign scores ranging from 0 to 20 for each anticipated response, with a maximum of five sticks per slot. The total number of sticks placed in the four slots determines the final score for each response, thereby converting qualitative opinions into quantitative data for further analysis. Each interview lasted approximately 1 h and 30 min per respondent.

To ensure reliability, interviewers clarified any misunderstandings during the process, and responses were double-checked at the end of each session.

2.2.5. Data Processing

The objective was to determine whether there were significant differences between factor levels corresponding to anticipated responses for each question, based on scores assigned by respondents. To achieve this, rigorous statistical tests were conducted.

First, the bootstrapping method [65,66] was applied to each stratum’s data. This technique involves generating multiple resamples of the same size as the original sample and then using the means of these resamples for further analysis. Bootstrapping was chosen for its ability to provide reliable estimates of population parameters, particularly variances, means, and quartiles, even with small sample sizes [67]. Additionally, this method is highly adaptable to multi-level stratified sampling systems, such as the one used in this study [65,68].

The second step involved verifying the normality of data distribution for each stratum and the overall sample using the Shapiro–Wilk normality test, which was selected for its high sensitivity in detecting deviations from normality, even in small datasets [69,70]. Once normality was confirmed, Levene’s test for homogeneity of variances was conducted for each stratum and the overall sample. Levene’s test was selected because it remains robust even when the data deviate from normality assumptions, making it suitable for real-world data characterized by variability in distribution [71].

Following this, Welch’s ANOVA [72] was used to test for differences between strata. This test is a variant of the classical ANOVA that is less sensitive to violations of homogeneity of variance, allowing for reliable analysis when variances across groups are unequal, a common feature in this study’s stratified data [73].

Finally, to conduct pairwise comparisons between anticipated responses (ARs) across strata, the Games–Howell post hoc test was applied. This test is well suited to handling heteroscedasticity and unequal group sizes, making it appropriate for data with varying variances and sample sizes. The Games–Howell test allowed for precise ranking of anticipated responses based on respondents’ scores, thereby enhancing the interpretation of significant differences between groups [74,75,76].

All tests used a significance level of 5% (p < 0.05) [77]. This threshold ensures a balance between Type I and Type II error rates, maintaining the statistical rigor of the analysis.

Table 2. Tests and hypotheses.

Test	Null Hypothesis H0	Alternative Hypothesis H1
Shapiro–Wilk normality test	The data follows a normal distribution	The data does not follow a normal distribution
Levene’s test for homogeneity of variances	The variances of the different groups are equal	At least one group variance differs from the others
Welch ANOVA for means comparison	The means of all groups are equal, regardless of variances	At least one group mean differs from the others
Games–Howell pairwise comparison	There is no significant difference between the means of the compared groups	There is at least one significant difference between the means of the compared groups

below presents the hypotheses formulated for each test.

Data from the study area were mapped using ArcMap 10.6.1 software [78]. Collected data were compiled in Microsoft Excel 2016 and bootstrapped using XLSTAT 2019 [79], an advanced statistical analysis software add-in for Microsoft Excel, which also facilitated additional statistical tests and graphical analysis.

3. Results and Discussion

3.1. Socio-Demographic Profile of the Studied Population

The socio-demographic analysis reveals a predominantly male agricultural population, with over 90% of respondents being men aged over 35 (Figure 5A,B). This composition raises concerns about the inclusion of women, who often remain underrepresented in the management of family farms despite their crucial roles [80] in labor-intensive tasks, such as excavation work for water conservation infrastructures (RWHPs). This situation underscores structural inequalities in women’s access to resources and land ownership in various regions [81], including Africa [82,83,84], Asia [85,86], America, and Europe [87], highlighting the need for their better inclusion. Strategic decision-making, especially regarding cash crops, is predominantly male-driven, limiting women’s autonomy in managing resources and participating in the agricultural economy. This gendered division of roles and decision-making reinforces disparities [88], curtailing women’s opportunities to adopt agricultural innovations and access essential resources like credit and new technologies [89].

Activity diversification (see Figure 5C) also stands out, with 35% of respondents involved in livestock farming and 20% in commerce, reflecting resilience strategies to stabilize households’ incomes. However, 47% of respondents report not engaging in any complementary activities, revealing untapped opportunities for diversification. Promoting income-generating activities such as aquaculture could enhance household economic autonomy [90] and provide a viable pathway for women’s involvement. Nonetheless, diversification presents its own challenges as resource and time management become more complex for family farmers juggling multiple responsibilities [91,92].

Finally, water access remains a significant issue, with 55% of respondents reporting shortages during the dry season (Figure 5D). Such shortages hamper productivity and increase the domestic burden on women, who are typically responsible for water collection. Addressing this constraint through improved water management solutions like RWHP [93,94] could mitigate these challenges and enable a more equitable distribution of tasks and resources, thereby advancing gender equality in the agricultural sector [83].

3.2. Survey Theme 1: Main Perceived Obstacles and Motivations

The primary objective of this initial Survey Theme (ST) was to prioritize the factors that discourage fish farming in RWHPs by ranking these barriers from the most to the least significant. Similarly, the survey aimed to identify and rank the key drivers that could encourage individuals to engage in fish farming. To address these themes, two survey questions were developed as follows:

• STQ1.1: What do you consider the greatest obstacle for using RWHPs for fish farming?

Respondents were presented with three predefined responses options to identify the primary obstacles: degradation of RWHPs (RWHPdegrad), insufficient water collection (InsufWatcol), and low additional benefits (Lowadbenef). Each option was designed to capture a specific challenge related to the RWHP system.

• STQ1.2: What would be your primary motivation for using RWHPs for fish farming?

To rank these motivations, respondents were asked to consider three potential motivating factors, each coded as follows: increased income (Inincome), opportunity for training in fish farming (OppTraining), and provision of material and financial resources (ProvMatFin).

• Global dataset distribution for STQ1.1 and STQ1.2

The illustration in Figure 6A highlights a significant difference among the STQ1.1 categories. The data for the variable InsufWatcol clusters around a higher median compared to the other factors. In contrast, the Lowadbenef category exhibits a lower median with greater variability, while RWHPdegrad is positioned between these two, showing a moderate spread with some outliers.

The distribution for STQ1.2 is shown in Figure 6B, where Incincome and OppTraining exhibit higher medians with relatively concentrated scores. In contrast, ProvMatFin displays a lower median with a broader range of values, indicating greater variability within this category. Further statistical tests will offer additional insights into the significance of these observed distributions.

• Shapiro–Wilk and Levene tests

The Shapiro–Wilk normality test for STQ1.1 and STQ1.2, conducted for each stratum and the overall samples, produced p-values ranging from 0.05 to 0.96, all above the significance level (α = 0.05). This result, as shown in

Table 3. Shapiro–Wilk and Levene tests for STQ1.1 and STQ1.2.

Test		Shapiro–Wilk p-Value					Levene Test p-Value
STQ	AR	E_NW	E_W	IE_NW	IE_W	Overall	Inter-Strata	Overall
STQ1.1	InsufWatcol	0.44	0.70	0.23	0.23	0.83	<0.0001	0.09
	Lowadbenef	0.11	0.48	0.86	0.50	0.26	0.10
	RWHPdegrad	0.46	0.10	0.15	0.90	0.96	<0.0001
STQ1.2	Incincome	0.053	0.17	0.11	0.88	0.18	<0.0001	<0.0001
	OppTraining	0.21	0.18	0.33	0.46	0.94	<0.0001
	ProvMatFin	0.38	0.13	0.05	0.75	0.53	0.01

, suggests that the examined variables are normally distributed. Regarding homoscedasticity, inter-stratum tests indicate heterogeneous variances for nearly all variables except Lowadbenef, which shows a Levene test p-value of 0.10, above the 0.05 threshold, suggesting failure to confidently reject the null hypothesis. However, when considering the overall sample, variance is homogeneous for STQ1.1 but remains heterogeneous for STQ1.2. Accordingly, the Welch ANOVA test is applied in the following table to compare means.

• Welch ANOVA test inter-strata results

The Welch ANOVA test yields p-values below 0.0001 for all variables, as shown in

Table 4. Welch ANOVA test for STQ1.1 and STQ1.2.

Parameters	Anticipated Response	F	Pr > F	Strata Means
				E_NW [95%CI]	E_W [95%CI]	IE_NW [95%CI]	IE_W [95%CI]
STQ1.1	InsufWatcol	19.51	**	16.01	15.86	15.56	14.83
				[15.76–16.27]	[15.61–16.2]	[15.30–15.81]	[14.57–15.08]
	Lowadbenef	187.92	***	9.55	5.76	9.54	10.96
				[9.25–9.84]	[5.47–6.06]	[9.24–9.38]	[10.66–11.25]
	RWHPdegrad	433.42	***	14.79	17.69	14.67	11.38
				[14.53–15.05]	[17.43–17.95]	[14.41–14.92]	[11.12–11.64]
STQ1.2	Incincome	101.87	***	16.44	13.78	16.24	14.27
				[16.15–16.72]	[13.50–14.07]	[15.95–16.52]	[13.98–1455]
	OppTraining	337.47	***	17.65	15.62	14.96	14.26
				[17.41–17.90]	[15.37–15.87]	[14.71–15.20]	[14.01–14.50]
	ProvMatFin	44.33	***	12.36	14.23	12.60	11.45
				[12.01–12.70]	[13.89–14.58]	[12.25–12.94]	[11.11–11.79]

Note: p-values are indicated by asterisks: ** p < 0.001, and *** p < 0.0001. The number of degrees of freedom (DF) for all analyses is 3.

, which are lower than the 5% significance level. This result indicates significant differences between strata across all variables of interest, highlighting notable variations in perceptions of obstacles and motivations for using RWHPs for fish farming. Consequently, interpretations are provided for each stratum separately to ensure a nuanced understanding of these dynamics. Table 4 presents the mean values and 95% confidence intervals of the studied variables by stratum for STQ2.1 and STQ2.2, laying the groundwork for further analysis with Games–Howell pairwise comparisons to explore specific differences between ARs inside strata (Figure 7 and

Table 5. Games–Howell pairwise comparison for STQ1.1 and STQ1.2.

Parameters	Anticipated Response	Games–Howell Pairwise Means and Groups
		E_NW		E_W		IE_NW		IE_W
		[95%CI]		[95%CI]		[95%CI]		[95%CI]
STQ1.1	InsufWatcol	16.01		15.86		15.56		14.83
		[15.72–16.31]	C	[15.61–16.2]	B	[15.33–15.79]	C	[14.53–15.13]	B
	Lowadbenef	9.55		5.76		9.54		10.96
		[9.25–9.84]	A	[5.52–6.01]	A	[9.31–9.77]	A	[10.66–11.26]	A
	RWHPdegrad	14.80		17.69		14.67		11.38
		[14.50–15.09]	B	[17.44–17.93]	C	[14.44–14.90]	B	[11.08–11.68]	A
STQ1.2	Incincome	16.44		13.78		16.24		14.27
		[16.14–16.73]	B	[13.41–14.16]	A	[16.02–16.45]	C	[13.99–14.54]	B
	OppTraining	17.65		15.62		14.96		14.26
		[17.36–17.95]	C	[15.25–15.99]	B	[14.74–15.17]	B	[13.99–14.53]	B
	ProvMatFin	12.36		14.23		12.60		11.45
		[12.06–12.65]	A	[13.86–14.61]	A	[12.39–12.81]	A	[11.18–11.72]	A

Note: This table presents the mean scores of the variables for each stratum, with classifications from the Games–Howell pairwise comparisons. For any given question, categories sharing the same letter are not significantly different, whereas those with distinct letters are significantly different.

).

• Games–Howell pairwise comparisons.

The findings for STQ1.1 (Table 5) indicate that perceived obstacles vary across strata depending on the characteristics of each stratum. For strata without access to an additional water source, such as E_NW (InsufWatcol, 16.01, group C) and IE_NW (InsufWatcol, 15.56, group C), water insufficiency is identified as the primary constraint, reflecting a dependency on local water resources. This highlights the critical role of water availability in sustaining small-scale aquaculture systems, a finding corroborated by the studies of [95,96]. In contrast, strata with water access report different challenges: experienced owners (E_W) prioritize infrastructure degradation as the main obstacle (RWHPdegrad, 17.69, group C), likely due to their awareness of structural requirements essential for sustainable aquaculture management [97]. For inexperienced owners with water access (IE_W), water insufficiency remains a substantial concern (InsufWatcol, 14.83, group B), indicating that even with water access, availability remains limited. An assessment of RWHP placement and seepage issues could help determine if water insufficiency stems from these factors, guiding targeted interventions. Findings from studies on impermeable lining solutions [45,98] provide promising insights into mitigating these challenges. In addressing pond degradation, establishing a system that allows pond owners to contract with a private maintenance provider who would visit the site every 2 to 3 months, or upon request from the owners, could greatly enhance the success of maintenance efforts. Additionally, training programs and awareness initiatives on regular upkeep could significantly reduce deterioration, ensuring sustainable use.

The motivations observed in STQ1.2 (Table 5) also vary by experience level. Experienced owners, E_NW and E_W, prioritize training opportunities (OppTraining, 17.65, group C for E_NW; 15.62, group B for E_W), underscoring a strong interest in skill development to optimize aquaculture practices. Previous studies [99,100,101] on farmers’ opinion also highlighted the lack of informations and technical knowledge on fish farming and ponds management as key barriers for fish farming adoption and improvement. In contrast, inexperienced owners, IE_NW and IE_W, primarily focus on income potential (Incincome, 16.24, group C for IE_NW). However, for IE_W, motivations are nearly balanced between Incincome (14.27, group B) and OppTraining (14.26, group B), indicating that respondents in this group may perceive both economic benefits and training as equally motivating factors. These findings align with the literature suggesting that financial incentives and skill acquisition [102,103,104] can jointly drive adoption in aquaculture. Awareness programs on the economic benefits of aquaculture [80] may encourage broader RWHP [105] adoption.

3.3. Survey Theme 2: Conditions and Environment

This section is dedicated to the factors that could most effectively facilitate the use of RWHPs as fish farming supports, as well as determining the most favorable season for this integration. To address these themes, two survey questions were developed as follows:

• STQ2.1: What do you think would better facilitate the practice of fish farming in your RWHP?

To assess the practical requirements, respondents were presented three predefined options. The proposed responses and their codes are as follows: additional groundwater supplied by pumping (SuppGroundWat); perfect lining of the RWHP (Perflining); and practical training on fish farming techniques (PractTrain&Tips). Each option represents a potential solution to improve the feasibility of fish farming in RWHPs.

• STQ2.2: Which season would be most suitable for you to practice fish farming in your RWHP?

This question sought to determine the seasonal preferences for fish farming based on water availability and climate conditions. Respondents were given three options to indicate their ideal timing for fish farming: during the rainy season (Rainseas); during the dry season provided there is an additional water source (DrySeasWatSource); or during any season (AnySeas).

• Global dataset distribution for STQ2.1. and STQ2.2

A distinct difference among the STQ2.1 categories is evident in Figure 8A, with SuppGroundWat displaying a higher median and a relatively narrow spread, highlighting its perceived importance as a facilitating factor. In contrast, Perflining and PractTrain&Tips exhibit lower medians with wider variability, reflecting mixed opinions on these options.

The distribution for STQ2.2, as shown in Figure 8B, reveals that RainSeas holds the highest median with clustered scores, indicating a strong preference for the rainy season. This is followed by DrySeasWatSource, which has a moderately high median and displays greater variability. In contrast, AnySeas shows the lowest median along with the widest range, suggesting a broad diversity of opinions.

Further statistical tests will clarify the significance of these observations.

• Shapiro–Wilk and Levene tests

The Shapiro–Wilk normality test for STQ2.1 and STQ2.2, conducted across each stratum and the overall sample, yielded p-values ranging from 0.05 to 0.98 (

Table 6. Shapiro–Wilk and Levene tests for STQ2.1 and STQ2.2.

Test		Shapiro–Wilk p-Value					Levene Test p-Value
STQ	AR	E_NW	E_W	IE_NW	IE_W	Overall	Inter-Stratum	Overall
STQ2.1	SuppGrounwat	0.22	0.25	0.53	0.43	0.67	<0.0001	0.59
	Perflining	0.89	0.34	0.05	0.90	0.34	<0.0001
	PractTrain&Tips	0.08	0.83	0.90	0.55	0.98	<0.0001
STQ2.2	RainSeas	0.07	0.46	0.40	0.22	0.31	0.02	0.18
	DrySeasWatSource	0.45	0.11	0.27	0.08	0.55	<0.0001
	AnySeas	0.51	0.47	0.91	0.59	0.58	<0.0001

). All values exceed the significance threshold (α = 0.05), indicating that the distributions of these variables generally align with normality assumptions. Therefore, for STQ2.1 and STQ2.2, all variables show acceptable normality.

Regarding homoscedasticity, Levene test p-values for inter-stratum comparisons indicate heterogeneous variances for nearly all variables (p < 0.0001), suggesting significant variance differences across strata. However, for the overall sample, both STQ2.1 (p = 0.59) and STQ2.2 (p = 0.18) demonstrate homogeneity of variances, supporting the use of ANOVA for mean comparisons within the overall sample. The Welch ANOVA test is then used for the inter-strata comparison in the following.

• Welch ANOVA test inter-strata results

The Welch ANOVA test yields p-values below 0.0001 for all variables in STQ2.1 and STQ2.2, which are lower than the 5% significance level. This outcome indicates significant differences between strata across all variables examined, revealing notable variations in respondents’ perceptions. Consequently, separate interpretations are provided for each stratum to ensure a nuanced understanding of these dynamics.

Table 7. Welch ANOVA test for STQ2.1 and STQ2.2.

Parameters	Anticipated Response	F	Pr >F	Strata Means
				E_NW [95%CI]	E_W [95%CI]	IE_NW [95%CI]	IE_W [95%CI]
STQ2.1	SuppGroundWat	218.29	***	19.00	14.51	18.50	16.89
				[18.75–19.24]	[14.26–14.76]	[18.26–18.75]	[16.65–17.14]
	Perflining	74.25	***	14.22	15.78	14.72	13.22
				[13.97–14.48]	[15.53–16.04]	[14.46–14.97]	[12.96–13.47]
	PractTrain&Tips	177.37	***	15.65	13.02	15.51	12.60
				[15.39–15.91]	[12.76–13.28]	[15.25–15.76]	[12.34–12.85]
STQ2.2	Rainseas	131.60	***	12.98	15.34	12.41	14.83
				[12.74–13.22]	[15.10–15.58]	[12.17–12.65]	[14.59–15.07]
	DrySeasWatSource	91.22	***	15.16	15.03	13.16	12.38
				[14.85–15.47]	[14.72–15.33]	[12.85–13.47]	[12.07–12.69]
	AnySeas	143.15	***	15.22	11.23	11.17	12.86
				[14.85–15.58]	[10.87–11.59]	[10.81–11.53]	[12.49–13.22]

Note: p-values are indicated by asterisks: *** p < 0.0001. The number of degrees of freedom (DF) for all analyses is 3.

presents the mean values and 95% confidence intervals of the studied variables by stratum for STQ2.1 and STQ2.2, laying the groundwork for further analysis with Games–Howell pairwise comparisons to explore specific differences between ARs inside strata (Figure 9 and

Table 8. Games–Howell pairwise comparison for STQ2.1 and STQ2.2.

Parameters	Anticipated Response	Games–Howell Pairwise Means and Groups
		E_NW		E_W		IE_NW		IE_W
		[95%CI]		[95%CI]		[95%CI]		[95%CI]
STQ2.1	SuppGroundWat	19.00		14.51		18.50		16.89
		[18.79–19.20]	C	[14.15–14.87]	B	[18.32–18.68]	C	[16.67–17.12]	C
	Perflining	14.22		14.78		14.72		13.22
		[14.01–14.43]	A	[15.42–16.14]	C	[14.54–14.90]	A	[12.99–13.44]	B
	PractTrain&Tips	15.65		13.02		15.51		12.60
		[15.44–15.86]	B	[12.66–13.38]	A	[15.33–15.69]	B	[12.37–12.82]	A
STQ2.2	Rainseas	12.98		15.34		12.41		14.83
		[12.71–13.26]	A	[14.91–15.76]	B	[12.15–12.67]	B	14.60–15.06]	C
	DrySeasWatSource	15.16		15.03		13.16		12.38
		[14.88–15.44]	B	[14.60–15.45	B	[12.90–13.42]	C	[12.15–12.61	A
	AnySeas	15.22		11.23		11.17		12.86
		[10.80–11.66]	A	[13.86–14.61]	A	[10.91–11.43]	A	[12.63–13.09]	B

Note: For interpretation of pairwise classifications, refer to the note below Table 5.

).

• Games–Howell pairwise comparison.

The findings for STQ 2.1 in Table 8 reveal that factors facilitating fish farming in RWHPs vary across strata based on water access and experience. Strata without access to additional water, E_NW (SuppGroundWat, 19.00, group C) and IE_NW (SuppGroundWat, 18.50, group C), prioritize groundwater support, highlighting their reliance on local water sources and the importance of groundwater as a primary water source for aquaculture, especially in regions with limited surface water. This dependency also aligns with studies emphasizing water access as crucial for aquaculture in resource-limited settings [106,107,108].

In contrast, for strata with water access, other priorities emerge. Experienced owners (E_W) emphasize perfect impermeabilization (Perflining, 14.78, group A), reflecting a focus on optimizing infrastructure. The issue of pond water seepage is well known and highly dependent on the nature of the underlying soil. During the investigation, several RWHP owners mentioned in their comments that they were able to retain water in their ponds from the end of the rainy season in October until December, achieving a retention period of up to three months. Several studies have explored effective and accessible local pond linings. In Ethiopia, combining clay and straw achieved retention times up to 45 days, compared to 5 days in untreated ponds [98] in that country. In Burkina Faso, assessed RWHP owners’ perceptions, conclude that concrete offers the most durable solution, although its high cost remains a challenge [45]. Expanding on these recommendations could address the concerns of farmers in this stratum.

Inexperienced owners with water access (IE_W) also prioritize groundwater support (SuppGroundWat, 16.89, group C), indicating that secure water availability remains important even for those with partial access [39,106].

For STQ2.2, preferences for fish farming seasons also differ, and distinct seasonal preferences emerge. For E_NW, preferences are nearly equal between the dry season provided there is a supplemental water source (DrySeasWatSource, 15.16, group B) and any season (AnySeas, 15.22, group A), suggesting adaptability to seasonal conditions. IE_NW leans towards the dry season provided there is a supplemental water source (DrySeasWatSource, 15.16, group B).

Among strata with water access, experienced owners (E_W) slightly prefer the rainy season (Rainseas, 15.34, group A) but also consider the dry season viable (DrySeasWatSource, 15.16, group B), reflecting flexibility. In contrast, IE_W prioritizes the rainy season (Rainseas, 14.83, group B), likely due to reduced water management challenges during this period which is also determining for adoption as shown by [108]. Ref. [109] also suggested fish stocking early in the rainy season and harvesting at the middle and the season end.

STQ2.2 responses demonstrate the willingness and desire of interviewees to engage in fish farming in RWHPs regardless of the season, as long as they have a reliable water source to support their efforts. These findings once again underscore that water scarcity is a primary limiting factor for activities on family farms and for the adoption of integrated systems.

3.4. Survey Theme 3: Perception and Practical Knowledge on the Reuse of FFE in Agriculture

In this third Survey Theme, the aim was to assess the level of knowledge and perception of RWHP owners regarding the reuse of fish farm effluent in agriculture, the comparison with conventional NPK, and the most effective application methods. Three (03) survey questions were therefore formulated.

• STQ3.1: What is the best fertilizer between FFE and NPK?

The three (03) anticipated responses proposed were as follows: fish farm drainage water is better (FFE); NPK is better (NPK); fish farm drainage water is equal to NPK (FFE = NPK); I have no idea (NoIdea).

• STQ3.2: Which application of FFE seems the best to you?

The three (03) proposed responses were as follows: trees (OnTrees); cereal crops (OnCereals); vegetable crops (OnVegetables).

• STQ3.3: Which water quality for fish farming do you consider the best for agriculture?

Three (03) different qualities of water were proposed: raw water (Rawwat); settled water, then dried sludge (DecWatDriedsldge); settled water then fresh sludge (DecWatFreshsldge).

• Global dataset distribution for STQ3.1, STQ3.2, and STQ3.3

A significant difference among STQ3.1 categories in Figure 10A FFE and NPK exhibit higher medians and relatively narrow ranges, suggesting these options are favored as fertilizers. In contrast, FFE = NPK displays a lower median with a wider spread, indicating more variability and less consensus. NoIdea falls between these extremes, with moderate variability, suggesting some indecision among respondents.

The distribution for STQ3.2 is shown in Figure 10B, where OnVegetables exhibits the highest median with a compact distribution, indicating a preference for using FFE on vegetables. OnCereals shows a slightly lower median with moderate spread, while OnTrees has the lowest median and greater variability, reflecting diverse opinions on FFE applications for trees.

A distinct difference among the STQ3.3 categories is evident in Figure 10C, with Rawwat displaying a higher median and narrow spread, suggesting a preference for raw water quality for agriculture. In contrast, DecWatDriedsldge and DecWatFreshsldge have lower medians and wider ranges, indicating more mixed views on these options.

Further statistical tests will provide additional insights into the significance of these observed distributions.

Shapiro–Wilk and Levene Tests

The Shapiro–Wilk normality test for STQ3.1, STQ3.2, and STQ3.3, conducted across each stratum and the overall sample, produced p-values ranging from 0.05 to 0.95. For STQ3.1 (

Table 9. Shapiro–Wilk and Levene tests for STQ3.1, STQ3.2 and STQ3.3.

Test		Shapiro–Wilk p-Value					Levene Test p-Value
STQ	AR	E_NW	E_W	IE_NW	IE_W	Overall		Overall
STQ3.1	FFE	0.77	0.95	0.67	0.30	0.67	0.00	<0.0001
	NPK	0.26	0.75	0.93	0.35	0.27	<0.0001
	FFE = NPK	0.06	0.12	0.05	0.16	0.25	<0.0001
	NoIdea	0.72	0.41	0.26	1.00	0.69	<0.0001
STQ3.2	OnTrees	0.44	0.62	0.45	0.99	0.81	<0.0001	0.78
	OnCereals	0.38	0.37	0.96	0.43	0.49	<0.0001
	OnVegetables	0.15	0.32	0.26	0.68	0.91	<0.0001
STQ3.3	Rawwat	0.30	0.40	0.36	0.96	0.55	<0.0001	0.14
	DecWatDriedsldge	0.30	0.74	0.63	0.50	0.91	<0.0001
	DecWatFreshsldge	0.30	0.74	0.63	0.50	0.30	<0.0001

), all variables meet normality assumptions, except for FFE = NPK in IE_W (p = 0.05) In STQ3.2, OnVegetables shows a p-value of 0.02 in IE_W, suggesting a slight deviation, while other categories align with normality assumptions. For STQ3.3, all variables generally conform to normality, though Rawwat and DecWatDriedsldge in some strata approach the threshold with values close to 0.05.

Regarding homoscedasticity, the Levene test indicates heterogeneous variances across most strata for all variables, with inter-stratum p-values below 0.0001, suggesting significant variance differences between groups. However, for the overall samples, STQ3.2 (p = 0.789) and STQ3.3 (p = 0.141) display homogeneity of variances, allowing for reliable mean comparisons within the overall sample.

• Welch ANOVA test inter-strata results

The Welch ANOVA test yields p-values below 0.0001 for all variables across STQ3.1, STQ3.2, and STQ3.3, as shown in

Table 10. Welch ANOVA test for STQ3.1, STQ3.2, and STQ3.3.

Parameters	Anticipated Response	F	Pr > F	Strata Means
				E_NW [95%CI]	E_W [95%CI]	IE_NW [95%CI]	IE_W [95%CI]
STQ3.1	FFE	154.21	***	13.14	9.45	8.07	10.69
				[12.80–13.48]	[9.11–9.78]	[7.74–8.41]	[10.36–11.03]
	NPK	126.11	***	6.83	12.77	7.92	7.72
				[6.45–7.20]	[14.40–13.14]	[7.55–8.29]	[7.35–8.09]
	FFE = NPK	120.50	***	5.84	5.79	5.17	3.36
				[5.55–6.12]	[5.51–6.08]	[4.89–5.45]	[3.08–3.64]
	NoIdea	120.58	***	7.59	4.76	8.90	6.96
				[7.19–7.99]	[4.36–5.16]	[8.49–9.30]	[6.55–7.36]
STQ3.2	OnTrees	17.81	***	10.16	10.13	11.10	10.04
				[9.81–10.51]	[9.78–10.48]	[10.75–11.45]	[9.69–10.39]
	OnCereals	68.67	***	10.55	14.30	13.03	12.88
				[10.24–10.86]	[13.99–14.60]	[12.72–13.34]	[12.56–13.19]
	OnVegetables	252.02	***	12.36	14.23	12.60	11.45
				[12.01–12.70]	[13.89–14.58]	[12.25–12.94]	[11.11–11.79]
STQ3.3	Rawwat	78.92	***	9.32	13.91	12.67	12.42
				[8.95–9.69]	[13.54–14.28]	[12.30–13.04]	[12.05–12.79]
	DecWatDriedsldge	111.64	***	11.97	9.39	7.52	10.05
				[11.60–12.34]	[9.02–9.76]	[7.15–7.89]	[9.68–10.42]
	DecWatFreshsldge	1.10	***	9.28	9.06	9.34	8.98
				[8.90–9.66]	[8.68–9.44]	[8.96–9.72]	[8.59–9.36]

Note: p-values are indicated by asterisks: *** p < 0.0001. The number of degrees of freedom (DF) for all analyses is 3.

, which are below the 5% significance level. This result indicates significant differences between strata for all assessed variables, highlighting distinct perceptions regarding fertilizer preferences, FFE application methods, and water quality for agriculture. Consequently, interpretations are provided for each stratum individually to capture these variations in perspective. The table presents the mean values and 95% confidence intervals for each variable by stratum, laying the groundwork for more detailed comparisons using Games–Howell pairwise classifications (

Table 11. Games–Howell pairwise comparison for STQ3.1, STQ3.2 and STQ3.3.

Parameters	Anticipated Response	Games–Howell Pairwise Means and Groups
		E_NW		E_W		IE_NW		IE_W
		[95%CI]		[95%CI]		[95%CI]		[95%CI]
STQ3.1	FFE	13.14		9.45		8.07		10.69
		[12.73–13.55]	C	[9.06–9.83]	C	[7.80–8.35]	B	[10.38–11.01]	D
	NPK	6.83		12.77		7.92		7.72
		[6.41–7.24]	B	[12.39–13.16]	D	[7.64–8.19	B	[7.41–8.03]	C
	FFE = NPK	5.84		5.79		5.17		3.36
		[5.42–6.25]	A	[5.41–6.18]	B	[4.90–5.45]	A	[3.05–3.67]	A
	NoIdea	7.59		4.76		8.90		6.96
		[7.18–8.00]	B	[4.37–5.14]	A	[8.62–9.17]	C	[6.64–7.27	B
STQ3.2	OnTrees	10.16		10.13		11.10		10.04
		[9.67–10.65	A	[9.86–10.40]	A	[110.87–11.33]	A	[9.80–10.28]	A
	OnCereals	10.55		14.30		13.03		12.64
		[10.06–11.04]	A	[14.02–14.57]	B	[12.81–13.26]	B	[12.40–12.88]	B
	OnVegetables	13.48		15.99		12.74		12.88
		[12.99–13.97]	B	[15.71–16.26]	C	[12.52–12.97]	B	[12.64–13.12]	B
STQ3.3	Rawwat	9.23		13.91		12.67		12.42
		[8.89–9.75]	A	[13.45–14.37]	B	[12.39–12.95]	C	[12.12–12.72]	C
	DecWatDriedsldge	11.97		9.39		7.52		10.05
		[11.54–12.40]	B	[8.93–9.85]	A	[7.24–7.80]	A	[9.75–10.35]	B
	DecWatFreshsldge	9.28		9.06		9.34		8.98
		[8.85–9.71]	A	[8.60–9.52]	A	[9.06–9.62]	B	[8.67–9.28]	A

Note: For interpretation of pairwise classifications, refer to the note below Table 5.

and Figure 11). The graphs in Figure 11 show, for each question, the mean score curves by stratum, and the final preferences within each stratum can be derived from Table 11 following the Games–Howell pairwise comparisons.

• Games–Howell pairwise comparison.

For STQ3.1, Table 11 data indicate that farmers’ opinions about fertilizers vary significantly across strata. E_NW farmers without additional water access prioritize fish farming effluents (FFE, 13.14, group C), indicating a preference for organic, locally sourced solutions. In contrast, E_W farmers favor NPK (12.77, group D), suggesting a reliance on commercial fertilizers for their consistent nutrient profile. IE_W also prioritizes FFE (10.69, group D), possibly due to limited experience with or access to NPK. Meanwhile, IE_NW shows no strong preference, divided among NoIdea (8.62, group C), FFE (8.07, group B), and NPK (7.92, group B), reflecting uncertainty and indicating a need for additional guidance on fertilizer options. These findings align with prior research [18,110,111,112,113] showing that organic fertilizers are often preferred in resource-limited agricultural contexts due to their potential benefits for soil quality and sustainability through the enhancement of soil structure, water retention, and microbial activity, leading to better crop yields. However, their adoption is influenced by factors like availability, economic considerations, and farmers’ knowledge [114,115,116]. Fish farming wastewater is not subject to the first two limiting factors due to its integration within the system. Its effectiveness as a fertilizer has been demonstrated in studies, including [117], who observed a 96% yield increase in tomato crops compared to NPK application. Other authors confirmed its efficiency in experiments on wheat, potatoes, and onions, resulting in up to a 40% reduction in NPK use and contributing to water conservation [16,118,119]. as the work of [120,121], also concluded that fish farming effluentscould partially or completely replacing NPK. Therefore, raising awareness of initiatives and practical training could motivate and guide farmers in maximizing the benefits of fish farming wastewater as a sustainable fertilizer option.

For STQ3.2, preferences for FFE application also vary. Farmers in the E_NW and E_W strata prioritize using FFE on vegetable crops rather than trees (OnVegetables, 13.48, group B for E_NW; 15.99, group C for E_W), suggesting that these respondents believe vegetable crops, due to their high nutrient demands, benefit more from organic fertilizers. In contrast, IE_NW and IE_W show a slight preference for FFE on cereal crops, with OnCereals marginally favored (13.03, group B for IE_NW) just above OnVegetables. For IE_W farmers, OnVegetables (12.88, group B) slightly surpasses OnCereals (12.64, group B).

This preference for cereals may reflect a greater interest in these crops. However, the small difference in scores between vegetable and cereal crops within these groups does not undermine FFE’s effectiveness for vegetables but simply indicates adaptations in choices based on experience and local farming practices.

These findings align with numerous studies that emphasize the efficacy of organic fertilizers across various crop types, including tomatoes [117,122], potatoes [123], onions [18,122], and cereals like maize [124] and wheat [119]. Additionally, basil and purslane, which have rapid nutrient absorption requirements, have shown similar positive responses to organic fertilization [125].

For question STQ3.3, on the most effective quality of fish farming effluent for agricultural use, responses varied across strata. Farmers in the E_NW and IE_NW strata prioritized irrigation with decanted water followed by dried sludge application (DecWatDriedsldge: 11.97, group B for E_NW; 12.39, group B for IE_NW), highlighting this application method’s capacity for gradual nutrient release, providing a sustained supply to crops. In contrast, farmers in the E_W and IE_W strata preferred raw water (Rawwat: 12.42, group C for IE_W; 13.91, group B for E_W), as it offers immediate nutrient availability without additional treatment, reflecting a preference for both accessible nutrients and simplicity in use.

These findings confirm that the surveyed population has a basic understanding of fish effluent reuse in agriculture, with simplicity of use being a critical factor to consider.

Previous research on the agricultural application of fish farming effluents has included experiments using raw effluent (water and sludge, including food particles and fish waste), dewatered sludge, dried sludge, and compost made from a mix of sludge and agricultural by-products. Studies by [37] demonstrated an agronomic efficiency of dried sludge of between 55% and 80%, largely due to its high phosphorus content. Other research [22] recommended using composted sludge on vegetables, following positive results with spinach. Regarding the most common applications (dewatered sludge and raw sludge), dewatered sludge often contains higher concentrations of essential nutrients, making it a valuable organic fertilizer, which can lead to improved crop yields [126]. For instance, the application of treated sludge has shown a crop yield increase of 14.6% to 49.1% at low dosages [127]. The work of [128] showed it is significantly lighter and less voluminous, reducing transportation expenses by up to 83% compared to raw sludge.

In contrast, the advantage of raw sludge is the possibility of its application directly without the need for processing, making it more accessible for farmers with limited financial resources [129]. In addition, it contains high levels of protein and essential fatty acids, which can be beneficial for soil health and crop production [130]. Raw sludge may provide a more balanced nutrient profile for certain crops, particularly in terms of immediate nitrogen availability, but can also introduce pathogens and heavy metals, which can be detrimental to crop health [127]. Therefore, while raw sludge is readily available and cost-effective initially, its potential for contamination and lower nutrient concentration may limit its long-term benefits. Thus, farmers must weigh immediate costs against future agricultural productivity and sustainability. This suggests that the choice between dewatered and raw sludge should be tailored to specific crop needs and soil conditions.

4. Conclusions

This study provides essential insights into the perceptions and preferences of runoff water harvesting pond (RWHP) owners in the Center and South-Center regions of Burkina Faso regarding the integration of fish farming with agriculture. The findings underscore significant interest in adopting integrated systems, although challenges and priorities differ significantly across strata, necessitating targeted interventions.

For RWHP owners without additional water sources (E_NW, IE_NW), water insufficiency emerged as the primary constraint, underscoring the necessity of improving pond retention through advanced waterproofing techniques and supplemental groundwater systems. Experienced owners with additional water sources (E_W) identified basin degradation as their key concern, highlighting the critical need for training in regular maintenance practices. Conversely, inexperienced owners with water access (IE_W) would benefit from foundational management training and awareness campaigns emphasizing the economic advantages of integration to drive adoption.

Participants expressed varied preferences for fish effluents as fertilizers. While some groups favored chemical fertilizers, others preferred fish sludge, emphasizing the need for adaptable application methods tailored to local conditions and specific crop requirements. These results reinforce the potential of fish effluents as a sustainable alternative to chemical fertilizers, promoting nutrient recycling and soil enrichment in Sahelianthe Sahelian context.

This study contributes to advancing sustainable aquaculture–agriculture systems by providing stratified, evidence-based recommendations. It demonstrates the utility of fish farming effluents as a valuable organic fertilizer, offering significant opportunities for improving soil fertility, reducing chemical fertilizer dependence, and enhancing environmental sustainability.

Future research should assess the performance of integrated aquaculture–agriculture systems with RWHP across diverse agroecological settings, focusing on both environmental and socio-economic outcomes. Investigating solar-powered groundwater solutions for dry season water access and integrating socio-economic analyses to understand community adoption dynamics are promising avenues for scaling this approach. Expanding these systems to other regions in the Sahel and beyond requires context-specific adaptations and continued stakeholder engagement to ensure feasibility and sustainability.