Adoption of Agroforestry Practices in and around the Luki Biosphere Reserve in the Democratic Republic of the Congo

Abstract

Despite the technical, socio-economic and environmental challenges, indigenous subsistence agroforestry, generally referred to as slash-and-burn agriculture or bush-fallow farming, is a common practice for local populations in the Democratic Republic of Congo. This study analyzed the proportion of adopters and non-adopters, together with other factors that influence farmers’ choices of adopting agroforestry or that discourage its adoption in the Luki Biosphere Reserve (LBR) area. Data were collected through a survey of 390 households using a structured questionnaire. A logistic regression model, with SPSS Statistics software was fitted to the data against a binary response (1 = adopt; 0 = not adopt). The proportion of adopters of agroforestry practices in the LBR area far exceeds (more than three-fold) that of non-adopters. Six factors exert a positive and significant (p-value = 5%) effect on peasant decisions to adopt agroforestry practices in LBR, including age (51 to 60 years old), marital status, education level, main activity, land tenure and farmers’ membership in a local association. Gender, other age categories, household size, number of years of agroforestry experience, number of assets, distance between residence and fields, and access to credit did not positively influence the adoption of these practices. The results of this study would help engage the indigenous community with different sectors and disseminate agroforestry as a sustainable practice appropriate to the real needs of local populations.

1. Introduction

Agroforestry is a relatively recent scientific discipline, but a very old practice in the humid tropics, where peasant agriculture combines annual crops, livestock, tree farming, and forest management [1,2,3]. Several studies argue that this innovative system, in both tropical and temperate regions, improves agricultural productivity, diversifies and increases farmers’ income, improves soil fertility, contributes to food security and climate change adaptation, maintains biodiversity, and contributes to the supply of wood and non-timber forest products [1,4,5]. In light of these agroecological, social and economic aspects of agroforestry, efforts are being made by both governments and non-governmental organizations in several countries to adopt agroforestry practices in sustainable development programs [6,7].

The adoption of a practice refers to a farmer’s or producer’s decision to apply an innovation over the long term based on an experience or series of events [8]. For [9], innovation (technique or technology) in adoption is not considered in the sense of novelty or invention. It is a process by which a practice, technique or technology is introduced, disseminated and adopted according to the conditions of the host environment. A practice becomes an innovation when it is adopted [8,9]. The adopter is the user of the innovation. The decision to adopt an innovation is dichotomous (0 or 1), whereby the adopter can decide whether or not to use the innovation or technology [10,11].

In the literature, the analysis of innovation adoption factors reveals the important role that is played by technical, socio-demographic and economic characteristics of the producer or adopter, including age, gender, marital status, household size, education level, labor force size, income, wealth, attitude towards risk, endogenous knowledge, membership in a local structure, and access to markets and credit [12,13]. These determinants of adoption can be described as intrinsic factors, i.e., specific to the user. In addition to these so-called endogenous factors, there are exogenous factors in the adoption of innovations such as physical or environmental factors of the farm, including the size of the farm, the distance between the farmer’s residence and his or her fields, and the topography of the land [14].

In the Democratic Republic of Congo (DRC), agroforestry has existed since time immemorial in the traditions of the populations of all provinces, so that public policies do not encounter many obstacles in the adoption of traditional agroforestry. According to the work of [15,16,17], modern agroforestry practices can be found in the Provinces of Kinshasa (in Mampu, Kinzono and Ibi village on the Batéké plateau), Tshuapa (in Monkoto) and Central Kongo (in Luki). To this end, in the Province of Central Kongo, the majority of farmers living in the Luki Biosphere Reserve (LBR) grow cassava (Manihot esculenta Crantz), groundnuts (Arachis hypogea L.), maize (Zea mays L.), beans (Phaseolus vulgaris L.), taro (Colocasia esculenta (L.) Schott), and rice (Oryza sativa L.), among other crops. Other annual crops are associated with these species, including banana (Musa sp.), pineapple (Ananas comosus (L.) Merr.) and vegetables [18,19,20]. Both modern and traditional agroforestry farmers are undertaking many initiatives, such as integrating perennial woody species (trees, shrubs, oil palms), small livestock rearing (poultry, pigs, small ruminants), apiforestry, and aquaforestry into their activities [18]. Bhattacharya et al. [19] highlighted the implementation in Central Kongo of agroforestry model farms since 2006 as part of a REDD+ project in and around the Luki Biosphere Reserve (LBR). These farms, which are a modern type, constitute reference sites for agricultural activities integrating woody species [21]. According to [19,22], agroforestry model farms are one of the key strategies for sustainable land use that would enable farmers to reduce their dependence on natural resources in the LBR, thereby contributing to the conservation of these resources. At the same time, traditional agroforestry farms are evolving according to farmers’ habits. Similarly, [23] highlighted the existence of traditional agroforestry systems with sylvo-banana, sylvo-coffee and sylvo-cacao plantations in the transition zone of the LBR from Belgian colonial times onwards.

There are some reports available on agroforestry in LBR, but no scientific study has been conducted regarding its adoption. Therefore, the adopters and non-adopters of this agrotechnology, together with the factors determining adoption, are not yet known. This study attempts to answer the following question: what factors positively influence the adoption of agroforestry practices in and around the LBR? Our hypothesis is that some individual technical and socio-demographic characteristics of LBR farmers positively determine the acceptability of agroforestry. The objective of this study is to analyze the proportion of adopters and non-adopters of agroforestry practices and the factors leading to their adoption in the LBR area.

2. Materials and Methods

2.1. Study Area

The study was approved by both the Research Ethics Committee of the University of Kinshasa and the Institut National pour l’Étude et la Recherche Agronomiques (INERA)-LUKI. It was conducted from 17 August to 5 September 2021 in the LBR area, which is an exemplar of the humid tropical rainforest ecosystem and one of 727 sites in the World Network of Biosphere Reserves of the UNESCO MAB Program. This protected area constitutes the unique relict forest ecosystem, given its potential in terms of rich and varied faunal, edaphic and hydric conditions and its floristic resources. It is located in the southwest region of the DRC, about 120 km east of the Atlantic coast and 30 km north of the City of Boma, Central Kongo Province. It is situated across three administrative sectors, i.e., Patu, Bundi and Moanda, which are integrated respectively into the territories of Lukula, Seke Banza and Boma Bungu. LBR covers an area of 33,000 hectares, which are delineated as 05°30′ to 05°43′ S by 13°14′ to 13°17′ E [24]. The climate of the LBR is humid tropical with a long rainy season of five months, a short dry season of one month, a short rainy season of two months and a long dry season of four months. The soils are of different textures (sandy, sandy clay, clay and silty). The LBR harbors primary and secondary forests and grassy, shrubby and wooded savannas. It is part of the Mayumbe region, which is occupied by low mountains, with a rugged relief, mostly plateaus with some plains, lowlands and hills ranging in elevation between 150 and 500 m a.s.l. In [24], the population living in the LBR and its surroundings was estimated at 176,818 individuals in 2020. They belong mainly to the Yombe ethnic group and allochthonous (immigrant) ethnic minorities. Figure 1 and Figure 2 show two agroforestry practices in LBR, namely, traditional home garden and aquaforestry.

2.2. Sampling

Apart from the 50 model agroforestry farmers from the REDD+ project that had been established in the LBR by the World Wide Fund (WWF)/Luki, the number of other agricultural and agroforestry households was not known in advance. Therefore, the sample size (denoted as n) was calculated according to the binomial distribution using the following formula, as suggested by [25]:

$$n=Z^2{}_{\mathsf{\alpha }}\frac{p·q}{d^2}$$

(1)

where Z_α , p, q, and d denote, respectively, (i) the value from the normal distribution that is related to the probability p at the 5% significance level (i.e., α = 0.05, Z_α = 1.96); (ii) the proportion or probability of individuals with the desired characteristics in the survey population; (iii) the proportion of individuals without these characteristics; and (iv) the desired precision measurement or margin of error.

In this study, the proportion or probability p of the adoption rate was selected to be 50% (p = 0.50), i.e., the one with the greatest variance, so that its complement q (with q = 1 − p) is also equal to 0.50. The margin of error d was set to the 0.05 level of significance. Therefore, the sample size n was calculated as:

$$n=\frac{{\left(1.96\right)}^2\times 0.50\times 0.50}{{\left(0.05\right)}^2}=384.4$$

This sample was rounded to the next highest decade, i.e., 390. The 390 farmers, who were surveyed using questionnaires, were distributed among 26 villages that were located in and around the LBR. The criteria for selecting a respondent were as follows: being a farmer, together with being a stock-breeder, beekeeper, or fish farmer, with trees (agroforestry) or without trees (non-agroforestry) that were dispersed on cropland.

2.3. Presentation of the Theoretical Model for the Adoption of Agroforestry Practices

2.3.1. Choice and Justification of the Logit Model

To study the adoption of innovative practices, the econometric models that are commonly used include linear regression and logistic regression (Tobit, Logit and Probit). For linear regression models, the dependent variable (response) is a linear function of one or more explanatory variables or predictors, and it follows a normal distribution. However, these models are not reliable because their probabilities can exceed 1. In contrast, logistic regression models are non-linear with a qualitative dependent variable with a dichotomous choice, i.e., with two value classes, which were either yes or no, and alternatively 1 or 0 [26,27]. For logistic models, the dependent variable follows a Bernoulli distribution. This dependent variable, which is the probability p of the event occurring, is a function of the linear combinations of the explanatory variables. Logistic regression models are preferred in this study despite their disadvantages of not being estimated by ordinary least squares, together with the difficulty in directly interpreting the model parameters.

In considering logistic regression models, the Tobit model is also known as the censored normal regression model; it is left-censored at zero, and frequently right-censored at 1. The model allows for the censoring of the adoption intensity data by assuming that both the determinants and effects of the determinants are identical for the probability of adoption and for the intensity of that adoption. This is not the objective of this study. At the same time, the Logit model is based upon the logistic distribution of probability, while the Probit (or Normit) model is based upon the cumulative distribution function of the standard normal distribution. Both models, with a qualitative dependent variable (adoption of an innovation), make use of probability calculations. The advantage of the Probit model over the Logit model is that the former has positive probabilities. Yet, these two models lead to similar results. There is no compelling reason to choose one model over the other. In the cases of many studies of adoption in agriculture and agroforestry, many researchers have preferred to use the binomial Logit (or binary logistic regression) when the dependent variable is binary, i.e., with two values, 0 and 1 [28,29,30]. In this study, the adoption of agroforestry practices is either 0 or 1, so the binary Logit model was chosen as the tool to analyze the factors determining adoption.

2.3.2. Mathematical Formulation of the Logit Model

The Logit model [31] is designed to determine the choice simultaneously between two alternatives, either 1 or 0. Considering the probability p of an individual i adopting practice y (innovation, technology or strategy) is 1 and 0 otherwise, the equation of the Logit model is described by the following model:

$$\mathsf{\pi }=p\left({\mathrm{y}}_i=1\right)=\frac{{\mathrm{e}}^{\mathsf{\alpha }+{\mathsf{\beta }\mathrm{ix}}_{\mathrm{i}}}}{1+{\mathrm{e}}^{\mathsf{\alpha }+{\mathsf{\beta }\mathrm{ix}}_{\mathrm{i}}}}$$

(2)

where:

• π = p(y_i) is the probability of an individual i adopting practice y, with p(y_i) = 1 if the practice is adopted and p(y_i) = 0 if it is not adopted;
• y_i= the explained variable, the adoption of the practice;
• e = the base of the natural logarithm (ln);
• x_i = explanatory variables, biophysical and socio-economic characteristics of the farmer;
• β_i = coefficients or parameters of the explanatory variables, the signs of which allow the interpretation of the results;
• α = constant.

This model (2) is converted to the linear form using ln-transformation [32,33], with SPSS Statistics 2020 (Armonk, NY, USA) [34].

2.3.3. Definition of the Variables Included in the Empirical Model

Most studies regarding adoption focus on endogenous (intrinsic) variables, such as the socio-demographic characteristics of the producer [11,12,13,14,15,16,17,19]. For each of these variables, assumptions are made about their influence on the producer or adoption decision. The dependent variable in this study is the adoption of agroforestry, with 12 independent variables (

Table 1. Definition of the variables specified in the binary logistic model.

Variables	Definition	Type of Measure	Expected Sign
Dependent
ADOPAF	Adoption of agroforestry	1 if yes; 0 if no	+ or −
Independent
GENDER	Gender of surveyed farmer	1 if male; 0 if female	+ or −
AGE	Age of surveyed farmer, in years	1 if 18–30 years; 2 if 31–40 years; 3 if 41–50 years; 4 if 51–60 years; 5 if + 60 years	+ or −
EDLEV	Education level of surveyed farmer	1 if can read and write; 0 if otherwise	+
MARSTAT	Marital status of surveyed farmer	1 if married; 0 if otherwise	+ or −
HOUSEHO	Household size of surveyed farmer	1 if < 3 persons; 2 if 3 to 5 persons; 3 if > 5 persons	+
MAINACT	Main activity of surveyed farmer	1 if other activities; 2 if agroforestry	+
EXPAF	Number of years of experience in agroforestry of surveyed farmer	1 if none; 2 if < 5 years; 3 if 5 to 10 years; 4 if > 10 years	+
MOLATE	Mode of land tenure of surveyed farmer	1 if owners; 2 if renters	+
PEWOF	Number of persons working in the farm	1 if 1 to 3 persons; 2 if 4 to 6 persons; 3 if > 6 persons	+
HTFD	Home-to-farm distance of surveyed farmer	1 if < 1 km; 2 if 1 to 3 km; 3 if > 3 km	-
FLOCASS	Surveyed farmer’s membership in a local association	1 if yes; 0 if no	+
ACCREDI	Access to credit of surveyed farmer	1 if yes; 0 if no	+

Source: Principal author’s estimates taken from the literature [11,12,13,14,15,16,17,19].

below).

According to the 1 dependent variable and 12 independent variables that are given, the empirical full model of this study is written as:

logit(p(ADOPAF)) = β₀ + β₁ × GENDER + β₂ × AGE + β₃ × EDLEV + β₄ × MARSTAT + β₅ × HOUSHO + β₆ × MAINACT + β₇ × EXPAF + β₈ × MOLATE + β₉ × PEWOF + β₁₀ × HTFD + β₁₁ × FLOCASS + β₁₂ × ACCREDI

(3)

where the acronyms are defined in Table 1, where β₀ (intercept or constant) and β_i (regression coefficients, i ranging from 1 to 12) are estimated from the data.

2.3.4. Multicollinearity Test

The presence of multicollinearity was verified in this study. In effect, collinearity exists between two exogenous variables when the linear correlation between these variables is high (example |r| > 0.8) [33]. For [33], collinearity between the exogenous variables reduces the precision and stability of the coefficient estimates (increases their variances, potentially alters their signs), thereby making model prediction more difficult.

2.3.5. Statistical Analysis of the Data

All statistical analyses, including descriptive statistics, model parameter analyses and the regression model, were performed with the SPSS Statistics 2020 release [34]. The ordinary least squares method was used in this study after linearizing the exponential function Logit model.

3. Results

3.1. Adoption of Agroforestry Practices

3.1.1. Proportion of Adopters and Non-Adopters

The proportion of adopters and non-adopters of agroforestry practices in the LBR area are summarize in

Table 2. Adopters and non-adopters (n = 390).

Variables	Modalities or Categories	Adopters (n = 302) %	Non-Adopters (n = 88) %
Gender	Female	9.6	26.1
	Male	90.4	73.9
Age	18–30 years	3.0	5.7
	31–40 years	18.2	18.2
	41–50 years	34.4	39.8
	51–60 years	23.5	20.5
	60 + years	20.9	15.9
Education level	None	11.9	53.4
	Educated	88.1	46.6
Marital status	Not married	21.5	43.2
	Married	78.5	56.8
Household size	<3 persons	21.5	15.9
	3 to 5 persons	68.9	71.6
	>5 persons	9.6	12.5
Main activity	Agroforestry	93.0	84.1
	Others	7.0	15.9
Experience in agroforestry	None	5.0	58.0
	<5 years	3.3	12.5
	5 to10 years	14.6	10.2
	>10 years	77.1	19.3
Mode of land tenure	Owners	93.0	88.6
	Renters	7.0	11.4
Number of persons working in the farm	1 to 3 persons	94.7	95.5
	4 to 6 persons	5.3	4.5
Home-to-farm distance	<1 km	2.3	3.4
	1 to 3 km	85.8	60.2
	>3 km	11.9	36.4
Membership of the farm in local association	Yes	78.8	85.2
	No	21.2	14.8
Access to credit or subsidies	Yes	70.2	95.5
	No	29.8	4.5

by variable.

Of 390 respondents, Table 2 shows that the proportion or relative frequency of the adopters of agroforestry practices in the study area is higher than that of non-adopters. These adopters are mainly men ranging in age from 31 to over 60 years old, who are mostly literate and married. They have a household size of three to five people and perform agroforestry as their main activity. In addition, most adopters have more than 10 years of experience, own land, and have one to three persons working in their fields. Almost all adopters mentioned that the distance between their home and their farm is 1 to 3 km. Most of these respondents belong to a local association and neither have access to credit nor have received a grant. In this regard, the key result is that the proportion of adopters of agroforestry practices in the LBR area far exceeds (more than three-fold) that of non-adopters.

3.1.2. Factors Determining the Adoption of Agroforestry Practices

The estimation of the empirical model, with socio-demographic (independent) variables being extensively tested, is shown in

Table 3. Estimation of the logit model for agroforestry adoption (Yes = 1, No = 0) for each level of the nominal and ordinal predictor variables.

Independent Variables	Β	SE	eβ (Odds-Ratio)	Signif. (p-Value)	Confidence Interval 95% for eβ
					Lower Bound	Upper Bound
Intercept	0.553	1.588		0.728
GENDER	0.166	0.776	1.181	0.830	0.258	5.402
AGE (category 1)	−0.767	1.932	0.465	0.692	0.011	20.505
AGE (category 2)	0.094	1.040	1.099	0.928	0.143	8.437
AGE (category 3)	−0.788	0.764	0.455	0.302	0.102	2.032
AGE (category 4)	−1.872	0.845	0.154	0.027 *	0.029	0.806
AGE (category 5)	0	-	-	-	-	-
EDLEV	1.316	0.604	3.728	0.029 *	1.140	12.188
MARSTAT	1.941	0.730	6.969	0.008 *	1.667	29.129
HOUSEHO (category 1)	−0.963	1.344	0.382	0.474	0.382	0.027
HOUSEHO (category 2)	−0.109	1.150	0.897	0.925	0.897	0.094
HOUSEHO category 3)	0	-	-	-	-	-
MAINACT	2.216	0.901	9.173	0.014 *	1.568	53.668
EXPAF (category 1)	−0.560	1.156	0.571	0.628	0.059	5.506
EXPAF (category 2)	−1.459	2.138	0.232	0.495	0.004	15.333
EXPAF (category 3)	−1.316	1.353	0.268	0.331	0.019	3.803
EXPAF (category 4)	0	-	0	-	-	-
MOLATE	−4.096	1.066	0.017	0.000 *	0.002	0.135
PEWOF (category 1)	−1.515	1.372	0.220	0.270	0.015	3.239
PEWOF (categories 2 and 3)	0	-	-	-	-	-
HTFD (category 1)	1.034	1.513	2.812	0.494	0.145	54.553
HTFD (category 2)	−1.062	0.671	0.346	0.114	0.093	1.289
HTFD (category 3)	0	-	-	-	-	-
FLOCASS	2.984	0.661	19.772	0.000 *	5.409	72.266
ACCREDI	0.146	0.987	1.157	0.883	0.167	8.002
	χ2 = 311.552, total degrees-of-freedom (df) = 19; Cox & Snell R2 = 0.550; Nagelkerke R2 = 0.838

Note: GENDER = gender, AGE = age, EDLEV = education level, MARSTAT = marital status, HOUSEHO = household size, MAINACT = main activity, EXPAF = experience in agroforestry, MOLATE = mode of land tenure, PEWOF = number of persons working in the farm, HTFD = home-to-farm distance, FLOCASS = farmer’s membership in a local association, ACCREDI = access to credit or subsidy; β and SE are regression coefficients and standard errors for β, respectively; p-values in boldface (*) are significant at the 5% level.

.

Table 3 shows that the adoption of agroforestry practices by peasants in the LBR is effectively determined by six significant factors or variables (p-value < 5%), including age (51 to 60 years old category), marital status, education level, main activity, land tenure and farmers’ memberships in a local association. Gender, other age categories, household size, number of years of agroforestry experience, number of assets, distance between residence and fields, and access to credit did not positively influence the adoption of these practices.

The pseudo-R² values, i.e., Cox and Snell R² (0.55) and Nagelkerke R² (0.838), which are close to the theoretical maximum of 1 and the calculated χ² (311.552) > theoretical χ² (30.143) at the 5% significance level (df = 19), show that the model used in this study is significant at the level given. In addition, Table 3 shows that some coefficients do not differ from zero for some levels of the independent variables, namely, AGE (category 5), HOUSEHO (category 3), EXPAF (category 4), PEWOF (categories 2 and 3) and HTFD (category 3).

In considering only the significant independent variables (Table 3), the empirical model can be expressed as follows:

logit(p(ADOPAF)) = −1.872 × AGE + 1.316 × EDLEV + 1.941 × MARSTAT + 2.216 × MAINACT − 409.6 × MOLATE + 2.984 × FLOCASS

(4)

Given that the βi coefficients in Table 3 have values greater than 1 (in absolute value), the odds ratios provide good estimates of the likelihood of adoption of agroforestry, as shown in

Table 4. Odds ratio, probability of adoption of agroforestry and confidence interval based on significant independent variables.

Independent Variables	β	SE	eβ	p = eβ/(1 + eβ)	Confidence Interval at 95% for eβ
					Lower Bound(LB)	1 – LB	Upper Bound(UB)	1 – UB
AGE (category 4)	−1.872	0.845	0.154	0.133	0.029	−0.71	0.806	−0.194
EDLEV	1.316	0.604	3.728	0.788	1.140	0.14	12.188	11.188
MARSTAT	1.941	0.730	6.969	0.875	1.667	0.667	29.129	28.129
MAINACT	2.216	0.901	9.173	0.902	1.568	0.568	53.668	52.668
MOLATE	−4.096	1.066	0.017	0.017	0.002	−0.998	0.135	−0.865
FLOCASS	2.984	0.661	19.772	0.952	5.409	4.409	72.266	71.266

. The regression coefficients for the reference value are set to zero (Table 3).

The analysis of marginal effects by statistically significant independent variables in Table 4 indicates that, all else being equal, the probability of respondents adopting agroforestry is 0.133 when respondents are in the 51–60 age range (AGE “category 4”), 0.788 when respondents are educated, i.e., can read and write (EDLEV), 0.875 when respondents are married (MARSTAT), 0.902 when respondents have agroforestry as their main activity (MAINACT), 0.017 when respondents are landowners (MOLATE), and 0.952 when respondents are members of a local association structure (FLOCASS).

With the exception of older respondents owning the land that they farm, Table 4 indicates that adopting respondents are more likely than non-adopters to incorporate agroforestry practices on the farm, on average, based upon 95% confidence intervals around each probability estimate (i.e., 1 − LB and 1 − UB).

4. Discussion

4.1. Analysis of the Proportion of Adopters and Non-Adopters of Agroforestry

The proportion of adopters of agroforestry practices in the LBR area far exceeds that of non-adopters (Table 2). This can be justified by the fact that respondents experience agroforestry in their activities. Indeed, working at INERA-Luki or WWF-Luki favors the instruction of some populations in agroforestry systems. In addition, the RBL is attractive because it is the only ecosystem in the Province of Central Kongo (DRC) that is full of forest potential with rich and varied floristic, faunal, edaphic and hydric resources. Therefore, the contact of local populations with researchers, extension workers, project agents and others would allow them to gain information and knowledge on innovations, positively influencing them to adopt agroforestry in large numbers. The results of this study are in accord with the work of [28], who argued that learning increases farmers’ perceptions of and preferences for innovation adoption. Yet, studies conducted by [35,36] showed that the percentage of agroforestry adopters is lower than that of non-adopters. Several constraints can hinder adoption [35], including the lack of willingness to plant trees and wait for two years to achieve the benefits of this innovative technology. Yet, agroforestry techniques integrate not only fruit trees, the harvesting of which can last for more than three years with traditional techniques, but also annual crops (vegetables, cereals) and livestock (short duration). Other perennial woody plants such as lianas, oil palms or non-fruit trees are also exploited in agroforestry. The last group can be used as caterpillar host species or for apiforestry (beekeeping in a treed environment) and aquaforestry (fish ponds or rice paddies dotted with trees). For [36], the constraints mitigating the adoption of agroforestry technologies by farmers include, among others, insufficient land for tree planting, the illegal felling of trees, the long developmental period of trees, the lack of technical assistance, the lack of planting materials, the lack of knowledge and skills, as well as competition between trees and arable crops on farmland.

4.2. Analysis of Factors Determining the Adoption of Agroforestry Practices

4.2.1. Influence of Age

This study indicates that only the age category of 51 to 60 years (Table 3 and Table 4) exhibited a positive and significant effect on peasant decisions to adopt agroforestry practices. This can be justified by the fact that as farmers grow older, the less they feel the effects of age limiting their vital functions, the more they can commit to innovations that are not constraints in terms of physical energy expenditure, as is the case for some agroforestry practices (e.g., sylvopasture, apiforestry, entomoforestry). Agrisylvicultural and agrisylvopastoral practices are components of agroforestry systems that are relatively restricted in the study area because, in addition to arboriculture, they include food crops that require much physical effort in their management. The same effect of age on the adoption of agroforestry innovations has been reported by other researchers [29,37,38]. However, the work of [30,39] has shown that younger farmers are more likely to adopt innovations than older ones. In this regard, it is useful to understand that the adoption of innovations is a choice with a certain probability of risk, so that younger, less experienced farmers may take more risks compared to older ones. It is clear from these analyses that age can influence the adoption of agroforestry or agricultural innovations either positively or negatively, depending upon the environment or the individual perceptions of farmers.

4.2.2. Influence of Marital Status

This study shows that marital status positively influences a farmer’s decision to adopt agroforestry practices in LBR (Table 3 and Table 4). The results contradict the work of [39], who demonstrated that marital status has no direct relationship with farmers’ decisions. Indeed, being married could modify spousal decision-making behaviors. For example, women may encourage or discourage their husbands, and vice versa, to adopt agroforestry. For the few women farmers who were surveyed, tree plantations and cultivated fields, such as rice, must be separated because they consider that trees can attract birds, and that birds have a negative effect on planted areas and crops. In contrast, certain men who were surveyed think that agroforestry is an activity reserved for physically hardy people, whereby women are excluded.

4.2.3. Influence of Education Level

This study found a statistically significant relationship between education level and the adoption of agroforestry practices in the LBR area (Table 3 and Table 4). It is justified by the fact that farmers with a high level of education have a better understanding of concepts underlying the innovations, and can make more informed decisions to engage in a new activity. The more education a farmer has, the better his or her ability to appreciate new technologies. This result is in agreement with several studies [37,40,41,42]. However, [43] determined that the level of education does not necessarily influence the adoption of agroforestry practices. From all of these analyses, we can see that the adoption of new technologies requires a certain level of education. From the analysis of our results, it appears that the adoption of new technologies requires a certain minimum level of education, but not one that necessarily focuses on knowledge that is acquired through formal education.

4.2.4. Influence of Main Activity

This study shows that the main activity of farmers positively influences the adoption of agroforestry practices (Table 3 and Table 4). Regarding this subject, two categories of main activities were observed among the farmers who were surveyed: (i) agroforestry (agriculture, livestock, beekeeping and fish farming, with tree exploitation) and (ii) other activities (agriculture, livestock, beekeeping and fish farming, without tree exploitation). These activities are complementary, meaning their co-adoption is not easy. In this regard, these results support the research findings of [44], who have demonstrated the limitations of research in predicting farmers’ decisions regarding whether to adopt an agroforestry practice or not, based upon the socio-economic characteristics of the farmer. Further exploration would be desirable, incorporating other endogenous factors as suggested by [28,45]. The psychology of the operator is worth mentioning too [41].

4.2.5. Influence of Land Tenure

Land ownership has a positive influence on the adoption of agroforestry practices in the LBR area (Table 3 and Table 4). In this regard, landowners enjoy their land rights by freely using the land for agricultural activities and planting trees. In contrast, tenants cannot plant trees on land that they use because they do not have the guarantee of keeping the land long enough to plant the trees or to undertake livestock activities to avoid conflicts over the management of these plant species. This result is similar to the findings of [13,46].

4.2.6. Influence of Local Association Membership

The results of this study show that membership in an association positively influences the adoption of agroforestry practices (Table 3 and Table 4). This can be justified by the simple fact that farmers belong to an association structure that allows peers to share and exchange information regarding the benefits of using innovative techniques. Similar findings were found by [13,26,47]. Indeed, association membership is an essential source of technical information and the farmer-to-farmer extension of best endogenous practices. Conversely, some other studies have shown that this variable did not positively influence the adoption and intensity of agroforestry practices because the main purpose of associations does not necessarily involve the diffusion of knowledge about agroforestry [30,48,49]. As regards this discussion, the sustainability assessment of agroforestry practices in and around the LBR is the prudent choice for all farms in the context of sustainable land management [50]. In a recent study on the adoption of agroforestry practices in Bangladesh, household size, age, education, training, extension visits, and market access influenced agroforestry adoption, while a lack of technical support, technical skill and training were the prime constraints [51].

5. Conclusions

The development of agroforestry in Central Kongo Province (DRC) demonstrates that it is possible to combine agricultural and forestry activities in the same space. The key results show that the proportion of adopters far exceeds that of non-adopters. In addition, six factors positively influence the adoption of agroforestry practices: age (the 51 to 60 years old category), marital status, education level, main activity, land tenure and membership in a local association. Other factors limited the adoption of these practices, including gender, other age categories, household size, number of years of agroforestry experience, number of assets, distance between the farmer’s home and his or her farm, and access to credit. Our results constitute a database that would better orient the methods of the diffusion and adoption of agroforestry practices towards the real needs of local populations. The idea is to inform and train as many people as possible across all villages with respect to the socio-economic and environmental benefits of agroforestry. In order to achieve these benefits, agroforestry should not be adopted by simply involving farmers in various projects, where the latter take a passive role. The diffusion of this technology could be increased by proactively working with potential adopters through established and expanded social networks. Such actions could serve to integrate other government institutions and extend the reach of agroforestry to all segments of the population. The greatest limitation of this research is that the data gathered during the survey were based on simple declarations. Nevertheless, the results may be useful in informing policy makers to guide the methods of the dissemination and adoption of agroforestry systems according to the realities of the study area. The Logit model that was constructed in this study could be further improved by introducing additional explanatory variables, such as farmer income, labor form size, farm size, and the benefits and costs of the new practice. Future research on the adoption of agroforestry practices should also include variables related to climate change and resilience.