Factors Driving the Adoption and Use Extent of Sustainable Land Management Practices in South Africa

The slow and discontinuous adoption of agricultural technologies is one of the major threats to low agricultural productivity in developing countries. These actions highlight the challenges encountered in the adoption and the continuous use of sustainable land management practices in addition to the choices regarding the type used. The study investigated factors influencing the adoption decisions of smallholder maize farmers and the intensity of adoption of sustainable land management practices. Empirical data were collected from 250 farmers through interviews using a structured questionnaire. The adopter group and non-adopter group were compared using t-test and chi-square statistics, while the double hurdle with the fractional outcome response model was applied to establish the factors responsible for the adoption and the extent of use of sustainable land management practices. The results indicated that socio-economic and institutional characteristics are determining factors responsible for the adoption of sustainable land management practices and the extent of its use. The study recommended that continuous adoption and extensive use can be fostered by encouraging farmers to join a social organisation where related and relevant information on sustainable land management practices is shared through trained agricultural extension officers. Furthermore, regular training and access to credit facilities should be offered.


Introduction
The inevitable damaging impact of climate variation threatens agriculture. The effects of climate change are evidenced in every nation's economy, social activities and the entire environment. Climate change is considered a central and powerful destructive force in sustainable agricultural production in sub-Saharan Africa (SSA). It has been noted that climate change has had a significant effect on global agriculture in the 21st century [1]. In addition, Arora [2] avow those variations in climate tend to inhibit economic growth and other facets of human lives and natural well-being. Notwithstanding the predictions that the effects of climate change on agricultural production and livelihoods are expected to increase greatly over time, researchers evince that the aftermath of climate fluctuations on farm products will increase in subsequent years and the consequential effects will vary across countries and regions [3]. Exposure of soils to climate change results in land degradation [4]. Heavy rainfall, floods and excessive temperatures are likely to increase the risk of soil erosion, which will lead to land degradation, unless adequate measures are taken to protect and restore the soils in order to enhance food security and mitigate against the effects of climate change [5,6]. Climate change and land degradation are considered major threats to the survival and livelihoods of millions of people in sub-Saharan Africa [7].
However, the concept of circular economy seems appropriate in this regard to combat the threat. A circular economy is an economic system that tackles global challenges and environmental degradation such as climate change, biodiversity loss, waste, and pollution [8]. Following this definition, sustainable land management practices could be said to be a circular economy system with the ability to reduce the impact of land degradation and effect of climate change on farm land. According to [9] the circular economy is now more important than ever due to global threat facing humanity resulting in to global environmental disruptions. Similarly, [10] explained that this concept of circular economy is considered essential to solve many of the existing global environmental and social challenges such as climate change, nature conservation. This principle of circular economy is based on three ideologies: design out waste and pollution; keep products and materials in use; regenerate natural systems [11]. This regenerative approach is in contrast to the traditional linear economy, which has a "take, make, dispose" model of production [12].
Accordingly, sustainable land management practices (SLMPs) play a vital role in the sustenance of food production by addressing the effect of climate change on soil and land use, and improving land degradation [13,14]. Apropos SLMPs, sustainable land management (SLM) is fundamentally concerned with how people look after the land for both present and future use. Thus, SLM is defined by Liniger et al. [15] as the adoption of land-use systems that, through appropriate management practices, enables land users to maximise the economic and social benefits from the land while maintaining or enhancing the ecological support functions of the land resources.
Correspondingly, SLMPs comprise a knowledge-based modus operandi or method that enables the integration of land, water, biodiversity and environmental management, in addition to the input and output externalities, to meet the escalating rate of food and fibre demands, while simultaneously allowing for the sustenance of ecosystem services and livelihoods [16]. Moreover, SLMP is a necessity that cannot be eschewed if a nation intends to meet the imperative or even the basic requirements of a growing population continually. It should also be noted that inappropriate land management will result in land degradation and an undeniable reduction in the production of food, food security and service functions [17].
Furthermore, the issue of improved land-use practices or sustainable land use has irrefutable relevance, particularly when considering the registered accumulated environmental problems. Some of these identified problems are land degradation; increased demand for food production; depleted natural resources; climate change; regional climate extremes and the threat of environmental pollution; biodiversity loss; disturbed landscape stability; economic globalisation, together with threats to energy security; water supply; and the increasing conflicts between socio-cultural, political-economic and environmental goals [18]. Regarding the above, SLMPs are advantageous because, among other advantages, they can stem land degradation, increase crop diversity, boost higher yields, reduce the costs of production, improve the micro-climate for plants and preserve organic carbon in the soil [19]. Additionally, SLMPs focus on alleviating the detrimental impacts of climate change on productivity and thus prevent degradation of natural resources.

Implementation of Sustainable Land Management Practices in South Africa
In addition to Argentina, China, Cuba, Senegal and Tunisia, South Africa is one of the six pilot countries participating in the global Land Degradation Assessment in Drylands (LADA) project. The goal of these countries is to enable a warranty that will ratify informed policy advice on land degradation. In South Africa, sustainable land use and management is cardinal and its advancement predominates, because of the various agricultural natural resources in the country and their complex nature and vulnerability to degradation. Hence, the Department of Agriculture, Forestry and Fisheries (DAFF), South Africa's LADA project coordinator, has supported the network World Overview of Conservation Approaches and Technologies (WOCAT), since 1998, in its effective methods of collecting, documenting, analysing and distributing information on best-management practices.
The creation of the LADA project affords the WOCAT and the LADA teams the opportunity to adopt the WOCAT questionnaire jointlythe questionnaire is now being used globally as the WOCAT-LADA questionnaire. The network WOCAT has been used in South Africa for many years and has led to the documentation of many approaches and technologies [17]. Additionally, numerous practices have evolved and are now used for the promotion and improvement of the sustainable management of natural resources. A summary of the available SLM approaches and technologies documented in South Africa is provided in the form of case studies. The summary also captures basic information regarding the location where the case studies were conducted and includes general facts about them.
Sustainable land management comprises measures and practices adapted to biophysical and socio-economic conditions aimed at the protection, conservation and sustainable use of resources (soil, water and biodiversity) and the restoration of degraded natural resources and their ecosystem functions. There are numerous examples of SLMPs that are categorised into land-use type and technology group.

Land-Use Types
Sustainable land management provides alternatives for all ecosystems and is an apt channel for addressing the causes of desertification, land degradation and drought (DLDD), because of its adaptability quality. Despite the profitable functionality of SLM, due to its all-encompassing nature, its broadness can also make it difficult to recognise, apply and share its technologies and practices. Sustainable land management can be used by applying it to specific land-use types. Land use is a description of a particular activity that is specially carried out on a unit of land in urban, rural and conservation settings [20]. Additionally, the land-use type is a way of classifying land by how it is used. Instances are SLM for cropland, SLM for forest/woodland, SLM for grazing land, SLM for mixed land and SLM for other land uses. It should be noteworthy here that Maponya [21] indicated that natural resources like land and water are the prerequisite for a smallholder farmer to engage in agricultural production.

Sustainable Land Management Technology Groups
Sustainable land management practices in the category of technology groups include integrated soil fertility management, vegetation management, water management, grazing pressure management, animal waste management, sustainable forest management, reduction of deforestation, afforestation/reforestation, forest restoration, agroforestry, agro-pastoralism, minimum soil disturbance, soil erosion control, fire, and pest and disease control.
However, despite enormous work done by WOCAT and LADA, most smallholder farmers dis-adopt SLMPs soon after the initial adoption. These actions highlight the challenges encountered in the continuous adoption of SLMPs and the choices regarding the type of SLMP used. It was evident that there is no clear understanding of the problems encountered by farmers in the adoption of recommended SLMPs. Thus, it is appropriate that the current situation is investigated.

Adoption of Sustainable Land Management Practices
The adoption of SLMPs is multidimensional, with numerous factors affecting the decision and behaviour of farmers regarding the choice of SLMP used. Several studies have been done on the adoption of SLM in which farmers either adopt at a lower rate or dis-adopt the practices [22]. Many of the results demonstrate the important influencing factors in the decision behaviour of households of smallholder farms towards various land conservation measures. For example, a study conducted by Adugna and Bekele [23], in the north-western part of Ethiopia, explained that economic variables such as land ownership, livestock units and size of the family influence the adoption of SLM.
Similarly, Adimassu and Kessler, Amsalu and De Graaff and Bekele and Drake [24][25][26] pose that socio-economic factors influence the adoption of SLMP in Ethiopia. Pender and Gebremedhin [27] concluded that male-headed households were more likely to use contour ploughing and to manure regularly. Muhammad-Lawal et al. [28] explain that the socio-economic factor of household size implies more labour to carry out land management practices. However, according to Holden and Yohannes [29], there is no relationship between family size and land management practices, although Shiferaw and Holden [30] confirmed a negative correlation between family size and land conservation practices.
According to the World Bank and Yirga [31,32], institutional factors such as land insecurity, access to credit, proximity to the all-weather road and market access influence the adoption of SLMP in Ethiopia. Deininger et al. [33] reported a negative relationship between farm size and certain land management practices such as tree planting. Similarly, there are indications that policy-related issues contribute to the adoption of SLMP. For instance, the findings of Nkonya et al. and von Braun et al. [34,35] indicate that the lack of strong policy action and the lack of an evidence-based policy framework are considered critical challenges for the effectiveness of SLMPs.
It is against this backdrop, that this study investigates the determinants for the continued use of SLMPs by smallholder farmers and the choices regarding the type of SLMP used adopted. The purpose of this study was to assess the socio-economic, institutional, psychological and biophysical factors that influence the adoption of SLMPs among smallholder farmers in the Mpumalanga province of South Africa. This research aligns with South Africa's national policy on climate change and food security, which conforms with the "integrated and inclusive rural economy", as stated in the National Development Plan Vision 2030 (chapter 6).
In addition, the research is within the scope of the sustainable development goals on food security and environmental sustainability. Policy-wise, the study seeks to provide key insights into food policy and national environmental sustainability, that is, the SLMPs that are prioritised by smallholder farmers, and to provide a strong basis for bridging the gap between adoption and extent of use of SLMP and aligning with the government's agricultural blueprints regarding the smallholder farmers as a centre of consideration.
Research question: What are the different factors that affect the adoption of SLMP and its intensity among smallholder maize farmers in the study area?

Description of Study Area
The current study was carried out in the Gert Sibande District of the Mpumalanga province. The province is situated in the east of the country and is bordered by Eswatini and Mozambique. It is encircled by the Limpopo province, far to the north, KwaZulu-Natal to the south, Gauteng to the west and the Free State province to the southwest. The Gert Sibande District covers an area of 31 841 km 2 , which accounts for about 6.5% of South Africa's land surface. The district consists of seven local municipalities, namely Govan Mbeki, Chief Albert Luthuli, Msukaligwa, Dipaleseng, Mkhondo, Lekwa and Dr Pixley ka Isaka Seme. Figure 1 shows the district and its local municipalities.
It is to be noteworthy that a report submitted to Food and Agriculture Organization of the United Nations by Agricultural Research Council -Institute for Soil, Climate and Water [36], compiled by Lianda Lötter, Liesl Stronkhorst, and Hendrik Smith, affirmed that Mpumalanga was part of the mapping area where the adoption of SLMP was introduced; hence, the province is selected as the study area. Similarly, the province was chosen as the study area because it is one of the three most, maize producing region in South Africa, which accounted for 23.5% of the total maize production in South Africa [37].

Data Types and Sources
The study used a quantitative research method and data were collected from both primary and secondary sources. Primary data were obtained from the sample respondents, who comprised smallholder maize farmers, while secondary data were sourced from bulletins, annual reports and published and unpublished documents.

Sampling Technique and Sample Size Determination
Data were collected in seven local municipalities. A list of small-scale maize farmers in the municipalities was obtained from DAFF. The Raosoft sample-size calculator was used to determine the size of the sample from the population of maize farmers in the study area [38]. This sample-size calculator takes into account the margin of error, confidence level and response distribution.
The calculation of the sample size n and the margin of error E is shown as follows: where N is the population size, r is the fraction of responses that you are interested in, and Z(c/ 100) is the critical value for the confidence level c.
The stratified random sampling technique was employed for the study. The farmer population was divided into strata, and the random-sampling method was used to select respondents from each stratum. Each stratum represented a local municipality, as shown in Table 1. A total of 250 questionnaires were administered across the municipalities.

Method of Data Collection
Permission was granted by DAFF, to collect data from the district, and ethical clearance was obtained from the University of South Africa. The data were collected in 2020 from the maize farmers using a questionnaire, as the survey instrument for data collection. The questionnaire consisted of a structured and logical flow of questions that were written in English. The questions covered land management practices and their adoption, socio-economics and farmbased characteristics. Validation and pre-testing of the questionnaire were done by an expert in the field of agricultural economics. Thereafter, face-to-face interviews were conducted in each local municipality with the help of extension agents, and each session lasted for 40 min. The questionnaires were completed anonymously, because the respondents' names, addresses and identification numbers were not required. In total, 250 questionnaires were administered in the study area.

Method of Data Analysis
The study used both descriptive statistics and the inferential model. The descriptive statistics included the use of frequency distribution, mean, standard deviation and percentages. The double-hurdle(D-H) model was adopted to examine the factors driving the adoption of SLMPs and the extent of adoption or the choice regarding the type used.

Descriptive Analysis
Mean value, percentages, frequency and standard deviation were used to analyse the socioeconomic characteristics of the smallholder maize farmers. Similarly, t-tests and chi-square tests were employed to compare the mean differences and the percentages between the two groups of SLMP adoption and SLMP non-adoption (Table 2).

Econometric Analysis
Double-Hurdle Model This research draws insight from the D-H econometric model. The D-H model has the ability to estimate precisely a dichotomous event or variables in the first hurdle, while the second hurdle ascertains the actual or observed level (extent) of the event. This model was specifically chosen due to the supposition that households make two sequential decisions regarding the adoption of innovation or technology. Consequently, the separation of the initial decision to adopt (y > 0 vs. y = 0) from the decision of how much to adopt (extent) becomes realisable.
In this study, the D-H model examined the determinants of incidences (adoption of SLMP) and the intensity (extent) of the adoption of SLMP. Each hurdle was shaped by the farmer's socio-economic features, together with the adopted SLMP. In the first hurdle, the probit regression model was explored, while in the second hurdle, the SLMP index was generated. Thereafter, a fractional outcome response model was used for modelling. The D-H model was structured to interpret the likelihood/unlikelihood of an event's occurrence so that, if it occurs, it assumes a continuous value. In relation to this study, the decision to adopt each practice is made first and is followed by the extent of its use (i.e. the level of adoption).
The first hurdle is stated as follows: Thus, the first hurdle is a probit model. The probit regression model is a multivariate technique that is appropriate when attempting to model a dichotomous dependent variable. Therefore, the decision of farmers to adopt SLMP is influenced by several socio-economic factors and institutional and farm characteristics. The model can be mathematically written as follows: Where Pr represents probability and Φ is the cumulative distribution function of the standard normal distribution. The maximum likelihood usually estimates the parameter β. Therefore, the probit model can be regarded as a latent variable model, assuming that an auxiliary random variable is present. Thus: where ε~N (0, 1). Y can then be regarded as an indicator if this latent variable is positive, given The model assumed that the probability of the ith farmer adopting SLMP (Y) is a function of the explanatory or independent variables (X) and the unknown parameter vector (e). Thus, it can be expressed as follows when the variables are fitted: Where: Y Binary response variable defined as the decision to adopt SLMP X 1 Gender X 2 Age X 3 Number of years spent in school X 4 Farm size X 5 Number of years farming X 6 Irrigation system X 7 Access to extension services X 8 Frequency of extension visits X 9 Member of social organisation X 10 Farming as a source of income X 11 Monthly income X 12 Access to credit facilities X 13 Awareness of SLMP Similarly, the second hurdle explored the use of a fractional outcome response. A fractional response requires the assumption of a functional form that imposes the desired constraints on the conditional mean of the dependent variable. It captures particular nonlinear relationships, especially when the outcome variable is near zero or one. The estimators fit models on continuous zero to one data using probit, logit, heteroskedastic probit and beta regression. Since the response variable that was generated by an index is a proportion that is naturally a fraction bounded between zero and one, this model is deemed fit. Fractional probit outcome was used to estimate the factors that influence the extent of use and the choice of the type of SLMP. The model is expressed as follows: The proportion of choice of channels used by a farmer E(y | x) is given by Where: y represents the dependent variable x is the explanatory variables θ is a vector of parameters G(.) is a cumulative distribution function of the standard normal distribution, which takes several forms such as the probit-G(xθ) ≡ Φ(xθ) or log-G(xθ) ≡ e −exθ functions [39].
Equation (12) can be estimated using a quasi-maximum likelihood method (QML), as suggested by Ramalho et al. [40] on the Bernoulli log-likelihood function: The marginal effects of the functional forms for the distribution of G(.) are given by Finally, the observed variable y i , is determined by the interaction of both hurdles as follows: If both decisions are made jointly (the dependent double hurdle), under a condition where the error term is assumed to have a bivariate normal distribution, it follows that the two decisions have been made together. Hence, it is stated as (u i , v i )~BVN (0, ψ) The composition of the two-stage decision suggests that the adoption of SLMP, the extent of its use and the choice regarding the type used can be estimated together in order to yield an efficient estimate. Since the factors affecting adoption decision and its intensity are different, the D-H model assumes that zero values can be reported in both the adoption decision and the intensity of the adoption stages. According to Cragg [41], the zeros reported in the first and second stages arise from non-adopters and, thus, there is no extent of choice (intensity).

Demographic Characteristics of Sampled Households
The descriptive statistics, shown in Table 3, reveal that among the 250 sampled respondents, 177 (70.8%) adopted SLMP and 73 (29.2%) did not. This indicates that the majority of the farmers adopted or used at least one SLMP at the time of the survey. Table 3 presents the results of the t-tests, which indicate the mean difference between the continuous variables, namely the age of the farmers, the number of years spent in school, farm size, the number of years farming, frequency of extension visits, and income. The average number of years spent in school was about 10 years, with the SLMP adopters indicating a mean of 11.15 years and the SLMP non-adopters a mean of 8.14 years. There was a statistically significant mean difference between the two groups at a 1% significance level.
This shows that the majority of the SLMP adopters are educated more that non-adopters, which could indicate that education plays a major role in the adoption of SLMP.
The number of years farming was found to range between 1 and 48 years, with an average of 10.828 years. The mean number of years farming for the farmers, who adopted SLMP, was 10.34 years and for the farmers, who did not, the mean was 12.00 years. The result shows that there was a statistically significant mean difference between the two groups at a 10% significance level. This revealed that most of the SLMP non-adopters have more farming experience than the SLMP adopters. This demonstrates that the more years of farming experience, the more sceptical or reluctant the farmers are in adopting SLMP.
Similarly, Table 3 indicates that the frequency of extension visits was statistically significant, with a 1% significance level. The total or pool was 2.288 visits, with the farmers, who adopted SLPM, demonstrating a mean of 2.49 visits compared with the non-adopters, who had a mean of 1.81 visits. This indicates that most farmers, who adopted SLMP, had more visits than their counterparts, who did not adopt SLMP. It is thus plausible that several extension visits play a vital role in constantly reminding farmers of the innovation and the adoption of technologies and ways to improve farming activities. Table 4 presents the chi-square results regarding the mean differences in the variables of gender, extension services, access to credit, access to the irrigation system and awareness of SLMP. Of the sampled farmers, 131 (52.4%) were male and 119 (47.6%) were female. In addition, 103 male farmers and 74 female farmers adopted SLMP, while 28 male farmers and 45 female farmers did not. The mean difference was found to be statistically significant at a 1% level. The results also demonstrate that more male farmers than female farmers adopted SLMP, which can be attributed to the fact that male farmers are more dominant and have more access to land than female farmers.
Similarly, the variable of extension services was found statistically significant at 1% for both groups of farmers. In total, 203 farmers (81.2%) received visits from extension services, while 47 farmers (18.8%) did not. Of these, 155 farmers (62.0%), who had adopted SLMP, received extension service visits, while 48 farmers (19.2%), who had not adopted SLMP, were visited. This shows that the majority of farmers, who had adopted SLMP, received visits from extension services and, thus, extension services play a significant role in the adoption of SLMP. This result is confirmed by [36], in their study carried out in the Kenyan drylands on the effect of participation in farmer groups on household adoption of SLMPs. The authors report that agricultural extension is an important entity in the adoption of SLMP [42]. Furthermore, the total number of sampled farmers, who had access to credit facilities, was 118 (47.2%). Of these, 67 farmers (26.8%) had adopted SLMP, while 51 farmers (20.4%) had not adopted SLMP. The difference between the group mean was found to be statistically significant at 1%. This shows that most farmers did not have access to credit; however, of the proportion of the farmers, who had access to credit, many of them had adopted SLMP. Consequently, access to irrigation system follows the same manner.
Furthermore, Table 4 reveals that the majority of the sampled farmers (93.2%) were aware of SLMP and, of these, 70.8% had adopted SLMP and 22.4% had not. The mean difference between the two groups was statistically significant at 1%. This explains that farmers' awareness of SLMP led to the adoption of at least one practice of the existing SLMPs in the study area. This is supported by Dayarathne et al. [43] who demonstrated a positive correlation between farmers' awareness and the adoption of SLMPs among tea smallholding farmers in Sri Lanka.
The existing SLMPs in the study area, presented in Table 5, are grouped into four categories, as shown in Table 6. Since most farmers had adopted more than one practice at the time of the survey, the correlation matrix, showing the association between the types of SLMPs adopted, is given in Table 7. Table 7 reveals that the grouped SLMP measures were positively associated with each other and were significant (p < 0.01) at a 1% level of confidence. Table 5 shows that the most frequently used practice is minimum soil disturbance, followed by mixed-cropping/inter-cropping. The least-used practice was agroforestry.

Determinants of Sustainable Land Management Practices
The first hurdle model estimated that the following variables were statistically significant and influenced the adoption of SLM among the maize farmers in the study area. Gender The gender of the farmers in the study area was found to be statistically significant (p < 0.01) and to influence the adoption of SLMP. The gender of households increases the probability of SLMP adoption by 16.5%, signifying that a male-headed household is likely to increase the adoption of SLMP. This is because male household farmers have access to farmland and participate more in farming activities than their female counterparts. This is confirmed by Oduniyi and Gebre et al. [44,45] who examined the impact of gender differences on maize productivity in Dawuro Zone, southern Ethiopia, and found that gender differentials between male-headed households and female-headed households are more pronounced at mid-levels of productivity.
Age The adoption of SLMP was influenced by the farmer's age. The results revealed that there is a positive correlation between age and the probability of SLMP adoption. This means that as the age increases, so does the adoption of SLMP. The marginal effect shows a 0.6% likelihood that farmers increase the adoption of SLMP per unit increase in age. The current study demonstrated that the majority of older farmers engage in practices such as indigenous conservation and agronomic practice, which are examples of SLMP. However, these are examples of indirect SLMP adoption. This was inconsistent with the findings of Song et al. [46] who report that older farmers usually retain their current practices because they are afraid of the unknown.

Years Spent in School
The education of the farmers plays an important role in the adoption of any technology. The first-hurdle estimated result in Table 8 reveals that there is a positive relationship between education and the probability of SLMP adoption. The higher the  Source: Sheng (1989) education attained, the more the probability of SLMP adoption. The marginal effect denotes that a unit increase of a year spent in school will result in a 2.5% probability of SLMP adoption. This is not surprising, since education facilitates the adoption of technologies. Educated farmers enjoy exploring and aim to be well informed regarding innovations. This result is confirmed by Lokonon and Mbaye [47] who report that education is important in the adoption of SLMP.
Years of Farming The number of years farming denotes the experience acquired by the farmer. The results show that the number of years farming or farming experience has a negative correlation with the adoption of SLMP. A year increase in farming experience among the farmers in the study area decreases the probability of adopting SLMP by 1.1%. Farmers with significant experience are usually risk-averse and may not want to change to a new technology that requires higher technical knowledge and investment. This result conforms to the findings of Agidew and Singh and Awoyinka et al. [48,49] which indicate that as farming experience increases, the decision to adopt SLMP decreases. However, this is inconsistent with the findings of Zeweld et al. [50] who explain that farming experience has a significant positive effect on SLMP adoption.

Frequency of Extension Visits
The number of extension visits received by the farmers influences the adoption of SLMP. The marginal report revealed that as access to extension visits increases by one visitation, the likelihood to participate in SLMPs increases by 19.9%. The number of visits received by a farmer relating to information regarding SLMP corresponds to the probability that the farmer adopting SLMPs. This is supported by Etsay et al. [51] who report that regular extension visits help in the adoption of agricultural technology.

Member of Social Organisation
Membership in a social organisation(s) among the maize farmers in the study area was found negative and statistically significant (p < 0.1), thus influencing the adoption of SLMP. The result revealed a negative correlation between being a member of a social organisation(s) and adopting SLMP. The result from the marginal effect demonstrates that being a member of a social organisation(s) decreases the probability of adopting SLMP by 12.5%. This could indicate that social organisations such as farmers' groups do not share information that is related to SLMP. However, according to Jagger and Pender and Babalola and Olayemi [52,53], being a member of an organisation is an important determinant that positively influences the adoption of SLPM.  Determinants of the Extent of Adoption of Sustainable Land Management Practices The extent of use or the intensity of SLMP was positively influenced and statistically significant (p < 0.01) by the number of years spent in school (education), farm size and awareness of SLMP. Farmers, who are educated, find it easy to explore and to adopt many SLMPs simultaneously in order to ameliorate land degradation. The more educated farmers are, the more likely they will be to adopt the various available types of SLMPs. Similarly, farm size positively and significantly influences the extent of the adoption of SLMP. The greater the farm size, the more a farmer is able to try different types of SLMPs on the same piece of land. This is confirmed by the findings of Sheng [54] who reports that cultivated farm size positively and significantly (p < 0.05) influences the use of land management practices. Consequently, farmers' awareness of SLMP promotes more choices simply by enhancing and informing new knowledge. However, access to credit facilities negatively influenced the extent of the adoption of SLMP. Moreover, access to credit facilities negatively influences the types of SLMPs adopted, since some practices require considerable money and investment to establish. Nkoya et al. [55] report that access to credit in combination with other assets is an important determinant of SLMP adoption by farmers.
Furthermore, the selection bias was estimated, representing the inverse Mills ratio in the second hurdle in Table 8. It was found to be positive and significantly significant (p < 0.01). This indicated that sample selection bias would have resulted if the outcome equation had been estimated without accounting for the decision to adopt SLM and thus, there is no evidence of a sample selection problem. The coefficients associated with the inverse Mills ratio were significant, justifying its inclusion. Thus, non-inclusion of the inverse Mills ratio would lead to biased results attributed to sampling selection bias. Table 9 presents the predictive margin that computed the standard errors of the means for the extent of SLMP adoption, which was found statistically significant (p < 0.01). This suggests how much the sample mean would vary if this study were to be repeated using new samples from within a single population. It also shows that there is a 20% probability for a farmer to adopt SLMP given all predictors held at mean values.

Conclusion and Recommendations
The study established the factors that influence the adoption and the choices of SLMPs in the study area. It revealed determinants of both continuous use and dis-adoption of SLMP among maize farmers in the Mpumalanga province of South Africa. Socio-economic and institutional characteristics were found to be determining factors responsible for the adoption and the extent of SLMP use. The main conclusion of this study was that gender of the farmer, age of the farmer, number of years spent in school by the farmer, number of years of farming or the experience of the farmers, the frequency of extension visits and being a member of a social organisation significantly affect the farmers' adoption of SLMP in the study area. Additionally, the extent of the use of types of SLMPs was influenced by the number of years spent in school by the farmers, farm size, access to credit facilities and awareness of SLMP. Furthermore, the study confirmed that there is no evidence of a sample selection problem predicted by the inverse mill ratio, which was found positively significant in the second-hurdle estimation in Table 8, which suggested that the results obtained are not spurious.
The study recommends that socio-economic characteristics and institutional engagement should be fostered. Joining a social organisation such as a farmers' group should be encouraged and sharing the related and relevant information regarding SLMP through a trained agricultural extension officer should be promoted. The channel of information dissemination should be improved to facilitate awareness and adoption of SLMP. Moreover, government should make credit facilities more accessible to the farmers in order to increase their SLPM choices, since this requires capital assets. Similarly, access to farmland should be facilitated not only among the male farmers, but the female farmers, to encourage them and to close the gap regarding gender differences in agriculture. Above all, regular training that serves as on-the-job training or education should be constantly provided for the farmers.