Farming with Trees for Soil Fertility, Moisture Retention and Crop Productivity Improvement: Perceptions from Farmers in Rwanda

Agroforestry is widely considered as a sustainable land management practice with multiple ecosystem benefits that support farmers’ livelihoods. While there is much evidence that the technologies are not being embraced as expected, there is knowledge gap about farmers’ perceptions to help devising extension strategies tailored to farmers’ needs in many countries, including Rwanda. This study determined factors that influence farmers’ perceptions on trees effects on soil fertility, moisture retention and crop productivity in semi-arid and humid regions of Rwanda. Logistic regression was used to analyze household survey data that covered 107 and 146 farmers in semi-arid and humid regions, respectively. Results showed that the proportion of farmers with fertile soil in the humid (92%) and the semi-arid (65%) regions was higher on farms with trees. In the semi-arid region, farmers’ perceptions of trees effect on soil fertility was positively influenced by education (P < 0.05) and negatively influenced by farm size (P < 0.1) whereas in the humid region, it was positively influenced (P < 0.1) by family size. Farmers’ perceptions of the effect of agroforestry on soil moisture were influenced by farmers’ knowledge (P < 0.1) in the semi-arid region and gender (P < 0.05) in the humid region. Farmers reported higher yields of potato, maize and beans on farms with trees in the humid region and beans in the semi-arid region. These factors should be taken into consideration to target agroforestry technologies more efficiently within farmer populations. However, constraints including lack of seedlings, tree-crop competition, droughts, termites and lack of mineral fertilizers for seedlings need to be addressed to increase farmers’ uptake of agroforestry technologies.


Introduction
Land degradation is a major impediment to agricultural productivity and the leading cause of food insecurity in sub-Saharan Africa (Stewart et al. 2020).Current estimates show that more than 35% of agricultural lands are degraded to varying degrees due to human-induced activities (Gupta 2019).In East Africa, the drivers of land degradation are unsustainable intensification of agriculture, overgrazing and deforestation (Kirui and Mirzabaev 2014).Among these factors, deforestation appears the most overarching cause of soil erosion and the decline of soil fertility (Kassa et al. 2017).In Rwanda, the rate of soil loss by rain water runoff is estimated to be 595 million tons per year with a rate of nutrient depletion of about 100 kg NPK ha −1 year −1 (Henao & Baanante 1999).Furthermore, agriculture in the region is mainly rainfed which constitutes a major hindrance to the development of the sector, especially in eastern Rwanda where drought events are frequent and severe (Karamage et al. 2017).
Given the strong linkage between production constraints and food insecurity, there is need to devise appropriate soil and water management technologies to reverse the effect of land degradation and improve the productivity of farming systems.Agroforestry, the integration of trees and crops and/or animals in the farming system is one of the sustainable land management options currently being used to reverse land degradation and restore productivity of farming systems (Pinho et al. 2012).Agroforestry can reduce soil erosion, raise soil organic carbon, regulate soil moisture and improve soil fertility resulting in improved farm productivity (Kuyah et al. 2019).Trees on farms can intercept rains by attenuating or neutralizing its kinetic energy and splash erosion, building a permeable barrier that slows down and intercepts water runoff (Li et al. 2019).Trees also improves soil structure through incorporation of organic matter, thus increasing structural stability and resistance to soil erosion (Pinho et al. 2012).Trees have been shown to reduce soil loss by forming barriers that slow runoff and capture sediments and protect the soil aggregates from direct raindrops (Angima et al. 2002).Soil loss equivalent to 3.6 t ha −1 of sediments was recorded on plots with sole maize against 0.3 t ha −1 when maize was integrated with Leucaena leucocephala (Lal 1989).By intercepting water runoff, trees improve soil infiltration characteristics, traps more water and enhance the soil water content.This makes soil moisture more available under trees than in the open.
Trees maintain soil fertility and improve the availability of soil nutrients.For example, in agroforestry systems the average N-fixation was found to be 246 kg N ha −1 year −1 , but much higher (300-650 kg N ha −1 year − 1) in improved fallows where trees are managed by pruning or coppicing (Nygren et al. 2012).It was also found that K, Ca, Mg and organic matter contents were respectively higher by 150-178, 2-270, 0-78% and 11-100% beneath Faidherbia albida than in the open (Boffa 1999;Kamara & Haque 1992).Beside, crop yields were higher under trees than on farms without trees.For example, maize grain yield was higher under faidherbia (2.0-2.7 Mg ha −1 ) and gliricidia (1.8-2.4Mg ha −1 ) compared to the yields obtained in farms without trees (Sileshi et al. 2012).It 1 3 Farming with Trees for Soil Fertility, Moisture Retention… was also reported that Prosopis cineraria, Tecomella undulate and Azadirachta indica increased crop yield by 86, 49 and 17% over the control (Tomar et al. 2021).
In Rwanda, agroforestry is now receiving increasing more attention as a sustainable land management option to increase agricultural production and restore the productivity of degraded lands; however, the uptake by smallholder farmers is slow (Kiyani et al. 2017).Getting agroforestry adopted and applied at scale requires concerted actions to ensure that the promoted technologies are tailored to farmers' needs.This can be achieved by involving farmers in the design of the technologies implemented on their farms.Their perceptions constitute an important basis for technology development as they reveal the real problem experienced day to day by farming communities.For example, in a study conducted in Ethiopia, farmers had positive perceptions about the effect of multipurpose trees on soil and water conservation, however, they revealed that adoption of the technologies was constrained by agronomic problems, low feed value and land shortage (Mekoya et al. 2008).The involvement of farmers at all stages of the design and implementation of an agroforestry technology is required to enhance adoption.
There are a large number of factors which influence or shape the perceptions of farmers and the decision to take up a new agroforestry technology.These are the characteristics of the farmer and the characteristics of the innovation, among others (Meijer et al. 2015).This study aimed to understand farmers' perceptions of the effect of agroforestry trees on soil fertility, moisture and crop productivity in semiarid and humid regions of Rwanda.This study addresses the following research question: what are the driving factors that influence farmers' perceptions of agroforestry trees and effect on soil fertility, moisture and crop productivity?Answering this question can help targeting appropriately agroforestry technologies or redesigning them to be locally relevant and sustainably solve farmers' problems.

Study Area
This study was conducted in two agroecological zones, namely semi-arid and humid regions, of Rwanda.The two contrasting regions were chosen as representatives of typical regions affected by land degradation and present significant differences in terms of climate, topography and farming systems.
In the humid region, the study covered the districts of Nyabihu (1.6955° S, 29.5572° E) and Rubavu (1.6479° S, 29.3483° E) in the Western Province of Rwanda while the semi-arid region was in Bugesera district (2.2315° S, 30.1127°E) in the Eastern Province of Rwanda (Fig. 1).The topography of the humid region is hilly and mountainous with an altitude lying between 1481 and 2445 m above sea level.The precipitation varies between 1200 and 1500 mm per year and has a temperate climate with an average temperature of 15 °C, favorable for the growth of a wide range of crops throughout the year.In the semi-arid region of Bugesera district, the landscape is characterized by a gently sloping topography and lies between 1100 and 1500 m of altitude with annual rainfall and temperature of about 855 mm and 20.8 °C, respectively.The area is characterized by a long dry season of 3-4 months.The rainfall is often unreliable and the effects of drought are aggravated by fragile soils with low water holding capacity.The soil is developed from fertile volcanic ashes and has a high potential for crop productivity (Cyamweshi et al. 2013).The soil is classified as oxisols and are generally poor in organic matter (Verdoodt & Van Ranst 2003).

Household Survey
The study sites were stratified as follows in four sectors, namely Karago and Bugeshi sectors located in Nyabihu and Rubavu districts respectively, and Gashora and Rweru sectors located in Bugesera district.These administrative sectors were selected following the coverage areas of the project called Trees for Food Security which funded this study.The sectors are composed of five to seven administrative cells, each with several villages.Three villages per cell were then purposively selected following the information from district Forest Officers indicating the villages where agroforestry is widely practiced.
Sample selection at village level was based on a list of 400 households provided by the extension workers.More than three villages were covered during the survey depending on the time available for the enumerator.The respondents were selected using systematic random sampling achieved by skipping every two names on the Farming with Trees for Soil Fertility, Moisture Retention… list until 32 persons per village were obtained, making a total sample size of 107 respondents in Bugesera district to capture farmers' perceptions in the semi-arid region and 146 in Nyabihu and Rubavu districts to assess the same in the conditions of the humid region.Data were collected using a semi-structured questionnaire.Participating farmers were asked questions about household characteristics, agroforestry systems and management, perceptions on the effects of agroforestry trees on soil fertility, moisture and crop yields.Beside this, information related to constraints that are likely to affect farmers' involvement in agroforestry were also collected.Prior to conducting the survey, a pre-test was carried out with 10 farmers and based on the responses some minor modifications were made prior to conducting the full survey.This allowed also to standardize the interview technique for all interviewers.

Data Analysis
To assess the negatively/positively factors influencing the perceptions of farmers on soil fertility, moisture and crop productivity due to the effect of trees, a binary logistic regression analysis was done and interpretation followed a conceptual framework (Fig. 2).The conceptualization focused on the perceptions from farmers.The demographic factors including age, education level, marital status, family size, land size, land tenure and source of knowledge are predicted to influence  the perceptions of farmers on the effect of trees on soil fertility, moisture and crop productivity.The explanatory variables for soil fertility as affected by trees include plant nutrients, biological nitrogen fixation and soil structure improvement.The farmers will rate their soil fertility as fertile, medium and infertile.Soil moisture will be perceived by farmers in terms of plant vigor, proliferation of small insects in the soil, fast decomposition of organic matter and soil softness when cultivating.The crop productivity will be perceived by farmers in consideration of how agroforestry trees will influence the performance of crops (yield level).
Collected data were in comparative perspective between farms with trees and farms without trees.
To analyze the perceptions of farmers on the effect of trees on soil fertility an ordered logistic regression helped to explain whether the soil is fertile, moderately fertile or infertile.The binary logistic regression was used to analyze farmers' perceptions on the effect of trees on soil moisture.
Data were presented by segregation where soil fertility and soil moisture were considered differently depending on the agroecological region (semi-arid, humid).
The perception of farmers (Yi) depends on various socio-economic variables ( X i ) and can be expressed as: where Y * i denotes the dependent variable (perception), X i is a vector of explanatory variables, denotes the vector of the parameter to be estimated, and i is the error term.
To identify the factors that influence farmers' perceptions on the effect of trees on soil fertility and soil moisture, a logistic regression model was selected and specified as follows: where: P i is the perception from farmers on the effects of trees on soil fertility or soil moisture.i:the th respondentse z i : the mathematical constant e raised to the power of z The equation can be expressed as: where Z i : explanatory variables.To meet the assumptions of the logistic models (binary and ordered), multicollinearity was checked using contingency coefficient tests.None of the explanatory variables was omitted and the multicollinearity was acceptable (VIF = 2.5).The expected signs of the explanatory variables are described in Table 1. (1) (3) Farming with Trees for Soil Fertility, Moisture Retention… The information provided by respondents about cultivated crops, farm size and yields were computed to obtain corresponding yields on a hectare land.The mean yields of farms with trees and farms without trees were then compared using t tests to ascertain the effect of trees on crop productivity.

Socioeconomic Characteristics of Farmers in Semi-Arid and Humid Regions of Rwanda
The socioeconomic characteristics of farmers in the study sites are summarised in Table 2.In both semi-arid and humid regions of Rwanda, the farming community was dominated by men (67%).The age of respondents varied between 35 and 60 years old and the majority of them (87%) were married and had primary school education (82%).The family size ranged between 5 and 10 persons in both regions and most of them practice agriculture on their own lands (86%) with acreage that varied between 0.5 and 1 ha in the semi-arid region (51%) and less than 0.5 ha in the humid region (57%).In the semi-arid region, farmers reported that they acquired knowledge on agroforestry from government agencies (58%) while in humid region the main sources of knowledge were government agencies (58%) and non-government organizations (33%) When comparing the two sites, there were significant differences in terms of the socioeconomic characteristics of farmers, except marital status, land tenure and the source of knowledge about agroforestry (Table 2).In particular, the proportion of male-headed households as well as the number of older farmers and those who attended primary school were significantly higher in the humid than in the semi-arid region.Besides, the proportion of farmers who owned < 0.5 ha was significantly higher in the humid region, while the number of 1 3 Farming with Trees for Soil Fertility, Moisture Retention… farmers with the land size land comprised between 0.5 and 1 ha and those with land size > 1 ha was significantly higher in semi-arid.The number of persons that varied between 1 and 5 persons per household was significantly (P < 0.001) higher in the humid region while the household size of 5-10 persons per family was significantly higher (P < 0.01) in semi-arid region.No significant differences were found in terms of marital status, land tenure, the source of knowledge about agroforestry.

Trees in the Farming Systems of Semi-Arid and Humid Regions of Rwanda
In the semi-arid region, the dominant agroforestry tree species are Grevillea robusta, Leucaena leucocephala, Mangifera indica and Cassia spectabilis and are mostly scattered (44%) within the farms.In the humid region, the most dominant tree species is Alnus acuminata (84%) and is mostly planted along contour hedges (46%).
Other tree niches in both regions are farm boundaries and combination of contour hedges and trees scattered across the farms (Table 3).

Effect of Agroforestry Trees on Crop Productivity in Semi-Arid and Humid Regions of Rwanda
Potato, maize, beans and pyrethrum were reported as the main crops cultivated by farmers in the humid region while maize and beans were the most cultivated by farmers in the semi-arid region.The yields of potato, maize and bean in the humid region and bean in the semi-arid region were significantly high in farms with trees.However, there was no significant effect of trees on maize and pyrethrum yields in the semi-arid and humid regions, respectively (Table 4).

Farmers' Perceptions of Trees Effects on Soil Fertility in Semi-Arid and Humid Regions of Rwanda
Trees on farms improved soils fertility when compared with farms without trees (Table 5).Farmers rated their soil as infertile, medium fertile or fertile following crop harvests obtained from their lands, e.g.poor harvests for infertile soils, moderate crop harvests for medium fertile soils and good harvests for fertile soils.In the semi-arid region, few respondents (28%) rated their soils as fertile on farms without trees while large number of respondents (65%) declared that trees improved soil fertility on their farms.Similarly, the proportion of respondents with fertile soil was higher on farms with trees (92%) compared to those who practiced agriculture on farms without trees (18%) in the humid region of Rwanda.
The results from the maximum likelihood estimation of the ordered logistic regression model for determinants of farmers' perceptions of soil fertility on farms with trees are reported in Table 6.The pseudo R-squared is about 0.137 and 0.103 for the semi-arid and humid regions, respectively.These values appear to be quite low but fortunately some of the independent variables are statistically significant.When considering each independent variable solely, it was found that in the semi-arid region, soil fertility was negatively influenced (P < 0.1) by household farm size and positively influenced (P < 0.05) by farmer education level.In the humid region, the most determinant factor of soil fertility was household size at 10% probability level (P < 0.1).

Farmers' Perceptions of the Effect of Agroforestry on Soil Moisture in Semi-Arid and Humid Regions of Rwanda
Thirty seven percent of the respondents in semi-arid and 9% of those in humid regions indicated that a good soil moisture is recognized by the vigor of plants grown on the land.Overall, 65.1% stated that a soil with good soil moisture becomes soft when cultivating, while others reported that a soil with good soil moisture is recognized by the fast decomposition of dead plants and insect proliferation in the soil (Table 7).
The maximum likelihood estimation of the binary logistic regression model for the assessment of factors that influence farmers' perceptions of trees effect on soil moisture are reported in Table 8.
The pseudo R-squared is about 0.137 and 0.103 for the semi-arid and humid regions, respectively.These values appear to be quite low but fortunately some of the independent variables are statistically significant.In the semi-arid region, farmers' perceptions about soil moisture were significantly (P < 0.1) influenced by the source of knowledge.In this region, the farmers who acquired knowledge on agroforestry were more likely to have positive perception on the effect of trees on soil moisture.In the humid region, gender was the most significant (P < 0.05) factor influencing their perceptions on soil moisture.The positive sign of the coefficient of the variable gender indicates that male farmers have higher likelihood of a positive perception of trees effect on soil moisture.
Farmers reported a number of constraints that hamper agroforestry expansion in the study areas among which are the lack of tree seedlings in both the semi-arid (90%) and humid (62%) regions (Fig. 3).Particular issues related to the failure of young trees reported in the semi-arid region were the lack of fertilizer inputs (40%) to enhance growth of planted seedlings, drought and termites.Other constraint, cited especially in the humid region, was related to tree-crop competition.

Discussion
The farming systems in the semi-arid and humid regions of Rwanda are characterized by various practices among which tree-crop is dominant.This might be due to household farm size, which is generally small in Rwanda and hence, farmers  prioritize crop production to ensure food availability and tree products such as firewood, stakes for climbing beans and timber.Land size has been shown to be an important determinant of agroforestry adoption and constitutes a fundamental factor that influence farmers' choice of farming technologies (Zerihun et al. 2014).Agroforestry tree species were more diverse in the semi-arid region with about six tree species in farms as compared to the humid region where Alnus acuminata was the most planted tree (84%).This may be due to colder climate of the region that does not favor growth of a variety of tree species in higher elevation sites.Growing alnus trees as a monoculture on farmland may result in reduced variety of tree products and services and more importantly may hamper the sustainability of the alnus trees in the case of any outbreak of pest and disease.Our results concur with Bucagu et al. ( 2013) who reported that agro-ecological conditions such as altitude and temperature have considerable influence on growth and development of different tree species.There is need to promote native tree species and to identify a variety of other tree species that can match the region's environment to enhance farm tree diversity in addition to Alnus acuminata.
Farmers pegged their perceptions of agroforestry trees to be soil fertility improving crop yields.This is consistent with the findings of Kuyah et al. (2019) who reported that agroforestry systems in Saharan Africa increase crop yields while maintaining delivery of ecosystem services.According to Isaac and Borden (2019), agroforestry trees can improve soil nutrient balance by reducing nutrient losses from soil erosion and leaching, and increasing nutrients inputs through nitrogen fixation; in addition they can also take up soil nutrients that are beyond the reach of crop roots and recycle them to the surface through litterfall (Pinho et al. 2012).This justifies better yields of most crops on farms with trees in both study areas.
The study analyzed factors that affect farmers' perceptions on soil fertility and soil moisture.The strong ties between farmers' education and improvement of soil fertility on farms with trees in the semi-arid region show that the education gained by farmers enhanced their skills and capabilities to manage trees in a way to achieve positive effect of the planted trees on soil fertility.This is consistent with Uematsu & Mishra (2010) who found that educated farmers have better access to information and knowledge, tend to possess higher analytic capability of the information and knowledge necessary to successfully implement new technologies and achieve expected results.Another factor, though with negative relationship with soil fertility on farms with trees, is the size of the farm.This means that the smaller the agricultural land, the more trees contribute to increasing soil fertility.This inverse relationship between farm size and soil fertility may be attributed to the fact that large farm might belong to people who do not only rely on farming activities for their livelihoods.Their lands are therefore less exploited and/or rented to other farmers who do not prioritize sustainable management of soil fertility.The results of this study concur with the findings of Akram (2007)who reported that large farmers may have enough alternatives to earn their livelihoods, which decreases their motivation to fully exploit the potential of their land and keep it for other than productive purposes.Besides, the long return on investment in agroforestry is also problematic, as many farmers do not have access to capital, credit or secure tenure for their lands.Promoting value chains for agroforestry products and services is an important action 1 3 Farming with Trees for Soil Fertility, Moisture Retention… to take (Parthiban et al. 2021).The negative correlation between farm size and farm productivity was also reported by Ansoms et al. (2008) in rural Rwanda.The positive relationship between the family size and soil fertility on land with trees in humid region suggests enough labor to cover the works required for better management of trees and crops on the same land to ensure the system productivity (trees, soil and crops) and household food security.Although most trees on farms need only seasonal or annual attention, compared to livestock and annual crops, the labor requirement is higher for agroforestry and the production typically does not pay off unless the labor is from within the family (Mutoko et al. 2014).These results are consistent with Mignouna et al. (2011) who reported that household size is used as a measure of labor availability and showed that larger households have the capacity to relax the labor constraints required during the introduction of new technologies.
The perception of farmers that agroforestry trees improve soil moisture was based on farmers' physical observation such as the humidity of the soil, the presence of insects in the soil (biological activity), plant vigor and fast decomposition of plant residues under trees.Ong et al. (2006) found that trees improve water storage and reduce evaporation and the rate of transpiration in agroforestry systems.Our results indicated that farmers' perceptions of the effects of trees on soil moisture was influenced by the source of knowledge and gender in the semi-arid and humid regions, respectively.Government agencies were rated as the main source of knowledge for farmers in this study.The positive and significant correlation of source of knowledge with farmer perceptions, reflects tremendous efforts of government agencies to improve farmers' knowledge and facilitate development of local skills to practice agroforestry in semi-arid region of Rwanda.This is in line with Cantore (2011) who reported the role of government agencies in building farmers' capacity through trainings to manage farm resources for improved agricultural productivity.
The positive influence of male farmers on the perceptions of the effects of trees on soil moisture in the humid region of Rwanda may be attributed to the fact that usually men have more access to resources, such as land and knowledge, which are fundamental towards their perceptions and participation in agroforestry.Moreover, our study indicated that men were more represented than women in agroforestry activities.Such factors place men in a better position to understand more of the benefits of trees as compared to women.Previous research has indicated that in Sub-Saharan Africa women have limited access to land and tree tenure rights (Kiptot et al. 2014) in addition to their low participation in extension services (Kilic et al. 2015).Therefore, promoting gender equity in agroforestry may generate greater benefits for both men and women for improved farm productivity and food security in Rwanda.
Although farming with trees is regarded by farmers as an effective practice that enhances soil fertility, moisture and crop productivity, the lack of seedlings appeared as an important constraint that is likely to hamper the expansion of agroforestry technologies.This may be due to various reasons including inadequate knowledge on seedling production.Therefore, organizing farmers' training is required to improve their knowledge on profitable agroforestry technologies such as tree seed and seedling production to enhance availability of planting materials.Farmers' awareness is one of the important factors that is positively associated with on-farm tree planting (Ajayi et al. 2003).The lack of tree seedlings in the study areas may also be due to the lack of capital and the nature of seedling markets in Rwanda that may limit the ability of farmers to have access to quality and affordable tree seedlings.Existing reports show that in Rwanda, tree seeds and seedlings system are largely controlled by the central government who, in collaboration with non-government organizations, distribute seedlings for free to farmers in the selected project sites while other farmers source seedlings from informal markets belonging to smallholder nursery owners where the quality is not guaranteed and the price of seedlings is not regulated to meet the purchasing power of farmers (Mukuralinda et al. 2016).Similar findings were also reported in Kenya where the lack of access to quality seedlings and their cost constituted aconstraint to adoption of agroforestry innovations by smallholder farmers (Wafuke 2012).To counter these constraints, there is need to develop a national strategy for germplasm supply through dialogues involving various actors such as the government, NGOs and the private sector.
The perceptions of farmers of the semi-arid region that most planted seedlings did not survive due to lack of mineral fertilizers may be associated to the lack of fertilization guidelines for young trees and more attention is given to food crops to ensure food security.Although fertilization of trees is not a common practice in Rwanda, there is a body of research that supports that fertilizer application can be used to improve the establishment of planted trees.Yuldashev et al. (2021) indicated that in Uzbekistan, application of mineral fertilizers increased the height (76 cm) of Picea mariana seedlings compared to the control (57 cm) and enhanced the growth and formation of saplings of Tilia cordata and Betula pendula.In view of farmers' needs to fertilize the planted seedlings, there is need to develop fertilization guide for agroforestry young trees to enhance rapid growth and survival rate of planted seedlings in semi-arid region of Rwanda.
Another cause of the mortality of planted trees as reported by farmers in the semi-arid areas of Rwanda is drought.In an agriculture quasi-rainfed areas such Rwanda, the mortality of planted trees is likely to persist unless irrigation measures are urgently taken by policymakers.Another possible solution is to identify drought tolerant tree species.The occurrence of termites as constraint to tree growth in the semi-arid region may be attributed to the warm climate of the region which is favorable to the proliferation of termites.This is supported by Debelo and Degaga (2017) who reported that in tropical and sub-tropical regions where rainfall is low, termite attacks appear most acute and cause serious problems in the development of nurseries and young tree plantations.
The constraint related to tree-crop competition as reported by surveyed farmers especially in the humid region, may be attributed to inadequate knowledge of tree management resulting in the loss of crop productivity.Ndoli et al. (2017) reported that when trees and crops were integrated, there was tree-crop competition for light, soil nutrient and water which affected crop productivity in smallholdings of Northern Rwanda.This phenomenon was reported by Gachie et al. (2020) who indicated that crop yields under pruned trees are generally higher than under unpruned trees.To help farmers obtain more benefits from tree-crop integration, there is need to improve their knowledge on tree management, especially branch and root pruning, so as to maintain the balance between tree and annual crop production.

3
Farming with Trees for Soil Fertility, Moisture Retention…

Conclusions
Agroforestry technologies have been promoted for several decades to restore the productivity of degraded lands to increase soil fertility, control erosion and improve water availability in Sub-Saharan Africa.There is much evidence that the technologies are not being embraced as expected.This study analyzed the factors that influence the perceptions of farmers on soil fertility, moisture and crop productivity in semi-arid and humid regions of Rwanda.The results indicated that factors influencing farmers' perceptions are not the same across the agro-ecological regions.Education and farm size influenced farmers' perceptions on trees effects on soil fertility while the source of knowledge acquired by farmers influenced their perceptions on soil moisture in semi-arid region.The household farm size and gender, respectively, influenced farmers' perceptions on trees effect on soil fertility and moisture in the humid region.Extension efforts should take these factors into consideration to target agroforestry technologies more efficiently within farmer populations.The improvement in soil fertility, moisture and crop yields on farms with trees was highlighted by farmers, however, field experimentations are needed to quantify the above benefits.Although agroforestry has a long history in Rwanda and is known to the smallholder farmers in the study regions there are several constraints that are likely to hinder farmers' willingness to plant more trees.The lack of tree seedlings is associated to the nature of the tree seed and seedling system which is centralized by the government in Rwanda.There is need to develop a national strategy for germplasm supply through dialogues involving various actors such as the government, non-governmental and private sectors.Tree-crop competition reported by farmers is an important constraint that can considerably depress crop yields and is thus likely to make trees undesirable to farmers.Farmers' training on tree management especially branch and root pruning can help to meet crop production goals in agroforestry systems.

Fig. 1
Fig. 1 Location of the study sites in Semi-arid and humid regions of Rwanda

Fig. 3
Fig. 3 Constraints that affect farmers involved in agroforestry in semi-arid and humid regions of Rwanda in 2021

Table 1
Description of variables used in the analysis and hypotheses on farmers' perceptions of soil fertility and soil moisture in 2021The sign ( +) indicates that the variable is expected to have a positive likelihood of influencing farmers' perceptions of soil fertility and soil moisture, while the sign (-) indicates that the variable is expected to have a negative likelihood of influencing farmers' perceptions of soil fertility and soil moisture

Table 2
Socioeconomic characteristics of farmers in semi-arid and humid regions of Rwanda in 2021 The figures in this table represent the number of farmers and the corresponding percentage are in parentheses.*** p < 0.01, ** p < 0.05

Table 3
Farming systems, dominant tree species and their locations in semi-arid and humid regions of

Table 4
Effect of agroforestry trees on crop productivity in semi-arid and humid regions of Rwanda in 2021

Table 5
Rating of soil fertility level on farms with trees and farms without trees in semi-arid and humid regions of Rwanda in 2021

Table 7
Farmer's indications of good soil moisture in farms with trees in semi-arid and humid regions of Rwanda in 2021 Standard errors are reported in parentheses.***p < 0.01, **p < 0.05, *p < 0.1