Biophysical characteristics of the research sites and socioeconomic characteristics of households
Two agro-ecological zones were compared, namely the Central Plateau (average altitude of 1,500–1,700 m a.s.l and annual rainfall of 1,160 mm) and the northern Buberuka highlands (average altitude of 1,800–2,650 m a.s.l and annual rainfall of 1,560 mm rainfall) both of which are considered to have good potential for agroforestry (Yamoah et al. 1989). The Central Plateau agro-ecological zone (AEZ) is located in south-west of Rwanda contrasting with the Buberuka highlands agro-ecological zone (AEZ) located in Northern part of the country. In the Central Plateau, Histosols and Cambisols are dominant in valleys and Cambisols, Acrisols and Leptosols dominant on hills. In the Buberuka highlands, soils are dominated by Cambisols, Nitisols and Leptosols in uphill areas and Histosols and Vertisols in wetland areas (Djimde 1988; Niang and Styger 1990). The Simbi sector was selected in the Central Plateau agro-ecological zone to represent a mixed cropping system with dominance of Phaseolus vulgaris, Manihot esculenta, Zea mays together with coffee (Coffea arabica) as a cash crop. Simbi is located at 1,634 m a.s.l with an average temperature of 20 °C. Umurera village (164 households, 1,324 inhabitants) was selected as a representative study site. Umurera village shares much of biophysical and socioeconomic variability with the central agro-ecological zone. Information collected through our own measurement or District official documents (Huye DDP 2007) indicate that population density, farm size, cattle ownership and other socio-economic features are comparable to those reported for the Central Plateau AEZ (Verdoodt 2002; Yamoah et al. 1989). Total rainfall averaged 1,061 and 1,044 mm in 2007 and 2008 respectively. In Buberuka highlands, Kageyo sector was selected to represent the typical farming system with dominance of wheat (Triticum sp.) and Irish potato (Solanum tuberosum). Kageyo is located at 1,736 m a.s.l with an average temperature of 15–16 °C, and average precipitation of 737 and 1,015 mm in 2007 and 2008, respectively. Mutobo village (94 households, 529 inhabitants) was purposely selected as study site because it has similar biophysical and socioeconomic features found in Buberuka (Gicumbi DDP 2007), be it in terms of population density, land use and most socio-economic indicators. In both locations, the periods from September to October and November to December 2007 season were exceptionally dry (Fig. 1).
In the two locations, wealth ranking allowed categorising local households into classes based on local farmer criteria including land size, the number of cattle, the type of house, the ability of the farmer to hire labour (adapted from Grandin 1988). Four farmer groups were identified: a wealthier farmer group, a moderately resourced farmer group, a poor farmer group and a landless farmer group. Wealthier farmers accounted for 2–7 % of the households, moderate farmers 8–30 %, poor farmers 66–84 % and landless farmers 1–2 %. The landless farmer group was not included in the study due to the fact that they had no land which they manage on their own. Table 1 gives an overview of the main socioeconomic characteristics of households at the two sites.
Inventory of current trees grown on farms
Before starting the inventory exercise, it was important to clearly define what “a tree” is. In an earlier study, a tree was defined differently depending on whether one uses the western or the Rwandan epistemology (den Biggelaar 1994). From the definition given by Kagame (1958 cited in den Biggelaar 1994) the term “tree” is understood as all plants that are not grasses (referred to as Rwandan-Bantu epistemology). The definition clearly differs from the western conception of a “tree” that only encompasses trees and shrubs. In the current study, we considered “a tree” based on the western epistemology, meaning woody and shrub vegetation excluding herbaceous species.
A formal survey was conducted with 65 farms in Simbi and 73 farms in Kageyo to identify which, where and to what extent different types of trees are currently grown on farm. The data were gathered separately for woodlots and croplands on individual farm types using a pre-tested and pre-coded questionnaire. Data included household characteristics such as farm identification and location, household status, education level, land area, the type and number of animals reared, and source of firewood. The second set of data related to farmer’s preferences for specific species and their management. Since the local language (Kinyarwanda) was used during the interview, tree names were given in local names and translated into scientific names. Names were cross-checked with a tree expert from ISAR/Agroforestry Department. The frequencies of the presence of tree species were recorded and used a proxy for identifying the most preferred tree species that were selected afterward for tree testing. Species richness (i.e. the total number of trees species on farm and tree density (the total number of trees per unit area) were recorded.
Testing farmers’ preferences
Farmers categorized into the three wealth categories were listed and 25 farmers per wealth group were selected based on a systematic sampling procedure by picking every second farmer on the list of farmers belonging to each wealth category. The trial was discontinued on 5 farms in Simbi and 3 farms in Kageyo due to various reasons including death, or farmers who had not planted any tree. A tree evaluation exercise was finally conducted with 20 farmers in Simbi and 22 farmers in Kageyo. Two species belonging to each of the most important tree classes were selected: timber trees (Eucalyptus urophyla, Grevillea robusta), legume shrubs (Calliandra calothyrsus and Tephrosia vogelii) and fruit trees (Persea americana and Citrus sinensis). Tree seedlings were obtained from the agroforestry nursery of the Rwanda Agricultural Research Institute, ISAR). Eucalyptus urophyla seedlings were supplied by ISAR (Rwanda Agricultural Research Institute/Forestry and Agroforestry Department). Grevillea, Calliandra and Tephrosia seeds were from Gisagara provenance (Southern Rwanda) and seedlings were produced by ISAR (Forestry and Agroforestry Department). Grafted fruit tree seedlings (Persea americana and Citrus sinensis) were produced and supplied by ISAR/Rubona station (Horticultural Department). A total of 60 trees (10 Eucalyptus urophyla, 10 Grevillea robusta, 10 Calliandra calothyrsus and 10 Tephrosia vogelii, 10 Persea americana and 10 Citrus sinensis.) were made available to each farmer for planting. A total of 2,520 tree seedlings were distributed across the two locations. Seedlings were 15–25 cm height at planting time. Farmers were free to choose which tree species to plant and where to plant them. Before planting, best tree planting practices were discussed. Farmers were advised to plant in pits of about 40 × 40 × 40 cm and apply manure and watering regularly for best results. Tree seedlings were planted at the start of the rainy season in September 2007. Trees, especially those planted on contours and home fields were weeded when this was done for adjacent crops. Fruit trees were mostly planted under banana crops near home compounds and were mulched. Some farmers watered trees at planting when a drought occurred. The chronological sequence of different farmer activities is provided in Fig. 1.
The number of trees effectively planted by each farmer was recorded after planting by counting the number of planted trees and expressing this as a percentage of the trees the farmer had received. Management practices were recorded and expressed in percentage of farmers that had conducted primary management practices for individual tree species. Height measurement was done using measuring poles. The tree survival rate and height were assessed at 4, 8 and 12 MAP (months after planting) in different tree niches on different farms. Only data at 12 MAP are reported. Assessment of productivity was limited to Eucalyptus urophyla, Grevillea robusta, Calliandra calothyrsus and Tephrosia vogelii since there was no fruit production recorded at 12 MAP. Tree productivity was expressed in terms of dry biomass of above-ground prunings, including leaves and twigs or sticks of or less than 2 m length. Tree species were carefully pruned and the fresh biomass was determined at 12 months after planting on a sample of 10 trees selected on each farm type and in each niche. Eucalyptus trees planted in woodlots were pruned and dry matter reported per unit area. For the tree species planted along contours or along paths, productivity per unit area was obtained by estimating the total biomass on 100 m contour length and squaring to estimate biomass on a per ha basis. To determine biomass dry matter content, a 1 kg sample of fresh leafy and twigs parts was collected for each species from the different farms and the average dry matter content determined after oven-drying at 103 °C to constant weight and weighted for dry matter content (Anderson and Ingram 1993).
A farmer evaluation was conducted through an inventory of the problems encountered during the tree testing exercise using a formal survey. The questionnaire used was designed after a focus group discussion with participant farmers. Farmers also evaluated the trees for a range of attributes. For this, a focus group discussion was conducted with farmers involved in the study together with randomly selected tree users (carpenters and charcoal makers) to identify key criteria farmers considered important for tree evaluation. Sampled farmers included a broad range of farmers: wealthier, moderate and poor farmers with both household sex groups fairly represented. Female households were 30–40 % of the participants. A total of 70–80 farmers and other tree users were involved at each study location. Farmers used different criteria for different tree species. For timber species, criteria were the ability of the tree species to provide poles, straightness, tree diameter, compatibility with other crops and coppicing ability. For legume species, the palatability for livestock, the ability to supply poles, the ability to coppice and the compatibility with other crops were the most important criteria for both locations. Other criteria were specific to sites. For instance, the durability of fire (the ability of firewood to keep burning for longer period), was an important criterion for the evaluation of timber species while the ability to contribute to soil fertility improvement was an additional important criterion to evaluate legume species in Simbi. For fruit trees, farmers focused on branching ability, adaptability to the site and growth vigour. Fruit trees were also assessed based on the early growth performance. Based on these criteria, an evaluation sheet was designed and only farmers who had planted trees as part of the study were asked to assess tree species using a scoring technique (Franzel 2001). The technique involves moving seeds or stones among pockets to score tree species on a scale of 1–5. In addition, an informal survey helped to assess the farmers’ future plans for agroforestry.
Data on the number of tree species, total number of trees per farm and per unit area basis were subjected to ANOVA using the mixed model procedure with site, farm type and farm location as fixed factors and farm (site) as the random factor in the Genstat statistical package (GENSTAT 2009). Data on the number of trees planted expressed as percentage of the total trees received per species, tree management activities, growth and productivity and farmers’ evaluation were presented as means over sites or tree species as no clear relationship with farmer resource status could be found.