Introduction

Increasing demand for food has caused natural resource degradation in many regions of the world as farming practices often reduce soil quality, biodiversity and ecosystem services (Power 2010). The problem is particularly important in Sub-Saharan African countries due to the high dependency on natural resources and limited access to mechanised farming, fertiliser, and irrigation. This calls for a viable alternative to sustain or increase agricultural productivity of smallholder farms in those countries (Jose 2009; Nair 2007).

In the Central Rift Valley (CRV) of Ethiopia, farmers grow annual crops in agroforestry parkland systems where naturally regenerated and scattered trees occur along with crops (Siriri et al. 2010). Faidherbia albida is one of the main tree species in these parkland systems. Farmers have retained F. albida in order to improve soil fertility, microclimate, and crop yield, and to provide branches for fencing and firewood. Competition for light, water, and nutrients can be minimised under F. albida trees that exhibit reverse phenology, i.e. when the tree sheds its leaves during the wetter crop growing season and regrows them between seasons during drier conditions. However, this phenomenon can be disrupted by factors such as climatic conditions in areas that have transitional single and double rainfall seasons, by drought (Boffa 1999) and by pruning.

Above- and below-ground competition in agroforestry systems can be minimised by crown pruning (Semwal et al. 2002). For example, crown pruning of alnus (Alnus acuminate) and calliandra (Calliandra calothyrsus) was required to sustain bean and maize production in agroforestry systems of Uganda (Siriri et al. 2010). Either under or at a distance from trees, crop production can be improved by the application of fertilisers.

Studies that involve field experiments on F. albida trees in parklands of the CRV have demonstrated positive effects of trees on crop yields, soil fertility, and microclimate. For instance, F. albida trees scattered in crop fields apparently improved some soil properties under their canopy as compared to adjacent open plots (Kamara and Haque 1992). However, no information is available in these studies on the effects on crop productivity of farm management practices such as pruning of F. albida tree crowns and the application of fertilisers, which are two common options currently used by farmers in this parkland system. The objective of this study was to determine the impacts of crown pruning and fertiliser application on maize production under F. albida in a parkland agroforestry system in Ethiopia. Research questions addressed in the study were (a) are there interactive effects of tree shade and fertiliser applications on maize yield, and (b) how are these interactions affected by crown pruning?

Materials and methods

Study site description

The study was carried out in the parklands of the CRV of Ethiopia at Adulala watershed, which is located approximately 104 km south-east of the capital city, Addis Ababa. Adulala watershed is situated at 8°29.5′N latitude, 39°20.5′E longitude and has an average elevation of 1,688 m above sea level.

The climate of the area is characterized by a bimodal rainfall distribution with a mean annual rainfall of 820 mm. The short rainy season lasts from March to May and the long rainy season extends from June to October. During the field study, annual rainfall was 482 mm for 2015 and 1103 mm for 2016, which included the two cropping seasons (approximately June-November) used in this research. Annual mean monthly minimum and maximum temperatures were 13.9 °C and 28.5 °C.

Soil in the study area is classified as a Fluvisol (Goma 2015). The top 20 cm layer of the cultivated soil is characterised by a pH of 7.6 with total N 0.056%, available P 19 ppm (Olsen method), organic matter 4.3%, and available water holding capacity 155 mg g−1 (Goma 2015).

Natural vegetation type of the Adulala watershed is Acacia woodland, dominated by tree species such as Acacia tortilis, A. seyal and Faidherbia albida (Argaw et al. 1999; Endale et al. 2017). Crop production is mainly rainfed, and F. albida is the main agroforestry tree species in the fields of major crops such as teff (Eragrostis teff), maize (Zea mays) and wheat (Triticum aestivum L. var aestivum). Diammonium phosphate and urea fertilisers are usually applied to teff, whereas compost and manure are applied to other crops.

Experimental design and crop establishment

A total of eighteen F. albida trees were randomly selected in farmers’ fields in the Adulala watershed. The trees were scattered across several adjacent farms; within a total area of approximately 91 ha. An on-farm experiment was conducted during the growing seasons of 2015 and 2016 as a split-plot design with six replications. Main plot treatments in each replicate included three levels of tree crown pruning, i.e., unpruned, 50% pruned and 100% pruned (Fig. 1), and a crop-only plot that was located about 30 m from any tree trunk. Applications of N and P fertiliser were included as sub-plot treatments. Tree crown pruning was done by cutting tree branches at 15–20 cm from their base. The 50% pruning was done by removing branches in the bottom half of the crown length, and the 100% pruning was done by cutting all the branches (Fig. 1). Pruning of the experimental trees was done in May 2015 during the start of the experiment and newly grown branches after the first pruning were removed prior to the second season (May 2016).

Fig. 1
figure 1

Examples of tree crown pruning levels, unpruned (a), 50% pruned (b) and 100% pruned (c), of F. albida used in the experiment at Adulala, CRV, Ethiopia

Full size image

A circular plot of 12-m diameter under each experimental tree was divided into four sub-plots (four quarters of the main plot that were oriented randomly). Urea (69 kg N ha−1) was added to one sub-plot, di-ammonium phosphate (23 kg P ha−1 and 9 kg N ha−1) to the second sub-plot, both urea and di-ammonium phosphate were added to the third sub-plot (78 kg N ha−1 and 23 kg P ha−1) and the fourth sub-plot was left as a control (no fertiliser applied). The four fertiliser treatments (Table 1) were allocated at random to the sub-plots under each tree and crop only plots. Fertilizers were applied by broadcasting either at sowing or after sowing, as below. Maize (Melkasa-2 variety) was sown in the third week of May during the 2015 and 2016 cropping seasons. Seeds were sown at a spacing of 0.75 m between and 0.30 m within rows in each sub-plot under each tree (a total of 18 trees, six trees per pruning level) and in the crop-only plots. DAP was applied only at sowing while urea was applied in two stages with different rates; the full rate (46 kg N ha−1) first at sowing and half the rate (23 kg N ha−1) 10 days after sowing (Table 1). Weeding was done manually in all the sub-plots every 2 weeks.

Table 1 Rates and timing of urea and DAP fertilisation in the four fertiliser treatments
Full size table

Measurements and data collection

Tree phenology

In order to estimate the relative amounts of leaves retained by tree crowns during the growing seasons, foliation (as a proportion of maximum foliation) of the unpruned trees of each treatment was observed and recorded. Maximum foliation was subjectively estimated as how much foliation there would be in a 100% foliated tree. The observation was made twice per month (May to October), in the first and third week of each month.

Light

Photosynthetically active radiation (PAR, µmol s−1 m−2) was measured under a total of nine trees (3 randomly selected trees from each of the unpruned, 50% pruned and 100% pruned trees) at different positions from the tree trunks (positions): 1, 3 and 5 m for the 0–2, 2–4, and 4–6 m positions, respectively, and in each crop-only plot. Measurements were located at the centre of each crop-only plot; and at four aspects (north, south, east and west) around each tree to provide an average value for each position. Measurements were taken after sowing at different times of the day: approximately 9:00 AM, 10:30 AM, 12:00 PM, 1:30 PM, 3:00 PM and 4:00 PM using a PAR sensor (AccuPAR model LP-80, Decagon Devices). PAR under each tree was measured for three consecutive days; the measurement under all trees was carried out between 1st June and 1st July 2016.

Soil nutrients analysis

In May 2015 prior to sowing, soil samples were taken under six randomly selected trees at two points in each tree position at depths of 0–20 cm, 20–40 cm, 40–60 cm and 60–80 cm using a core sampler. The two samples for each combination of tree position and depth were mixed to make a total of 12 composite soil samples from under each tree. At the same depths, soil samples were also taken at two randomly selected points from each of the six crop-only plots and combined to make a total of four composite samples per crop-only plot. Soil samples were air dried and analysed for organic carbon (Walkley & Black), total nitrogen (Kjeldahl) and available phosphorus (Olsen).

Soil moisture

Gravimetric water content was measured in the soil that had been sampled from three randomly selected trees from each pruning treatment (unpruned, 50% and 100% pruned trees) and crop-only plots during the growing season of 2015 at the flowering stage and in 2016 at the sowing, flowering, and physiological maturity stages. The soil samples were collected at two points in each tree position and at randomly selected two points in the crop-only plots at depths of 0–20 cm, 20–40 cm, 40–60 cm and 60–80 cm. The two soil samples for each tree position and depth as well as the samples for each crop-only plot and depth were bulked and oven dried for 24 h at 105 °C.

Crop yield and biomass

Maize total above-ground biomass and grain yield were determined by manual harvesting of all the plants from 1-m2 quadrats located randomly in all replicates of sub-plots under each tree position and crop-only plot. Harvesting was done at maturity, between the first and the third weeks of November 2015 and 2016. Grain moisture was measured using an electronic moisture tester, and grain yield was adjusted to 12% moisture content.

Statistical analysis

Differences between treatments for each parameter measured (light intensity, soil properties, crop biomass and yield) were analysed using the mixed procedure of SAS 9.4 (SAS Institute 2013). Using a split-plot design, pruning intensity, position from tree trunk and fertiliser applications were considered as fixed treatment effects while individual trees and open-plots were treated as random effects. Years were analysed separately. Significant differences (at p < 0.050) between treatment means were determined by the Tukey multiple range test.

Results

Tree phenology

Unpruned trees maintained 35–60% of maximum foliation during each growing season (Fig. 2). The pattern was similar in both seasons, increasing from May until a peak of foliage maintained in August, which then declined. Thus, foliation followed a similar pattern to rainfall in the region.

Fig. 2
figure 2

Relative foliation of unpruned trees in 2015 and 2016 growing seasons in parklands at Adulala, CRV, Ethiopia (n = 6). Relative foliation is defined as the leaves retained by trees relative to those when fully foliated. Bars indicate standard error

Full size image

Light transmission

The amount of PAR received in the open (control) plots (52.7 ± 0.04 mol m−2 day−1) was significantly greater (p = 0.001) compared to PAR under tree crowns, which was reduced by 34%, 23% and 9% under unpruned, 50% and 100% pruned trees, respectively (Fig. 3). The amount of PAR increased with pruning and distance from the base of trees to the outer position. For example, there was an increase of PAR from 30.0 mol m−2 day−1 at the 0–2 m position of unpruned trees to 47.1 mol m−2 day−1 at the 4–6 m position of the 100% pruned trees (Fig. 3).

Fig. 3
figure 3

Daily integrated radiation as affected by pruning and position from tree, and in the control (crop-only) plots in the parkland at Adulula. The LSD bar is based on the pruning-by-position interaction

Full size image

Soil nutrient status under tree canopies

The effect of position from F. albida trees on soil properties was significant for organic carbon, available P, and total N (p = 0.001), with greater values in at least one position under trees relative to open areas down to 60 cm depth (Table 2). The difference in soil nutrient properties in the upper 0–20 cm depth among tree positions was not significant for N or P; however organic carbon was significantly greater under the 0–2 m position (1.7 ± 0.06) relative to the 2–4 and 4–6 m positions. All soil nutrients at this soil depth were greater under tree canopies compared to open areas (Table 2).

Table 2 Mean (n = 6) soil total nitrogen (N), extractable phosphorous (P), and organic carbon (OC) concentrations in relation to position from F. albida trees and soil depth in the parkland agroforestry system at Adulula
Full size table

Soil moisture

Soil moisture content ranged from 10 to 20% and generally decreased with depth (Fig. 4). Soil moisture content at the flowering stage (in 2015 and 2016) and at harvest in 2016 was significantly greater under trees compared to open areas, regardless of pruning levels for the 0–40 cm soil depth (Fig. 4, data shown for soil moisture at flowering stage in 2015 and 2016). There was no significant difference at 40–80 cm soil depth. Pruning did not significantly affect soil moisture content at any depth when measured during either growing season.

Fig. 4
figure 4

Gravimetric soil moisture content in relation to tree position and depth at flowering in 2015 (a) and 2016 (b). The error bars are standard error bars

Full size image

Crop yield and biomass production

Crop yield and biomass were significantly greater under trees (2–4 and 4–6 m positions, p = 0.001) compared to crop-only plots (Fig. 5, Table 3) in both growing seasons, regardless of pruning level. Yield increases in the outer position (4–6 m) were relatively larger (45–51% in 2015 and 75–76% in 2016) than the innermost positions (0–2 m) (2% in 2015 and 12% in 2016). Tree pruning significantly increased crop yield only under the 2–4 m tree position in 2015 (Fig. 5c). However, pruning significantly increased crop yield under all positions (0–2, 2–4 and 4–6 m) in 2016 (Fig. 5d), with the lowest yield (2253 kg ha−1) obtained under unpruned trees (0–2 m position).

Fig. 5
figure 5

Crop biomass as affected by pruning and position from tree trunk in 2015 (a) and 2016 (b), and crop yield under those trees in 2015 (c) and 2016 (d)

Full size image
Table 3 Tests of ANOVA fixed effects on maize yield for 2015 and 2016 growing seasons
Full size table

Fertilisation with urea and DAP significantly increased yields under trees (2–4 and 4–6 m) relative to crop only-plots in both 2015 (p = 0.055) and 2016 (p = 0.001). The highest yield was obtained in 2016 from the Urea + DAP fertiliser combination (78 kg N/ha and 23 kg P/ha) under the 4–6 m tree position (Fig. 6).

Fig. 6
figure 6

Crop yield under trees as affected by fertiliser treatment and position from tree trunk in 2016

Full size image

Discussion

Reverse phenology of F. albida trees is promoted as a major advantage of growing this tree species with crops (Hadgu et al. 2009; Kho et al. 2001). However, close observation of their canopies during this study indicated that trees did not totally defoliate during the cropping season, maintaining 35–60% of maximum foliation. Lack of defoliation during the cropping season can be attributed to phenological disruption due to canopy pruning (Barnes and Fagg 2003). Farmers in the study area totally prune tree branches (pollarding) at intervals of 3–4 years such as recommended by Orwa et al. (2009). Reducing total canopy volume by about 35% before the onset of the rainy season, i.e. moderate pruning, might not appreciably reduce tree growth (Boffa 1999).

Results of this study demonstrate that canopy pruning results in increased PAR transmitted under trees. Lower crop biomass and yield recorded under unpruned trees (particularly under 0–2 m position) relative to the totally pruned and 50% pruned trees could partially be attributed to reduced PAR available to crops in the absence of leaf fall during the cropping seasons. The result is in agreement with Jama and Getahun (1991), who speculated that low maize yields under F. albida trees compared to crop-only plots could be the result of shading by the trees. However, the trees in that study were much younger (5 years old) than those studied in the present research (probably 35–45 years old), suggesting that the benefits associated with mature trees may take decades to develop. Faidherbia albida trees need 20–40 years to grow to a size that can significantly improve yields of understory crops (Poschen 1986). A study conducted by Suresh and Rao (1998) in a semiarid India region showed that PAR intercepted under F. albida trees was facilitated by leaf shedding (defoliation) during the crop growing period that in-turn lead to the highest sorghum (Sorghum bicolor) yield being obtained under this tree species compared to other nitrogen fixing trees that retained leaves during the cropping season such as Acacia ferruginea and Albizia lebbeck.

Significantly greater maize biomass and yield were found under F. albida tree crowns 2–6 m from the trunk compared to crop-only plots (Fig. 5). Lower or lack of yield benefit near to tree trunks (under 0–2 m positions) may be attributed to competition for resources between crops and trees in that position (Jose et al. 2000). Greater yield under trees, particularly in the 2–6 m positions could be attributed to improved soil nutrient concentrations and moisture levels associated with greater organic matter concentrations under tree canopies than in the open fields away from trees that outweighed any negative effects of shading. A similar study in the Hararghe highlands of eastern Ethiopia reported that grain yields of sorghum and maize grown under F. albida trees were increased by 76% and 36%, respectively, compared to crops grown in the open fields away from trees (Poschen 1986). The study asserted that the increase in crop yield was a result of improved soil chemical and physical conditions under tree canopies (Poschen 1986). In West Africa, a study in Niger showed that millet yield under the F. albida tree canopy was about 36% greater than the yield obtained from open fields (Kho et al. 2001). Although such cases of positive effects of F. albida tree on crop yields and biomass are well documented (Chamshama et al. 1998; Vandenbeldt and Williams 1992), some studies reported negative or no effects on some crops. In Cameroon for instance, Harmand and Njiti (1992) reported a reduction in yields of groundnut (Arachis hypogaea L.), cotton (Gossypium sp.) and sorghum by 34%, 11% and 40%, respectively. Such differences in F. albida effects on crops can be expected because of the variations and complex interactions of factors such as tree age, soil characteristics, water regime and climate (Barnes and Fagg 2003).

Fertilizer application increased crop yield on plots under tree canopies and on crop-only fields; maize yield increase in 2016 (with high rainfall) was bigger than the increase in yield in 2015 (with low rainfall). This result is consistent with nutrients being relatively less limiting than water during dry seasons (Kho et al. 2001). In the same study, Kho and colleagues reported a 36% increase in dry matter production of pearl millet (Pennisetum glaucum) under tree canopies compared to open crop-only plots. However, no increase due to trees was observed with a high rate of N fertiliser (180 N kg ha−1, Kho et al. 2001). This result suggested that the effect of F. albida on crop production is more pronounced in conditions of low soil fertility (Sileshi 2016) as nutrients are less limiting to crops at greater fertility levels. In the present study, recommended rates of fertilisers under tree canopies led to greater yield increase in 2016 (with high or average rainfall) and maize response to fertiliser was larger under the trees than in the crop-only plots. This result could be due to greater availability of water and soil nutrients other than N and P (e.g. K) under F. albida trees that were therefore less limiting to growth and thereby enabled a larger response to N and P fertiliser (Hadgu et al. 2009; Kamara and Haque 1992). This complexity is symptomatic of the response of crops to fertilisers being determined by a range of soil physical and chemical conditions (Baligar et al. 2001; Tittonell et al. 2008). Also in the present study, we note that low N and P availability limited maize growth, and it is unlikely that the N and P fertiliser rates used would have maximised growth.

Studies in Ethiopia report that concentrations of organic carbon, available P and total N decrease with increasing distance from F. albida trees and with increasing soil depth (Hadgu et al. 2009; Kamara and Haque 1992). The present study indicated that the values for those soil parameters were greater for 0–40 cm depth under trees relative to open plots and the differences for deeper 40–80 cm depth were not significant, with the exception of organic carbon, which was significantly greater under trees to a depth of 60 cm. These results are in agreement with Saka et al. (1994), who found greater levels of surface (0–15 cm) organic matter, N and Ca under tree canopies than in open areas, and no significant difference at the 30–45 cm depth. Similarly, in the Ethiopian highland Vertsols, Kamara and Haque (1992) reported greater concentrations of soil organic matter, total N, available P and exchangeable K under F. albida trees than in open fields away from trees. A study on other F. albida based systems of northern Ethiopia also revealed that soil organic matter, total N and available P were greater under tree canopies than outside canopies (Hadgu et al. 2009). Comparable results were also reported in other studies (Kho et al. 2001; Rhoades 1995; Umar et al. 2013). The mechanisms that improve soil nutrient concentrations under F. albida have been widely debated (Barnes and Fagg 2003). Some studies suggest that the factors for increased soil fertility under F. albida include deep capture and recycling of nutrients, improvement in soil biological activity, and symbiotic and asymbiotic N fixation (Rao et al. 1997; Rhoades 1995; Umar et al. 2013). Other authors have argued that improved soil fertility under tree canopies is due to lateral redistribution of nutrients, by domestic and wild animals including birds, tree roots, and wind erosion, or due to pre-existing greater soil fertility conditions favoured by establishing tree seedlings (Geiger et al. 1994).

An important effect of trees relates to soil moisture. This study show that at the flowering and maturity stages of maize, soil moisture in the top 40 cm depth was 22% and 24% greater than in the open field, respectively. Similarly, Rhoades (1995) reported increased soil moisture to a depth of 15 cm under F. albida trees compared to open areas but found no significant difference at 15–30 cm depth. Increased levels of soil moisture under tree canopies compared to open fields can be attributed several processes: (1) F. albida tree roots can take up water from deep in some soil profiles (Dupuy and Dreyfus 1992; Roupsard et al. 1999) resulting in hydraulic lift (movement of water from deeper to shallower soil layers) (Bayala et al. 2014); (2) reduced evapotranspiration compared to crop-only conditions; and (3) increases in soil organic matter that improve soil water holding capacity and moisture availability (Lal 2006; Sileshi 2016). For example, Makumba et al. (2006) reported greater (50%) soil moisture retention in gliricidia–maize than in sole maize because of a 65% increase in soil organic matter in the gliricidia–maize intercropping system in Southern Malawi.

Tree density in the CRV of Ethiopia is sparse, about 4.2 trees per hectare (Sida et al. 2018a), due to browsing by free-grazing livestock that kills naturally regenerated or planted tree seedlings (Endale et al. 2017). In addition, there is a preference by small-holder farmers to allocate their land to food crops in order to maximise food production for home consumption or sale (Abebe et al. 2013). Increasing tree population density in the study area could be a viable tree management strategy to enhance crop production in the long-term, as maize yield and biomass were greater under tree crowns compared to crop only plots. The results of the present study suggest that maize productivity under F. albida trees could be further improved by pruning and the application of fertilisers, particularly during high or average rainfall seasons. However, use of chemical fertilisers for smallholder maize production is limited by the high cost of fertilisers and the associated financial risk (Bacha et al. 2001).

Field experiments, conducted at particular points in time and space, can provide valuable information on system performance, but they are expensive, time consuming and can be of limited value in the quantitative transfer of experiences between sites. Alternatively, simulation models offer a means of quantitatively integrating complexity to evaluate system performance (Kassie et al. 2014; Smethurst et al. 2017) and to conduct virtual experiments (Luedeling et al. 2016). Therefore, we recommend that further agroforestry research in the parklands of Ethiopia includes simulation of biophysical productivity and socio-economic factors, e.g. tree-crop interactions at the plot scale, and fertilizers, manures, wood, fuel, livestock, labour, and other socio-economic factors at the farm scale. At the plot scale, the APSIM (Agricultural Production Systems sIMulator) tree-crop daily simulation model could potentially be adapted to parkland systems as it can integrate a range of biophysical factors including water, N and microclimate (Dilla et al. 2017; Sida et al. 2018b; Smethurst et al. 2017). The biophysical outputs of APSIM could then be used as inputs to farm-scale livelihood modelling options such as Simile (Muetzelfeldt and Massheder 2003), APSFarm (Rodriguez et al. 2006) and Farm-Safe (Graves et al. 2011).