Introduction

For the past three decades, the integration of fast-growing timber trees in smallholder farming systems in the Philippines has been extensively promoted to diversify farm output and produce timber for household use and sale. As a result, trees planted on farms are today an important source of raw materials for the local timber industry, and income for smallholders.

One of the unique advantages that smallholders have in tree production is the practice of intercropping: the continuous land cultivation, weeding and fertilization for crops improves tree survival and promote faster tree growth by preventing weed infestation and improving site conditions (Garrity et al. 1996; Kapp and Beer 1995). In Mindanao, Philippines diameter at breast height (dbh) and total height of 2-year-old intercropped Falcata (Paraserianthes falcataria (L.) I. C. Nielsen) were, respectively, 33 and 21% greater than non-intercropped trees (Nissen et al. 2001). Growth of associated crops may also benefit by the presence of trees which reduce weed invasion and growth (Gajaseni and Jordan 1992). Miah (1993) reported that weed infestation and weed dry matter yield in an upland rice (Oryza sativa L.)-tree association were respectively 30–38% lower than in the sole rice plots.

Planting trees and crops in association can also produce direct financial benefits. In Latin America, it has been estimated that the costs of soil preparation, weeding, and pest and fire control were 51–68% lower in an intercropping system than in pure reforestation plantings (Beer et al. 2000). Nissen et al. (2001) found that in the first 2 years after planting management costs of intercropped Falcata were less than half the costs of Falcata monocultures. For all these reasons, the century-old system of Taungya reforestation in which intercropping is practiced during the first few years after tree planting, is a popular strategy for tree establishment and survival, to reduce reforestation costs and to produce timber for on-farm use and sale (Jordan et al. 1992).

In spite of the above advantages, there is substantial evidence that competition effects in intercropping systems may reduce or override overall productivity gains and financial returns compared with tree monocultures. When fast-growing timber trees are combined with light-demanding annual crops, the growth of the understorey crop could be inhibited as a result of competition between trees and crops for both above- and below-ground resources (Ong et al. 1996). With few exceptions, the common timber tree species promoted for farm forestry have been reported to depress yields of those associated crops which are generally cultivated under full sunlight. The genetic potential of trees for rapid growth makes them more ‘aggressive’ and hence successful competitors for site resources. In Guatemala, for example, 4 years after planting trees at 3 × 2 m, the yields of maize (Zea mays L.) and green bean (Phaseolus vulgaris L.) intercrops were reduced by 35% by Casuarina equisetifolia J. R. & G. Forst, 83% by Eucalyptus globulus Labill and 91% by Alnus acuminata Kunth compared to the first year crop (Leiva and Borel 1994). In Uganda, Okorio et al. (1994) found that of 17 timber trees intercropped with maize and beans, only A. acuminata did not have a negative effect on annual crop yields, probably because of minimal shading and nitrogen-rich litter. Interestingly, A. acuminata was the most competitive species in the Guatemala study (Leiva and Borel 1994). Across five seasons, the maximum average reduction in annual crop yield was 60%. In India several studies quantified the substantial decline of annual crop production due to intercropping with eucalypt trees (Ahmed 1989; Malik and Sharma 1990; Saxena 1991). Consequently serious concerns have been raised over the sustainability and appropriateness of tree farming for resource-poor farmers (Shiva and Bandyopadhyay 1987).

When water and nutrients are freely available, as in areas in the wet tropics with well-distributed rainfall and where fertilizers are commonly used, light availability may be the most important limitation to production of understorey annual crops (Ong et al. 1996). Branch pruning is effective in reducing light interception by the tree canopy, and thus prolonging the period of intercropping (Watanabe 1992). Miah (1993) found that the yields of rice and mungbean [Vigna radiata (L.) R. Wilczek] planted in alleys between lines of severely pruned multipurpose trees [Gliricidia sepium (Jacq.) Walp., Acacia auriculiformis A. Cunn. ex Benth., and Acacia mangium Willd.] were comparable with those of the sole crop plot. In a hedgerow agroforestry system with Gmelina (Gmelina arborea R.Br.) planted at 1 × 6 m, the grain yield of rice in association with severely pruned trees increased three-fold over the yield in the unpruned plot (Gonzal 1994). Thus, in the Philippines, farmers often practice severe branch pruning every season before the planting of crops to reduce tree–crop competition as well as to improve tree shape (Bertomeu 2004). In Indonesia, some small-scale timber farmers start severe branch pruning (retaining live crown ratios of 40% or less) at 6 months to reduce tree-annual crop competition, improve tree shape, and reduce wind damage to trees (Roshetko et al. 2004).

While intensive pruning benefits the understorey crops, the practice may reduce the profitability of tree farming as it slows tree growth (Smith 1962), reducing tree dbh and final timber yield and resulting in lower timber revenue. Miah (1993) reported that the total biomass of intensively pruned 2-year-old trees was 34% smaller than that of unpruned trees. Gonzal (1994) found that 2-year-old intensively pruned trees had a significantly smaller stem dbh (7.38 cm) than unpruned trees (9.83 cm).

Farmers instinctively anticipate crop yield losses as trees grow, and positive crop increases with severe pruning. However, it is unlikely that they are able to accurately predict the period of profitable intercropping and the net financial returns of alternative management regimes across a full tree rotation. On-farm trials were initiated in Claveria, Philippines, to investigate the effect of several pruning regimes on tree growth and maize yield and their implications for farmers in terms of financial returns. These trials were part of a larger study undertaken to examine the appropriateness, profitability and technical feasibility of planting timber trees at wide spacing in smallholder farming systems (Bertomeu 2004; Bertomeu 2006).

Materials and methods

The study site

The field research was conducted in Claveria, an upland municipality of the Philippines located 42 km northeast of Cagayan de Oro City, in northern Mindanao. The municipality covers an area of 112,175 ha, and has a mountainous (390–2,000 m.a.s.l.) topography with 62% of the area having slopes >18%. Soils are derived from volcanic parent material and classified as deep acidic oxisols with pH of 3.9–5.2 and texture ranging from clay to silty clay loams, with low available P, low cation exchange capacity (CEC), high Al saturation and low exchangeable K (Magbanua and Garrity 1988). The average rainfall is 2,500 mm with the wet season from June to December (>200 mm rainfall per month) and a short dry season from March to April (<100 mm rainfall per month) (Kenmore and Flinn 1987). Temperatures vary little throughout the year, with an average monthly maximum of 28.6°C and average monthly minimum of 21.3°C.

The average farm size in Claveria is 2.5–3 ha, with farmers commonly cultivating two or more parcels of land. At lower elevations (400–700 m.a.s.l.), maize is the dominant crop, cultivated twice a year or in rotation with cassava (Manihot esculenta Crantz) or upland rice. Typically, a crop planted at the onset of the rainy season (May) is followed by a dry season crop planted in September or October. Tomatoes and other vegetable cash crops are commonly grown at the higher elevations (700–900 m.a.s.l.).

Research plot set-up and management

The study consisted of on-farm trials with experimental plots laid out in a randomized complete block design with four treatments and four replications, established at two farms (two replications at each site). Both farms were located in the same village (Cabacungan), at the same elevation (around 400 m a.s.l.), with similar slope (20–30%) and orientation (14 degrees north), but differed notably in their land management histories. Before the establishment of the experimental plots, the farm at site 1 had been used for maize cropping and contained 3-year old natural grass strips. The farm at site 2 was pasture land, grazed by goats, with evidence of rill erosion. Before the establishment of trial plots, soil samples from each farm were taken with a soil auger. One composite sample from the upper, middle, and lower part of the slope was derived from several sub-samples. All soil samples were analyzed at the International Rice Research Institute (IRRI) at Los Baños, Philippines. The farms differed in some physical and chemical soil properties (Table 1). The soil at site 1 had a slightly higher pH, lower clay content, and notably greater CEC and exchangeable K, Ca and Mg than the soil at site 2.

Table 1 Physical and chemical soil properties at the trial sites before the start of the experiment in Claveria, Philippines
Full size table

Plots were 300 m2 (15 × 20 m) containing three lines of Gmelina planted at 1 × 10 m, i.e., 1,000 trees per hectare (tph), with 16 trees per line (i.e., 48 trees per plot), and 15 rows of maize planted for six cropping seasons in each of the 10 m-wide alleys. The 10-m tree interrow spacing was chosen based on field observations of agroforestry systems in Claveria where fruit and timber trees are planted widely spaced on contour lines 6–8 m apart. This planting design was also considered most appropriate for smallholder timber production systems by Santiago (1997).

Four pruning regimes were chosen: (a) T 1 (control): retaining a live crown ratio (LCR) (i.e., the percentage of total tree height retaining live branches) of 60–70%; (b) T 2: retaining a LCR of 40–50%; (c) T 3: retaining a LCR of 30–40%; and (d) T 4: retaining a LCR of 20–30%.

In the last week of September and first week of October 1997, Gmelina seedlings were planted in 40 × 40 × 40 cm holes manually cultivated at the trial sites. Dead trees were replaced at the end of December 1997. From January to May 1998, trees were watered twice a month due to severe drought conditions. In July 1998, following the drought, dead seedlings were replaced to maintain homogenous plot conditions. Replacement trees were not included in the calculations of tree growth parameters, except for the assessment of stem shape at the end of the experiment.

Contours of natural grass were established in the research plots by leaving strips of grass unplowed. Trees were planted immediately uphill from the grass strips. Maize cropping commenced in May 1998 and continued for six consecutive cropping seasons with the last harvest in January 2001. Every year, a wet season maize crop was planted in May and harvested in early September, followed by a dry season crop sown in early October and harvested in January. Draught animal power was used for land preparation, consisting of two ploughings and one harrowing operation. Maize fertilizing and weeding were performed manually following local practices. Every cropping season, a hybrid maize variety (Pioneer 3014) was sown into furrows at a spacing of 30 cm along each row and 60 cm between rows. Each maize crop was fertilized with the recommended dose of 80–30–30 kg N–P–K ha−1. Phosphorus (solophos 0–18–0) and potassium (muriate of potash 0–0–60) fertilizer and the insecticide–nematicide furadan 3G were applied at sowing. Maize re-sowing occurred 5–7 days after emergence (DAE). Nitrogen (urea 46–0–0) was applied as equal split doses by side dressing at 15 and 30 DAE. Nitrogen application was followed by interrow cultivation to cover the fertilizer with soil and control weeds. Manual weeding of the maize crop was also conducted as needed, usually one to 2 weeks after the second interrow cultivation at 30 DAE.

Weeding around trees was conducted at planting. Subsequent weed suppression operations were conducted twice per cropping season in the first and second year.

Removal of double stems and form pruning were conducted when trees were 1-year-old to retain a single stem and improve shape. From May 1999 to October 2000, four branch pruning operations were performed before or immediately after the planting of maize. A 50% stem thinning was conducted at 30 months after planting by removing the smaller and suppressed trees.

Data collection and analysis

Maize grain yield data were recorded row by row from a 6 m-wide centred net plot. At harvest, fresh grain and total biomass were measured and two plant samples randomly taken from the five maize rows of each of the upper, middle and lower alley zones. Grain yield at 14% moisture content was obtained after oven-drying the sample.

Tree dbh and total tree height were recorded twice a year from the trees in the net plot (i.e., excluding border trees) until the age of 42 months. The mean annual increment (MAI) in dbh was estimated as the average of the two annual increments in dbh during the period 18–42 months. Analysis of variance (ANOVA) from General Statistic 11 edition (Genstat) program was used to analyse variations of tree dbh increment, maize grain yield and tree height increment across the four pruning regimes and research sites. The least significant difference (LSD) test was used to identify means differences. At the end of the experiment, stem shape was assessed by visual inspection. Trees were rated as: A = trees with crooked or knotty stem; B = trees with medium stem shape; and C = trees with excellent, straight and nearly cylindrical stem shape. This rating was comparable to the three grading categories of Gmelina sawn timber used by local timber traders reported by Bertomeu (2008).

The financial net benefits of the maize–Gmelina agroforestry system under the four pruning regimes were assessed in terms of the land expectation value (LEV) per hectare and the net returns to labour, as this indicator is relevant for labour-constrained farmers and those with off-farm jobs that compete for their labour time.

Returns to labour were estimated as noted by Franzel et al. (2002)

Returns to labour = discounted net benefits to labour (1)/discounted labour days (2)

$$ \left( 1 \right) = \sum\nolimits_{j = 1}^{n} {\left[ {\left( {{\text{B}}_{j} - I_{j} } \right)/\left( {(1 + r)^{j} - 1} \right)} \right]}, $$

where B j  = benefits in year j, j = 1, 2,…, n, and I j  = input costs in year j, j = 1, 2,…, n.

$$ \left( 2 \right) = \sum\nolimits_{j = 1}^{n} {\left[ {{\text{WD}}_{j} /\left( {\left( {1 + r} \right)^{j} - 1} \right)} \right]}, $$

where WD j  = labour work-days.

For each pruning regime, six scenarios are presented by assuming two discount rates (15 and 20%) and three scenarios for the average dbh at harvest: (a) 30, 29, 28 and 27 cm for T 1, T 2, T 3 and T 4, respectively (scenario 1); (b) 30 cm for T 1, 29 cm for T 2 and T 3 and 28 cm for T 4 (scenario 2), and; (c) 30 cm for T 1 and T 2 and 29 cm for T 3, and T 4 (scenario 3). The annual discount rates of 15 and 20% were assumed based on the cost of borrowing capital in the study area and farmers’ perception of the risk of the agroforestry practice.

For each treatment, four pruning operations (one form pruning and three lift pruning) and two thinnings (each at 50% intensity) were considered in the financial calculations (Table 2). Based on experience at the site, pruning labour rates were assumed as: (a) Form pruning: 1-man-day ha−1 for all treatments; (b) First lift pruning: 6-man-day ha−1 for T 1, 8-man-day ha−1 for T 2, 10-man-day ha−1 for T 3 and 13-man-day ha−1 for T 4; (c) Second lift pruning: 4-man-day ha−1 for T 1 (500 tph), 10-man-day ha−1 for T 2 (1,000 tph), 13-man-day ha−1 for T 3 (1,000 tph) and 15-man-day ha−1 for T 4 (1,000 tph); (d) Third pruning: 3-man-day ha−1 for T 1 (250 tph), 7-man-day ha−1 for T 2 (500 tph), 17-man-day ha−1 for T 3 (1,000 tph) and 19-man-day ha−1 for T 4 (1,000 tph).

Table 2 Schedule for maize cropping, tree pruning and tree thinning operations used in the financial calculations
Full size table

A discounted cash flow analysis (DCF) was made assuming an 8-year tree rotation and a final tree–crop of 250 tph. Annual maize yields used in the DCF analysis are those reported in Table 3 which are above the break-even yield of 5 t ha−1 year−1 (3 t ha−1 for the wet season crop and 2 t ha−1 for the dry season crop) found by Bertomeu (2006) for a similar Gmelina–maize system with the same level of inputs and management as in this study. Costs, prices and revenues used in the DCF are also based on Bertomeu (2006).

Table 3 Effect of four pruning levels on maize production (grams per linear meter) at two farms in Claveria, Philippines
Full size table

Results

Maize grain yield

There was a significant difference in maize production between sites, treatments and crops (F prob = 0.048) (Table 3). Maize production was generally greater at site 2, most notably in crop 2 and 4 (dry season crop of the first and second year). This may indicate that during the dry season water was more limiting in site 1 than in site 2. Compared to the first year (crop 1 and 2), maize production in the third year (crop 5 and 6) was substantially reduced at both sites (crop 1 was around 50% greater than crop 5 at both sites, and crop 2 was 3–34% greater at site 1 and 37–54% greater at site 2 than crop 6) due to competition from Gmelina. This reduction in maize production as trees grew occurred in all treatments but was more pronounced in maize under T 1 than under T 4. Differences in grain yield between T 1 and T 4 were clearly significant after the first year (except in crop 6 at site 2). In the second year (crop 3 and 4) maize grain yield under T 4 was around 23–52% greater at site 1 and 20% greater at site 2 than under T 1. In the last year (crop 5 and 6) maize grain yield in T 4 at both sites was 30–40% greater than under T 1 (Table 3).

In each year, the pattern of the maize grain yield across the alley conformed to a bell-shaped curve and maize yields differed significantly (P < 0.05) among the pruning regimes and with the distance from the tree line (Table 4). During the first year, yields under pruning regime T 4 were greatest for all maize rows except for rows 3, 11, 12 and 14. However, only the yield of the first maize row under T 4 (398 g lm−1) was significantly different from that of T 1 (272 g lm−1). In the first year, differences in grain yield between T 4 and T 1 ranged from 5 to 14% in rows 7 and 9 (center of the alley) up to 32% in row 1. Across the alley, in all pruning regimes the grain yield of rows next to the trees (rows 1 and 15) were significantly different from the rows in the middle of the alley (rows 5–10).

Table 4 Effect of pruning on grain yield (grams per linear meter) of maize rows intercropped on a 10-m wide sloping alley between rows of Gmelina arborea (upper tree row at x = 0 m; lower tree row at x = 10 m) in Claveria, Philippines
Full size table

In the second year, as trees grew taller, the bell-shaped pattern of the maize grain yield in the alley became more marked due to competition from Gmelina: maize grain yield of the three rows next to the trees (rows 1, 2, 3, 13, 14 and 15) were 40–90% lower than the grain yields at the center of the alley (maize row 8). The differences between treatments also became more marked and regular. Yields under pruning regime T 4 were greatest for all maize rows except in row 3 (in which the yield under T 4 was equal to the yield under T 2), and rows 5 and 6 (in which yields under T 3 were greatest). Compared to T 1, grain yields under T 4 were 40–50% greater in the two rows next to the tree lines and 25–35% greater in the center of the alley, although only maize rows 2, 9, 11 and 12 differed significantly.

In the third year, the bell-shaped curve of maize grain yield became less pronounced, indicating the dispersion of competition effects across the alley as trees grew taller. Yields under pruning regime T 4 were highest for all maize rows, except for rows 3, 11 and 13 which showed greater yields under pruning regime T 3, but only the grain yield of maize rows 3, 4, 5, 14 and 15 under T 4 differed significantly from the grain yield under T 1.

The analysis of maize grain yield in each pruning treatment and cropping season showed that the wet season (first crop) yield was consistently greater than that of the dry season crop (second crop) (Table 5). In the first year, no statistically significant differences (P < 0.05) in maize grain yields across the treatments was detected. But as trees grew, grain yield under T 4 became significantly greater (P < 0.05) compared to those under T 1. The cumulative difference in grain yield between T 1 and T 4 across the six cropping seasons was 3.58 t ha−1. In the first year, all pruning treatments had annual maize yields greater than the break-even yield of 5 t ha−1 year−1. In the second year, however, only the maize yields under T 3 and T 4 (5.23 and 5.49 t ha−1, respectively) were above the break-even yield.

Table 5 Effect of pruning regime of Gmelina arborea on grain yield of intercropped maize in Claveria, Philippines
Full size table

Tree growth

During the first 18 months after planting, trees grew notably faster at site 1 than at site 2. The average dbh of 18-month-old trees was 4.8 cm for T 1, 4.9 cm for T 2, 4.0 cm for T 3, and 3.9 cm for T 4 at site 1, and 1.1 cm for T 1, 1.2 cm for T 2, 1.1 cm for T 3, and 1.1 cm for T 4 at site 2. However from the second year (month 24) until the end of the experiment (month 42), tree dbh increment was consistently larger at site 2 than at site 1. The difference in dbh increment between sites was statistically significant during the periods 30–36 and 36–42 months (Table 6). The large difference in early tree growth between site 1 and site 2 may be explained by the differences in soil physical and chemical properties, whereas the subsequent larger increment in tree dbh at site 2 may have been due to improved soil physical and chemical conditions as a result of ploughing and fertilization.

Table 6 Mean diameter increment (cm) of Gmelina arborea under the four pruning regimes in Claveria, Philippines
Full size table

Tree dbh increment was highest under pruning regime T 1 and lowest under T 4 (Fig. 1). The effect of pruning on tree dbh increment was statistically significant (P < 0.001) only in site 1 during the 18–24 month and 30–36 month periods. At site 2, trees under T 1 consistently showed greater dbh increment than all other treatments. However, observations were not statistically significant, probably because of variable soil conditions within the site. This assumption is supported by analysis that demonstrates that the site-treatment interaction at site 2 was highly significant. Pairwise comparisons of treatments means showed that the difference in mean diameter increment between T 1 and T 4 was statistically significant in both sites and in all periods except at site 1 during the period of 36–42 month. Comparisons between T 2 and T 4 showed that differences in mean dbh increment were significant only at site 1 during the periods of 18–24 and 30–36 month, and at site 2 during the period of 36–42 month (Table 6).

Fig. 1
figure 1

Mean diameter increment (cm) of trees under four pruning regimes at the two experimental sites in Claveria, Philippines

Full size image

The MAI in dbh was greatest for trees under moderate pruning (T 1). At site 1, MAI in dbh was 4.6 cm year−1 for pruning regime T 1, 4.5 cm year−1 for T 2, 4.1 cm year−1 for T 3 and 3.8 cm year−1 for T 4. At site 2, the MAI in dbh was 5.4 cm year−1 for pruning regime T 1, 4.9 cm year−1 for T 2, 4.8 cm year−1 for T 3 and 4.4 cm year−1 for T 4. Mean maize grain yield was greatest under T 4 pruning regime, with an average difference between T 1 and T 4 of 0.56 ton ha−1 at site 1 and 0.63 ton ha−1 at site 2 (Figs. 2 and 3).

Fig. 2
figure 2

Mean annual increment at diameter breast height (cm) and annual maize grain yield (t ha−1) under four tree pruning regimes at site 1 in Claveria, Philippines

Full size image
Fig. 3
figure 3

Mean annual increment at diameter breast height (cm) and annual maize grain yield (t ha−1) under four tree pruning regimes at site 2 in Claveria, Philippines

Full size image

There was no significant difference in tree height increment among treatments throughout the observation periods. There was, however, a significant difference (P < 0.001) in tree height increment between sites, with more growth at site 1 than at site 2, probably as a result of differences in soil properties. Mean height increment ranged from 3.4 to 4 m year−1 at site 1 and 2.7–3.1 m year−1 at site 2. At the end of the experiment, the average tree height was 14.0 m for T 1, 14.1 m for T 2, 12.7 m for T 3 and 11.8 m for T 4 at site 1, and 9.8 m for T 1, 9.6 m for T 2, 10.9 m for T 3 and 11.0 m for T 4 at site 2.

No significant difference was found in stem shape between treatments. About 50% of the trees assessed in each treatment presented crooked or knotty stems, around 46–47% presented medium stem shape and only 3–4% were rated as excellent in shape.

The results of the financial assessment showed that for a 15% discount rate, moderate tree pruning regimes (T 1 and T 2) were more profitable than high pruning regimes (T 3 and T 4) if the difference in average dbh at the end of the rotation was 2 cm (11% difference in timber yield) (Table 7). However, in all scenarios the pruning regime T 4 showed the most returns to labour, indicating that greater maize yields as a result of reduced crop suppression effects could compensate for lower timber yields. The returns to labour of T 1 (at a 15% discount rate) would be equal to that of T 4, only if the average dbh at harvest of T 4 was 24 cm (a difference of 6 cm), which is equivalent to a timber yield of 50 m3 ha−1. The results of this study, however, did not indicate such a large difference in dbh between trees under T 1 and T 4.

Table 7 Returns to land and labour of agroforestry with Gmelina arborea and maize intercropped across an 8-year tree rotation under four pruning regimes and two timber yield scenarios in Claveria, Philippines
Full size table

Discussion

These on-farm trials revealed that intensive pruning was an effective practice to increase the yield of maize intercropped with Gmelina and prolong the period of profitable intercropping. However, intensive pruning also slowed tree growth, resulting in reduced diameter growth and lower timber yields. These results were consistent with other studies conducted in the Philippines and Indonesia (Gonzal 1994; Manurung et al. 2009; Miah 1993). The current study provided further evidence that while intensive pruning was beneficial for grain production, the practice may reduce the profitability of timber production below levels acceptable for farmers choosing to grow commercial timber. Financial analysis showed that under intensive pruning, grain yield compensated for reductions in timber yields of up to 6% (1 cm difference in average dbh at harvest). Even if average dbh were reduced by 3 cm (or a corresponding 16% reduction in timber volume) as a result of intensive pruning, the returns to labour would be greater than systems with moderately pruned trees (T 1). The returns to labour under moderate pruning will be the same as that of the high pruning treatment only in the unlikely event that the average dbh at the end of the timber rotation under intensive pruning was 24 cm (6 cm less than the dbh under T 1). In the context of this study intensive pruning of timber trees during the 2-year period of intercropping provided greater returns to the labour and greater profitability providing that timber yields are not excessively reduced.

The intensively-pruned Gmelina–maize agroforestry systems of this study required a total of 24 man-days ha−1 of labour more than the systems with moderate pruning. As farmers need to provide this labour during the cropping season (before the emergence of maize plants), labour availability may be the main impediment for farmers who want to adopt intensive pruning. To overcome this, and reduce the costs of tree establishment and management, it may be advisable to plant trees at final or quasi-final spacing (250–400 tph). Analysis using modelling of native timber species with maize in the Philippines supports this proposition (Martin and van Noordwijk 2009). However, timber trees planted at wider spacing will probably require more labour for pruning than trees at higher densities, as more and larger branches are likely to grow. Moreover, producing quality timber by minimizing the number of trees planted and applying formative pruning may fail, as shown by Kerr and Morgan (2006) with a study of 4 temperate timber trees planted at densities ranging from 600 to 1,370 t ha−1. Future research should assess the trade-offs of planting timber trees at final spacing.

Another option for farmers to reduce pruning labour is to plant timber species that have an architecture more favourable for intercropping (i.e., narrower crowns). In a survey of tree management practices conducted in Claveria, growers of Swietenia macrophylla King. (mahogany) and Eucalyptus deglupta Blume (bagras) responded that they save considerable labour because those species require less intensive pruning than Gmelina (Bertomeu 2004). Tree farmers cited the narrow crown and smaller branches of mahogany and the straight stem and self-pruning habit of bagras as the most notable advantages of those species over Gmelina.

In the Philippines smallholder-produced timber has become an important source of raw material for the local wood industry (Bertomeu 2008). However, current timber management practices (no thinning, intensive pruning and inappropriate pruning methods) result in undersized, and low quality timber. Over the past 10 years the market has been saturated with such timber, resulting in low prices for farm-grown trees. A market survey indicated that local timber traders and wood processors were willing to pay a premium for quality timber (Bertomeu 2008). Therefore managing for larger, better quality trees is recommended to enhance the financial returns from small-scale tree farming. In traditional forestry, moderate pruning is a common practice to improve tree form (taper and stem shape), to reduce wood defects and thus, produce high yields of quality timber. But in the context of resource-poor farmers (with a priority to grow food crops) intensive pruning increased profitability of maize-timber production systems by enhancing crop yield without excessively reducing timber yields. The question remains now whether this pruning strategy can result in quality timber that commands a higher price. Our opinion is that high and frequent pruning (during the 2 or 3 years of intercropping) should be compatible with knot-free quality timber production as long as pruning operations are properly implemented. Future research should study this issue, as well as assessing what impedes farmers from adapting proper pruning ratios and recommended pruning for different timber species and different smallholder timber production systems.

The observed slower early growth but subsequent faster dbh increment of trees at site 2 also highlighted the importance of intensive management to enhance tree growth and compensate for poor site properties. It also demonstrated the need to develop better tree growth evaluation methods by integrating site quality information with general spatial information as identified by Martin et al. (2010). Finally, there is also a need to conduct long-term studies, throughout a full timber rotation, to better assess final timber yields and the potential benefits of integrating shade-demanding crops or animals on overall system productivity and profitability.

Conclusion

When fast-growing timber trees are intercropped with light-demanding annual crops, intensive pruning (i.e., retaining LCR of 20–30%) before crop production generates greater returns than moderately pruned trees (as recommended in classical forestry text books). The gains in yield of annual crops resulting from reduced shading compensate for labour costs associated with pruning and the detrimental effect on tree growth resulting from intensive pruning. Farmers whose immediate objective is to produce food crops, but who are also interested in producing short rotation (8–12 years) commercial timber, should prune trees intensively in order to prolong the period of profitable intercropping. Once grain yields decrease below the break-even point, intercropping should be discontinued and trees should be managed for quality timber production.