Agroforest Syst (2009) 77:18 DOI 10.

1007/s10457-009-9244-8

Tree seedling establishment in living fences: a low-cost agroforestry management practice for the tropics
B. E. Love E. W. Bork D. Spaner

Received: 21 August 2007 / Accepted: 4 July 2009 / Published online: 22 July 2009 Springer Science+Business Media B.V. 2009

Abstract Establishing trees in pastures can have production and conservation benets, but is complicated by the presence of livestock. The need to protect seedlings from livestock increases tree establishment costs, which in turn, can deter landowners from planting trees. Living fences are a ubiquitous feature of pasture landscapes in the tropics that could help protect newly planted trees by preventing livestock trampling and browsing. This study quantied the effectiveness of a living fence in protecting tree seedlings during the rst 2 years after planting. The four native tree species evaluated were: Cedrela odorata L., Pachira quinata (Jacq.) W.S. Alverson, Samanea saman (Jacq.) Merr., and Tabebuia rosea (Bertol.) A. DC. Results show that the living fence provided protection from livestock except in cases where tree species were highly palatable as forage (i.e. P. quinata). Trees planted into the living fence generally had greater survival (62 vs. 28%), relative growth (10.3 times initial height vs. 5.8 times initial height), and nal height (191 cm vs. 108 cm) compared to those planted in open pasture after 2 years. However, survival and growth of trees planted into the fence remained lower than that observed at a nearby plantation with no livestock, regular weeding and no

living fences. This study indicates that use of living fences as a protective barrier could be an effective low-cost approach for establishing trees in tropical pasture landscapes. Keywords Growth Livestock damage Native species Tropical Silvopastoral Survival

Introduction Introducing trees into pastures is an important production and conservation objective (Long and Nair 1999). Theoretical benets include diversied production and income; increased total productivity; maintenance of tree biodiversity; the provision of shade and browse for livestock; and the provision of a refugia for biodiversity. In Panama, pasture landscapes are especially prevalent having replaced 70% of the native forest (Ledec 1992). Landowner initiated establishment of trees in pastures can be difcult. Tree seedlings in grazed pastures can be damaged by cattle browsing and trampling (Pitt et al. 1998). Physical barriers (Beetson et al. 1991), abrasive substances, and chemicals (Eason et al. 1996) can be used to protect trees from livestock damage. All these protection strategies increase the costs of tree establishment, and physical barriers can result in poor seedling development (Beetson et al. 1991). In developing countries, landowners prefer low-cost methods

B. E. Love E. W. Bork (&) D. Spaner Department of Agricultural, Food, and Nutritional Sciences, University of Alberta, 4-10 Ag/For Building, Edmonton, AB T6G 2P5, Canada e-mail: edward.bork@ualberta.ca

123

Agroforest Syst (2009) 77:18

for establishing trees (Arnold and Dewees 1998) and increased establishment costs may deter tree planting. Living fences are a common feature of pastures in Latin America (Harvey et al. 2005). Living fences consist of closely-spaced trees that delimit a eld boundary to which fencing material (usually barbed wire) may be attached (Budowski 1993). Generally, living fence stakes are vegetatively propagated to form fences (Zahawi 2005), although pre-existing trees may be incorporated into the fence line. Living fences in Central America have been found to account for as much as 45% of all fencing, and cover up to 50.5 linear meters per hectare (Leon and Harvey 2006). Generally, landowners are accepting of establishing trees in living fences (Borel and Romero 1991). This may be because planting trees along eld edges instead of within the interior of existing elds interferes less with eld crops. Living fences can provide low-cost protection to tree seedlings planted directly underneath it during establishment by acting as a robust physical barrier to livestock (Lehmkuhler et al. 2004). However, because animals seek boundaries upon entering elds and rest in shade during the day (Stuth 1991), saplings near the eld edge may continue to be at risk of damage. Additionally, tree species palatability is variable (Kaitho et al. 1996) and certain species may be more prone to browsing. Living fences may also compete with seedlings for above- and below-ground resources, as demonstrated by competition studies in alleycropping systems (Jose et al. 2004). While natural recruitment of trees into living fences is common, little eld research has investigated deliberate strategies to employ living fences as a protective environment into which tree seedlings may be planted. The objectives of this study were to (1) evaluate the survival and growth of several tree species planted either within or outside living fences, and (2) to assess the impact of livestock on tree seedlings.

an average annual precipitation and temperature of 2,000 mm and 27C, respectively. There is a pronounced 5-month dry season that begins in January. The pasture where the experiment was situated is *4.5 ha in size, topographically at, and bound on all four sides by a living fence with a total length of 900 m. The pasture consisted of a pure sward of African star grass (Cynodon nlemfuensis Vanderyst). The pasture was grazed periodically throughout the year. Stocking densities averaged 5 cattle ha-1 during the rainy season, and 3.3 cattle ha-1 during the dry season, which are stocking levels representative of pastoral operations in the region. Overall, grazing intensities were considered light during the dry season and light to moderate during the rainy season. At no time did the pasture sward show signs of overgrazing (e.g., bare ground, erosion, pugging). At the beginning of each rainy season (JuneJuly) the pasture was sprayed with a single application of 2,4-D to control broadleaf weeds using a backpack sprayer. A 1 m buffer zone adjacent to all planted tree seedlings was left unsprayed. The living fence used in this investigation consisted largely of living stakes that were pollarded once annually during the dry season. Specically, of the 120 living stakes neighboring seedlings planted within living fences, there were 62 Spondias mombin L., 51 Jatropha curcas L., 3 Gmelina arborea Roxb. ex Sm., 2 Guazuma ulmifolia Lam., and one each of T. rosea and Cordia alliodora (Ruiz and Pv.) Oken. On average, living stakes were 106 cm distant from planted seedlings and had an average basal area and height of 48 cm2 and 310 cm, respectively. Characteristics of the living fence in this study were typical of living fences observed in the region. Experimental design Single-tree seedling replicates (Huxley 1987) were arranged in a completely randomized design. The study was a factorial experiment with four native tree species planted at each of two locations (i.e. inside and outside living fences) within the pasture. Fifteen replicates were established for each species both within the living fence and again within open pasture (n = 120). Previous plantation studies in Central America have shown that samples of 15 trees provide adequate growth estimates (Piotto et al. 2003). The tree species evaluated were: Cedro AmargoCedrela

Materials and methods Study site A eld experiment was established in El Cacao township of Tonos Province in the Republic of Panama. The study area was located 50 m.a.s.l., in Holdridges (1967) humid tropical life zone, and has

123

Agroforest Syst (2009) 77:18

ordorata L. [family MELIACEAE], Cedro Espino Pachira quinata (Jacq.) W.S. Alverson [family BOM BACACEAE], GuachapalSamanea saman (Jacq.) Merr. [family LEGUMINOSAEMIMOSOIDEAE], and RobleTabebuia rosea (Bertol.) A. DC [family BIGNONIACEAE]. Field methods: living fence planting Tree seedlings, produced in a nursery 3 months prior to establishment, were planted on 10 August, 2004 during the early part of the wet season. Average seedling heights (95% condence interval) at planting were 26 2.6 cm, 19 1.8 cm, 20 2.2 cm, and 18 0.7 cm for C. odorata, P. quinata, S. saman, and T. rosea, respectively. Seedlings were planted at regular intervals (*3 m) in a linear fashion into both a living fence (directly below barbed wire) and again within open pasture at a 10 m distance from the living fence. No physical or chemical protection was provided to seedlings planted in open pasture. All seedlings were assigned completely at random to planting sites. Fertilizer (32 g each of 12-24-12 fertilizer and 0-46-0 super-phosphate) was applied to the bottom of each planting hole (*14 cm diameter 921 cm depth) and covered with soil before planting. Where planting sites at 3 m intervals conicted with a living fence stake they were shifted along the fence up to 15 cm, to avoid planting directly on top of living stakes. A square-area of 1 m2 was manually weeded to clear African star grass from around each seedling prior to planting. At planting, carbofuran (2,3-dihydro-2,2dimethyl-7-benzofuranyl methylcarbamate) was surface applied 2 cm from the stem to control white grub (Phyllophaga spp.), a common insect pest of the area. Fertilizer application and white grub control mimicked standard timber plantation establishment protocols for the region. Areas around tree seedlings were subsequently manually clean-weeded (1 m2 around seedlings) of all herbaceous vegetation, 5 mo and 10 mo after planting. Manual weeding is consistent with plantation establishment throughout the region. Tree seedlings were measured six times: at planting, and 1 mo, 4 mo, 10 mo, 17 mo, and 2 years after planting. Survival, livestock damage (i.e. the presence or absence of trampling, defoliation, and rubbing), and seedling height were recorded during all measurement periods.

Basal diameter was taken for the last three measurement periods only. Hoof impressions in the soil combined with bark scarring on the lower bole were used to identify trampling. Bitten leaves and stems were used as indications of browsing and were easily distinguished from insect herbivory. Abrasion and hair deposits on the upper bole were used as evidence of animal rubbing. At the end of the experiment, two experts evaluated seedlings for acceptable timber production characteristics (i.e. height over 1.5 m and straight, single-stemmed boles). In order to characterize the living fence used in the experiment and aid in interpreting tree seedling responses within this treatment, the closest living fence stake on each side of every planted seedling was measured at the end of the experiment. Measurements for living stakes included: distance to seedling, species, basal area, and height. Height was measured to the tip of the tallest branch prior to pollarding after 1-year of growth. Basal area was calculated from basal diameter assuming a circular bole. Field methods: plantation establishment The same tree species described above were planted in mono-specic plantation plots at a nearby site (*0.5 km away) at the same time in 2004. The plantation was part of the Native Species Reforestation Projects1 on-farm trial involving 24 landowners at two locations within Panama. These plots received plantation type management (manual clean weeding three times a year), were free of competition from living fences, and were protected from livestock with fencing. The plantation site was on a slope that varied from 7 to 21% and was bordered by secondary forest. Despite its near proximity to the living fence experiment, plantation plots were established on lighter colored soils that were well-drained. Trees were established using the same procedures described previously, except seedlings were planted in a 3 m 9 3 m grid pattern (1,111 trees ha-1). Sixty trees from each mono-specic plot were measured after 2 years of growth.

http://research.yale.edu/prorena/about_prorena.htm.

123

Agroforest Syst (2009) 77:18

Statistical analysis Descriptive statistics (simple means and 95% condence intervals) were calculated to describe living fence characteristics and tree growth under plantation conditions. Mixed model analyses based on maximum likelihood estimation were used to model variance heterogeneity in order to: (1) compute least squared means and 95% condence intervals for relative growth and nal measurements of height and basal diameter for surviving trees, and (2) assess the signicance of differences in relative growth and nal size measurements using Tukeys test (Day and Quinn 1989). Signicance testing through mixed model analysis was applied only to the controlled experiment involving pasture and living fence environments. Data from the plantation site were examined separately for anecdotal comparison only. Relative growth was calculated as the nal size measurement divided by the initial size at time of planting (e.g. Relative growth in height = Final height/Initial height). Contingency tables and a Fishers two-way exact test were employed to assess the statistical signicance of observed differences in count data (survival, acceptable tree seedlings for timber production, cattle damage) and the MantelHaenzel statistic was employed to assess overall responses to planting location using the FREQ procedure in SAS (Stokes et al. 1995). A two-tailed t-test was used to evaluate the signicance of differences in growth between seedlings damaged by cattle and those that were not. Statistical Analysis System (SAS Institute 2004) software was used to conduct all analyses.

greater (P \ 0.05, n = 30) in fences, but only for C. odorata, P. quinata, and T. rosea. While survival of S. saman tended to be higher within living fences, this difference remained non-signicant (P [ 0.05) (Fig. 1). Overall, tree survivorship at the end of the trial was markedly greater within the living fence (P \ 0.01, n = 60 per location), with the odds of tree seedlings surviving being 9 times higher for the 60 seedlings planted within the living fence compared to those in open pasture. Despite this, survival of trees in the fence remained 16% lower than that observed at the nearby plantation (Table 1), and was particularly poor for P. quinata (33 versus 75%, respectively). Despite respectable tree seedling survival rates within the living fence, C. odorata and P. quinata produced no trees of acceptable conformation for future high quality timber production in either the
Table 1 Tree seedling survival and means (95% condence interval) for nal height and nal basal diameter after 2 years of growth for each of four native tree species established in a living fence, open-pasture, or a monoculture plantation Planting locations and tree species Living fence C. odorata P. quinata S. saman T. rosea Open-pasture C. odorata P. quinata S. saman T. rosea All species Living fence Open-pasture Plantation C. odorata P. quinata S. saman T. rosea All species 50 75 93 93 78 310 41 294 26 286 26 487 145 352 46 7 0.7 7.6 0.7 5.5 0.4 7.6 0.4 6.9 0.3 62 28 191 45a 108 18b 3.8 0.7a 2.9 0.8a 7 0 56 47 66 174a 3.8 3.5a 0 121 52a 98 23a 0 2.9 1.2a 2.9 0.8a 40 33 87 87 72 15b 44 10c 297 83a 3.1 0.7ab 1.2 0.2b 4.6 1.4a Survival (%) Mean nal Mean nal basal height (cm) diameter (cm)

196 41ab 4.0 0.8a

Results Seedling survival and acceptability About 4 months after planting there was a sharp decline in tree seedling survival, particularly among trees in open pasture, after which survival stabilized with only slight declines in both the living fence and open pasture until the end of the experiment. After 2 years, 62% of the 60 trees planted within the living fence had survived compared to only 28% of those planted in open pasture (Table 1). Survival also differed between planting locations on the basis of tree species (Fig. 1). Tree survival was

Means within a column for each comparative category (living fence, open-pasture, all species) with different lower case letters differ (P \ 0.05). No statistical tests were applied to the plantation data as they were not part of the experimental design. Means are least-squared means except for the all species and plantation categories, which are unadjusted means

123

Agroforest Syst (2009) 77:18

1.0
Open-pasture Living Fence

5
25

0.8

Relative growth in height

* **

Open-pasture Living fence

20

Proportion surviving

0.6

15

0.4

10

0.2

*
C. odorata P. quinata S. saman T. rosea ALL SPECIES

0.0
C. odorata P. quinata S. saman T. rosea ALL SPECIES

Tree Species

Species

Fig. 1 Seedling survival after 2 years for each of four native tree species planted within either open pasture or a living fence in Panama, 20042006. An * indicates signicant differences between the proportion surviving in open pasture versus a living fence (* P \ 0.05; ** P \ 0.01). Total sample size for each species was n = 30 (n = 15 per planting position); for combined data the total sample size was n = 120 (n = 60 per planting position)

open pasture (n = 15 for each species) or the living fence (n = 15 for each species). In contrast, T. rosea and S. saman had signicantly more trees suitable for timber production (P \ 0.01, n = 15 for each species) within the fence. Specically, T. rosea produced no trees suitable for timber production in the pasture compared to 53% (8 of 15) in the fence, while 27% (4 of 15) of S. saman trees in the pasture were suitable for timber production versus 80% (12 of 15) within the fence. Absolute and relative growth performance Relative growth in tree height differed by tree species (P \ 0.01, F3,47 = 20.36) and planting location (P \ 0.05, F1,47 = 6.84) (Fig. 2). Similarly, nal tree heights differed by tree species (P \ 0.01, F3,47 = 28.29) and planting location (P \ 0.01, F1,47 = 7.89) (Table 1). Trees of all species except C. odorata were taller within the fence than in open pasture at the end of the trial. Additionally, most tree species (all but S. saman) planted within the fence remained shorter than those observed at the nearby plantation (Table 1). All species except C. odorata exhibited greater height growth during the study period when planted within the fence (Fig. 2) relative to the open pasture. Mean nal basal diameter differed by tree species (P \ 0.01, F3,46 = 30.54), but not by planting location

Fig. 2 Mean relative height growth (95% condence interval) after 2 years for each of four native tree species planted within either a living fence or open pasture in Panama, 2004 2006. An * indicates a signicant difference in relative height growth between planting positions (P \ 0.05). Total sample size varied by species (C. odorata, n = 7; P. quinata, n = 5; S. saman, n = 22; T. rosea, n = 20; All species, n = 54). Sample sizes per planting position varied with survivorship. With the exception of the all species category, means are least-squared means

(P = 0.31, F1,46 = 1.06) (Table 1). Similar to the height responses, the nal basal diameters of trees planted within the living fence were lower compared to those observed in the nearby plantation with the exception of S. saman (Table 1). Cattle damage Livestock damage to trees during the experiment included trampling, browsing, and rubbing. Over the course of six observation periods, a total of 44 trampling, 28 browsing and 3 rubbing events were recorded, affecting a total of 35, 13 and 3% of all planted seedlings, respectively. Most of the trampling damage (77%) occurred during the rst month, while browsing took place throughout the study, and rubbing was only observed during the last measurement when trees were taller. P. quinata was the most frequently browsed species, accounting for 89% of all browsing events. Of the tree seedlings experiencing damage from cattle, more (40 of 59, or 68%) damaged seedlings were found within the pasture rather than within the living fence (P \ 0.01). Overall, the odds of trees within the pasture being damaged were 5 times greater than for those planted within the living fence.

123

Agroforest Syst (2009) 77:18

While trampling was highly associated with planting in open pasture (P \ 0.01, n = 44), no association was detected between browsing and planting location for P. quinata (P = 0.71, n = 25). Finally, trees planted within the living fence that were damaged by cattle during the experiment produced less relative growth compared to undamaged trees (P \ 0.05, df = 35, T-value = 2.0). On average, undamaged trees were 12 times taller than their initial height after 2 years, while damaged trees were on average only 6 times taller.

Discussion Although the living fence did not altogether prevent the mortality of tree seedlings, greater survival was consistently observed within the living fence compared to open pasture for all tree species. These results indicate that the living fence provided a more favorable environment for the survival of seedlings. Moreover, the growth in height of seedlings that survived was greater in the living fence compared to the open pasture for 3 of the 4 timber species tested. Enhanced survival and growth of seedlings within the living fence is attributed to reduced cattle damage throughout the duration of the experiment, particularly trampling. The living fence appeared to provide protection from trampling because cattle could not access seedlings unless they physically contacted the overlying fence. Trampling damage was most pronounced early in the experiment within the open pasture when seedlings were numerous and small, and is consistent with other studies assessing livestock damage to establishing trees (Lewis 1980). In contrast to trampling, browsing occurred throughout the experiment both in open pasture and the living fence because tree heights under 200 cm and the close physical proximity of the fence barrier to trees did not altogether prevent access by cattle. Indeed, the poor performance of P. quinata within the living fence appeared to be related to browsing. Concentration of browsing on P. quinata was likely due to its salty tasting foliage, which local ranchers recognize as being palatable to cattle. Observations from the adjacent plantation suggest that trees established within the fence had lower growth and higher mortality compared to trees established and grown under plantation conditions.

Thus, while the living fence acted as a favorable nurse crop in grazed pasture, it may also impose competition and reduce the growth of seedlings compared to ungrazed conditions free of living fences. Samanea saman was the only species to exhibit height and basal diameter growth within the living fence comparable to that observed at the nearby plantation. These results are not surprising given that plants are known to experience complex interactions with neighboring vegetation that can include facilitative and competitive effects simultaneously (Callaway 1995). In the pasture environment of this experiment, although competition may exist between tree seedlings and the overlying living fence for resources such as light, water and nutrients, the benet of protection from livestock appears to more than offset any observed decline in performance due to competition. Despite exposure to livestock and potential competition with living stakes, and limitations in the ability of tree seedlings to produce high quality timber after 2 years, both S. saman and T. rosea trees within the living fence had average heights of 300 and 200 cm, and were 16 and 11 times taller than their initial heights, respectively. These respectable growth rates coupled with the low management and opportunity costs of establishing trees in a living fence may compensate for survival and growth that is lower than that achieved under plantation style management. The total costs for planting 60 seedlings in a living fence amounted to *US 42 per tree ($15 for seedlings, fertilizer, and pesticides; plus two workdays at US $5/ day), with no further expenditures expected for stand maintenance. In contrast, the cost of the rst 3 years alone for plantation establishment in Panamas canal watershed varied from US $2.50$5.80 per tree given planting densities of 1,111 trees ha-1 (Craven et al. 2008). Moreover, this strategy of timber production required no modication of pasture management by landowners. Plantation costs were higher due to higher establishment and management costs (e.g., land preparation, weeding, etc.). Competition with other agricultural land-uses, including grazing, is a key reason why timber plantations are not more widely adopted (Hyde et al. 1996). The potential of systematically using living fences as an alternative to producing timber in plantations is evidenced by landowners continuing to opportunistically harvest timber trees out of living fences (pers. obs.). This is done despite losing the section of tree bole that

123

Agroforest Syst (2009) 77:18

7 Budowski G (1993) The scope and potential of agroforestry in Central America. Agrofor Syst 23:121132. doi:10.1007/ BF00704910 Callaway RM (1995) Positive interactions among plants. Bot Rev 61:306349. doi:10.1007/BF02912621 Craven D, Hall J, Verjans J-M (2008) Impacts of herbicide application and mechanical cleanings on growth and mortality of two timber species in Saccharum spontaneum grasslands of the panama canal watershed. Restoration Ecol (Published On-line) Day RW, Quinn GP (1989) Comparisons of treatments after an analysis of variance in ecology. Ecol Monogr 59:433463. doi:10.2307/1943075 Eason WR, Gill EK, Roberts JE (1996) Evaluation of anti-sheep tree-stem-protection products in silvopastoral agroforestry. Agrofor Syst 34:259264. doi:10.1007/BF00046926 Harvey CA, Villanueva C, Villacis J et al (2005) Contribution of live fences to the ecological integrity of agricultural landscapes. Agric Ecosyst Environ 111:200230. doi:10.1016/ j.agee.2005.06.011 Holdridge LR (1967) Life zone ecology. Tropical Science Center, Costa Rica Huxley PA (1987) Agroforestry experimentation: separating the wood from the trees? Agrofor Syst 5:251275. doi: 10.1007/BF00119125 Hyde WF, Amacher GS, Magrath W (1996) Deforestation and forest land use: theory, evidence, and policy implications. World Bank Res Obs 11:223248 Jose S, Gillespie AR, Pallardy SG (2004) Interspecic interactions in temperate agroforestry. Agrofor Syst 61:237 255. doi:10.1023/B:AGFO.0000029002.85273.9b Kaitho RJ, Umunna NN, Nsahlai IV et al (1996) Palatability of multipurpose tree species: effect of species and length of study on intake and relative palatability by sheep. Agrofor Syst 33:249261. doi:10.1007/BF00055426 Ledec G (1992) New directions for livestock policy: an environmental perspective. In: Downing T, Hetch SB, Pearson HA et al (eds) Development or destruction: the conversion of tropical forest to pasture in Latin America. Westview Press, Colorado, pp 2765 Lehmkuhler JW, Felton EED, Schmidt DA et al (2004) Tree protection methods during the silvopastoral-system establishment in Midwestern USA: cattle performance and tree damage. Agrofor Syst 59:3542. doi:10.1023/ A:1026184902984 Leon MC, Harvey CA (2006) Live fences and landscape connectivity in a neotropical agricultural landscape. Agrofor Syst 68:1526. doi:10.1007/s10457-005-5831-5 Lewis CE (1980) Simulated cattle injury to planted slash pine: combinations of defoliation, browsing, and trampling. J Range Manage 33:340345. doi:10.2307/3897879 Long AJ, Nair PKR (1999) Trees outside forests: agro-, community, and urban forestry. New For 17:145174. doi: 10.1023/A:1006523425548 Piotto D, Montagnini F, Ugalde L et al (2003) Performance of forest plantations in small and medium-sized farms in the Atlantic lowlands of Costa Rica. For Ecol Manage 175:195204. doi:10.1016/S0378-1127(02)00127-5 Pitt MD, Newman RF, Youwe PL et al (1998) Using a grazing pressure index to predict cattle damage of regenerating tree

retains the ingrown and embedded fencing wire and must therefore be left behind as a rooted fence post. Our study suggests the need for further research into a larger suite of tree species planted into living fences and pastures of varying characteristics in different agro-ecological zones. The microsite environment experienced by seedlings in a living fence can be highly variable. Large-scale assessment of the use of living fences as a protective barrier for establishing timber species could provide landowners with useful rule-sets for successfully establishing timber trees within fences given their particular conditions. The results of our study lead us to conclude that: (1) living fences can protect seedlings from damage by livestock, resulting in increased survival and growth, (2) tree species that livestock prefer to browse (i.e. P. quinata) are not appropriate for planting into living fences bordering pastures, (3) trees planted into living fences may have poorer growth than trees managed under ideal plantation conditions, and may be limited in their ability to produce high quality timber, (4) S. saman was the only species in the study to perform close to plantation standards, and (5) low capital and opportunity costs may make living fence plantings acceptable to landowners seeking low-cost timber production strategies (e.g. resource limited farmers).
Acknowledgments A Natural Sciences and Engineering Research Council Canada Graduate Student Scholarship and the John G. Bene Fellowship in Community Forestry offered by the International Development Research Centre supported B. Love during this research. The Native Species Reforestation Project (PRORENA): Smithsonian Tropical Research Institute, Roosevelt Ave., Tupper Building401, Balboa, Ancon, Panama, Republica de Panama, provided tree seedlings and comparative plantation data. We thank Shane Matias of Peace Corps Panama for his assistance during data collection, and Dr. Scott Chang of the University of Alberta together with 2 anonymous reviewers, for helpful suggestions during manuscript development.

References
Arnold M, Dewees P (1998) Rethinking approaches to tree management by farmers. Nat Resour Perspect 26:19 Beetson TB, Taylor DW, Nester MR (1991) Effect of treeshelters on early growth of four Australian tree species. Aust For 54:6065 Borel R, Romero F (1991) On-farm research in a silvopastoral project: a case study. Agrofor Syst 15:245257. doi:10.1007/ BF00120191

123

8 seedlings. J Range Manage 51:417422. doi:10.2307/ 4003326 SAS Institute (2004) Sas/stat 9.1 users guide. SAS Institute, Cary, N.C Stokes ME, Davis CS, Koch GG (1995) Categorical data analysis using the sas system. SAS Institute Inc., Cary, NC

Agroforest Syst (2009) 77:18 Stuth JW (1991) Foraging behavior. In: Heitschmidt RK, Stuth JW (eds) Grazing management: an ecological perspective. Timber Press, Portland, OR, p 259 Zahawi RA (2005) Establishment and growth of living fence species: an overlooked tool for the restoration of degraded areas in the tropics. Restor Ecol 13:92102. doi:10.1111/ j.1526-100X.2005.00011.x

123