Biomass Dynamics of Erythrina Lanceo Lata As Influenced by Shoot-Pruning Intensity in Costa Rica

Biomass dynamics of Erythrina lanceo lata as influenced by shoot-pruning intensity in Costa Rica

1 2 Berninger Frank ,* and Salas Eduardo ,3 lDepartment of Forest Ecology, 00014 University of Helsinki, P.O. Box 27, Finland; 2Escuela de Fitotecnia, Universidad de Costa Rica, San Pedro de Montes de Oca, Costa Rica; 3Escuela de. Ciencias Agrafias, Universidad Nacional, Apdo 86·3000 Heredia, Costa Rica; *Author for correspondence (e~mall: [email protected]; fax: +3589 191 58100)

Received 5 April 2001; accepted in revised fono 27 May 2002

Key words: Biomass production, Defoliation, Nitrogen, Non·structural carbohydrates, Nutrientcycling.

Abstraet

Pruning of agroforestry trees, while reducing shade of the crops, usually reduces both biomass production and nitrogen fixation. Short pruning cyc1es are often not sustainable 00 the long run, because tree production declines over subsequent pruning cycles. We compared biomass and labile carbohydrate dynamics of Erythrina lanceolata Standley (Papilionaceae) shade trees under total and partial pruning regimes in a vanilla (Vanilla planifolia L.) plantation in South-western Costa Rica. The highest biomass production was rneasured in the unpruned control, followed by trees with 50% of the 1eaf pruned every three months, while total pruning every six months resulted in the lowest biomass pruductioo. In the more productive treatments, a higher proportion of the production was in branches. Because, the N content of woody branches was high, they were important for nitrogen cycling. In the partial pruning treatrnent more nitrogen was returned to the soil ftom litter and woody branches than fram pruned leaf. Sugar concentrations were not different between treatments and the dynamics of non-structural car bohydrates (sugar and starch) seems to depend more on plant phenology than pruning treatment. However, the starch concentrations in the total pruning were lower than in the other treatments.

Inlroduclion of legume tree based agroforestry systems rely largely

on the idea that N2 fixation and rapidly decomposing In many agroforestry systems, trees and crops com litter provide a cheap source of nutrients to the crop pete inevitably for light, nutrients and other resources. (Nair et al. 1999). Pruning of the tree component is, particuIarly in alley In addition, pruning treatments affect the total cropping systems, a powerful approach to regulate amount of firewood, green manure andJ or forage pro this competition. Therefore, pruning is a cornmon duced by the system. On the other hand pruning re practice in a many traditional and modern agrofor gimes are flexible: pruning intervals and intensities estry systems. However frequent pruning can lead to can be modified to match various production goals. decreasing production and slow death of the trees After a complete pruning, trees left without leaves (Dugurua et al. 1988; Romero et al. 1993). depend on their non-structural carbohydrate reserves The effects of pruning, especially belowground, for regrowth. Conceptual models of tree growth atter are not well understood. Nygren and RarnÍrez (1995) pruning c1aim that the arnount of available carbohy found that in leguminous Erythrina poeppigiana drates will strongly influence the regrowth of the trees (Walp.) O.F.Cook pruning induced dieback of dinitro (Erdmann et al. 1993; Nygren et al. 1996; Berninger gen fixing nodules and the N2·fixing E. poeppigiana et al. 2000). In other words, trees with abundant caro trees were actually non- fixing for- 5 months during a bohydrate reserves will regrow vigorously. Too fre year. Nevertheless, scientific justifications for the use quent pruning will deplete the reserves and tree , 4

20 growlh will decline (i.e., the trees will he 'cIose to 900,------, starvation'). 800 111997/1998 In the present study, we compare biomass produc E700 DAverage tiaD and non-structural carbobydrate dynamics in S" 600 pruned Erythrina lanceo/ata Slandley, a tropical agro e .2500 forestry tree used as shade and support plant for va :§ 400 niUa (Vanilla planifolia L.) in Soulh-western Costa 'c. Rica. Trees of genus Erythrina L. (Papilionaceae: '0 300 Phaseoleae) are cornmon in Central American agro a.l!! 200 forestry systems, where they are used for shade, green 100 manure production, living fences, forage and as me O -I-'--I--'--'r-""r--""I"'- dicinal plants (Kass 1994). E. lanceolata is one of the 12 1 2 3 4 5 6 7 8 9 1011 less studied yet cornmonly used species of the genus. Month It is a small tree adapted to humid climates with a Figure l. RainfaIl in the experimental area (Quepas, Costa Rica; natural distribution from Honduras to Panama (Hold 9°26' N, 84°09' W, 20 ID a.s.l.) during the season 1997/1998. (Av ridge and Poveda 1975). erages refer lO the 19 year perlad 1980-1999 from the Damas Me teorological station about 10 km from the site).

Materials and Methods and the trees in the variety trial plot were pruned with a lower intensity. The pruning treatrnents started 6 Study site month befare the start of the experiment. In June 1997, the plantalion was divided into four The field research was conducted in the experimental plots with about 100 trees in each plot. Four pruning farm of the Costa Rican Ministry of Agticulture and regirnes of E. lanceo/ata were assigned to the plots: Animal Husbandry (MAG) close to Quepos, Puntar (i) complete pruning every six months (T-6), (ii) com enas, Costa Rica (9°26' N, 84°09' W, ~ 30 m a.s.I.). plete prurting every three months (T-3), (iii) partial, The cIimate is hot and humid most of the year but ca. 50%, pruning every tbree months (P-3), and (iv) there is a three months dry season. According to the an unpruned control (C) (pruned the last time 2 years database ,of Instituto Meteorológico Nacional, San before the start of lhe experiment). The biomass dy José, Costa Rica, (www.imn.ac.cr on 24 Sep. 1999), namics of the trees were followed during 11 monlhs The monthly average of maximum temperatures var from December 1997 to November 1998. Sampling ied fmm 30 oC (November) to 32 oC (February was concentrated in a core-plot of 12 trees growing through April) and minimum from 21°C (January) to in lhe centre of each plot. Around each core-plot there 23 oC (April lhrough June). Most of the fieldwork were at least three rows of trees with identical prnn was conducted in 1998, when the dry season was ex ing treatrnents. The experiment was unreplicated in ceptionally long and dry compared to lhe 15 years' order to have plots big enough to avoid edge effects average (Figure 1) (an 'El Niño' year). The soil was a between the treatments, and henee statistieal data re sandy loam of alluvial origino The water labl" was present pseudoreplieations whieh have to be treated below two meters depth in the wet season. The sur with eaution. face soil was acidic (pH = 5.3) and had a low nitm gen and phosphorous availability (1.5 g N kg-1 using Fine roor and nodule sampling the Kjeldal method, and 9.1 mg P 1-1 using lhe Olsen J;l1ethod). The soil type was an Inceptisol. The fine root and nodule biomass of E. lanceolata E. lanceolata was planted as a shade and support was sampled monthly from December 1997 through tree in 1993. Trees were planted as large stakes. The Nov. 1998 except in July and Oct. 1998. Fine roots trees were planted at lhe density of 2 x 3 m. The « 2mm) were sampled according to the methodology 2 whole plantalion area was 2,300 m • About a quarter of Scbrolh aud Zeth (1995). In each sample plot, 12 of the afea was reserved fOI testing vanilla varieties eores were taken in the field with a stainless steel new to the region and the rest was used for cornmer sampling tube (with an internal diameter of 6.5 cm). cial vanilla cultivation. The trees in the cornmercial The random sampling localions were selected by plantation were partially pruned once ay twice ayear, --' 21

throwing a non-rolling object with closed eyes into a Root cartography random direction. Each core was -divided into three depths (0-15, 15-30 and 30-45 cm). Por each depth, Spatial distribution of fine roots was measured twice the 12. samples were joined and the soil was we11 in the border of the sample plol. One sampling was mixed. Thereafter, fine roots were separated from a done during the dry season (March 1998) !he other subsample 1/3 of the soil (defined by mass) for esti two months after the onset of the rains (late June mating the root biomass in the soil (Schroth and 1998). The ditches were 1.5 m deep and had !he Kolbe 1994). length of half a row interval (1.5 m). The ditches were The nodules were sampled using a stratified sam dug immediately before !he sampling. Root impacts pling approach. The plot was divided into two sam were counted on the wall going from the tree to the pling areas: A 1 m wide strip along the tree rows and middle of the inter-row. We placed a metal net with a the 2 m wide inter-row space. Fifteen random samples mesh-size on 18 x 18 mm on !he wall. Por each cell per sampling zone and plot were taken with a 6.5 cm the presence or absence of life roots was detennined. auger to the depth of 15 cm. The soil samples were Results presented here are for fine roots (diameter 0-5 pooled by sample area and washed on a 2.5 mm sieve. mm). The recovered nodules were oven-dried at 70 oC for 48 h. Nodule biomass per unit area was calculated Carbohydrate measurements using the surface area weighted average for both suata. Carbohydrate concentrations -were measured for branches, sterns' aod coarse roots. Sarnples were cut Aboveground biomass (branches, sorne coarse roots) or taken with a Pressler bOfer (sorne coarse roots, sterns) from the tree. A sub Litter was collected weekly in each plot from three sample of the chipped (using a garden chipper), ho baskets (or 0.25 cm each) per sample plot for estimat mogenized branch material was used for the carbohy ing the litterfall. Leaf and branch biomass was calcu drate analysis. Samples were refrigerated as sooo as lated using a regression of branch leaf and branch possible (less !han 2 hours after sampling) and were wood biomass on branch diameter. During each sam stored in 99% ethanol in less than 8 hours afier sam pling the diameter of a11 branches of the sample plot pling. Analysis followed !he procedure for analysis of was measured. In December 1997, March 1998, June food carbohydrates used in Costa Rica. In shor!: fresh 1998, and November 1998, a subsample of branches samples were homogenized in a vortex mixer. Soluble in each plot was destructively harvested for determin sugars and starch were analysed separately. Sugars ing the regressions. The branches were selected ran were extracted using hot water. Chlorophyll and other domly and contained all branch diameter elasses. Leaf pigrnents were precipitated using potassium ferrocya was defined as leave blades plus the leaf pelioles. nid and zinc acetate. Sugars were analysed spectro Branch wood ineluded green and non-lignified parts photometrically using the anthrone reaction. From the of the branch. In order to eliminate biases caused by residues of the sugar extractions starch was hydroly heteroscedasticity, the data was log-transformed for sed using hot HCI, and the resulting glucose was anal both diameter and biomass values. General linear ysed spectrophotometrically using the anthrone reac models were used to account for sampling date and tion. Concentrations were converted from fresh to dry treatment effects in the data and the final models were mass using the water content of parallel samples. selected using the adjusted r 2 criterion. A11 calcula tions were done in SAS (SAS Institute, Cary, North Statistical analyses Carolina, ©1995(1999). The regressions were applied for completely non Data was analysed using the SYSTAT (Kruscall Wal destructive estimation of leaf and branch biornass lis test) and SAS statistical packages. General linear (Nygren et al. 1993), i.e. we applied the regression to models and analysis of variance were made using the a11 branches in the stand. The estimated biomasses procedure GLM !hat corrects for unbalanced designs. were surnmed up to stand biomass. Leaf production Given mean values are weighted least square means, was calculated as change in leaf biomass plus litter that are corrected for unbalanced designs. fal1. Results were expressed per unit area and time. 22

Resulls 400·,------,

Aboveground biomass 350 300 The aboveground biomass equations differed between :§ • Dcc 97 mcasurcmenls 11 Mar 98 measllremenls dates and treatments for leaf biomass, while there 111 250 f11 Jllne. 98 meaSllrements o ..'" Nov 98 measuremcnts were no differences between dates and treatments for ~ 200 hranch biomass. The leaf biomass r 2 values were only :¡; moderate. The final r2 was 0.68. For the leaf biomass, '; 150 o inclusion of. sampling date and pruning treatment im .3 O 100 o 2 prbved the adjusted r • The best-fit leaf biomass model had the form: 50 o W ~(a +a +a )Db,+b, f· o s tr O 2 3 4 5 Diameter (cm) where W( is the leaf biomass, Qo is a constant (8.95), Q" is the pruning treatment effect (-0.37 for the con 1600 trol, 1.19 for the partial treatrnent, 0.42 and totalprun o ing), as is the seasonal effect to account for the dif 1400 B) ferent sampling times (1.64 for 12/97,0.677 for 3/98, 1200· o 1. 11 for 6/98 and 1.0 for 11/98), b o is a constant :§ - (2.60), and b, is the seasonal effect of the nonlinear :: 1000 ity of the curve(-2.018 for 12/97, -1.142 for 3/98, .. E - -0.08 I for 6/98 and O for 11/98). Sampling time and o :¡; 800· pruning treatment effects were incorporated or not in .s::. o 2 u - corporated depending Oil changes in the adjusted r • r:: 600 ••••o Figure 2a displays leaf biomass as a function braoch .. lb - -[JI diarneter. '" 400 For branches sampling time and treatrnent effects failed to improve the adjusted r 2 The best-fit model 200· o for the branch biomass (W B) was: O O 2 3 4 5 Diameter (cm) Figure 2. Leaf (a) and woody (b) biomass in a branch of Elyth where a b and b b are pararneters with the values 9.44 rina lanceolata as a function of branch diameter (Quepos, Costa 2 g and 3.311, respectively. The r of the equation was Rica). 0.93. Figure 2b displays the branch biomass as a func tioo of hranch diameter. trol was much higher Ihan in the olher treatrnents. The During the dry season the allometric relations gave trees under T-3 pruning regime died after !he pruning lower leaf biomasses (Figure 2). EspecialIy the expo in March 1998, or in less than ayear afler establish nent of the equation changed strongly, indicating !hat ing the treatment. We assume that the extreme dry largor branches had less leaf in Ihe dry Ihan in !he wet seaso~ in combination with the stress from pruning season. This is probably linked to !he phenology of was responsible for the death of the trees, because E. lanceo lata, which has a lower leaf biomass during similar pruning treatments were previously used in the dry season. the region (A. Chacón pers. comm.). Aboveground biomass growth was slow during the Litter production was a small proportion of a11 !he dry season. After the onset of rains there was a con productivity in all treatrnents (Figure 3). The leaf bio siderable growlh of leaf and branch biomass (Fig mass in Ihe partialIy pruned trees increased faster and ure 3). Leaf biomass was higher in the control and was approximately equal to the control prior to the P-3 treatrnents Ihan in Ihe T-6 pruning regime during September 1998 pruning. the whole experiment. The branch biomass in the con- • 4

" 3000 I~===;;==:======:==¡------' 200 2500 A) I-II-c -+-P3 -+-T61 ~ 180 A) =;- 2000 ,;; 160 • .c E 1500 01 140 ~ 1000 e • 120 500 '0 J 01 ~ O~~~~~ g 100 80 2112 3111 114 3115 30n 2819 27111 i!! '" 60 _30000~------, " i! 40 B) z ".~25000 20 ';;20000 o ~ 15000 o Control Partial Total iD 10000 ~ g 5000 ~ O~~~~~~~~~ 400 2112 3111 114 31/5 30n 2819 27/11 350

35.00 ~ 300 30.00 C) ~ 250 •o 2500 01 0 g-; 20.00 m 200 ~" E ~ ~ 15.00 .2 150 • m al § ~ 1~:~~ 100 .=z 0.00 50 2/12 3111 114 31/5 2819 27/11 son O Control PartiaJ Total y; 50 Figure 4. A) Nitrogen returned 10 the soil from different aboye '. O) ~40 ground blOmass components in Quepas, Costa Rica. In the tbree ~ pruning treatments Control, Partial pruning and Total pruning. (B) Nitrogen in tbe standing aboye graunG biomass at the end of the experiment.

Root mapping 1/4 31/5 30n 28/9 27/11 Date The number of root impacts was higher in the wet Figure 3. Dcvelopment of standing ¡eaf biomass (a), standing season than in the dry season and differed between branch biomass (b), litter production Ce), and leaf production (d) of treatments: the highest number of impacts was in the Erythrina lanceolata shade trees during the one year in a vanilla control (22.2 imp,acts m 2 dry season, 64 impacts m 2 plantation in Quepas, Costa Rica. wet season) followed by the partial pruning (12.4 im pacts m 2 dry season, 21.3 impacts m 2 wet season). N concentrations were on the average 0.032 The number of impacts in total pruning treatment was g[N] g-l [DM] and 0.011 g[N] g-I [DM] for Jeaf and lower (6,7 impacts m 2 dry season, 8.9 impacts m 2 wet branch wood, respectively. The N concentration of season). Root mapping also indicated that there were litter was 0.018 g[N] g-l [DM] (Figure 4). The high a large number of fine roots in the lower soH layers N concentrations in branches made branch wood an below the depth of cores (Figure 5). The increase in importantstock of N. Also pruning residues of N root impact counts was much higher, also in relative were an important source of N: in the pruning treat tenns, in lhe control and partial pruning treatrnents. mene' 38% (P3) and (22%) of lhe N flux from the More lhan half of lhe increase took place in the top trees to the soil was from branch wood (Figure 4). 40 cm of the soil. In the control there was sorne in crease in lhe lowest part of the soil proliJe (Figure 5). 24

600TA~)------~------~'

"i;oo I..... C-+-PS ... T61

31/1 1/4 31/5 30/7 28/9 27/11

50,------, l~b4d B) 9 27 45 63 81 99 117 135 143 Depth (cm) Figure 5. Raot density as measured by impact points of Erythrina lanceolata llnder thrce pruning treatments as a function of depth in March 1998 (dry season) (a) and lune 1998 (wet season) (b) in Quepas, Costa Rica. o~~~~~~~~~~~~ 2/12 3111 1/4 31/5 30(7 28/9 27/11 Fine root biomass Date Figure 6. Root (a) and nodule biomass (b) dcvelopment in Eryth Fine root biomass in the upper 45 cm of !he soil pro rina lanceolata as a function of tIme in Quepas, Costa Rica. file was higher in !he P-3 and control treatment than in the T-6 treatment. Fine root biornass increased at 0.6 -----;:;:::;:::::;:::;::::::;=;;-1 1 C -+-P3 the beginning of the wet season. However, the timing I..... -..-T61 Di 0.5 of the increase was slightly different for the control e and P-3 treatments (Figure 6). Average wet season ,g 0.4 m (May to November) fine root biomass was 202 a: I • kg ha- Nodule biomass ranged from O to 56 kg ha-' •m 0.3 (Figure 6). Surface fine root to leaf ratios are shown "1;; ~ 0.2 in Figure 7. -' b o a: 0.1 Carbohydrate concentrations o~~~~~~~~~~~~~ Sugar cQncentrations were significantly different be 2/12 31/1 1/4 31/5 30(7 28/9 27/11 tween dates (p < 0.001) and between biomass com Date partments (root, stem, branch). Further, there were 1 Figure 7. Fine root to Ieaf biomass ratios (kg kg- ) in Erythrina significant differences between biomass compart lanceolata under three pruning treatments and thcir development ments. No clear reduction in sugar concentrations fol over time in Quepos, Costa Rica. lowing a pruning was ohserved in either pruning treatment. Sugar concentrations were highest at the !he explained variance of the ANOVA was 0.37. The beginning of !he wet season, decreased till Septem least square means of starch concentrations decreased ber and increased towards November (Figure 8). in the order control, P-3, T-6 and T-3. Differences in Least square means of the sugar concentrations were the starch concentration between the P-3 and the con 4.8% for roots, 4.3% for branches and 3.1% for trol were not significant using a t-test, but differences stems. There were no significant differences between between the T-6 and P-3 and control treatments were treatments in the sugar concentration. significant (p < O.O! for control, p < 0.05 for the par Starch concentrations differed significantIy be tia! pruning). The least square means for the plots tween treatments (p < 0.05), bioinass compartments were 72.8 mg Starchlg for the Control, 69.2 mg!g for (p < 0.001) and dates (p < 0.01). The proportion of the partial pruning and 57.3 mg Starch /g for!he tota! i¡ ~-.~ "

Branch 10 160 9 Sugar 140 Starch 8 ---l 120 7 100 6 5 80 4 60 3 40 2 20 1 O O 16103 15/05 14107 12/09 11/11 15/01 16103 15/05 14107 12/09 11/11 15/01

Stem -... 10 --1 120 ';" 9 I 100 Starch I Ol 8 Sugar Ol 7 I 80 S 6 r:: 5 60 .~ 4 ... 40 l!! 3 ... 2 20 r:: 1 r::~ O O O 15/01 16103 15/05 14107 12/09 11/11 15/01 16103 15/05 14107 12/09 11/11 (,) Root 10 120 9 8 100 Starch 7 80 6 5 60 4 40 3 2 20 1 O O 15/01 16103 15/05 14107 12/09 11/11 15/01 16103 15/05 14107 12/09 11/11

I-tl-c __ P3 __ T61 Date Figure 8. Time developmcnt of the concentrations of earbohydrates in branehes (a),'stems (h) and coarse roots (e) in Quepas, Costa Riea. pruning. The least square means for the different plant Discussion parts were 45.8 mg Starch Ig dry matler for roots, 64.3 mg Starch Ig for the stem and 80.0 mg Starchlg Pruning treatments resulted in drastic changes in de for branches. velopment of biomass production and allocation pat tems. Total production was the highest in the control, folJowed by !he P-3 treatrnent. Altogether, biomass production was in the lower rauge of the values re ported by Nair et al. (1999) for tropical alley crop- 26

" I ping systems. Decreases in biomass production ties decreased in al1 samplings from fhe surface lo 45 caused by pruning have been observed previously: cm. This illustrates that root sampling using soH cor Miah et al. (1997) showed thal the biomass of Acacia ing requires additional methods to check for root mangium, A. auriculifonIDs and Gliricidia sepium depth distributions, such as rool mapping. These was 27-38.5% 10wer than Ihe control in pruned trees mefhods shou1d preferably be repeated during differ after two years of trealment. Sanginga el al. (1994) ent seasons. We, for example, decided OUT sampling observed biomass decrease by 46% in young G. depth based on a preliminary sampliog in November sepium due to pruning. 1997. At that time fhe results indicaled thal it was N concentrations in the leaf of E. lanceolata were improbable fhat there were maoy fine roots be10w 45

quite high (3.5%). when compared to non-N2-fixing cm deplh. trees. Pa1m and Sanchez (1990) reported similar val Surface fine root to leaf biomass ratios were gen ues for an unidentified Erythrina sp., whi1e Nygren erally lower during the dry season, probab1y, because and Campos (1995) measured higher values we failed to estimate deep fine root biomass. Grow (4.6-5.1 %) for E. poeppigiana. N concentrations in ing season surface fine root to leaf biomass ratios the branches were about a third of the 1eaf concentra were rather constant in the control and T-6 pruning tians. This means that branches were an important regimes, but increased in the P-3. Simulation studies, sink of N (Figure 4). Leaf li!ter N concentration was using a growth model based on transport-resistance about 40% 10wer than in pruned 1eaves. In other approach (Berninger et al. 2000), suggesl a similar words, apart from 1eaf prunings, branches and litter behaviour. Altogefher surface fine root and leaf bio had quantitative importance for N cycling in the sys mass were well correlated with each other. This temo Cut woody branches (exc1uding 1eaf) conttibuted agrees wifh fheories on balanced growth in p1ants like 38% and 22% respectively of the N cyc1ed through the functional balance approach (Davidson 1969). fhe trees in fhe P3 and T6 treatmenls, respective1y The fact tiJat leaf and surface fine root biomass are while leaf recyc1ed 44% of Ihe N lo the soil in the P3 closely related indicates that also the competitive ca and 58% in the T6 treatment. pacity of trees for above- and belowground resources During the dry season, tree growfh and production correlate c1osely. Pruning diminishes the competition were 10w. Especial1y afier fhe December 1997 prun both be10w- and aboveground al the expense of a ing, very low production rates were observed. This lower production of the shade trees. The root to foli may be caused by an especially severe dry season. age Tatios were similar to those observed previously After fhe onset of the first rains production peaked in (i.e., Nygren and Campos (1995)). However, Nygren the control (Figure 3). et al. (2000) observed fhat the input of branch bio Smface fine root biomass was lowest in the dry mass was reduced in G. sepium when pruned partially seaSOll. It seems that most of the roots were in deep every two months, while differences in the input of soil Iayers. However, surface fine root biomass in leaf to the soil were not pronounced. creased after the onset of the rmoy seasoll. Root mOf The control and P-3 treatment produced more bio tality started between the September and November mass during the experiment. There also were qualita sampling. Excavations in November 1998 showed tive differences in growth; aboveground production in that sorne, but few, roots were growing below 2 m the P-3 and T-6 treatments was half leaves and haIf depth. Low surface root biomass has been observed branches, bul aboul 2/3 of the production in fhe con in G. sepium duríng the dry season (Lehmann and trol consisled of branch growth. This agrees with tiJe Zech 1998). Nygren and Campos (1995) showed thal oretic¡¡l investigations, based on the pipe model fhe pruning depressed fine root biomass in E. poeppigi OfY, which suggest tiJat in 1arger trees production is ana. geared lowards woody growth (MiikeHi 1988; Nikin During root sampling, we found only a few nod maa 1992; Berninger and Nikinmaa 1997). However, ules below 0.15 m, although during the excavation of in agroforestry research there are empirical results root systems (Salas et al. unpublished) we found nod supporting and nol supporting these ideas (e.g., Du u1es at 1.5 m depth. However, we assume fhat only guma et al. (1988) and Sanginga el al. (1994)). few nodules were below our sampling depths. On fhe Branches and litter were important components of the other hand, most of the fine root biomass escaped N cycling of the system. Leaf litter and braoches sampling in the dry season, because it was located composed 10-18% and 20-38%, respeclively, of the below our sampling depth. Fine root biomass densi- N returned lo fhe soi! (Figure 4). The branches were 27 important inputs of N and C to the soi!, espeeialIy in Acknowledgements the P-3 treatment Altogether, the amount of nitrogen eyc1ed through the trees was considerable, 90-230 We wish to express OUT sincere thanks to aH the peo kg[N] ha-1 a-1 Nygren et al. (2000) deseribed simi pIe that made this work possible: Dr Pekka Nygren, lar amounts of N accumulation in their experiments Praf Carlos Rarrúrez, Mr José Mattey Fonseea, Dr with G. sepium. Petra Schmidt, Mr GuiIJermo Savorio, Mr Asdrúval Sugar eoneentrations did not differ signilieantly Chacón, and Dr Bruno Rapidel. The work was fi between pruned and unpruned trees in abaut a month naneed through the grant 28203 from the Finnish after pruning. Sugar concentrations were relatively Aeademy, This projeet would have been impossible constant throughout the observation perlad and without institutional support from the University of seemed to follow more a yearly eyc1e than pruning Costa Rica and the Ministry of Agtieulture and Live treatments. stock (Costa Rica), We thank two anonymous review Erdmann et al. (1993) and Nygren et al. (1996) ers foy their fruitful comments. hypothesised that stareh reserves limit the tree re growth afier pruning, We observed that stareh eon centrations were significantly lower in the less pro References duetive T-6 treatment, but no depletion of stareh concentrations was observed after pruning. The stud Berninger E and Nikinmaa E. 1997. Differences in pipe model pa ied trees were relatively big, and thus the total stareh i.nneters determine differences in growth and affect response of Scots pine to cUmate change. Funct Ecol 11: 146-156. pool in a tree may have been so large that remobili Berninger E, Nikinmaa E., Sievanen R. and Nygrcn P. 2000. Mod sation of stareh to initial regrowth did not affeet sig elling reserve carbohydrate dynamics, regrowth and nodulation nifieantly the eoneentration, García et al, (2001) have in a N 2 fixing tree managed by periodic prunings. Plant Cell shown that the size of the total starch pool is more Envlron 23: 1025-1041. important for regrowth of Gliricidia sepium than high Davidson R 1969. The effect of rootlleaf temperature 00 rootJshoot ratio of sorne pasture grasses and clover. Ann Bot 33: 561-569. starch concentratioll. The lowest starch concentration Duguma B., Kang B.T. and Okali D.U.U. 1988. Effect of pmoiog in the T_6 treatment of E. lanceolata, whieh produeed intensities of three woody leguminous species grown in alley less biomass, may he a result of too intensive prun croppiog with maize and cowpea on an alfiso1. Agrofor Sys 6: iug. In G. sepium, the least productive pruning inten 19-35. sity also depleted most stareh eoneentration (García Erdmann T.K., Nair P.K.R. and Kang B.T. 1993. Effects of cutting et al. 2001). frequency and cuttiog height on reserve carbohydrates in Gliri cidia sepium (Jacq.) Walp. For Ecol Manag 57: 45--60. The research showed that Erythrina lanceolata is GarcÍa H., Nygreo P. and Desfootaines L. 2001. Dynamics of noo a potential agroforestry tree, and its slow growth structural carbohydrates and biomass yield in a fodder legume make it a good altemative as a support tree for climb tree under different harvest intensities. Tree Physiol 21: 523- ing, shade tolerant eraps like blaek pepper or vauilla. 53!. Holdridgc L.R. and Poveda L.J. 1975. Arboles de Costa Rica. Trop Partial pruning seems to be a suitable management ical Science Center, San José, Costa Rica, 546 p. practice for the species, leading to higher biomass Kass D.C.L. 1994. Erythrina species - pantropical multipurpose production and N yield. Further, sorne shade is rnain tree legumes. In: Gutteridge RC. and Shelton R.M. (eds), For tained all the time. Partial pruning also seems to be age Tree Legumes in Tropical Agriculture. CAB International, less disruptive for the dynamics of other biomass Walliogford, pp. 86-94. Lehmann J. and Zech W. 1998. Fine root turnover of irrigated compartments. While partial pruning may be prob hedgerow intercropping in Northcrn Kenya. Plant Soil 198: lematic in association with annual crops, it may be a 19-31. feasible alternative for shade trees in perennial crop Makelii A. 1988. A carbon balance model of growth and self-prun ping systems, Due to the rather high N eoneentration ing in trees based on structural relationships. Forest Sci 43: in branehes, pruned branehes might be an important 7-24, Miah M.G., Aragon M.L. and Garrity D.P. 1997. Growth, Biomass souree of N, especially under partial pruning regimes, production and distribution of three multipupose tree speces in an agroforestry system as affected by pruning. J Trop For Sci !O: 35-49. Nair P.K.R, Buresh R.J., Mugendi D.N. and Latt c.R. 1999. Nu trient cycling in tropical agroforestry systems: myth and sci- 28

eoce. In: Buck L.E., Lassoie l.P. and Fernandes EC.M. (eds), Palm C.A. and Saochez P. 1990. Decomposition and nutrient re Agroforestry in Sustainable Agricultural Systems. CRe, Boca tease pattems 01' the leaves of three tropical Jegumes. Biotro Raton, pp. 1-31. pica 22: 330--338. Nikinmaa E. 1992. Analysis of the growth of Scots pine: matching Romero F., Monlenegro J., Chana C., Pezo D. and Borel R. 1993. patterns with function. Acta Forestalia Fennica 235: 1-68. Cercas vivas y bancos de proteína de Erythrina beteroana Nygrell P. and Campos A. 1995. Effed of foliage pruning 00 fine manejados para la produccíon de biomasa comestible en el rool biomass of Erythrina poeppigiana (Fabaceae). In: Sino trópico húmedo de Costa Rica. In: Westley S.E. and Powell quet H. and Cruz. P. (cds), Ecophysiology of Tropical Intcr M.R. (eds), Erythrinas in the New and Old Worlds. 205-210 empping. INRA, París, France, pp. 295-302. Nitrogen Fixing Tree Research Reports. Westley S.B. and Pow Nygren p" Cruz P., Domenach A.M. and Vaillant V. 2000. Inftu en M.H. (eds) Erythrinas in the New and Old Worlds. 205-210 eoce of forage harvesting regimes on dynamics of biological Nitrogen Fixing Tree Research Reporls. dmitrogen fixation of a tropical woody legume. Tree Physiol Sangmga N., Danso S.K.A., Zapata F. and Bowen G.D. 1994. In 20,41-48. flueoce of pruning management io P and N distribution and use

Nygren P., Rebottaro S. and Chavarría R. 1993. Application of the efficiency by N2 fixing and non-Nz fixing trees used in altey pipe model theory to nOll-destructive estimation of leaf biomass cropping systems. Plaot Soil167: 219-226. and leaf area in pruned agroforestry trees. Agroforestry Systems Schróth G. and Kolbe D. 1994. A method of processing soil core 23,63-77. samplcs for root studies by subsampling. Biol Fertil Soils 18: Nygren P., Kiema P. and Rebottaro S. 1996. Canopy development, 60-62.

CO2 exchangc and carbon balance of a modelled agroforestry Schroth G. and Zech W. 1995. Root lenglh dynamics io agrofor tree. Tree Physiol16: 733-745. estry with Gliricidia sepium as comparcd to sole cropping in

Nygren P. and Ramírez C. 1995. Production and turnovcr of N 2 the semi deciduous rainforest zone of West Africa. Plant Soil fixing oodules in retation to foliage in periodicalIy pruned 170, 297-306. Erythrina poeppigiana (Leguminosae) trees. Forest Eco} Manag 73: 59-65.