Conversion of Secondary Forest Into Agroforestry and Monoculture

Forest Ecology and Management 163 (2002) 131±150

Conversion of secondary forest into agroforestry and monoculture plantations in Amazonia: consequences for biomass, litter and soil carbon stocks after 7 years GoÈtz Schrotha,b,*, Sammya Agra D'Angelob, Wenceslau Geraldes Teixeirac, Daniel Haagd, Reinhard Liebereia aInstitute of Applied Botany, University of Hamburg, P.B. 302762, 20355 Hamburg, Germany bBiological Dynamics of Forest Fragments Project, National Institute for Research in the Amazon (INPA), C.P. 478, 69011-970 Manaus-AM, Brazil cEmbrapa AmazoÃnia Ocidental, C.P. 319, 69011-970 Manaus-AM, Brazil dInstitute of Soil Science and Land Evaluation, University of Hohenheim, 70599 Stuttgart, Germany Received 1 February 2001; accepted 4 April 2001

Abstract

Large areas of primary forest in Amazonia have been cleared for cropping or pasture, thereby releasing carbon into the atmosphere. Part of this carbon is re-assimilated by secondary forest after the land has been abandoned. Agroforestry and tree crop plantations are options for the economic valorization of previously cleared land in the humid tropics; however, for evaluating the consequences of these land uses for regional carbon ¯ows when established on secondary forest land, information is needed on carbon accumulation in the biomass and soil of these land use systems and of the successional vegetation that they replace. Above- and belowground biomass and litter accumulation were measured for three multistrata agroforestry systems and ®ve tree crop monocultures seven years after their establishment on secondary forest land on a xanthic Ferralsol in central Amazonia. The biomass of the tree crop systems was compared with that of the 14-year-old secondary forest that would have covered the area in their absense. The agroforestry systems were studied at two fertilization levels. Allometric relationships were developed for estimating the aboveground biomass of eight tree crop species, and root systems were excavated to determine root-shoot-ratios. Depending on species composition and fertilizer input, the multistrata systems had an aboveground biomass of 13.2±42.3 t per ha, a belowground biomass of 4.3±12.9 t per ha, and a litter mass of 2.3±7.2 t per ha. The monocultures had an aboveground biomass of 7.7±56.7 t per ha, a root biomass of 3.2±17.1 t per ha and a litter mass of 1.9±5.6 t per ha. The combined biomass and litter was highest in the peach palm for fruit (Bactris gasipaes) monoculture, followed by two of the multistrata systems. The 14-year-old secondary forest had a combined biomass and litter stock of 127 t per ha. The soil carbon stocks tended to be lower in the agricultural systems than under adjacent primary forest in the topsoil, but not when summed over the soil pro®le to 2 m depth. Multistrata agroforestry had several advantages over monocultures as it allowed to combine high and long-term biomass accumulation with early generation of income from annual and semiperennial intercrops, increased growth and earlier yields of certain tree crops, long-term accumulation of capital in larger trees, and more complete occupation of the soil than in common tree crop monocultures of the region.

* Corresponding author. Present address: Biological Dynamics of Forest Fragments Project, National Institute for Research in the Amazon (INPA), C.P. 478, 69011-970 Manaus-AM, Brazil. E-mail address: [email protected] (G. Schroth).

Trees with low litter quality seemed to have a favourable effect on soil carbon even when associated with species with high litter quality and could be used as an insurance against soil organic matter loss under tree crop agriculture. # 2002 Elsevier Science B.V. All rights reserved.

Keywords: Bactris gasipaes; Bertholletia excelsa; Carbon cycling; Citrus sinensis; Cocos nucifera; Hevea sp.; Humid tropics; Pueraria phaseoloides; Theobroma grandi¯orum; Tree crop agriculture

1. Introduction of these land use systems and of the vegetation that they replace. Furthermore, as much of the carbon in During the past decades, large areas of primary terrestrial ecosystems is stored in soil organic matter, forest land in Amazonia have been cleared for changes in the carbon content of the soil as in¯uenced cropping or pasture, thereby releasing carbon into by different land use systems and vegetation types the atmosphere, and have often been abandoned after under humid tropical conditions need to be understood variable periods of time under agricultural use (SerraÄo (van Noordwijk et al., 1997). In Amazonia, research et al., 1996). The secondary forest that develops on on these issues has so far mainly concentrated on such abandoned land plays an important role for the primary and secondary forests and pastures (Fearnside regional carbon budget, as it re-assimilates part of the and Guimaraes, 1996; Koutika et al., 1997), with carbon that was released when cutting and burning the relatively few data being available for tree crop original forest vegetation (Houghton et al., 2000). agriculture and agroforestry (Woomer et al., 2000). Agroforestry and tree crop agriculture are two Here we provide information on above- and possible options for the valorization of previously belowground biomass and litter stocks in three cleared forest land in the humid tropics (Fujisaka and multistrata agroforestry systems and ®ve tree crop White, 1998). Where these land use systems succeed monocultures of 7 years age that were planted after in maintaining soil fertility on a satisfactory level and clearing secondary forest on a previously abandoned increasing farmers' income, further clearing of upland site in central Amazonia. The agroforestry primary forest and accompanying carbon emissions systems were studied at two fertilizer levels. Soil may be reduced. Furthermore, timber trees and tree carbon stocks to 2 m depth of agricultural systems, crops in these systems accumulate carbon in their fallow and primary forest as in¯uenced by individual biomass and, if planted on degraded areas, possibly tree species are also presented. In combination with also in the soil, and they may provide ®rewood and data from other land use systems and vegetation types charcoal as substitutes for fossil fuel (Unruh et al., in the region, the presented information can help in the 1993). On the other hand, when agroforestry systems analysis of regional carbon budgets as in¯uenced by or tree crop plantations are established on previously predicted increases in forest clearing and land cleared fallow or secondary forest land, carbon is occupation in Amazonia (Laurance et al., 2001) and released from the fallow vegetation, and the succes- in the evaluation of different land use options for the sional processes, which would have led to progressive region according to their respective environmental further accumulation of carbon in biomass and litter, impacts. are interrupted. Instead, tree crops, timber trees, annual crops and eventually cover crops are established, whose growth and development are in¯uenced 2. Materials and methods by management practices such as fertilizer application and suppression of spontaneous vegetation through 2.1. Study site weeding. For predicting the effects of agroforestry and tree The study was carried out on the research station crop agriculture on the carbon budgets of humid of Embrapa AmazoÃnia Ocidental near Manaus in tropical regions such as Amazonia, information is Brazilian Amazonia. The climate is lowland humid needed on rates of carbon accumulation in the biomass tropical with mean annual precipitation of 2600 mm, G. Schroth et al. / Forest Ecology and Management 163 (2002) 131±150 133 mean air temperature of 26 8C and mean atmospheric personal communication). Biomass data of the humidity around 85% (mean values 1971±1993, Cabral secondary forest presented further below indicate that and Doza, unpublished). The driest months are July to fallow regrowth at this site was somewhat slower than September, and the wettest months are February to after shifting cultivation in other tropical regions, April. The soil is a xanthic Ferralsol according to the which can be explained with the naturally infertile classi®cation of FAO±UNESCO (1990) with clayey soil, mechanical land clearing and more thorough texture and kaolinitic mineralogy. It is acidic and poor removal of tree regrowth in the plantation. It was in P and nutrient cations (for detailed soil chemical data similar to or faster than fallow regrowth after aban- see Schroth et al., 2000a). doned pasture in Amazonia, where soil conditions are The study site was ®rst cleared from primary forest often severely degraded (Fearnside and Guimaraes, in 1980, using heavy machinery to pile up vegetation 1996). Most deforested areas in Amazonia have been debris and for removal of tree stumps. In 1981, an cleared either for cattle pasture or for shifting cul- experiment with rubber trees (Hevea brasiliensis) was tivation (SerraÄo et al., 1996). established and was abandoned in 1986 because of heavy disease attack. The developing secondary forest 2.2. Experimental design was manually cleared in 1992 for the establishment of the tree crop and agroforestry plots which were used in The experiment was established in February and the present study, and the vegetation was burnt on the March 1993 on an area of about 13 ha, consisting of site. On one side of the experiment a 1 ha plot of the different arrangements of important tree crops of the then 7-year-old secondary forest was retained as a region (Table 1). The following tree species were control and was included in this study. planted in monoculture: peach palm (Bactris gasi- Our study site was directly representative for paes) for fruit production, at 4/4 m; peach palm for approximately 75,000 ha of rubber plantations which palmito (heart-of-palm) production, at 2/2 m; cupuacËu were established on rainforest land in Amazonia dur- (Theobroma grandi¯orum), at 6.4/7 m; rubber (H. ing the 1970s, pro®ting from government incentives, brasiliensis, grafted with Hevea pauci¯ora), at 4/8 m; and were mostly abandoned after a few years because and orange (Citrus sinensis), at 4/8 m. For the two of disease problems and unsatisfactory development peach palm treatments the same planting material was of the trees (Gasparotto et al., 1997). The site con- used, but the management was different. For palmito ditions also approximated those of former shifting production, the main stem of the palms was cut at cultivation sites, which are left to fallow after a few 18 months to stimulate the formation of offshoots, years of use. When the 7-year-old secondary forest which were then harvested two to three times per was cleared in 1992, its aspect was approximately year, whereas the fruit palms were allowed to grow similar to fallows of the same age on shifting cul- freely. Some fruit palms also developed offshoots, but tivation land in the region (Gasparotto and MaceÃdo, these were not harvested. The fruit palms had an

Table 1 Tree species and number of trees per hectare in three multistrata agroforestry systems and ®ve monoculture plantations in central Amazonia

System Peach palm Peach palm CupuacËu Rubber Brazil nut Annatto Citrus Coconut Total for fruit for palmito palm

Multistrata 1 125 250 78.1 125 578.1 Multistrata 2 78.1 156.3 93.3 93.3 156.3 577.3 Multistrata 3 34 108.7 72.5 54.3 269.5 Peach palm fruit 625 625 Peach palm palmito 2500 2500 CupuacËu 223.2 223.2 Rubber 312.5 312.5 Citrus 208.3 208.3 134 G. Schroth et al. / Forest Ecology and Management 163 (2002) 131±150 understorey of palmito palms at a spacing of 2/2 m, trees per unit area. In system 1, the higher stratum was but these were suppressed by shading and are not formed by peach palm for fruit and rubber trees, and further considered here. the lower stratum by peach palm for palmito and In addition, there were three different multistrata cupuacËu. During the ®rst 3 years, papaya (Carica agroforestry systems in the experiment (Table 1 and papaya) was grown as a temporary intercrop in single Fig. 1). Systems 1 and 2 had about the same number of rows between the tree rows. System 2 had fewer palms than system 1. The upper stratum was formed by peach palm for fruit and Brazil nut trees, and the lower stratum by peach palm for palmito, cupuacËu and annatto (Bixa orellana). Annatto was cut back once per year at 150 cm to increase fruit set, and the prunings were left to decompose under the trees. During the ®rst year, cassava (Manihot esculenta) was grown as temporary intercrop between the trees. Multistrata system 3 had a much lower tree density than systems 1 and 2. The upper stratum was formed only by the rubber trees, as the coconut palms belonged to a dwarf variety. Initially, a fast-growing timber tree species, Schizolobium amazonicum, had been planted between the cupuacËu trees, but it had to be removed after 3 years because the stems and branches were often broken by wind and caused damage to the smaller trees. During the ®rst year, cowpea (Vigna unguiculata), maize (Zea mays) and cassava were grown between the trees. The intercrops were fertilized and limed according to local recommendations. The monoculture plots were fertilized according to the tentative recommendations for the fertilization of each tree crop in the region (``full fertilization'', Table 2). During the ®rst 2 years after planting, between 0.2 and 1.5 kg of dolomitic lime were applied to each tree, depending on the species, and in 1996, 2.1 t per ha of dolomitic lime were broadcast in all plots of this fertilization level. In the three multistrata systems, there was a second input level, ``low fertilization'', where only 30% of the fertilizer and lime of the former treatment was applied. From the third year onward (1996), this latter treatment was developed into a ``minimum input'' treatment, where no nitrogen fertilizer was applied to see whether the leguminous cover crop (and the mineralization of soil organic matter) could satisfy the N requirements of the tree crops. Furthermore, no dolomitic lime was applied in this treatment in 1996. Depending on the tree species, Fig. 1. Layout of the three multistrata agroforestry systems in planting density, and fertilization level of trees and central Amazonia. R, rubber trees; P, peach palm for palmito; C, cupuacËu; Pf, peach palm for fruit; B, Brazil nut trees; A, annatto; intercrops, average annual fertilization rates differed Coc, coconut palms; Cit, citrus (for temporary intercrops see text). between cropping systems (Table 3). Note particularly G. Schroth et al. / Forest Ecology and Management 163 (2002) 131±150 135

Table 2 Average annual fertilization with mineral N, P and K of seven tree crop species (gram per plant per year) between 1993 and 1999 at two input levels

Peach palm CupuacËu Rubber Brazil nut Annatto Citrus Coconut

Nitrogen Full fertilization 68 71 38 45 61 126 215 Low fertilization 13 8 8 7 7 19 45 Phosphorus Full fertilization 18 43 30 27 38 90 75 Low fertilization 5 13 9 8 11 27 22 Potassium Full fertilization 63 86 39 63 110 132 331 Low fertilization 19 26 12 19 33 45 95 the high fertilizer application to the closely spaced different size classes of each species for the devel- peach palm monocultures and the relatively low fer- opment of allometric relationships. Trees in the plot tilization of the other monocultures, which was an boundaries were neither included in the inventory nor effect of relatively low planting densities of the trees selected for biomass determination. These activities and the absence of fertilized intercrops. were carried out between December 1999 and May The plots were arranged in a randomized complete 2000. The selected trees were cut at ground level and block design. Three replicate blocks were included separated into stem, branches of diameters >2, 1±2 and in the study. Plot size was 48 m Â 32 m, except for <1 cm, leaves (subdivided into old, medium and the peach palm monocultures, where plot size was young leaves as required for each species) and in¯o- 24 m Â 32 m. rescences. Fruits were not included in the measurements because these are regularly removed from the 2.3. Estimation of aboveground biomass and plots and do not contribute to the accumulated bio- litter mass mass. Each fraction of the trees was weighed and representative samples collected for dry matter deter- The estimation of the aboveground biomass of the mination by drying at 65 (leaves) or 105 8C (woody trees was based on an inventory of the trees (Table 4) structures) to constant weight. Tree mortality was and the destructive harvesting of 7±10 trees from determined for each species and plot and the biomass corrected accordingly. Mortality was less than 10% for all species except peach palm for fruit in monoculture Table 3 Average annual fertilization with mineral N, P, K and dolomite (kg (11%) and peach palm for palmito (11±14%, depend- per ha per year) of three agroforestry systems and four monoculture ing on the cropping system). plantations between 1993 and 1999 The aboveground biomass of the ground vegetation and the litter mass were quanti®ed in May to June N P K Dolomite 2000. In each plot, the ground vegetation and litter Multistrata 1 Full fertilization 112 32 68 541 were collected in the central part of the plots from an Low fertilization 28 10 21 73 area that was representative for the plot with respect to Multistrata 2 Full fertilization 41 24 51 324 the in¯uence of the different tree species and their Low fertilization 7 7 15 8 respective proportion in the system. The size of the Multistrata 3 Full fertilization 32 22 47 835 sample areas per plot varied between 12 m2 in the Low fertilization 6 7 14 161 citrus monocultures and 46 m2 in multistrata system 3. Peach palm 170 49 158 369 For example, in multistrata system 2 the sample area CupuacËu 16 12 19 314 was one quarter of the area between a row of peach Rubber 12 12 12 320 palm, a row of annatto, a Brazil nut and a cupuacËu tree Citrus 26 21 28 342 (Fig. 1). In system 3, several subplots were collected 136 G. Schroth et al. / Forest Ecology and Management 163 (2002) 131±150

Table 4 Number of plots for each land use system and number of trees per plot which were included in the inventory for the estimation of plot biomassa

System No. of plots Trees per plot

Peach palm Peach palm CupuacËu Rubber Brazil nut Annatto Citrus Coconut for fruit for palmito

Multistrata 1 6 6 6 6 6 Multistrata 2 6 6 6 4 6 6 Multistrata 3 6 3 6 3 4 Monocultures 3 Per species 24 18 12 24 12

a Because of missing or dead trees the actual number of measured trees in a speci®c plot could be smaller than indicated. which neighbored a coconut palm, a citrus tree, a reported by Martius et al. (1999a) was used. It was rubber tree and a cupuacËu tree, respectively. The litter measured by randomly collecting the litter from and ground vegetation were weighed separately and twenty circular samples of 21 cm diameter from a dry weight was determined on subsamples. To correct 40 m Â 40 m area. The litter was dried at 65 8C for 4 for contamination of the litter with topsoil, subsamples days. To this value (24.7 t per ha), the result of a single of the litter were ashed at 600 8C and the ash weight estimate of the coarse woody litter in the same area was subtracted from the total dry weight. from November to December 1997 was added (2.3 t The biomass of the now 14-year-old secondary per ha). It was obtained by measuring the diameter of forest, which had been retained when clearing the site all dead wood above 3 cm diameter and 40 cm length for the establishment of the experimental plots, was on the ground. The average diameter was converted to estimated from an inventory of the diameters of all volume assuming a cylindric shape of the wood trees at 130 cm height (dbh) in an area of 60/60 m, (Martius et al., 1999b). For the conversion to weight leaving a boundary of 20 m on all sides (the total area we used an estimated dead wood density of 0.4 g cmÀ3 of the plot was 100 m Â 100 m). The biomass was (Woomer et al., 2000) instead of 0.69 g cmÀ3 as in the estimated from dbh measurements with allometric original paper. relationships developed by Nelson et al. (1999) for a secondary forest with very similar site history and soil 2.4. Estimation of root mass in the region. For trees that had split into several stems at less than 130 cm height the relationships were Root mass of trees and herbaceous vegetation was applied to each individual stem, although this may estimated from their aboveground biomass and root- lead to a certain overestimation of the tree biomass shoot-ratios (RSRs) which had been determined in one (Nelson, personal communication). During the inven- of the plots of multistrata system 2 in 1996 (Haag, tory, the dominating trees were identi®ed to the 1997). For this, the root system of two or three species, and species-speci®c allometric relationships individuals of each tree species in the system was were used if available (Nelson et al., 1999). For the excavated to a depth of 100 cm, and all roots >1 mm other trees, the generic relationship given by these diameter were collected (Table 5). The excavated area authors was used (using this generic relationship for was 2/2 m with the tree in the middle for the dicot all trees in the plot only changed the estimated plot trees, and either 2/2, 1/2, or 1/1 m (with the tree in one biomass by 0.7%). To obtain an estimate of the spatial corner) for the peach palms. For the ground cover, pits variability of the biomass, the secondary forest plot of 1/1 m were excavated. To quantify the mass of was subdivided into 36 quadrats of 100 m2 each, and lateral roots which extended beyond the limits of the means and standard errors were calculated from these excavation pits of the trees, radial transects of 0.5 m 36 subplots. width were excavated from one individual of each As litter mass of the secondary forest, the average of tree species until no further roots of this species monthly measurements of the litter stock between were found. For ®ne roots, cylindric soil samples August 1997 and March 1999 in this forest plot as were collected during the excavation, and additional G. Schroth et al. / Forest Ecology and Management 163 (2002) 131±150 137

Table 5 Ranges of aboveground biomass (kg dry matter per tree) and RSRs of 3-year-old tree crops in central Amazonia, Brazil, and RSRs used for root mass estimates

Species No. of trees Aboveground Root-shoot Ratio used for excavateda biomassa ratioa estimates

Peach palm (fruit) 2 16.0±33.5 0.29±0.49 0.3 Peach palm (palmito) 2 3.6±8.6 0.86±1.88 1.0 Brazil nut 3 7.8±12.1 0.27±0.28 0.25 CupuacËu 3 0.9±3.4 0.32±0.39 0.3 Annatto 3 6.0±13.5 0.29±0.43 0.3

a From Haag (1997). samples were collected from the depth 100±150 cm mass of the rubber trees, as these authors found a on the bottom of the pits. The roots were extracted decrease of the RSR of this species in Vietnam from from subsamples of the cores (Schroth and Kolbe, 0.34 at 3 years age to 0.16 at 7 years age. It is however 1994) by washing over a 0.5 mm sieve. To adjust for possible that RSRs of older trees are underestimated in mineral soil adhering to the roots, the C content of excavation studies by incomplete recovery of the root subsamples was measured with a CN-analyzer and systems. the dry weight of the roots was adjusted to 45% C The cover crop Pueraria phaseoloides was the content. The aboveground dry matter of the excavated dominating ground cover species in most of the plots trees and ground vegetation were determined as of this study (Table 8). For three plots of 1 m2 described above. dominated by this species, the excavations had yielded For peach palm for fruit, Brazil nut, cupuacËu and RSRs of 0.26±0.71, omitting a stray value (0.08) from annatto, the RSRs of the excavated trees are given in a fourth plot (Haag, 1997). This range was in accord Table 5. As RSRs of trees usually decrease with with results from P. phaseoloides fallows of 17±53 increasing age (Fearnside and Guimaraes, 1996; months age in the Peruvian Amazon region (RSR PolinieÁre and van Brandt, 1967), we used ratios close 0.23±1.24, with higher values for the older fallows; to the lower limit of the observed range rather than Szott et al., 1994). Accordingly, we used an RSR of 0.5 average ratios for the root mass estimates (Table 5). for estimating the root mass of the cover crop. As no For peach palm for palmito, the RSRs were very RSR was available for spontaneous ground cover high and differed widely between the two excavated species (Table 8), we used the Pueraria value also for individuals, with a larger value for the smaller plant. these. This was obviously a result of the regular cutting of For the 14-year-old secondary forest, an RSR of the larger offshoots. We used an RSR of 1.0 for this 0.25 was used as given by Fearnside and Guimaraes species, which is closer to that of the larger palm, (1996) for secondary forests of this age in different because the larger plants contribute more to the plot tropical regions. biomass than the small individuals. No RSRs from the study site were available for 2.5. Measurement of soil carbon stocks coconut palms, rubber trees and citrus trees. For the coconut palms we used the RSR of the peach palms for Soil samples were collected between July and fruit (0.3); for citrus we used that of cupuacËu and October 1996, when the experiment was about 3.5- annatto, two small trees of approximately similar size year-old, from multistrata system 2, the monocultures as the citrus trees (equally 0.3); and for the rubber trees of peach palm for fruit and for palmito, the cupuacËu we used the RSR of Brazil nut (0.25) which was the monoculture and a fallow with similar site history as only other large dicot tree species in the experiment. the agricultural plots (including slashing and burning According to PolinieÁre and van Brandt (1967), this in 1992) whose vegetation was dominated by Vismia RSR may have resulted in an overestimate of the root spp. (three replicate plots for each treatment). In the 138 G. Schroth et al. / Forest Ecology and Management 163 (2002) 131±150 multistrata plots, one sample per plot was collected variance for a randomized complete block design. If at 50 cm distance from a representatively chosen indi- an F-test proved signi®cant at P < 0:05, then mean vidual of peach palm for fruit, peach palm for palmito, differences were compared by least signi®cant cupuacËu, Brazil nut and annatto, and one sample difference tests at the same level of signi®cance. under the Pueraria cover crop. In the monocultures The primary and secondary forest data were compared and the fallows, two samples per plot were collec- only qualitatively with those from the agricultural ted. In addition, samples were collected in adjacent plots, because they were not spatially interspersed primary forest at 50 cm distance from six trees with these plots and inferential statistics would not belonging to two species, Eschweilera sp. (Lecythi- have been valid (Hurlbert, 1984). daceae) and Oenocarpus bacaba (Arecaceae). The secondary forest was not included in the sampling program. Soil was collected from the following 3. Results and discussion depths: 0±10, 10±30, 30±50 and then in 50 cm inter- vals down to 2 m. After air-drying and sieving the soil 3.1. Estimation of tree biomass to pass 2 mm, total C was measured with a CNS analyzer. For the ®ve dicot species, cupuacËu, rubber, citrus, To calculate carbon stocks per hectare, the carbon annatto and Brazil nut trees, basal area was an concentration in the soil was multiplied with the bulk adequate predictor of total aboveground dry matter density of the respective soil layer. The bulk densities (Fig. 2). In all cases, second-order polynomial to 2 m soil depth were measured with cylinders of equations gave highly signi®cant approximations of 100 cm3 in one of the plots of multistrata system 2, and the measured data points, with highest r2 values for the the bulk density of the topsoil (1±6 cm) was measured two largest tree species, rubber and Brazil nut in all positions were soil samples had been collected (Table 6). Annatto differed from the other species (see above). Under each sample tree, three cylinders in so far as it reached a saturation value of about 23 kg were collected and were averaged to give one value of dry matter per tree for basal areas above 130 cm2, per tree per plot for the statistical analysis. The presumably as a result of the complete annual pruning average bulk densities under the different species lay at 150 cm height which limited the potential biomass between 0.82 g cmÀ3 under O. bacaba in the primary accumulation of these trees (Fig. 2). For cupuacËu, forest and 1.01 g cmÀ3 under grass weeds in the cup- rubber and citrus, trees from monoculture and uacËu monoculture plots. All other values lay between agroforestry plots could be described with the same 0.86 and 0.96 g cmÀ3. Comparing all positions, the regression equations, indicating that the different species effect on topsoil bulk density was signi®cant at levels of shading in these systems did not affect the P 0:08. After removing the values from the primary allometric relationships. forest, there was no detectable species effect for the The aboveground biomass of individual offshoots of agricultural and fallow plots (P 0:28). We therefore peach palm for palmito could also be estimated from used separate topsoil bulk densities for the primary basal area alone, because prior to the development of a forest (0:84 Æ 0:02 g cmÀ3) and for the agricultural true stem these young shoots grew simultaneously in and fallow plots (0:91 Æ 0:01 g cmÀ3) for the calcula- diameter and height. However, the largest offshoots tion of soil carbon stocks. In the lower soil depths, we showed a more-than-proportional increase in biomass used for all positions the bulk densities from the soil due to the beginning formation of a woody stem, and pro®le, i.e. 0.98 (10±30 cm), 0.92 (30±50 cm), 0.95 this resulted in a sigmoid shape of the regression (50±100 cm), 0.94 (100±150 cm), and 1.00 g cmÀ3 equation (Fig. 2). For very small offshoots the equa- (150±200 cm). tion overestimated the biomass; therefore, the aboveground dry matter of the smallest measured offshoot 2.6. Statistical analysis (0.76 kg, basal area 1.5 cm2) was used for all offshoots with a basal area smaller than 2 cm2. This included a Statistical comparisons of biomass and soil data for proportional section of the cone of aerial roots at the the experimental plots were calculated by analysis of base of the offshoots. G. Schroth et al. / Forest Ecology and Management 163 (2002) 131±150 139

Fig. 2. Allometric regressions for eight fruit tree species grown under agroforestry and monoculture conditions in central Amazonia after 7 years in the ®eld (for regression equations see Table 6). Data points marked ``M'' indicate trees taken from monoculture plots, all other trees were taken from multistrata plots.

For peach palm for fruit, the biomass could not be tall palms. For the dwarf coconut palms, in contrast, the estimated from basal area alone, but was closely related stem contributed only 32 Æ 2% to the aboveground dry to the product of basal area and height (as an matter (n 7). To estimate the biomass of these palms, approximation of the stem volume; Fig. 2). This was it was therefore necessary to develop an index which because the stem contributed 80 Æ 2% (mean and S.E. included both the number of leaves and the height and of the 10 measurement trees) to the total weight of these basal area of the stem (Fig. 2 and Table 6). 140 G. Schroth et al. / Forest Ecology and Management 163 (2002) 131±150

Table 6 Regression equations for the estimation of aboveground dry matter (kg per tree) of eight fruit tree species in central Amazonia after 7 years in the ®elda

2 Species Equation Units HBA (cm) r Peach palm fruitb DM 40.6 121 BAH 391 (BAH)2 BAH, m3 130 0.97 Peach palm palmitoc DM 0.97 0.078 BA À 0.00094 BA2 0.0000064 BA3 BA, cm2 100 0.98 CupuacËuDMÀ3.9 0.23 BA 0.0015 BA2 BA, cm2 30 0.93 Rubber DM À3.84 0.528 BA 0.001 BA2 BA, cm2 150 0.99 Citrus DM À6.64 0.279 BA 0.000514 BA2 BA, cm2 30 0.94 Annatto DM À21.7 0.505 BA À 0.00143 BA2 BA, cm2 20 0.95 Brazil nut DM À18.1 0.663 BA 0.000384 BA2 BA, cm2 130 0.99 Coconutd DM À35.7 1.73 SLI See footnote 100 0.98

a For number of data points and range of validity of the equations see Fig. 2. DM, aboveground dry matter; BA, basal area; BAH, basal area 2 times height; SLI, stem-leaf-index; HBA, measurement height of basal area (cm). All r were signi®cant at P < 0:003 or higher. b Equation includes trees and their offshoots if present. c Equation is for individual offshoots and corresponding section of the cone of aereal roots. d The stem-leaf-index (SLI) is calculated from the basal area (BA, in cm2), height (H, in cm) and number of leaves (L) according to: SLI 0:000206 BAH 1:4L.

3.2. Effect of fertilizer level and cropping system grown, showed a substantial enrichment with nutrients on tree biomass such as available P, Ca and Mg (Schroth et al., 2001a). There was also a bene®cial effect of the papaya on the The aboveground biomass of the tree crop species leguminous cover crop, P. phaseoloides, which had varied between less than 10 kg per tree for peach palm shown a more vigorous development in system 1 than for palmito and more than 200 kg per Brazil nut tree in the other treatments during the ®rst years of the (Fig. 3). The tree species differed in their response to experiment, presumably also as a side-effect of the fertilization. On the average of the different multi- fertilizer applied to the intercrop. As a consequence, strata systems, signi®cant increases in biomass at the grasses and other weeds were effectively suppressed in higher fertilizer level were measured for cupuacËu this system during the early development of the tree (P 0:009), rubber (P 0:019) and annatto trees crops, when these were particularly sensitive to weed (P 0:048). The Brazil nut trees also tended to be competition (Schroth, personal observation). larger at the higher fertilizer level. The effect was not The trees in the monocultures tended to be smaller signi®cant for the data presented here (P 0:13), but than those in the multistrata systems at the same was signi®cant for a larger data set from the same site fertilizer input per tree, irrespective of whether they (P 0:041, data not shown). were canopy trees such as rubber and peach palm for Peach palm for fruit did not respond to the fertilizer fruit, or understorey trees such as cupuacËu and citrus level, but showed signi®cantly better growth in (Fig. 3). For the canopy trees this could re¯ect less multistrata system 1 than in system 2 (P 0:004, competition for light in association with smaller trees, mean of both fertilizer levels). The growth of the but the understorey trees were clearly more shaded in rubber trees was also signi®cantly faster in multistrata the multistrata systems than in monoculture. The system 1 than in system 3 (P 0:019). The relatively likely reason for the relatively slow growth of the trees fast growth of these two tree crops in system 1 was in the monocultures was the relatively low tree density presumably an effect of residual fertilizer from the and absence of fertilized intercrops, which resulted in papaya intercrop which had been grown in the lower per-hectare fertilizer inputs than in the multi- interrow spaces during the ®rst 3 years. Because of strata systems at full fertilization level (Table 3). As the fertilizer applied to the papaya, the per-hectare the fertilizer was applied to the individual trees, the fertilizer input in multistrata system 1 was substan- young tree crops in the monocultures developed in tially higher than in systems 2 and 3 (Table 3), and the small fertilized islands surrounded by unfertilized and soil between the trees, where the papaya had been often grass-invaded soil. This explanation does not G. Schroth et al. / Forest Ecology and Management 163 (2002) 131±150 141

Fig. 3. Aboveground dry matter of eight fruit tree species grown under agroforestry and monoculture conditions in central Amazonia after 7 years in the ®eld as affected by two fertilizer levels (mean and S.E.). apply to the peach palm for fruit monoculture, which per ha, a belowground biomass of 4.3±12.9 t per ha, received the highest fertilizer input from all systems, and a litter mass of 2.3±7.2 t per ha, depending on yet the palms were smaller than in multistrata system 1 species composition and fertilization (Fig. 4). In (P 0:052, Fig. 3). This can be explained with the accord with the effect of fertilizer on the growth of the high planting density and intraspeci®c competition individual tree species, the biomass and litter between the palms in monoculture. accumulation of the three systems was signi®cantly higher at full than at low fertilization (P 0:013). 3.3. Biomass and litter accumulation in different With 46.3 and 58.4 t per ha at low and full fertilization, land use systems and secondary forest respectively, system 2 had the highest biomass and litter accumulation from the three multistrata systems, During 7 years in the ®eld, the multistrata systems followed by system 1 (45.2 and 51.1 t per ha). System accumulated an aboveground biomass of 13.2±42.3 t 3 had only 23.4 and 31.6 t per ha of biomass and litter, 142 G. Schroth et al. / Forest Ecology and Management 163 (2002) 131±150

Fig. 4. Above- and belowground biomass and litter accumulation in three multistrata agroforestry systems and ®ve monocultures in central Amazonia 7 years after their establishment in the ®eld, and of the 14-year-old secondary forest that would have occupied the area in the absence of the plantations (mean and S.E.). For the three multistrata systems (MS), the left column indicates the low and the right column the full fertilization level. The difference between land use systems (without secondary forest) was signi®cant at P < 0:001. The error bar shows one LSD (P 0:05). Pf, peach palm for fruit; P, peach palm for palmito; C, cupuacËu; R, rubber trees; Cit, citrus; Sec, secondary forest. re¯ecting its lower tree density compared with the the monocultures of cupuacËu, rubber and citrus other two systems. The open structure of this system received less fertilizer than multistrata system 2 at allowed a vigorous development of the ground vege- full but not at low fertilization (Table 3). They were tation, which, together with the large coconut fronds, rather an effect of planting density, individual tree led to the highest litter mass from all systems (7.2 t per growth rates and, especially for the palmito palms, tree ha at full and 5.9 t per ha at low fertilization). management. Although system 3 also received the lowest fertilizer None of the systems had reached the biomass and input per hectare, the high biomass of system 2 was litter accumulation of the secondary forest that they not an effect of high fertilization (Table 3) but rather of had replaced (127.5 t per ha, Fig. 4 and Table 7). In its species composition, speci®cally the presence of addition to a total aboveground biomass of 80.4 t per the large Brazil nut trees (see below). ha, i.e. 1.4 times more than that of the peach palm for In the monoculture systems, the range of observed fruit monoculture, the secondary forest had a very biomass and litter accumulations was larger than in the large litter mass (27 t per ha, including 2.3 t per ha of multistrata systems, with 7.7±56.7 t per ha of above- coarse woody litter) (Martius et al., 1999a,b). The ground biomass, 3.2±17.1 t per ha of root biomass and biomass of the secondary forest was lower than that 1.9±5.6 t per ha of litter (Fig. 4). Comparing all expected from fallow growth after shifting cultivation systems and fertilization levels, the peach palm for in various tropical regions, but was similar to or higher fruit monoculture had the largest biomass and litter than that reported for secondary forests growing after accumulation (78.8 t per ha), followed by multistrata pasture in Amazonia (Fearnside and Guimaraes, systems 2 and 1. The remaining systems had all less 1996). The dominating trees belonged mostly to the than 40 t per ha, with lowest values for the mono- genera Miconia, Bellucia and Vismia. In addition, cultures of citrus (15.3 t per ha) and cupuacËu (20.0 t rubber trees from the previous plantation contributed per ha). These differences could not be explained with 11% to the total biomass of the secondary forest fertilizer inputs, as the palmito monoculture received (Table 7). Primary forests growing under the same one of the highest fertilizer levels from all systems and pedoclimatic conditions in the region have a mean G. Schroth et al. / Forest Ecology and Management 163 (2002) 131±150 143

Table 7 Characteristics of a 14-year-old secondary forest, developed from a 5-year-old, abandoned rubber plantation on a xanthic Ferralsol in central Amazonia

Trees per ha 6070 Æ 280a Stems per ha 7180 Æ 300a Rubber trees per hab 206 Æ 34a Basal area (130 cm; m2 per ha) 16.9 Æ 0.9a Dominating tree species Miconia pyrifolia Burseraceae Vismia japurensis Clusiaceae Vismia cayennensis Clusiaceae Vismia guianensis Clusiaceae Vismia gracilis Clusiaceae Bellucia grossularioides Melastomataceae Bellucia dichotoma Melastomataceae Rollinia exsucca Annonaceae Simaruba amara Simaroubaceae Vochysia vismiifolia Vochysiaceae Goupia glabra Celastraceae H. brasiliensisb Euphorbiaceae

a Mean and S.E. of 36 plots of 100 m2. b Remains of the previous rubber plantations. aboveground biomass of about 350 t per ha (Laurance Achyranthes sp.), which did not stand out in the other et al., 1999). plots, whereas the leguminous cover crop P. phaseoloides was largely shaded out. With their open cano- 3.4. Contribution of different tree species and pies, the rubber and cupuacËu monocultures allowed ground cover to biomass accumulation the formation of a denser ground cover, and in the citrus monoculture 60% of the biomass of the plots Most of the biomass of the multistrata systems was consisted of a vigorous layer of P. phaseoloides. concentrated in the large trees, especially the fruit palms, rubber and Brazil nut trees (Table 8). The 3.5. Soil carbon biomass of system 2 consisted of almost 60% of Brazil nut trees, although these contributed with only 16% to Fig. 5a shows the carbon concentrations in the soil the number of trees in the plots. The peach palms for to a depth of 200 cm in multistrata system 2, three palmito and the annatto trees, which were regularly cut monoculture systems, fallow and primary forest. In back, only made a small contribution to the respective the top 10 cm of soil, the carbon concentration was plot biomass. The ground cover added about 5% to the higher in the forest and lower in the fallow than in biomass in multistrata systems 1 and 2, and 20% in the the agricultural plots, suggesting that conversion of opener system 3. In systems 1 and 3 it consisted primary forest and subsequent slashing and burning of mainly of the planted cover crop P. phaseoloides, and secondary vegetation was accompanied by some in system 2 of the spontaneous Clidemia sp. (see carbon loss from the topsoil. Lower carbon concen- footnote to Table 8). trations in the top 50 cm of soil under a tree crop plan- The wide range of relative contributions of trees and tation (oil palm) than under adjacent primary forest ground cover to the biomass of the monocultures have also been reported from another central Ama- re¯ects the contrasting character of these agroeco- zonian site with similar soil (Schroth et al., 2000b). systems (Table 8). The palm monocultures, either for Similarly, van Noordwijk et al. (1997) reported fruit or for palmito, were completely dominated by the decreasing carbon contents in the topsoil from primary tree crops. Under the dense shade of the palms, the forest to secondary forest to perennial crops in ground vegetation was characterized by an apparently Indonesia. However, the carbon concentrations below shade-tolerant species of Amaranthaceae (probably 10 cm were very similar under all vegetation types Table 8 Contribution of different tree species and ground cover in percent to the total above- and belowground biomass of three multistrata agroforestry systems and ®ve monoculture plantations in central Amazonia (mean and S.E.)a

System Peach palm Peach palm CupuacËu Rubber Brazil nut Annatto Citrus Coconut palm Ground coverb for fruit for palmito

Multistrata 1 48.0 Æ 3.0 (22) 5.1 Æ 0.4 (43) 10.0 Æ 1.1 (14) 32.2 Æ 2.4 (22) 4.7 Æ 0.8 Multistrata 2 18.6 Æ 1.1 (14) 2.6 Æ 0.5 (27) 8.6 Æ 0.9 (16) 59.0 Æ 2.3 (16) 6.3 Æ 0.5 4.9 Æ 0.7 Multistrata 3 10.2 Æ 0.9 (13) 34.1 Æ 4.9 (40) 14.2 Æ 2.8 (27) 21.9 Æ 1.9 (20) 19.5 Æ 2.1 Monocultures Peach palm fruit 99.3 Æ 0.6 0.7 Æ 0.6 Peach palm palmito 91.2 Æ 1.4 8.8 Æ 1.4 CupuacËu 63.4 Æ 3.4 36.6 Æ 3.4 Rubber 82.3 Æ 4.8 17.7 Æ 4.8 Citrus 39.3 Æ 2.8 60.7 Æ 2.8

a Values in parenthesis denote the percent contribution of the respective species to the total number of trees in a multistrata system (see Table 1). In the monocultures this contribution is always 100%. b In multistrata systems 1 and 3 and the citrus monoculture the ground cover was dominated by the planted cover crop, P. phaseoloides; in system 2 by Clidemia sp.; in the rubber and cupuacËu monoculture by a mixture of P. phaseoloides, Clidemia sp. and grass (mainly in the monoculture of cupuacËu); and in the peach palm monocultures by an Amaranthacea (probably Achyranthes sp.) and Clidemia sp. G. Schroth et al. / Forest Ecology and Management 163 (2002) 131±150 145

Fig. 5. Carbon concentration in a xanthic Ferralsol to 2 m depth under (a) different vegetation types and (b) different tree and cover crop species in multistrata agroforestry system 2 (means of two fertilization levels). Error bars depict 1 S.E. of the mean difference for the respective depth interval.

(Fig. 5a), indicating that the effect of forest conversion The average carbon stock in the top 10 cm of soil and site management on soil carbon concentrations varied between 22:5 Æ 1:3 t per ha in the primary was largely restricted to the topsoil. forest, 21:6 Æ 0:6 t per ha in the multistrata systems, Within the multistrata plots, there was a tendency 19:8 Æ 0:5 t per ha in the monocultures and 17:2 Æ for higher carbon concentrations in the topsoil under 1:6 t per ha in the fallow (mean and S.E.), re¯ecting the two tree species with recalcitrant leaf litter, the different carbon concentrations in the topsoil and, cupuacËu and Brazil nut, compared with the species in the forest, the slightly lower bulk density. However, with more readily decomposable litter, peach palm, the carbon stocks for the complete pro®les to 200 cm annatto and Pueraria (Fig. 5b). Under the former two depth were very similar under all vegetation types, species the carbon concentrations in the topsoil were with 107 Æ 2 t per ha in the primary forest, 113 Æ 2t in fact very similar to those under primary forest per ha in the multistrata plots, 111 Æ 2t perhain (Fig. 5a). Although the differences between species in the monocultures of peach palm and cupuacËu, and the multistrata plots were not signi®cant (P 0:13), 112 Æ 2 t per ha in the fallow. There were also no they tended to con®rm results by Lehmann et al. signi®cant differences between tree species within the (2001) who found more particulate organic matter in multistrata plots (data not shown). This re¯ected the the soil under cupuacËu than under Pueraria and (non- lack of major differences in soil carbon concentrations signi®cantly) peach palm in the same plots and below 10 cm depth (Fig. 5). Also, sampling positions explained this with the low quality of the cupuacËu with high carbon contents in the topsoil tended to have litter (the soil under Brazil nut was not included in this lower carbon contents in the subsoil and vice versa study). (see forest and fallow in Fig. 5a and cupuacËuin 146 G. Schroth et al. / Forest Ecology and Management 163 (2002) 131±150

Fig. 5b), so that differences were canceled out when carbon stock would increase by half the annual rate of summing over the entire pro®les. carbon accumulation for every year it is left growing The soil carbon stocks could not be determined (2 t per ha per yr; Palm et al., 2001), e.g., 30 t per ha for all study plots, as there were no samples from for a 15-year rotation. Regrowth periods of succes- multistrata systems 1 and 3, the rubber and citrus sional vegetation are highly variable in central monocultures and from the secondary forest. How- Amazonia, with up to about ten years, but very often ever, the similarity of soil carbon stocks for such less than ®ve years, for shifting cultivation fallows in contrasting vegetation types as primary forest, young settlement areas to decades if agricultural land is fallow and diverse agricultural systems suggests that abandoned (G.F. da Souza, pers. comm.). differences were also unlikely for the other systems. The peach palm for fruit monoculture accumulated For the present comparison of carbon pools in carbon faster than the forest (5.1 t per ha per yr) due to perennial, tree-dominated vegetation types, the soil the close spacing of a fast-growing variety and carbon pool could thus be regarded as essentially substantial input of fertilizers. However, with an constant. average height of the palms of 12 m after seven years, harvesting the fruit bunches became increasingly 3.6. Effects of land use options on carbon dif®cult and damaging to the fruits. For this reason, accumulation a plantation would have to be regenerated at a maximum age of ten years, i.e. with a time-averaged Seven years after the establishment of several carbon stock of about 25 t per ha. Its establishment on multistrata agroforestry systems and tree crop mono- secondary forest land would thus lead to net carbon cultures on an abandoned, secondary forest site in sequestration if the regrowth period of the forest was central Amazonia, none of the systems had reached the less than 12±13 years (many fallows) and to net carbon combined biomass and litter stock of the secondary release if it was longer (abandoned areas). forest that they had replaced, whereas the soil carbon The other monocultures accumulated carbon at stock seemed to be largely unaffected by the land use much lower rates, i.e. 1.0 t per ha per yr for citrus transformations. On the short term, an environmental (a large part of which was ground vegetation), cost of the forest conversion into cash crop plantations 1.3 t per ha per yr for cupuacËu and 2.5 t per ha per yr was thus the net release of different amounts of carbon for rubber. Citrus plantations hardly reach an age of into the atmosphere. Assuming a carbon content of the more than ten years in central Amazonia because of biomass and litter of 45%, it ranged from 22 t per ha the high disease pressure, and their carbon stocks thus for the peach palm for fruit monoculture to about remain low. CupuacËu trees can become very old as 50 t per ha, i.e. 80±90% of the carbon in biomass and long as witches' broom disease (Crinipellis perni- litter of the secondary forest, for the monocultures of ciosa) is controlled; however, this requires regular peach palm for palmito, cupuacËu and citrus. On the pruning of affected crown parts and the trees thus have longer term, or if seen over larger areas with fallows to be kept small. The carbon stock of the palmito and plantations of different ages, the net effect of monoculture of 11±12 t per ha had probably already forest conversion depends on the time-averaged been reached a few years after planting and further carbon stocks for the whole growth cycle of the forest accumulation was prevented by the periodic harvests or fallow and the land use systems by which it is of the large offshoots. Substitution of successional replaced (see Palm et al., 2001). It is a function of the vegetation by any of these monocultures would thus rate of biomass accumulation (which may change over lead to net carbon release unless this vegetation was time) and the rotation length, i.e. the time until an area managed in a very short fallow rotation. Rubber is clear-felled and replanted. plantations have an economic life of several decades The secondary forest accumulated carbon in and may accumulate large carbon stocks (Palm et al., biomass and litter at a rate of about 4 t per ha per yr; 2001; PolinieÁre and van Brandt, 1967), but the an exact value cannot be given because of the presence cultivation of this tree crop in central Amazonia is of the older rubber trees (Table 7). Assuming a cons- still severely hindered by the endemic fungal disease, tant rate of carbon accumulation, its time-averaged South American leaf blight (Microcylus ulei). G. Schroth et al. / Forest Ecology and Management 163 (2002) 131±150 147

From the viewpoint of the carbon balance, multi- species because of yield reductions in the understorey strata agroforestry was an interesting alternative to the at higher tree densities. Brazil nut trees, however, form monocultures. The fastest-growing system 2 accumu- rather large crowns if grown under agroforestry lated 3.8 and 3.0 t per ha per yr of carbon in the full conditions. In multistrata systems where the main and the low-fertilization treatment, respectively. The economic output depends on the understorey, there is corresponding values for system 1 were 3.3 and thus a trade-off between high carbon accumulation 2.9 t per ha per yr. Carbon accumulation rates of and economic production. 3.6 t per ha per yr have been reported for cocoa- and An alternative scenario would be that the develop- rubber-based agroforests in Cameroon and Indonesia, ment of the multistrata systems is seen as a succ- although these did not include root carbon (Palm et al., essional process, in which yield depressions of more 2001). These relatively high rates compared to most shade-tolerant understorey trees (cupuacËu) are accep- monocultures were due to a relatively high tree density ted, and shade-sensitive understorey trees (annatto) (Table 1) and to the association of the smaller tree are progressively removed, as their production is crops, such as cupuacËu and peach palm for palmito, increasingly complemented by that of the overstorey. with larger and faster-growing trees, i.e. peach palm This would include the rubber trees, the second for fruit, rubber and Brazil nut trees. On the longer generation of peach palm for fruit, and the Brazil nut term, the lower carbon accumulation rate of the trees many of which started to fruit in the eighth year multistrata systems compared with the peach palm for after planting. With this view, the overstorey could be fruit monoculture would tend to be compensated by left growing more or less unchecked, except for the longer economic life cycle of some species in the pruning of the Brazil nut trees to improve the quality agroforestry systems, especially the Brazil nut trees of their timber. The possibility to accumulate high in system 2 and the rubber trees in system 1. The other carbon stocks in multistrata systems without negative species could be regenerated on a species-per-species impacts on total economic output depends thus on the basis without ever clear-felling the plots. combination of shade-tolerant understorey with high- The future growth rates and rotation lengths of the value overstorey trees. multistrata systems, and thus their time-averaged From an agronomic point of view, the association of carbon stocks, would however depend on a number of crops with different production cycles in the multi- factors, including the production objectives and strata systems proved to be particularly advantageous relative economic values of over- and understorey during the early years after their establishment. In trees, and are therefore dif®cult to predict. In both system 1, the temporary papaya intercrop allowed not systems 1 and 2, the peach palms for fruit would soon only a more ef®cient use of the space and the early have to be regenerated, which would reduce the generation of income, but also an improved vegetative system carbon stocks (Table 8) but could stimulate the development of the longer-living tree crops (Fig. 3). In growth of associated species. Biomass accumulation the case of cupuacËu, this association also allowed in system 2 depended very much on the vigorous signi®cantly higher fruit yields than in monoculture growth of the Brazil nut trees, which had reached an during the ®rst three harvests (Schroth et al., 2001b). average dbh of 20.2 cm and a height of 10±15 m seven In some of the monocultures, in contrast, a consider- years after planting. However, fruit production of the able part of the space, and therefore light and soil understorey trees, cupuacËu and annatto, declined in resources, remained underutilized by the tree crops the last years of the experiment, at least partly because still seven years after the establishment of the of shading by the Brazil nut trees. To maintain plantations. This is illustrated by a contribution of satisfactory yields in the understorey, these would, as a the ground cover of more than 60% to the total plot minimum, have to be drastically pruned to allow more biomass in the citrus monoculture, far more than in light entry into the lower strata, but eventually the any of the multistrata systems (Table 8). It has been overstorey might also need to be thinned out. Beer shown previously that under the pedoclimatic condi- (1992) found that in shaded coffee and cocoa tions of the region, insuf®cient occupation of the soil plantations the number of mature timber trees should by the root systems of the young tree crops leads to be limited to 100 per ha even for small-crowned leaching of nitrate into the subsoil of the intertree 148 G. Schroth et al. / Forest Ecology and Management 163 (2002) 131±150 spaces, and consequently to cation loss and further Alternatively, the total carbon stock in the soil to 2 m soil acidi®cation (Schroth et al., 1999; Schroth et al., depth could only be less sensitive than the carbon 2000b). A more complete utilization of the space in content of the topsoil as a measure for soil organic young tree crop plantations cannot only be achieved matter loss over a relatively short time period. In this through crop associations as in agroforestry, but also case, the data in Fig. 5a could indicate the progressive by planting the tree crops in monoculture at increased release of carbon from soil organic matter as a long- density and thinning the stand in later years (Nakasone term consequence of tree crop agriculture (and and Paull, 1998). However, this automatically implies agroforestry) on forest land, as reported earlier from the loss of fruit trees, and thus of the money and labour other regions (Ahenkorah et al., 1987; Ollagnier et al., that was invested in their establishment, and seems, 1978; van Noordwijk et al., 1997). The data seem to therefore, to be less adapted to the requirements of support the former rather than the latter explanation, tropical smallholders than temporary or permanent but the question certainly requires further study. These intercropping and agroforestry. results also underline the need for adequate sampling Increased fertilizer input and the integration of depth when investigating the effect of land use on soil fertilized intercrops increased the biomass accumula- carbon stocks (Hamburg, 2000). tion in the systems (Fig. 4), beside improving the A potentially signi®cant result of the present study yields of certain tree crops such as annatto (Schroth is that the tree crops with low litter quality, cupuacËu et al., 1999) and cupuacËu (Schroth et al., 2001b). It is and Brazil nut, were able to build up and maintain clear that for a full carbon balance of the systems, organic matter levels in the topsoil comparable to energy investments into the production and transport those in the primary forest, even when they were of such external inputs would also have to be grown in association with tree and cover crops which considered. However, this is presently hindered by produced easily decomposable litter. When integrated uncertainties about the economically optimal fertili- into multistrata agroforestry systems, such tree crops zation levels for most if not all of the tree crops in this could act as an insurance against soil organic matter study under Amazonian conditions. Of particular loss, in addition to their direct production role. interest for the energy and carbon budgets of the land use systems is, however, that several tree crops did not respond to mineral nitrogen fertilizer at this site 4. Conclusions (Schroth et al., 1999), nor at another site with similar soil in Amazonia (Schroth et al., 2000b). On an infertile upland soil in central Amazonia, the In all of the investigated plantation systems, there successional vegetation accumulated carbon in above- was more than twice as much carbon in the soil orga- and belowground biomass and litter at a rate of about nic matter than in the biomass and litter combined. 4 t per ha per yr. The tested tree crop monocultures Changes in the soil organic matter stocks could, differed widely in their ability to compensate for the therefore, be of crucial importance for the net carbon release of this carbon, with highest values for peach effect of land use transformations (Sombroek et al., palm for fruit (due to its fast growth) and rubber trees 2000). However, in this study we found no effects of (due to their long potential life cycle, provided that vegetation types and plant species on the organic diseases can be controlled). Other monocultures had matter stocks of the soil pro®le to 2 m depth, although much lower growth rates than the successional vege- there were indications for effects on the carbon content tation. Multistrata agroforestry systems combined of the topsoil. There are two possible explanations for relatively high growth rates with a long economic this. First, the conversion of primary forest into life cycle of the system, which would allow them to different tree crop plantations could have affected the accumulate more biomass than several common tree distribution of carbon in the soil, but not its total crop monocultures of the region. quantity. Such changes could arise through an altered Whether the establishment of a multistrata (or other distribution of root mass in the soil pro®le, or through tree crop-based) system reduces or increases total car- differences in the abundance and activity of burrowing bon stocks depends of course on the vegetation that is soil fauna between vegetation types and plant species. replaced. Ideally, such systems should be established G. Schroth et al. / Forest Ecology and Management 163 (2002) 131±150 149 on sites with low standing biomass, such as degraded Beer, J., 1992. Production and competitive effects of the shade trees pastures or other degraded areas, whereas older and Cordia alliodora and Erythrina poeppigiana in an agroforestry system with Coffea arabica. Ph.D. Thesis. University of vigorously growing secondary forests should be con- Oxford, Oxford. served. However, Amazonian pastures are often char- FAO±UNESCO, 1990. Soil Map of the World. Food and acterized by topsoil compactions and soil erosion Agriculture Organization of the United Nations, Rome, 119 (Fearnside, 1985; Koutika et al., 1997), and the growth pp. (revised). and yields of tree crops would probably be reduced on Fearnside, P.M., 1985. Agriculture in Amazonia. In: Prance, G.T., Lovejoy, T.E. (Eds.), Key Environments: Amazonia. Pergamon, such sites. New York, pp. 393±418. Aiming at a rational use of land and biological Fearnside, P.M., Guimaraes, W.M., 1996. Carbon uptake by resources in a tropical forest environment implies that secondary forests in Brazilian Amazonia. For. Ecol. Manage. whenever forest needs to be converted for agriculture 80, 35±46. to satisfy the needs of the local population, the resul- Fujisaka, S., White, D., 1998. Pasture or permanent crops after ting land use systems should be productive in the long slash-and-burn cultivation? Land-use choice in three Amazon colonies. Agrofor. Syst. 42, 45±59. term, i.e. sustainable. Multistrata agroforestry based Gasparotto, L., Santos, A.F., Pereira, J.C.R., Ferreira, A.F., 1997. on site-adapted tree crops is a promising option to DoencËas da Seringueira no Brasil. Embrapa, BrasõÂlia, 168 pp. achieve this. Although the availability of sustainable Haag, D., 1997. Root distribution patterns in a polycultural system land use options does not automatically lead to less with local tree crops on an acid upland soil in central Amazonia. Masters thesis, University of Bayreuth, Bayreuth, clearing of primary forest, it is certainly a precondition 88 pp. for it. Offering the perspective of long-term agricul- Hamburg, S.P., 2000. Simple rules for measuring changes in tural production from the same piece of land may be ecosystem carbon in forestry-offset projects. Mitigation the most signi®cant contribution of agroforestry and Adaptation Strategies Global Change 5, 25±37. tree crop agriculture to the regional and global carbon Houghton, R.A., Skole, D.L., Nobre, C.A., Hackler, J.L., Lawrence, K.T., Chomentowski, W.H., 2000. Annual ¯uxes budgets (Sanchez et al., 1997; Unruh et al., 1993). The of carbon from deforestation and regrowth in the Brazilian effect of forest conversion into tree crop-based sys- Amazon. Nature 403, 301±304. tems on soil organic matter stocks and the potential of Hurlbert, S.H., 1984. Pseudoreplication and the design of certain tree crops to build up and maintain high orga- ecological ®eld experiments. Ecol. Monogr. 54, 187±211. nic matter levels in the soil deserve further attention. Koutika, L.-S., Bartoli, F., Andreux, F., Cerri, C.C., Burtin, G., ChoneÂ, T., Philippy, R., 1997. Organic matter dynamics and aggregation in soils under rain forest and pastures of increasing age in the eastern Amazon Basin. Geoderma 76, 87±112. Acknowledgements Laurance, W.F., Fearnside, P.M., Laurance, S.G., DelamoÃnica, P., Lovejoy, T.E., Rankin-de Merona, J., Chambers, J.Q., Gascon, This research was funded by the German Ministry C., 1999. Relationships between soils and Amazon forest of Education and Research (BMBF) together with the biomass: a landscape-scale study. For. Ecol. Manage. 118, Brazilian Conselho National de Desenvolvimento 127±138. CientõÂ®co e TecnoloÂgico (CNPq) as part of the SHIFT Laurance, W.F., Cochrane, M.A., Bergen, S., Fearnside, P.M., DelamoÃnica, P., Barber, C., D'Angelo, S.A., Fernandes, T., program, project 0339641/ENV 23/45. Rachel Rodri- 2001. The future of the Brazilian Amazon. Science 291, gues and Lailton ArauÂjo made the secondary forest 438±439. inventory, Luiz Coelho the botanical identi®cations, Lehmann, J., Cravo, M.S., Zech, W., 2001. Organic matter and Michaela Schaller the soil carbon analyses. The stabilization in a xanthic Ferralsol of the central Amazon as Ã affected by single trees: chemical characterization of density, Embrapa Amazonia Ocidental provided the site and aggregate, and particle size fractions. Geoderma 99, 147±168. infrastructure for the experiment. Martius, C., Garcia, M.V.B., Hoefer, H., Hanagarth, W., 1999a. Litter production, litter stocks and decomposition coef®cients in a central Amazonian rain forest, a secondary forest and References agroforestry systems. In: Soil Fauna and Litter Decomposition in Primary and Secondary Forests and a Mixed Culture System Ahenkorah, Y., Halm, B.J., Appiah, M.R., Akro®, G.S., Yirenkyi, in Amazonia. Final Report 1996±1999, Staatliches Museum fuÈr J.E.K., 1987. Twenty years' results from a shade and fertilizer Naturkunde, Karlsruhe, pp. 61±70. trial on Amazon cocoa (Theobroma cacao) in Ghana. Expl. Martius, C., de Morais, J.W., Garcia, M.V.B., 1999b. Dead wood Agric. 23, 31±39. volume in primary forest, secondary forest and an agroforestry 150 G. Schroth et al. / Forest Ecology and Management 163 (2002) 131±150

plantation system in central Amazonia. In: Soil Fauna and under perennial cropping, fallow and primary forest in central Litter Decomposition in Primary and Secondary Forests and Amazonia. Eur. J. Soil Sci. 51, 219±231. a Mixed Culture System in Amazonia. Final Report 1996± Schroth, G., Rodrigues, M.R.L., D'Angelo, S.A., 2000b. Spatial 1999, Staatliches Museum fuÈr Naturkunde, Karlsruhe, pp. 71± patterns of nitrogen mineralization, fertilizer distribution and 73. roots explain nitrate leaching from mature Amazonian oil palm Nakasone, H.Y., Paull, R.E., 1998. Tropical Fruits. CAB Interna- plantation. Soil Use Manage. 16, 222±229. tional, Wallingford, 445 pp. Schroth, G., Salazar, E., da Silva Jr., J.P., 2001a. Soil N Nelson, B.W., Mesquita, R.C.G., Pereira, J.L.G., de Souza, S.G.A., mineralization under tree crops and a legume cover crop in Batista, G.T., Couto, L.B., 1999. Allometric regressions for multi-strata agroforestry in central Amazonia: spatial and improved estimate of secondary forest biomass in the central temporal patterns. Expl. Agric. 37, 59±73. Amazon. For. Ecol. Manage. 117, 149±167. Schroth, G., Lehmann, J., Rodrigues, M.R.L., Barros, E., MaceÃdo, Ollagnier, M., Lauzeral, A., Olivin, J., Ochs, R., 1978. Evolution J.L.V., 2001b. Plant±soil interactions in multistrata agroforestry des sols sous palmeraie apres deÂfrichement de la foreÃt. in the humid tropics. Agrofor. Syst., in press. OleÂagineux 33, 537±547. SerraÄo, E.A.S., Nepstad, D.C., Walker, R., 1996. Upland agricul- Palm, C.A., van Noordwijk, M., Woomer, P.L., Alegre, J.C., tural and forestry development in the Amazon: sustainability, Arevalo, L., Castilla, C., Cordeiro, D.G., Hairiah, K., Kotto- criticality and resilience. Ecol. Econ. 18, 3±13. Same, J., Moukam, A., Parton, W.J., Ricse, A., Rodrigues, V., Sombroek, W.G., Fearnside, P.M., Cravo, M.S., 2000. Geographic Sitompul, S.M., 2001. Carbon losses and sequestration assessment of carbon stored in Amazonian terrestrial ecosys- following land use change in the humid tropics. ASA Special tems and their soils in particular. In: Lal, R., Kimble, J.M., Publication, in press. Stewart, B.A. (Eds.), Global Climate Change and Tropical PolinieÁre, J.P., van Brandt, H., 1967. Production en matieÁre seÁche et Ecosystems. CRC Press, Boca Raton, pp. 375±389. autres caracteÁres veÂgeÂtatifs de greffes d'heÂveÂa en fonction de Szott, L.T., Palm, C.A., Davey, C.B., 1994. Biomass and litter leur age. Bois ForeÃts Trop. 112, 39±51. accumulation under managed and natural tropical fallows. For. Sanchez, P.A., Buresh, R.J., Leakey, R.R.B., 1997. Trees, soils, Ecol. Manage. 67, 177±190. and food security. Phil. Trans. R. Soc. London B 352, 949± Unruh, J.D., Houghton, R.A., Lefebvre, P.A., 1993. Carbon storage 961. in agroforestry: an estimate for sub-Saharan Africa. Clim. Res. Schroth, G., Kolbe, D., 1994. A method of processing soil core 3, 39±52. samples for root studies by subsampling. Biol. Fertil. Soils 18, van Noordwijk, M., Cerri, C.C., Woomer, P.L., Nugroho, K., 60±62. Bernoux, M., 1997. Soil carbon dynamics in the humid tropical Schroth, G., da Silva, L.F., Seixas, R., Teixeira, W.G., MaceÃdo, forest zone. Geoderma 79, 187±225. J.L.V., Zech, W., 1999. Subsoil accumulation of mineral Woomer, P.L., Palm, C.A., Alegre, J.C., Castilla, C., Cordeiro, nitrogen under polyculture and monoculture plantations, fallow D.G., Hairiah, K., Kotto-Same, J., Moukam, A., Ricse, A., and primary forest in a ferralitic Amazonian upland soil. Agric. Rodrigues, V., van Noordwijk, M., 2000. Slash-and-burn effects Ecosyst. Environ. 75, 109±120. on carbon stocks in the humid tropics. In: Lal, R., Kimble, J.M., Schroth, G., Seixas, R., da Silva, L.F., Teixeira, W.G., Zech, W., Stewart, B.A. (Eds.), Global Climate Change and Tropical 2000a. Nutrient concentrations and acidity in ferralitic soil Ecosystems. CRC Press, Boca Raton, pp. 99±115.