Nutrient fluxes in rainfall, throughfall and stemflow in tree-based land use systems and spontaneous tree vegetation of central Amazonia

Nutrient fluxes in rainfall, throughfall and stemflow in tree-based land use systems and spontaneous tree vegetation of central Amazonia

Agriculture, Ecosystems and Environment 87 (2001) 37–49 Nutrient fluxes in rainfall, throughfall and stemflow in tree-based land use systems and spon...

176KB Sizes 1 Downloads 108 Views

Agriculture, Ecosystems and Environment 87 (2001) 37–49

Nutrient fluxes in rainfall, throughfall and stemflow in tree-based land use systems and spontaneous tree vegetation of central Amazonia G. Schroth a,∗ , M.E.A. Elias b , K. Uguen c , R. Seixas b , W. Zech a b

a Institute of Soil Science and Soil Geography, University of Bayreuth, D-95440 Bayreuth, Germany Empresa Brasileira de Pesquisa Agropecuaria–Amazônia Ocidental, C.P. 319, 69011-970 Manaus-AM, Brazil c Institut de Recherche pour le Developpement, 32 ave Henry Varagnat, 93143 Bondy Cedex, France

Received 6 March 2000; received in revised form 25 July 2000; accepted 12 October 2000

Abstract The quantification of nutrient fluxes is an important step in the development of sustainable land use systems, especially on low-fertility soils of the humid tropics. Nutrient concentrations in rainfall (RF), throughfall (TF) and stemflow (SF) were measured for ten rainfall events in a polyculture (multi-strata agroforestry system) composed of five tree crops, three tree crop monocultures, spontaneous tree fallow and two rainforest tree species in central Amazonia and nutrient fluxes were calculated for a 1 year period. Nutrient inputs in wet deposition during 1 year were 5.5 kg ha−1 of total N, of which 42% were in the organic form, 0.07 kg ha−1 of total P, of which 71% were in the organic form, 2.6 kg ha−1 of K, 0.8 kg ha−1 of Ca and 0.3 kg ha−1 of Mg. The nutrient concentrations in TF and SF were influenced by tree species, land use system, nutrient status of the trees and size of the rain events. The rainforest species had high N but low P concentrations in their TF and SF. The highest P concentrations were measured in SF of annatto (Bixa orellana), which was 115 times richer in total P and 400 times richer in phosphate-P than RF for small rain events. Higher fertilizer application increased the concentrations of P and Mg in TF and SF. On the plot level, the nutrient fluxes in TF and SF were greatest in the systems with the highest plant density and crown cover (fallows and palm monocultures). On the species level, strong increases of the nutrient fluxes in TF and SF were observed close to the stem of certain trees compared with the plot average (more than 10 fold for P, K and Mg under peach palm Bactris gasipaes). In the polycultures and the fallow, the recycling of N, P, Ca and Mg in TF and SF was about 5–10% of the total recycling including litterfall, but was 50–53% for K (77% in the peach palm monocultures). Throughfall and SF are most relevant for K cycling and by influencing small-scale patterns of nutrient inputs into the soil. Stemflow is especially important in vegetation with high stem density, such as certain fallows and in systems dominated by palms. These results can help to devise measurement programs for nutrient cycling in tree-dominated land use systems and spontaneous vegetation in the humid tropics. Possible effects of concentrated nutrient solutions on microbial processes in soil and litter merit further investigation. © 2001 Elsevier Science B.V. All rights reserved. Keywords: Agroforestry; Amazonia; Bactris gasipaes; Bertholletia excelsa; Bixa orellana; Eschweilera sp.; Nutrient cycling; Oenocarpus bacaba; Perennial crops; Theobroma grandiflorum



Corresponding author. Present address: Biological Dynamics of Forest Fragments Project, National Institute for Research in the Amazon, C.P. 478, 69011-970 Manaus-AM, Brazil. Fax: +55-92-642-2050. E-mail address: [email protected] (G. Schroth). 0167-8809/01/$ – see front matter © 2001 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 7 - 8 8 0 9 ( 0 0 ) 0 0 2 9 4 - 2

38

G. Schroth et al. / Agriculture, Ecosystems and Environment 87 (2001) 37–49

1. Introduction For the sustainable agricultural use of humid tropical regions with soils of low fertility, such as central Amazonia, it is important to develop land use systems that are characterized by efficient nutrient cycling. Nutrient fluxes in agricultural systems can be divided into exchanges between the system and its environment (e.g., additions with atmospheric deposition and biological N2 -fixation and losses with deep leaching and erosion) and internal recycling. The latter includes nutrients in litterfall and those leached from the biomass into the soil. For increasing the nutrient efficiency of a system, it is important that these internal fluxes are as closed as possible, i.e., the respective nutrients are effectively recycled by the vegetation and are not lost from the system, e.g., through leaching into the subsoil. Specifically, patterns of nutrient availability and demand should coincide within a land use system both temporally (“synchrony”) and spatially (“synlocation”, Myers et al., 1994). This is especially relevant under conditions of high rainfall and permeable soils, which are typical for many humid tropical regions. For designing land use systems characterized by optimum correspondence between nutrient supply and demand and thus closed internal nutrient cycles, it is necessary to dispose of information not only about the magnitude of the respective nutrient fluxes, but also about their spatial patterns and the factors that determine these patterns. Nutrient fluxes with throughfall and stemflow have repeatedly been measured in the humid tropics (Jordan, 1978; Brinkmann, 1983; Hölscher et al., 1998), although rarely in agricultural and agroforestry systems (Imbach et al., 1989; Leite and Valle, 1990; Opakunle, 1991). These latter studies were carried out in relatively homogeneous systems composed of one tree crop species and shade trees and the emphasis was on the quantification of nutrient fluxes on the plot level. Similar studies in heterogeneous associations of several tree crops which are common in the humid tropics (Torquebiau, 1992; Michon and de Foresta, 1999) have not yet been reported. Also, little is known about the effect of tree species differing in crown architecture, size and nutrient concentrations in the biomass on these nutrient fluxes. As reported previously, certain tree species such as palms can have very high stemflow and this may increase nutrient leaching,

especially when fertilizer is applied close to their stem (Schroth et al., 1999b). Here, nutrient fluxes in rainfall, throughfall and stemflow in central Amazonian land use systems differing widely in species composition and structure during 1 year are reported. Included in the study was a polyculture (multi-strata agroforestry system) with five tree crops at two fertilization levels, three tree crop monocultures, spontaneous tree fallow and two rainforest tree species. Comparisons were made both on the systems level and on the species level to identify factors that influence nutrient fluxes in throughfall and stemflow. These nutrient fluxes were also compared with those in litterfall. The hydrological conditions of the investigated species and vegetation systems have been reported previously (Schroth et al., 1999b). 2. Materials and methods 2.1. Study site The study was carried out on the research station of Embrapa Amazônia Ocidental near Manaus in Brazilian Amazonia (3◦ 8 S–59◦ 52 W, 40–50 m a.s.l.). The climate is humid tropical with an annual precipitation of 2622 mm, air temperature of 26◦ C and atmospheric humidity around 85% (mean values 1971–93, O.M.R. Cabral and C. Doza, unpublished). The soils are Xanthic Ferralsols according to FAO–Unesco (1990) with a topsoil pH around 4.5 and generally low nutrient contents. Detailed soil data are given in Schroth et al. (2000). The following vegetation types were included in the study (Table 1): A polyculture with peach palm (Bactris gasipaes) for fruit and for palmito (palmheart) production, cupuaçu (Theobroma grandiflorum), Brazil nut (Bertholletia excelsa) and annatto (Bixa orellana, see Schroth et al., 1999a, for plot layout); a monoculture of peach palm for palmito, planted at 2 × 2 m2 ; a monoculture of peach palm for fruit, planted at 4×4 m2 with an understorey of peach palm for palmito at 2 × 2 m2 spacing; and a monoculture of cupuaçu, planted at 7 × 6.4 m2 . Plots with spontaneous vegetation of the same age as the agricultural plots, dominated by Vismia spp., were included in the study. In addition, two tree species from a nearby primary rainforest were selected: Eschweilera sp., a dicotyledoneous tree and Oenocarpus bacaba, a palm. Both species are relatively frequent in

G. Schroth et al. / Agriculture, Ecosystems and Environment 87 (2001) 37–49

39

Table 1 Characteristics of polyculture, monoculture, fallow and forest plots in central Amazoniaa System and plant species

Plants (ha)

Height (m)

m2 per tree Polyculture Peach palm fruit Peach palm palmito Cupuaçu Brazil nut Annatto Peach palm fruit monoculture Peach palm palmito monoculture Cupuaçu monoculture Fallow Primary forest Eschweilera Oenocarpus

78 156 93 93 156

8.9 3–4 2.6 5.4 ∼3

20 4 3.5 30 13

625 2500 223 19500

10.3 3–4 1.9 ∼5

20 4 1.3 0.5

19 17

69 70

– –

Throughfallb

Crown area % of plot

Stemflowc

40 cm

150 cm

62 46 95 103 115

91 96 97 83 95

1.80 0.62 0.13 0.85 0.05

100 100 3 100

50 52 82 77

63 82 99 –

1.48 0.82 0.12 0.10

– –

76 96

90 80

0.05 0.46

15 6.3 3.3 27 20

a

From Schroth et al. (1999b). Throughfall measured at 40 and 150 cm stem distance, in percent of open-area rainfall. c Stemflow in liter per millimeter of open-area rainfall per tree. b

this forest and are of commercial interest, Eschweilera for its wood and Oenocarpus for its fruits. The measurement plots with the exception of the rainforest sites, but including the fallow plots, were arranged in a randomized complete block design with three replications. For the primary forest species, three individuals of each species were chosen in a forest adjacent to the experimental area. Plot size was 24 × 32 m2 in the peach palm monocultures and 48 × 32 m2 in all other treatments. All plots had been planted in February/March 1993, and the trees had been 3 21 –5 years in the field when the measurements were carried out. The polyculture was studied at two fertilization levels: full fertilization and liming according to local recommendations and 30% of this fertilization level without N-fertilization and lime (low fertilization; for fertilization rates and management see Schroth et al., 1999a). The monoculture plots received the higher fertilization level. 2.2. Sample collection and nutrient analyses Open-area rainfall (RF) was measured at about 70 cm above the soil with six polyethylene-collectors which were placed at about 4.5 m distance from the nearest trees in the three cupuaçu monoculture plots (two per plot). The collectors had a diameter of 7.4 cm (9.3 cm

after February 1997). A narrow bottleneck reduced evaporation of collected water. Throughfall (TF) was measured with similar collectors under six representative trees per species and cropping system (one tree per species and plot in the polycultures and two trees per plot in the monocultures). The collectors were installed at 40 and 150 cm distance from the stem. In the polycultures, two additional TF collectors per plot were positioned in areas not covered by tree crowns. In the dense fallow vegetation, two collectors per plot were positioned at random in the plots. Stemflow (SF) was measured at the same trees as TF with polyurethane collars as described by Likens and Eaton (1970). Regular measurements of RF, TF and SF were conducted between March 1996 and March 1997. For further details see Schroth et al. (1999b). Between August 1996 and July 1997, samples from all RF, TF and SF collectors were analysed following nine individual rain events. An additional sample was collected in March 1998 to increase the number of rain events >30 mm (see below). Before exposing the collectors in the field, they were thoroughly washed with tapwater followed by deionised water. On the morning following a rain event, the water samples were filtered through a pre-washed paper filter and stored at about 10◦ C. Inorganic P and N forms were analysed within 1 or 2 days after the collection of samples and the other

40

G. Schroth et al. / Agriculture, Ecosystems and Environment 87 (2001) 37–49

elements were analysed soon afterwards. Ammonium, nitrate, total N, phosphate and total P were measured colorimetrically with a segmented flow analyser. Total N and P were measured after on-line UV digestion in the presence of K-peroxodisulfate and for P, sulfuric acid. Potassium, Ca and Mg were measured by atom absorption spectrometry. Electrical conductivity (EC) was measured with an EC-sensor and pH was measured with a glass electrode after addition of KCl to the samples to increase ionic strength. 2.3. Calculation of nutrient fluxes with rainfall, throughfall and stemflow Calculations were made for a reference period from April 1996 to March 1997, the period of the detailed hydrological study reported in Schroth et al. (1999b). Total rainfall during this period was 2672 mm. Nutrient inputs with wet deposition were calculated from the nutrient concentrations in the RF samples and daily RF as measured at the Embrapa weather station. First, correlations were calculated between the mean concentration of each nutrient per sampling date and the corresponding RF volume. For nutrients whose concentration in the samples was independent of the RF volume, average values from the 10 sampling dates were multiplied with total RF for the measurement period to obtain annual wet deposition. For nutrients whose concentrations decreased significantly with RF volume, regression curves were adjusted to the measured data to predict concentrations as functions of RF size. First-order polynomial, second-order polynomial and logarithmic regression functions were compared and the function was chosen which explained the greatest proportion of the observed variability (adjusted r2 ). From the regression functions and the daily RF values during the reference period, nutrient inputs per rain event were calculated and these were summed to obtain annual values of wet deposition. Nutrient concentrations in SF and TF also generally decreased with increasing RF volume. However, the concentrations were more variable than those in RF and regression equations did not give satisfactory fits. The RF events were thus grouped into three classes, <15 mm, 15–30 mm, and >30 mm and mean concentrations per class were calculated for each sampling position. These were weighted according to the amount of RF in each class (834.3, 887.6 and 949.9 mm, re-

spectively) to obtain mean concentrations for the whole year. To obtain total nutrient fluxes in TF and SF for the study year, nutrient inputs per millimeter of RF were calculated from nutrient concentrations and sample volume for every collector and sampling event and the mean values per RF class were multiplied by the annual RF in this class. The resulting values per class were added to obtain nutrient inputs per collector per year. For peach palm for palmito, the deposition values in SF were multiplied by two because the plants usually had two larger offshoots, plus sometimes a variable number of very small offshoots which produced only insignificant amounts of SF (Schroth et al., 1999b). Nutrient fluxes per plot were calculated by assuming that the TF collectors at 40 cm from the trees represented an area of 2 m2 around the trunk (radius 80 cm), and the TF collectors at 150 cm from the trees represented the remaining area of the respective tree crown (Table 1). For the plot areas not covered by tree crowns, the data from the open-area collectors within the plots were used. 2.4. Foliar nutrient concentrations and litterfall Leaf samples from peach palm for palmito, cupuaçu, Brazil nut and annatto in the polycultures at full and low fertilization were collected in November 1997, January 1998, May 1998 and August 1998. After drying at 70◦ C, the samples were ground and digested according to Novozamsky et al. (1983). Nitrogen and P were measured colorimetrically with a segmented flow analyser and K, Ca and Mg were measured by atom absorption spectrometry. Litterfall was measured under the same tree species in the polyculture and the fallow plots between September 1997 and August 1998. In each polyculture plot, triangular collectors made of 1.5 mm nylon mesh mounted on wooden frames were placed under five trees per species, covering 10% of the crown area of cupuaçu, peach palm and annatto and 6.3% of the crown area of Brazil nut. In each fallow plot, five collectors of 1 × 1 m2 were installed at random positions. The litter was collected every two weeks. After drying and weighing of the total samples, subsamples were analysed for main nutrients. 2.5. Statistical analysis The effect of the fertilization level on the nutrient concentrations in TF and SF within the polyculture

G. Schroth et al. / Agriculture, Ecosystems and Environment 87 (2001) 37–49

plots was tested by ANOVA for a randomized complete block/split plot design with the fertilization level as main plot factor and the tree species as subplot factor. Nutrient concentrations and nutrient fluxes in TF and SF were compared for the different species and vegetation types, including the five tree species in the polycultures, the monocultures, fallow and the two rainforest trees. For this analysis, a fully randomized design was used because the rainforest trees were not part of the experimental blocks. In case of significance of the F-test at p < 0.05, species means were compared with Duncan’s multiple range test at the same probability level.

3. Results and discussion 3.1. Wet deposition of nutrients The concentrations of ammonium-N, phosphate-P, organic P, total P, Ca and Mg as well as the pH of the RF

41

samples were independent of the RF volume whereas the concentrations of nitrate-N, organic N, total N, K and electrical conductivity decreased significantly with increasing RF (Fig. 1). Nutrient concentrations in RF were generally low, resulting in an average electrical conductivity of only 10.2 ␮S cm−1 (Table 2, last column). Especially the inorganic and organic P concentrations were always close to the detection limit and occasionally negative values were obtained (i.e., concentrations lower than in the deionised water that was used as reference). Negative concentrations were also occasionally measured for K (Fig. 1) and ammonium-N (not shown). The annual wet deposition values were low (Table 3, last column). Of the total N inputs of 5.5 kg ha−1 per year, 42% were in the organic, 33% in the ammonium and 25% in the nitrate form. The inputs were in good agreement with measurements in the nearby Ducke forest reserve of 5 kg ha−1 per year of total N and 3.9 kg ha−1 per year of organic N (Brinkmann, 1983), and are within the reported range of other lowland

Fig. 1. Nutrient concentrations in open-area rainfall in central Amazonia as influenced by rainfall volume (means and S.E.). Regression equations 2 = 0.46, p = 0.019); organic N = 225 − 8.41 rain + 0.083 rain2 (r 2 = 0.81, r 2 were: 1g nitrate − N = 2.13 − 0.0197 rain (r 2 = 0.52, radj. adj. 2 = 0.31, = 0.75, p = 0.003); total N = 383−6.92 rain (r 2 = 0.69, radj 2 = 0.65, p = 0.003); K = 225−7.38 rain+0.067 rain2 (r 2 = 0.41, radj. 2 2 −1 −1 p = 0.09); 1g EC = 1.061 − 0.0104 rain (r = 0.46, radj. = 0.38, p = 0.046) with nutrients in ␮g l , EC in ␮S cm and rain in mm.

42

G. Schroth et al. / Agriculture, Ecosystems and Environment 87 (2001) 37–49

Table 2 Nutrient concentrations in stemflow of five tree crops in polyculture, fallow and two rainforest species in central Amazonia in comparison to open-area rainfall (weighted means of 10 sampling events)a PFb

PPb

Cupb

Brb

Anb

Falb

Eschb

Oenb

RFb

NH4 -N (mg l−1 ) NO3 -N (mg l−1 ) Org. N (mg l−1 )

0.06 b 0.08 bc 0.56 c

0.07 b 0.07 bc 0.21 ef

0.10 b 0.08 bc 0.48 cd

0.08 b 0.05 c 0.49 cd

0.14 b 0.16 b 0.80 b

0.14 b 0.03 c 0.50 cd

0.44 a 0.37 a 1.04 a

0.15 b 0.12 bc 0.49 cd

0.06 b 0.05 c 0.09 f

Total N (mg l−1 )

0.69 cd

0.35 ef

0.66 cde

0.61 cde

1.10 b

0.67 cde

1.85 a

0.76 cd

0.20 f

(␮g l−1 )

PO4 -P Org. P (␮g l−1 )

59 b 31 bcd

5c 10 de

28 bc 35 bcd

8c 18 cde

144 a 72 a

26 bc 41 bc

16 bc 27 bcd

3c 17 cde

1c 2e

Total P (␮g l−1 )

90 b

15 cd

63 bc

26 cd

215 a

67 bc

43 bcd

20 cd

3d

(mg l−1 )

K Ca (mg l−1 ) Mg (mg l−1 )

4.2 ab 0.43 a 0.35 a

0.8 cd 0.12 c 0.05 de

1.0 cd 0.35 ab 0.11 de

2.6 bc 0.18 bc 0.08 de

4.6 a 0.47 a 0.23 bc

1.5 cd 0.36 ab 0.10 de

1.1 cd 0.14 c 0.05 de

0.8 cd 0.12 c 0.14 cd

0.1 d 0.03 c 0.01 e

pH

5.5 a

5.0 bc

5.0 bc

5.2 b

5.4 a

5.1 bc

4.9 cd

4.6 d

5.0 bc

Values followed by the same letter within a row are not significantly different at p < 0.05 (Duncan’s multiple range test). PF: peach palm for fruit; PP: peach palm for palmito; Cup: cupuaçu; Br: Brazil nut; An: annatto; Fal: fallow; Esch: Eschweilera; Oen: Oenocarpus; RF: open-area rainfall. a

b

tropical forest regions (Bruijnzeel, 1991). Franken and Leopoldo (1984) reported inputs of 6 kg ha−1 per year of NH4 -N alone in the Ducke reserve, which is high compared with the values measured in the present study and by Forti and Moreira-Nordemann (1991) at the same site (0.49 kg ha−1 per year). Total P inputs were very low (0.07 kg ha−1 per year) and were mostly in

the organic form (71%). In the Ducke reserve, inputs of 0.4 kg ha−1 per year of total P (Brinkmann, 1983) and 0.2 kg ha−1 per year of phosphate-P (Franken and Leopoldo, 1984) have been measured. Other reported P inputs into tropical ecosystems are highly variable but are generally also somewhat higher than the values from the present study (Bruijnzeel, 1991;

Table 3 Nutrient fluxes in stemflow and throughfall in polyculture at full and low fertilization, three monocultures and fallow in comparison to open-area rainfall in central Amazonia during 1 year (in kg ha−1 per year)a Poly full

Poly low

Mono PFb

NH4 -N NO3 -N Org. N

1.4 cd 2.0 a 3.5 bc

1.4 cd 2.0 a 3.9 bc

1.0 d 0.9 c 2.5 c

2.5 a 2.0 a 5.1 b

1.5 bcd 1.5 ab 2.2 c

Total N

6.9 cd

7.3 bc

4.4 d

9.6 ab

5.2 cd

PO4 -P Org. P

0.10 ab 0.14 bc

0.04 b 0.12 bcd

0.21 a 0.12 bcd

0.10 ab 0.17 b

0.02 b 0.06 d

0.15 ab 0.28 a

0.02 b 0.05 d

Total P

0.24 bc

0.16 bc

0.33 ab

0.27 ab

0.08 c

0.42 a

0.07 c

K Ca Mgc

7.9 cd 1.5 abc 0.6

9.4 bcd 2.0 ab 0.9

13.5 bc 1.1 bc 0.6

17.9 ab 1.5 abc 0.6

Mono PPb

Mono Cupb

2.7 d 0.8 c 0.2

Fallow 2.2 ab 1.9 ab 7.2 a 11.3 a

25.5 a 2.3 a 1.0

RFb 1.8 abc 1.4 bc 2.3 c 5.5 cd

2.6 d 0.8 c 0.3

Values followed by the same letter within a row are not significantly different at p < 0.05 (Duncan’s multiple range test). Values differing significantly from RF are in italic face. The RF nutrients as determined by regression from daily rainfall (see Section 2) were unreplicated and could not be included in the analysis of variance, but were included in the Duncan’s test based on the error term from the trees. b PF: peach palm for fruit; PP: peach palm for palmito; Cup: cupuaçu; RF: open-area rainfall. c For Mg, the difference between systems was significant at p = 0.07. a

G. Schroth et al. / Agriculture, Ecosystems and Environment 87 (2001) 37–49

Newman, 1995). The probable reason is that in the present study the collectors were not continuously exposed to the atmosphere between rain events. This reduced the risk of contamination, but also the deposition of atmospheric dust (dry deposition) in the collectors. The K inputs were higher than values by Forti and Moreira-Nordemann (1991) (0.38 kg ha−1 per year), but were in good agreement with Franken and Leopoldo (1984) (2.1 kg ha−1 per year). Magnesium and Ca inputs of <1 kg ha−1 per year in the study region were also reported by other authors (Brinkmann, 1983; Forti and Moreira-Nordemann, 1991). Studies from other humid tropical regions often found higher inputs of these cations, possibly due to inclusion of dry deposition and/or closer distance to the sea (Waterloo et al., 1997; Hölscher et al., 1998). 3.2. Factors affecting nutrient concentrations in SF and TF The nutrient concentrations in SF and TF were influenced by tree species, cropping system, fertilizer application and rainfall size. The nutrient concentrations in SF were often drastically higher than in RF and there was a significant effect of the tree species for all nutrients (Table 2). The highest N concentrations were measured for the rainforest tree Eschweilera, presumably because of the large crown and thus contact area for SF and TF. In contrast, by far the highest P concentrations were measured in SF of annatto, in agreement with the high P concentrations in the foliage of this species (Table 4). Both N and (for most species) P in SF were predominantly in the organic form (Table 2). The nutrient concentrations in TF also differed

43

significantly between tree species and were in most cases higher than in RF, although the increase was often not significant (Table 5). Pronounced increases in concentration were especially observed for N under the rainforest species, for P under annatto, for Ca under Brazil nut and for Mg under the rainforest palm Oenocarpus bacaba. Calcium leaching from the Brazil nut crowns may reflect relatively high foliar Ca concentrations of this species (Table 4). In total, 55% of total N and 70% of total P were in the organic form (67 and 80%, respectively for the rainforest species). The nutrient concentrations in TF were often higher in the peach palm monocultures than for the same species in polyculture (Table 5). At 40 cm stem distance, total P in the fruit monoculture was 34 ␮g l−1 (15 ␮g l−1 in polyculture) and K was 1.2 mg l−1 (0.5 mg l−1 ). At 150 cm stem distance, total N in the fruit monoculture was 0.39 mg l−1 (0.26 mg l−1 ) and total N in the palmito monoculture was 0.36 mg l−1 (0.23 mg l−1 ). Potassium in the fruit monoculture was 0.5 mg l−1 (0.2 mg l−1 ). In SF, the nutrient concentrations did not usually differ between poly- and monocultures, with the exception of phosphate and total P for peach palm for fruit (130 vs. 59 ␮g l−1 and 176 vs. 90 ␮g l−1 in mono- and polyculture, respectively). The reason for these differences was obviously the denser plant stand and consequently the higher leaf area in the monoculture systems (Table 1). The fertilization level had a significant (p = 0.020) effect on the Mg concentrations in SF (0.18 mg l−1 at full and 0.14 mg l−1 at low fertilization, mean of all tree species) and tended to increase phosphate-P in SF of peach palm for fruit (88 vs. 29 ␮g l−1 ) and annatto (174 vs. 113␮g l−1 ). In TF at 150 cm stem distance,

Table 4 Nutrient concentrations (mg g−1 ) in leaves of four tree crop species in a polyculture system at full and low fertilization in central Amazonia (means of four collection dates during 1 year)a

N P K Ca Mgc a

Fertilization level

Peach palm palmito

Cupuaçu

Brazil nut

Annatto

Meanb

31.8 (0.5) 1.8 (0.03) 9.2 (0.3) 3.1 (0.2) 2.6 (0.2) 1.4 (0.1)

18.9 (0.3) 1.0 (0.02) 4.0 (0.2) 4.8 (0.2) 3.0 (0.1) 1.9 (0.1)

18.4 (0.2) 0.8 (0.01) 4.3 (0.2) 9.3 (0.6) 3.7 (0.2) 2.3 (0.1)

31.3 (1.3) 2.1 (0.13) 13.2 (0.4) 7.8 (0.5) 4.0 (0.2) 2.0 (0.1)

Meanb Meanb Meanb Full Low

Values in brackets are standard errors. Mean of full and low fertilization. c The Mg levels differed significantly between fertilization levels for all species at p < 0.05 (LSD test). b

44

G. Schroth et al. / Agriculture, Ecosystems and Environment 87 (2001) 37–49

Table 5 Nutrient concentrations in throughfall at 40 and 150 cm stem distance under five tree crops in polyculture, fallow and two rainforest species in central Amazonia in comparison to open-area rainfall (weighted means of 10 sampling events)a Distance (cm)

PFb

Total N (mg l−1 )

40 150

0.34 cd 0.26 d

Total P (␮g l−1 )

40 150

15 bc 4 bc

K (mg l−1 )

40 150

0.5 bcd 0.2 cde

Ca (mg l−1 )

40 150

Mg (mg l−1 )

40 150

PPb

Cupb

Brb

0.34 cd 0.21 d

0.37 cd 0.29 cd

9 bc 4 bc

9 bc 5 bc

0.3 cd 0.2 bcde

0.3 cd 0.1 de

0.7 abc 0.4 abcd

0.7 bc 0.3 bcde

0.08 b 0.06 ab

0.06 bc 0.04 bc

0.07 bc 0.04 bc

0.14 a 0.09 a

0.05 abc 0.02 bc

0.02 cd 0.02 bc

0.03 bcd 0.01 c

0.06 ab 0.04 b

0.35 cd 0.23 d 12 bc 4 bc

Anb

Eschb

Oenb

RFb

0.81 a 0.60 a

0.59 b 0.61 a

0.20 e 0.20 d

11 bc 9 ab

8 bc 8 abc

3c 3c

1.0 ab –

0.7 bc 0.6 a

0.8 ab 0.7 a

0.1 d 0.1 e

0.06 bc 0.05 bc

0.07 b –

0.10 b 0.06 abc

0.08 b 0.09 a

0.03 c 0.03 c

0.03 bcd 0.02 bc

0.04 bcd –

0.04 abc 0.04 b

0.07 a 0.07 a

0.01 d 0.01 c

0.31 de 0.26 d 22 ab 12 a

Falb 0.42 cd – 12 bc –

a Values followed by the same letter within a row are not significantly different at p < 0.05 (Duncan’s multiple range test). The pH was between 4.9 and 5.1 in all sampling positions. b PF: peach palm for fruit; PP: peach palm for palmito; Cup: cupuaçu; Br: Brazil nut; An: annatto; Fal: fallow; Esch: Eschweilera; Oen: Oenocarpus; RF: open-area rainfall.

annatto had significantly (p < 0.05) higher concentrations at full than at low fertilization of both phosphate-P and total P (7 vs. 1 ␮g l−1 and 17 vs. 6 ␮g l−1 , respectively). All tree species except Brazil nut had higher concentrations of organic P in TF at 150 cm stem distance with full than with low fertilization (5–10 vs. 2–5 ␮g l−1 ; p = 0.015). Higher fertilization thus increased the concentrations of Mg and P in SF and TF, apparently through increased leaching from the tree crowns. For Mg this was in agreement with increased foliar concentrations at the higher input level (Table 4). Stemflow produced by small RF events was generally much richer in nutrients than that produced by large events (Fig. 2). A similar tendency was observed for TF (data not shown). Apparently, most nutrients were leached from the tree crowns at the beginning of RF events. This may be explained by dust deposited on the crowns and with release of nutrients and organic substances from epiphytes upon rewetting (Coxson, 1991). 3.3. Nutrient fluxes in SF and TF and their spatial patterns The nutrient fluxes in SF and TF in the different land use systems depended on the crown cover of the vegetation (Table 3). With a crown cover of 3% (Table 1), cupuaçu in monoculture had no significant effect on any of the nutrients, as little RF came into contact with

the tree foliage. The most pronounced effects on nutrient fluxes were measured in the palm monocultures and the fallow which had crown covers of approximately 100%. The crowns of the palmito monoculture and the fallow increased total N fluxes by 4–5 kg ha−1 per year, mostly in the organic form and added substantial amounts of mineral and organic P to the rainwater. The K flux in the rainwater was increased tenfold by the crown passage in the fallow and 5–7 fold in the palm monocultures. The peach palm for fruit monoculture was the only system that significantly decreased the flux of one nutrient (ammonium-N), presumably through uptake by leaves and epiphytes, in agreement with results from rainforest in the region (Forti and Moreira-Nordemann, 1991). The polyculture systems, which had a crown cover of about 70% (Table 1), were intermediate between the cupuaçu and the palm monocultures in their effect on the nutrient fluxes (Table 3). Because of the close spacing of the trees in the palmito monoculture and especially in the fallow, TF and SF reached the soil in a relatively homogeneous spatial distribution. This was not the case in the polyculture plots, where the trees were planted at wider spacing and where the tree species differed significantly in nutrient composition and volume of TF and SF. Due to the combination of relatively high nutrient concentrations in SF and an exceptionally high SF volume, the fruit palms increased the fluxes of P, K and Mg into

G. Schroth et al. / Agriculture, Ecosystems and Environment 87 (2001) 37–49

45

Fig. 2. Nutrient concentrations in stemflow of five tree crops, fallow and two rainforest species in comparison to open-area rainfall as influenced by rainfall volume in central Amazonia. Bars indicate least significant differences for the respective rainfall class.

the soil surrounding their stem by a factor 11–13, Ca by a factor 7, and N by a factor 3 in comparison with the plot average (Table 6). Brazil nut caused 3–4 fold increases in the K, Ca and Mg fluxes to the soil in its proximity due to its substantial SF volume and also relatively high concentrations of these nutrients in TF at 40 cm stem distance. Annatto had little SF, but high K and especially P concentrations in SF and TF, which

caused an increase of the fluxes of these elements to the soil close to the trees by a factor 3–4. 3.4. Importance of nutrient fluxes in SF and TF and consequences for their measurement The increase of the nutrient load of SF and TF in comparison with RF may have two principally

46

G. Schroth et al. / Agriculture, Ecosystems and Environment 87 (2001) 37–49

Table 6 Nutrient fluxes in stemflow and throughfall into a 2 m2 area around five tree crop species in polyculture during 1 yeara Peach palm fruit

Peach palm palmito

Cupuaçu

Brazil nut

Annatto

0.8 (0.4) 116 (12)

0.9 (0.4) 131 (19)

1.4 (0.6) 200 (17)

1.1 (0.5) 155 (29)

Total N

g m−2

per year % of plot av.b

2.3 (0.9) 327 (25)

Total P

g m−2 per year % of plot av.b

0.22 (0.06) 1061 (208)

0.02 (0.00) 103 (7)

0.03 (0.01) 155 (26)

0.03 (0.00) 184 (29)

0.08 (0.03) 359 (95)

K

g m−2 per year % of plot av.b

9.9 (2.2) 1179 (253)

1.1 (0.2) 128 (30)

1.0 (0.2) 120 (28)

3.5 (0.7) 396 (58)

2.9 (1.1) 294 (74)

Ca

g m−2 per year % of plot av.b

1.13 (0.17) 667 (104)

0.17 (0.02) 102 (7)

0.18 (0.02) 108 (16)

0.50 (0.03) 298 (38)

0.18 (0.03) 108 (16)

Mg

g m−2 per year % of plot av.b

0.90 (0.15) 1308 (250)

0.07 (0.02) 99 (10)

0.08 (0.01) 106 (18)

0.21 (0.02) 289 (37)

0.09 (0.02) 116 (23)

a b

Values in brackets are standard errors. Two square meters is the assumed infiltration area for stemflow, see Schroth et al. (1999b). Mean values in percent of the plot average as given in Table 3.

different origins: leaching from plants and epiphytes of nutrients which have previously been taken up from the soil; and dry deposition from the atmosphere on the vegetation (including biological N2 fixation by epiphytes). Only the second source is an input into the system. In a study in the nearby Ducke forest reserve, Forti and Moreira-Nordemann (1991) concluded that during the wet season, most of the K, Ca, Mg and ammonium in TF were leached from the vegetation and represented no atmospheric inputs. The observed relationships between nutrient concentrations in SF/TF and in the leaves in the present study tend to confirm this. Furthermore, the analysis of aerosols in the Ducke reserve showed that, although certain elements are transported into central Amazonia from the sea and even the Sahara, most of the P and K and about half of the Ca in the aerosols are of biogenic origin (Artaxo and Hansson, 1995), i.e., they can equally not be considered inputs to the site. It is thus assumed that the measured differences between the nutrient fluxes under the canopies and in RF reflect mainly the recycling of nutrients at the site, and not external inputs. In this respect, the nutrient fluxes in TF and SF are similar to those in litterfall and it is thus interesting to compare these two nutrient recycling mechanisms. The comparison in Table 7 is only approximate because the litterfall of the fruit palms was not measured, and it was thus assumed that the litterfall of one row of fruit palms was similar to that of one row of palmito palms, which may be an underestimation. Also, the litterfall measurements were not carried out in the same year

as the TF and SF measurements. With these reservations, it can be concluded that in the polyculture and the fallow, the net returns of N, P, Ca and Mg to the soil in TF and SF were only about 5–10% of the total returns including litterfall. In contrast, the K returns in TF and SF were equal to or even slightly higher than those in litterfall. These calculations only include spontaneous processes, i.e., without considering nutrient cycling resulting from the management of the trees, such as pruning and harvesting. The similarity of the relative importance of TF and SF between the two systems in Table 7 is remarkable in view of the large absolute differences in nutrient fluxes. A similar comparison can be made for the palmito monoculture, assuming that litter production per individual palm was similar in this system and in the polyculture. This comparison gives almost the same contributions of TF + SF − RF to the total returns of P, Ca and Mg as in the polyculture in Table 7, but the contribution of TF + SF − RF to the total returns is 14% for N and as much as 77% for K (data not shown). This particularity of K, which is obviously related to its leachability from biomass, is confirmed by results from a Bornean rainforest, where the K returns in TF were 61% of the total returns in TF and litterfall, whereas the contribution of TF was only 6% for N and Ca, 10% for Mg and 19% for P (Burghouts et al., 1998). In a Mexican dry forest, TF contributed 47% of the total K returns, but only 1% of the Ca returns and 16% of the Mg returns (Campo et al., 2000). It can thus be concluded that, for a coarse estimate of the spontaneous nutrient returns in tropical forests,

G. Schroth et al. / Agriculture, Ecosystems and Environment 87 (2001) 37–49

47

Table 7 Net nutrient returns in stemflow and throughfall as compared with litterfall in polyculture and fallow in central Amazoniaa SF + TF − RFb (kg ha−1 per year) Polycultured N P K Ca Mg Fallow N P K Ca Mg

Litterfallc (kg ha−1 per year)

Total returns (kg ha−1 per year)

SF + TF − RFb (% of total)

1.6 (0.3) 0.13 (0.02) 6.1 (1.0) 1.0 (0.2) 0.5 (0.1)

19.3 (0.7) 1.7 (0.1) 5.7 (0.3) 12.1 (1.0) 3.7 (0.5)

20.9 (0.7) 1.8 (0.1) 11.8 (1.2) 13.0 (1.0) 4.1 (0.5)

8 (1) 7 (1) 50 (3) 8 (2) 12 (4)

5.8 (1.4) 0.35 (0.10) 22.9 (1.1) 1.5 (0.4) 0.7 (0.2)

77.7 (6.0) 6.2 (0.6) 20.7 (1.8) 40.5 (3.1) 10.6 (0.8)

83.5 (6.8) 6.6 (0.7) 43.6 (2.9) 42.0 (3.3) 11.3 (1.0)

7 (1) 5 (1) 53 (1) 4 (1) 6 (2)

a

Values in brackets are standard errors. SF: stemflow; TF: throughfall; RF: open-area rainfall. c Including leaves, flowers, twigs and litter fragments (trash), but no harvest residues and prunings. d Mean of two fertilization levels. b

fallows or tree crop systems on the plot level, it would suffice to measure litterfall, with the exception of K which would also have to be measured in SF and TF. However, when studying nutrient fluxes on a smaller spatial scale, the local importance of SF and TF may be greater because of the concentration of nutrient returns in TF and SF close to the stem of certain species, especially palms (Table 6). Although the litter of small trees with low crowns such as cupuaçu also accumulates in a limited area around the tree stem, that of larger trees such as Brazil nut or, in the forest, Eschweilera is distributed on a much greater area corresponding to the crown diameter and height of these trees (Table 1). In fertilized tree crop systems, the nutrient quantities in SF and TF are probably of little importance, especially as the fertilizer is usually also applied close to the stems of the trees. In unfertilized systems and natural vegetation, however, high SF may be a mechanism by which certain trees increase their access to the nutrients in wet deposition and also those that were leached from their own biomass. The question if such extreme SF rates as those observed for the fruit palms improve the access to nutrients or rather reduce it by increasing nutrient leaching in the proximity of the stem would however merit further study. Beside the spatial distribution, a further difference between litter and TF/SF as nutrient sources is the immediate availability of much of the nutrients in the

solutions, whereas the litter especially of cupuaçu and Brazil nut decomposes slowly and accumulates on the soil. The availability of organic N and P forms in TF and SF for plants and microbes in tropical ecosystems is insufficiently known. The observation that dissolved organic N forms can be used by certain plant species when infected with mycorrhiza (Turnbull et al., 1995; Näsholm et al., 2000) may be relevant especially for the rainforest species in view of their high N concentrations in SF and TF. Some studies of nutrient fluxes in TF in tropical ecosystems have not included SF measurements because its contribution to these fluxes was considered too small to justify the effort (Imbach et al., 1989). In the present study, the importance of SF for the nutrient fluxes under the tree crowns depended strongly on the respective species. For the dicot species Brazil nut, cupuaçu and annatto, SF contributed <15% to the total nutrient fluxes in SF and TF under the area covered by the tree crowns (Table 8). Although the SF volume of the Brazil nut trees was substantial, the associated nutrient fluxes became relatively unimportant when related to the area covered by the crowns of these large trees. Opakunle (1991) also reported a contribution of only 7.4% of SF to the nutrient fluxes in TF in a shaded cocoa plantation in Nigeria. The situation was different for the palms and the fallow. Under peach palm for fruit, SF contributed between 20 and 65% to the

48

G. Schroth et al. / Agriculture, Ecosystems and Environment 87 (2001) 37–49

Table 8 Nutrient fluxes in stemflow and throughfall under the crown area of five tree species in polyculture and fallow in central Amazonia during 1 year, and contribution of stemflow PFa

PPa

Cupa

Fala

g per tree SFa (%)

16.1 (0.9) 20 (2)

3.1 (0.2) 21 (3)

Total P

g per tree SFa (%)

0.61 (0.14) 59 (7)

0.08 (0.01) 32 (4)

K

g per tree SFa (%)

27.0 (4.4) 65 (8)

3.3 (0.4) 47 (8)

2.5 (0.5) 7 (1)

34.9 (9.1) 8 (2)

11.1 (2.7) 5 (2)

5.1 (0.7) 15 (5)

Ca

g per tree SFa (%)

5.0 (0.7) 42 (6)

0.6 (0.1) 31 (5)

0.5 (0.1) 12 (3)

7.7 (1.5) 2 (0.4)

1.5 (0.1) 2 (0.3)

0.5 (0.1) 30 (7)

Mg

g per tree SFa (%)

2.8 (0.4) 63 (9)

0.3 (0.1) 38 (6)

0.2 (0.03) 10 (2)

3.4 (0.9) 3 (1)

0.5 (0.1) 3 (0.4)

0.2 (0.04) 24 (7)

0.07 (0.01) 13 (1)

24.1 (1.9) 3 (1)

Ana

Total N

a

2.7 (0.3) 4 (1)

Bra

0.44 (0.05) 6 (2)

8.5 (0.9) 1 (0.1)

2.3 (0.2) 17 (5)

0.39 (0.11) 5 (1)

0.08 (0.02) 31 (5)

PF: peach palm for fruit; PP: peach palm for palmito; Cup: cupuaçu; Br: Brazil nut; An: annatto; Fal: fallow; SF: stemflow.

nutrient fluxes in SF and TF. Under peach palm for palmito, the contribution of SF to the nutrient fluxes was less, but still substantial. The relatively high nutrient fluxes in SF in the fallow were due to the tree density of almost 2 stems per m2 (Table 1). From the data in Table 8 it follows that in systems with high tree density, such as the fallow, and especially in systems based on palms, SF is an important and for some elements even the dominating component of the nutrient fluxes in TF and SF and cannot be ignored in respective studies. Whereas the quantity of nutrients added with SF and TF may be most relevant for the nutrient supply of plants, microorganisms living in the soil or the litter layer and epiphytes living on the tree trunks may be influenced principally by the nutrient concentration in these solutions. The SF produced by rain events <15 mm under annatto, for instance, was 5 times richer in total N, 46 times richer in K, 115 times richer in total P and more than 400 times richer in phosphate than RF (Fig. 2). Access to these solutions should give a significant advantage to microbes living on or close to such trees, and this may affect microbial processes in litter and soil (Lodge et al., 1994).

4. Conclusions Wet deposition adds only small amounts of nutrients to central Amazonian ecosystems. These cannot

sustain agricultural production unless when allowed to accumulate over extended fallow periods but may be of relevance for forest ecosystems (Waterloo et al., 1997). The nutrient fluxes with TF and SF were higher in land use systems with a closed canopy than in open vegetation as it would be expected. On a plot basis, the recycling of most macronutrients with TF and SF was only approximately 5–10% of the total nutrient recycling including litterfall, although TF and SF returned approximately as much K to the soil as litterfall. However, close to the stems of certain trees, especially the fruit palms, nutrient fluxes in SF and TF were several times increased over the plot average. Organic forms of N and P made a substantial contribution to the total quantities of these nutrients in SF and TF and need to be included in respective measurement programs. The effects of the high nutrient concentrations in SF and TF of some tree species, especially during small rain events, on soil and litter microbes merits further study.

Acknowledgements This work was funded by the German Ministry of Education and Research (BMBF) together with the Brazilian Conselho Nacional de Desenvolvimento Cient´ıfico e Tecnológico (CNPq) as part of the SHIFT program, project 0339641/ENV 45. L.F. da Silva helped in the collection of the solution samples. M.E.A. Elias received a CNPq fellowship while working on this study.

G. Schroth et al. / Agriculture, Ecosystems and Environment 87 (2001) 37–49

References Artaxo, P., Hansson, H.-C., 1995. Size distribution of biogenic aerosol particles from the Amazon basin. Atmos. Environ. 29, 393–402. Brinkmann, W.L.F., 1983. Nutrient balance of a central Amazonian rainforest: comparison of natural and man-managed systems. In: Keller, R. (Ed.), Hydrology of Humid Tropical Regions with Particular Reference to the Hydrological Effects of Agriculture and Forestry Practice. International Association of Hydrological Sciences, Wallingford, pp. 153–163. Bruijnzeel, L.A., 1991. Nutrient input-output budgets of tropical forest ecosystems: a review. J. Trop. Ecol. 7, 1–24. Burghouts, T.B.A., van Straalen, N.M., Bruijnzeel, L.A., 1998. Spatial heterogeneity of element and litter turnover in a bornean rain forest. J. Trop. Ecol. 14, 477–506. Campo, J., Maass, J.M., Jaramillo, V.J., Yr´ızar, A.M., 2000. Calcium, potassium, and magnesium cycling in a Mexican tropical dry forest ecosystem. Biogeochem. 49, 21–36. Coxson, D.S., 1991. Nutrient release from epiphytic bryophytes in tropical montane rain forest (Guadeloupe). Can. J. Bot. 69, 2122– 2129. FAO–Unesco, 1990. Soil map of the world, revised legend. Food and Agriculture Organization of the United Nations, Rome, Italy, 119 pp. Forti, M.C., Moreira-Nordemann, L.M., 1991. Rainwater and throughfall chemistry in a terra firme rain forest: central Amazonia. J. Geophys. Res. 96, 7415–7421. Franken, W., Leopoldo, P.R., 1984. Hydrology of catchment areas of central-Amazonian forest streams. In: Sioli, H. (Ed.), The Amazon — Limnology and Landscape Ecology of a Mighty Tropical River and its Basin. W. Junk, Dordrecht, pp. 501–519. Hölscher, D., Sá, T.D.A., Möller, R.F., Denich, M., Fölster, H., 1998. Rainfall partitioning and related hydrochemical fluxes in a diverse and in a mono specific (Phenakospermum guyanense) secondary vegetation stand in eastern Amazonia. Oecologia 114, 251–257. Imbach, A.C., Fassbender, H.W., Borel, R., Beer, J., Bonnemann, A., 1989. Modelling agroforestry systems of cacao (Theobraoma cacao) with laurel (Cordia alliodora) and cacao with poro (Erythrina poeppigiana) in Costa Rica 4. Water balances, nutrient inputs and leaching. Agrofor. Syst. 8, 267–287. Jordan, C.F., 1978. Stem flow and nutrient transfer in a tropical rain forest. Oikos 31, 257–263. Leite, J.O., Valle, R.R., 1990. Nutrient cycling in the cacao ecosystem: rain and throughfall as nutrient sources for the soil and the cacao tree. Agric. Ecosys. Environ. 32, 143–154. Likens, G.E., Eaton, J.S., 1970. A polyurethane stemflow collector for trees and shrubs. Ecology 51, 938–939.

49

Lodge, D.J., McDowell, W.H., McSwiney, C.P., 1994. The importance of nutrient pulses in tropical forests. Trends Ecol. E 9, 384–387. Michon, G., de Foresta, H., 1999. Agro-forests: incorporating a forest vision in agroforestry. In: Buck, L.E., Lassoie, J.P., Fernandes, E.C.M. (Eds.), Agroforestry in Sustainable Agricultural Systems. Lewis Publishers, Boca Raton, pp. 381–406. Myers, R.J.K., Palm, C.A., Cuevas, E., Gunatilleke, I.U.N., Brossard, M., 1994. The synchronisation of nutrient mineralisation and plant nutrient demand. In: Woomer, P.L., Swift, M.J. (Eds.), The Biological Management of Tropical Soil Fertility. Wiley, Chichester, pp. 81–116. Näsholm, T., Huss-Danell, K., Högberg, P., 2000. Uptake of organic nitrogen in the field by four agriculturally important plant species. Ecology 81, 1155–1161. Newman, E.I., 1995. Phosphorus inputs to terrestrial ecosystems. J. Ecol. 83, 713–726. Novozamsky, I., Houba, V.J.G., van Eck, R., van Vark, W., 1983. A novel digestion technique for multi-element plant analysis. Commun. Soil Sci. Plant Anal. 14, 239–248. Opakunle, J.S., 1991. Contribution of stemflow to nutrient transfer in a cacao field in south western Nigeria. Acta Oecol. 12, 305–315. Schroth, G., da Silva, L.F., Seixas, R., Teixeira, W.G., Macêdo, J.L.V., Zech, W., 1999a. Subsoil accumulation of mineral nitrogen under polyculture and monoculture plantations, fallow and primary forest in a ferralitic Amazonian upland soil. Agric. Ecosys. Environ. 75, 109–120. Schroth, G., da Silva, L.F., Wolf, M.A., Teixeira, W.G., Zech, W., 1999b. Distribution of throughfall and stemflow in multi-strata agroforestry, perennial monoculture, fallow and primary forest in central Amazonia, Brazil. Hydrol. Process. 13, 1423–1436. Schroth, G., Seixas, R., da Silva, L.F., Teixeira, W.G., Zech, W., 2000. Nutrient concentrations and acidity in ferralitic soil under perennial cropping, fallow and primary forest in central Amazonia. Eur. J. Soil Sci. 51, 219–231. Torquebiau, E., 1992. Are tropical agroforestry home gardens sustainable? Agric. Ecosys. Environ. 41, 189–207. Turnbull, M.H., Goodall, R., Stewart, G.R., 1995. The impact of mycorrhizal colonization upon nitrogen source utilization and metabolism in seedlings of Eucalyptus grandis Hill ex Maiden and Eucalyptus maculata Hook. Plant Cell Environ. 18, 1386– 1394. Waterloo, M.J., Schelleken, J., Bruijnzeel, L.A., Vugts, H.F., Assenberg, P.N., Rawaqa, T.T., 1997. Chemistry of bulk precipitation in southwestern Viti Levu. Fiji. J. Trop. Ecol. 13, 427–447.