Agriculture, Ecosystems and Environment 75 (1999) 109–120

Subsoil accumulation of mineral nitrogen under polyculture and

monoculture plantations, fallow and primary forest in a ferralitic
Amazonian upland soil
Götz Schroth a,∗ , Luciana Ferreira da Silva b , Rosangela Seixas b , Wenceslau Geraldes Teixeira b ,
Jeferson L.V. Macêdo b , Wolfgang Zech a
a Institute of Soil Science, University of Bayreuth, D-95440 Bayreuth, Germany
b Empresa Brasileira de Pesquisa Agropecuaria (Embrapa) – Amazônia Ocidental, C.P. 319, 69011-970 Manaus-AM, Brazil
Received 25 May 1998; received in revised form 2 December 1998; accepted 20 April 1999

Abstract
Central Amazonia is characterized by high and intensive rainfall and permeable soils. When rainforests are cleared for
agricultural use, the efficient nutrient recycling mechanisms of the forests are disrupted and the nutrient availability in the
topsoil is increased by fertilization, thereby increasing the potential for nutrient leaching. In this study, the distribution
of mineral N in the upper two meters of a ferralitic upland soil was evaluated as an indicator for nutrient leaching and
for the potential contribution of the subsoil to crop nutrition. A perennial polyculture system with four tree crops and
a leguminous cover crop at two fertilization levels was compared with a monoculture plantation of peach palm (Bactris
gasipaes), spontaneous fallow and primary rainforest. Mineral N accumulated principally as nitrate in the subsoil under all
agricultural crops and also under the primary forest, although to a lesser extent. Within the polyculture system, there were
significant differences in N accumulation between the tree crop species, and for one of the species (Theobroma grandiflorum)
also between fertilization levels. The principal sources of subsoil N were mineral fertilizer and presumably N from the
mineralization of leguminous biomass and soil organic matter. The N losses from the agricultural systems and the absence of
yield responses of the tree crops to N fertilization indicated that agricultural production was not limited by N at this site, or
that N was too rapidly leached to be taken up efficiently by the crops. None of the tree crop species seemed to be efficient in
capturing leached N. Strategies are discussed for reducing N losses from agricultural systems with perennial crops, including
the development of site- and species-specific fertilizer recommendations, closer tree spacing, and the encouragement of lateral
and vertical tree root development. ©1999 Elsevier Science B.V. All rights reserved.
Keywords: Bactris; Bertholletia; Bixa; Cover crop; Nutrient leaching; Pueraria; Single-tree patterns; Theobroma; Tree crop

1. Introduction

∗
Ferralsols (oxisols) account for 50% of the soils
Corresponding author. Present address: Embrapa Amazô-
of humid tropical America and for 35% of the soil
nia Ocidental, C.P. 319, 69011-970 Manaus-AM, Brazil. Tel.:
+55-92-622-2012; fax: +55-92-622-1100 cover of the humid tropics of the world (Szott et al.,
E-mail address: schroth@internext.com.br (G. Schroth) 1991). In the Amazon basin, the natural vegetation of

0167-8809/99/$ – see front matter ©1999 Elsevier Science B.V. All rights reserved.
PII: S 0 1 6 7 - 8 8 0 9 ( 9 9 ) 0 0 0 6 8 - 7
110 G. Schroth et al. / Agriculture, Ecosystems and Environment 75 (1999) 109–120
these soils is typically rainforest. By definition, Fer-
ralsols are very poor in weatherable minerals and rich
in caolinitic clay and sesquioxides, resulting in a gen-
erally low chemical fertility (van Wambeke, 1992).
Where the soils are deep and the plant roots cannot
reach the underlying rock or its less completely weath-
ered residues, the nutrient supply of the vegetation
depends strongly on the humus-enriched topsoil and,
notably, on the nutrients in the biomass of the rain-
forest vegetation which are effectively recycled from
decomposing organic materials by the dense, superfi-
cial roots and their mycorrhiza (Herrera et al., 1978;
Stark and Jordan, 1978; Cuevas and Medina, 1988).
When a rainforest is cleared for agricultural use,
these efficient nutrient recycling mechanisms are dis-
rupted, and nutrients released by burning the origi-
nal vegetation may be rapidly leached. As a conse-
quence, multiple nutrient deficiencies develop early
in annual crop rotations (Cravo and Smyth, 1997).
Where the population density is too high for sustain-
able shifting agriculture, land use systems with peren-
nial crops, such as plantation agriculture and agro-
forestry that provide a permanent soil cover and, to
some extent, mimic the natural forest vegetation, are
considered more suitable for these conditions than an-
nual cropping systems (Ewel, 1986). However, be-
cause of the low chemical fertility of the soils, the cor-
rection of high acidity and initial nutrient deficiencies
(e.g., phosphorus, calcium and micronutrients) as well
as the replacement of nutrient exports in the harvested
biomass will normally also be necessary in permanent
agricultural systems (Szott and Kass, 1993).
As the nutrient availability in the topsoil is increased
by fertilization, intensive rainfalls may wash mobile
nutrients, such as nitrates, into the subsoil, despite
the presence of the perennial root systems of the tree
crops. Some of this nitrate can be retained in the sub-
soil through sorption to positively-charged surfaces,
thereby remaining within the reach of deep-rooting
plants (Cahn et al., 1992; Hartemink et al., 1996). The
amount of nutrient leaching and the degree to which
these nutrients can be recycled by the vegetation de-
pend obviously on the species of plants present (e.g.,
their root distribution) and their management, espe-
cially the fertilization rate and timing.
The objective of this study was to identify com-
binations of perennial crop species and management
practises that minimize nutrient losses into the sub-
soil and/or increase nutrient uptake from the subsoil,
and that might be of use for the development of
resource-efficient, permanent land use systems for the
region. The distribution of mineral N in the first two
meters of a ferralitic upland soil of central Amazonia
was evaluated at the beginning of the rainy season.
Within a perennial polyculture system, the soil under
four common tree crop species of the region and a
leguminous cover crop was studied at two fertiliza-
tion levels. The polycultures were compared with
a monoculture plantation, spontaneous fallow and
primary rainforest. The study was part of a larger
research program on the recovery of abandoned land
with perennial polycultures in central Amazonia.
2. Materials and methods
2.1. Study site
The study was conducted on the research station of
Embrapa Amazônia Ocidental near Manaus in Brazil-
ian Amazonia (3◦ 80 S, 59◦ 520 W, 40–50 m a.s.l.). The
climate is of the Köppen Am type, with an annual
precipitation of 2622 mm, air temperature of 26◦ C
and atmospheric humidity around 85% (mean val-
ues 1971–1993, O.M.R. Cabral and C. Doza, unpub-
lished). The driest months are July to September, and
the wettest months are February to April. The soil
is a Xanthic Ferralsol, according to FAO/UNESCO
(1990). It is well-drained despite a clay content of
about 800 g kg−1 , has a very low cation exchange
capacity and, in the natural state, high aluminium sat-
uration (see fallow and primary forest profiles in com-
parison to the limed agricultural plots in Table 1).
Available phosphorus contents are very low except in
the top 10 cm in the fertilized plots. Total C and N de-
crease strongly with increasing soil depth, but are still
detectable at 2 m depth (Table 1).
The study site was first cleared from primary for-
est in 1980, using heavy machinery for windrowing
and the removal of tree stumps. In 1981, an experi-
ment with rubber trees (Hevea brasiliensis) was es-
tablished, which was abandoned in 1986 because of
heavy disease attack. The developing secondary forest
was manually cleared in 1992 and the vegetation was
burnt on the site.
 G. Schroth et al. / Agriculture, Ecosystems and Environment 75 (1999) 109–120 111

Table 1
Soil fertility data from the study area on a ferralitic upland soil in central Amazonia under agriculture, fallow and primary forest (means
of three replicate profiles)
Depth Bulk densitya C total N total pH pH P avail. ECECb Acidityc
(cm) (Mg m−3 ) (g kg−1 ) (g kg−1 ) (H2 O) (KCl) (mg kg−1 ) (cmolc kg−1 ) (%)

Agricultural plot (Brazil nut, full fertilization)

0–10 0.88 25.9 2.2 4.5 4.0 20.9 2.33 82.1
10–30 0.98 13.2 1.2 4.4 4.1 2.8 1.68 94.2
30–50 0.92 9.5 1.0 4.4 4.1 0.8 1.54 96.4
50–100 0.95 5.3 0.7 4.5 4.3 0.2 1.15 93.0
100–150 0.94 3.6 0.5 4.7 4.5 0.1 0.63 76.5
150–200 1.00 2.7 0.4 5.0 4.7 0.1 0.34 62.2
Fallow
0–10 18.9 1.6 4.5 4.0 6.3 2.12 92.4
10–30 12.0 1.1 4.4 4.1 1.4 1.64 95.7
30–50 8.5 0.9 4.7 4.2 0.4 1.30 96.8
50–100 5.5 0.7 5.0 4.3 0.2 0.82 97.1
100–150 3.5 0.5 5.1 4.4 0.1 0.55 97.3
150–200 2.6 0.4 4.7 4.5 0.1 0.32 94.4
Primary forest (Oenocarpus bacaba)
0–10 27.8 2.0 4.2 3.6 3.9 2.89 95.1
10–30 10.8 1.0 4.4 4.1 0.9 1.62 96.8
30–50 7.7 0.8 4.7 4.2 0.4 1.31 96.8
50–100 5.0 0.6 4.9 4.3 0.2 0.86 97.8
100–150 3.0 0.4 5.0 4.4 0.1 0.56 97.6
150–200 2.3 0.4 5.4 4.6 0.1 0.33 92.2
a Measured in a soil profile under Pueraria in one of the agricultural plots.
b Effective cation exchange capacity.
c Exchangeable acidity in percent of ECEC.

2.2. Experimental plots

The total experimental area was about 13 ha. The

experimental plots were planted in February/March
1993. The centerpiece of the study was a polyculture
system with four tree crop species: the palm Bactris
gasipaes (peach palm, Arecaceae) for the production
of palmito (heart of palm); Theobroma grandiflorum
(cupuaçu, Sterculiaceae), a small tree related to cacao
whose fruit pulp is widely used in (and increasingly
outside) the region for the preparation of juice, ice
cream and sweets; Bertholletia excelsa, the Brazil nut
tree (Lecythidaceae), which, beside the well-known
nuts, produces excellent wood; and Bixa orellana (an-
Fig. 1. Layout of a polyculture plot with peach palm (P), cupuaçu
natto, Bixaceae), which is widely cultivated in the (C), annatto (A) and Brazil nut (B) on a ferralitic upland soil
tropics for its non-toxic red dye. The trees were grown in central Amazonia. The soil between the trees is covered by
in rows with 4 m spacing between the rows (Fig. 1). A Pueraria phaseoloides.
row with peach palm (at 2 m spacing within the row)
alternated with a mixed row of cupuaçu and Brazil nut of cupuaçu and Brazil nut, after which the next row
(at 6.7 m spacing within the row), a row of annatto (at of peach palm followed. Between the trees, Pueraria
4 m spacing within the row) and again a mixed row phaseoloides (tropical kudzu, Fabaceae) was sown as
112 G. Schroth et al. / Agriculture, Ecosystems and Environment 75 (1999) 109–120

Table 2
Fertilization of four tree crops on a ferralitic upland soil in central Amazonia with N, P and K in May 1996, December 1996 and May
1997 (the last three applications before the sample collection for the present study)
Plants per ha Na (g plant−1 ) Pa (g plant−1 ) Ka (g plant−1 )
May December May May December May May December May
1996 1996 1997 1996 1996 1997 1996 1996 1997

Polyculture, full fertilization

Peach palm 312.5 42 45 21 0 20 6 20 50 13
Cupuaçu 93.3 31 23 47 0 34 39 23 25 63
Brazil nut 93.3 23 23 21 0 20 11 42 0 25
Annatto 156.3 26 45 42 0 34 20 43 50 75
Sum (kg ha−1 ) 22 25 19 0 16 9 19 26 24
Polyculture, low fertilization minus N
Peach palm 312.5 0 0 0 0 6 2 9 15 4
Cupuaçu 93.3 9 0 0 0 10 12 7 8 19
Brazil nut 93.3 8 0 0 0 6 3 13 0 8
Annatto 156.3 8 0 0 0 10 6 13 15 23
Sum (kg ha−1 ) 7 0 0 0 5 3 7 8 7
Polyculture, low fertilization plus N
Peach palm 312.5 13 14 6 0 6 2 9 15 4
Cupuaçu 93.3 9 7 14 0 10 12 7 8 19
Brazil nut 93.3 7 7 6 0 6 3 13 0 8
Annatto 156.3 8 14 13 0 10 6 13 15 23
Sum (kg ha−1 ) 7 8 6 0 5 3 7 8 7
Monoculture
Peach palm 2500 42 45 21 0 20 6 20 50 13
Sum (kg ha−1 ) 105 113 53 0 49 14 49 125 31
a Fertilizer forms: N in 1996 as urea (45% N) and in 1997 as ammonium sulfate (21% N); P as triple super phosphate (22% P), except

in December 1996 (North Carolina Phosphate ‘Atifos’, 13% P); K as potassium chloride (50% K).

a cover crop, or developed from residual seed from the
previous rubber plantation. This system was studied
at two fertilization levels — full fertilization accord-
ing to local experiences and 30% of this fertilization
level with no more N fertilizer applied after May 1996
(‘low minus N’, Table 2). The latter treatment was in-
troduced to test if the leguminous cover crop alone
was able to supply sufficient amounts of N for the tree
crops in the system. In a further fertilization treatment
— ‘low plus N’ — the same fertilization was applied
as in the ‘low minus N’ plots plus 30% of the N fer-
tilizer applied in the full fertilization plots. The latter
treatment was not included in the soil N study, but
yield levels are given in Table 2. The fertilizer was
applied in two doses per year in November/December
(beginning of the rainy season) and May/June.
For comparison, a peach palm monoculture, planted
at 2 × 2 m, and plots with spontaneous fallow vege-
tation of the same age as the agricultural plots were
included in the study. The fallow plots were domi-
nated by the tree Vismia spp., which is a characteristic
genus in the vegetation of young fallows and degraded
lands in the region. The fallow vegetation was very
dense (1.95 ± 0.17 stems per m2 ) and approximately
5 m high. The peach palm monoculture was fertilized
at the same rate per tree as the peach palm in the poly-
culture plots with full fertilization (Table 2), and the
fallow plots were unfertilized. In November/December
1996, one year before the study, the fully fertilized
plots and the monoculture, but not the low fertilization
plots, were limed with 2.1 Mg ha−1 of dolomitic lime.
In addition, two tree species from a nearby primary
rainforest were included in the study: Eschweilera sp.
(Matá-matá, Lecythidaceae), a dicot tree, and Oeno-
carpus bacaba (Bacaba, Arecaceae), a palm. Both
species are relatively frequent in this forest and are
of commercial interest — Eschweilera for its wood
and Oenocarpus for its fruits that are collected by the
 G. Schroth et al. / Agriculture, Ecosystems and Environment 75 (1999) 109–120
local population for the preparation of juice. For each
species, three well-developed, adult specimens were
included in the measurements.
The measurement plots with the exception of the
rainforest sites, but including the fallow plots, were
arranged in a randomised complete plot design with
four replications, three of which were used in the N
study. Plot size was 24 m × 32 m in the peach palm
monoculture and 48 m × 32 m in the polycultures and
the fallow plots.
The peach palm in both polycultures and monocul-
tures was managed by cutting the main stem about 1.5
years after planting and harvesting the palmito three
times per year when the offshoots reached a diameter
of 8 cm at 1 m height. The harvest residues were al-
ways left in the plot to decompose. The annatto was
cut back at about 1.5 m height once per year after the
harvest between March and May to increase fruit pro-
duction, removing all the leaves and small branches
that were also left to decompose under the trees.
Hydrological information for the investigated tree
species and vegetation systems are given by Schroth
et al. (1999).
2.3. Sample collection
Soil samples were collected between mid-October
and mid-November 1997, i.e., at the end of the dry
season. In the six polyculture plots (two fertilization
levels, three replications), five samples per plot were
collected. One sample was collected at 50 cm from the
trunk of a typical tree of each of the four species, and a
further sample was collected under the Pueraria cover
crop in the middle between a cupuaçu and a Brazil nut
tree, i.e., at maximum distance from the trees in the
plots. The objective of this sampling design was not
to measure an average situation within the polyculture
systems, but rather to characterize the extreme points
within the plots and to relate these to the properties and
the management of the different tree species present
(single-tree patterns). In the monoculture plots, one
sample per plot was collected at 50 cm from a typical
tree. In the fallow plots, one sample per plot was taken
at a random position because of the close spacing of
the trees. In the primary forest, one sample was taken
at 50 cm from each of the six tree individuals (two
species, three replications).
113
In each of the 42 sampling positions, soil was col-
lected from the following depths: 0–10 cm, 10–30 cm,
30–50 cm, 50–100 cm, 100–150 cm and 150–200 cm.
For the first depth, a cylindric root corer was used to
avoid loss of the fragile surface soil and to guaran-
tee representative sampling of all subhorizons. The re-
maining depths were sampled with an Edelman auger.
2.4. Soil analysis
The samples were stored in plastic bags in the shade
and were taken to the laboratory within a maximum of
three hours. In the laboratory, from every soil sample
two subsamples of 20 g were taken for N extraction,
and two samples of about 50 g were taken for the deter-
mination of the water content by drying at 105◦ C for
two days. For N extraction, 150 ml of 1 M KCl solu-
tion was added to the samples, the solution was mixed
with the soil by manual shaking and the samples were
left standing overnight to ensure complete wetting of
aggregates, which facilitated the extraction. The next
morning, the samples were extracted by mechanical
shaking for 30 min. The extraction solution was col-
lected with a pipette without filtration after a sedimen-
tation time of 1–3 h. Filtration of the subsoil samples
had been found almost impossible during preliminary
tests because of the high clay content of the soils, and
was also a source of ammonium contamination.
The extraction solutions were either analysed the
same day or were kept below 0◦ C (but not frozen) for
a few days until the analysis. Ammonium and nitrate
(after reduction to nitrite) were analysed photomet-
rically with a segmented flow analyzer (Skalar, The
Netherlands). The N contents were related to the dry
soil, taking into account the water content of the ex-
tracted soil samples.
The samples for the general soil characterization
(Table 1) were collected by the same method as the
samples for N extraction in the same plots one year
earlier (dry season 1996). Carbon and N were mea-
sured with a CNS analyzer. Phosphorus and basic
cations were extracted by the Mehlich III solution at
a soil : solution ratio of 1 : 10 (Tran and Simard, 1993)
and were measured photometrically with a segmented
flow analyzer (P) or by atomic absorption spectrome-
try. Exchangeable acidity was extracted with 1 M KCl
at a soil : solution ratio of 1 : 15 by 10 min shaking, and
a subsample of 25 ml was titrated with NaOH against
114 G. Schroth et al. / Agriculture, Ecosystems and Environment 75 (1999) 109–120
a phenolphthaleine indicator. The cation exchange ca-
pacity was calculated by summing the basic cations
and the exchangeable acidity. The pH was measured
by glass electrode at a soil : solution ratio of 1 : 2.5 in
distilled water and 1 M KCl.
For the calculation of N stocks per hectare, the
bulk density of the soil was measured with volumet-
ric cylinders in a soil pit to 2 m depth under the cover
crop in one of the polyculture plots. Five cylinders
of 100 cm3 were collected at 10 cm intervals in the
top 1 m and at 20 cm intervals between 1 and 2 m
depth. The results are included in Table 1 in the Brazil
nut profile. In addition, cylinders were collected from
the topsoil (1–6 cm) at 40 cm from the stem of one
tree per species in three polyculture plots (three cylin-
ders per species and plot), three monoculture plots
(three cylinders per plot), three fallow plots (three
cylinders per plot), and the primary forest (six trees
from two species, nine cylinders per tree). The mean
bulk density of the topsoil of all species lay between
0.82 and 0.96 Mg m−3 (mean = 0.89 Mg m−3 ), with-
out significant differences between species (F = 1.20,
p = 0.361). We, therefore, assumed that there were also
no significant species effects on bulk density at greater
soil depths and applied the bulk densities from the
profile to all sampling points of the N distribution
study.
2.5. Statistical analysis
Statistical comparisons were conducted separately
for the different soil depths and for the sum of all
depths. For summing the soil depths, the bulk den-
sity was taken into account, as explained above. Sepa-
rate analyses were calculated for (a) ammonium-N, (b)
nitrate-N and (c) mineral N (the sum of a and b). First,
the five sampling positions (four tree species and the
cover crop) within the polyculture systems at the two
fertilization levels were tested for significant differ-
ences, without taking the other systems (monoculture,
fallow and forest) into account (‘species comparison’).
For this, a two-factorial analysis of variance for a ran-
domized complete block/split plot design was calcu-
lated with the fertilization level as the main plot factor
and the plant species as the subplot factor (Little and
Hills, 1978). In case of a significant F-test at p < 0.05,
mean separations were computed by least significant
difference tests at the same level of significance.
The system types (or main plots of the experiment,
i.e., polyculture plots at full and low fertilization,
monoculture and fallow) were compared in a separate
analysis for a randomised complete plot design (ref-
ered to as ‘systems comparison’ below). For this anal-
ysis, it was necessary to calculate a mean value from
the different sampling positions in the polyculture
plots, which could be compared with the values from
the monoculture and the fallow. The sampling scheme
had been designed for detecting extreme values within
the plots and did not allow the characterization of the
area around a tree (or, for the cover crop, between the
trees) for which each sample was representative. For
this reason, it was not possible to calculate precise
area averages for the different variables measured for
the whole polyculture systems. Instead, the simple
arithmetic mean of the five sampling positions within
a plot was calculated for each variable and these val-
ues were used in the comparison with the monoculture
and fallow plots. As before, means were separated by
LSD test in case of a significant F-test at p < 0.05.
The sampling positions in the primary forest were
only compared qualitatively with the other positions
and were not included in the analyses, because they
were not part of the original experimental design and
were not spatially interspersed with the experimental
plots (Hurlbert, 1984).
3. Results
3.1. Crop yields and tree growth
The crop yields in the polyculture and monoculture
plots for the year 1997, during which the soil N study
was carried out, are given in Table 3. From the fertil-
ization levels in Tables 2 and 3, only the levels ‘full
fertilization’ and ‘low fertilization without N’ were
included in the N distribution study (see above).
The cupuaçu yields were slightly higher in the full
fertilization than in the low fertilization treatment, but
the difference was not significant. The comparison of
the low fertilization treatment with and without N fer-
tilization showed that N alone had no effect on the
yields. Annatto produced significantly higher yields in
the full fertilization than in the low fertilization treat-
ment, but the comparison of the low fertilization treat-
ments with, and without, N fertilizer also showed that
 G. Schroth et al. / Agriculture, Ecosystems and Environment 75 (1999) 109–120 115

Table 3
Crop yields in polyculture at three fertilization levels and in mono-
culture on a ferralitic upland soil in central Amazonia during the
study year 1997
System Polycultures Monoculture F

Fertilization Low −N Low +Na Full Full

Peach palmb 0.62 0.67 0.62 0.68 0.50

Cupuaçuc 2.98 2.20 3.52 0.30
Annattod 0.59 bc 0.37 c 0.90 a 6.78e
a Treatment not included in the soil N study.
b In kg palmito cream per tree.
c In kg fresh fruit per tree.
d In kg dry seeds per tree.
e p < 0.05; values followed by the same letter within a row are

not significantly different at p < 0.05 (LSD test).

the yield response was not to the applied N. Peach

palm did not respond to higher fertilization. The Brazil
nut trees had a mean height of 6.5 m and a mean diam-
eter of 10.6 cm at breast height without significant dif-
ferences between fertilization treatments (C. Azevedo,
unpublished data). These data indicate that N fertil-
ization did not increase plant growth and yield at this Fig. 2. Profiles of mineral N under four different tree crop species
site. and a cover crop in a polyculture system at two fertilization
levels on a ferralitic upland soil in central Amazonia (means and
S.E.). The area between the y axis and the circles represents the
3.2. Distribution of mineral N in the polyculture soil ammonium-N, and the area between the circles and the squares
(species comparison) represents the nitrate-N.

Table 4
Fig. 2 shows the profiles of mineral N for the five Mineral N forms in the soil under four tree crop species and
sampling positions within the polyculture systems at a leguminous cover crop in a polyculture system on a ferralitic
two fertilization levels. In all cases, most of the min- upland soil in central Amazonia (means of two fertilization levels)a
eral N in the profiles was nitrate. Significant amounts Nitrate-N Ammonium-N Mineral N
of ammonium were measured only in the surface soil; (g m2 )
they decreased to low values at 10–30 cm depth. This Peach palm 14.2b 1.1 15.3b
shows that nitrification was rapid despite the low pH Cupuaçu 26.2a 1.2 27.5a
values, in agreement with measurements at a nearby Brazil nut 13.2b 1.3 14.5b
site under rainforest (Vitousek and Matson, 1988). Annatto 14.1b 1.4 15.4b
Pueraria 26.1a 1.5 27.7a
Fertilization level and sampling position (i.e., tree or
cover crop species) had no significant influence on the F (Fertil. level) 0.66 7.17 0.61
F (Species) 3.98b 0.50 4.10b
ammonium concentrations in any of the six soil layers
F (Fertil. × Species) 2.11 0.76 2.20
(data not shown), or in the whole profile (Table 4).
a0–2 m soil depth.
The nitrate concentrations in the soil decreased in bp < 0.05; values followed by similar letters in the same column
all cases from the topsoil to a depth between 50 and
are not significantly different at p < 0.05 (LSD test).
150 cm, as would be expected. With further increasing
soil depth, however, there was an increase in the ni- As the ammonium concentrations did not differ be-
trate concentrations under all species (Fig. 2). Table 5 tween sampling positions, the results for nitrate alone
gives the results of the statistical analysis for mineral N (data not shown) were very similar to those for min-
(ammonium-N + nitrate-N) in the different soil layers. eral N. The fertilization effect was not significant in
116 G. Schroth et al. / Agriculture, Ecosystems and Environment 75 (1999) 109–120

Table 5
Mineral N in the soil under four tree crop species and a leguminous cover crop in a polyculture system on a ferralitic upland soil in
central Amazonia per soil depth (means of two fertilization levels)
Soil depth [cm] 0–10 10–30 30–50 50–100 100–150 150–200
Mineral N (mg kg−1 )a

Peach palm 17.0 8.5 5.3b 4.2 6.1b 12.5

Cupuaçu 12.0 7.3 5.5b 8.8 18.8a 21.9
Brazil nut 27.4 6.9 3.1b 3.0 5.3b 12.5
Annatto 15.0 7.1 5.6b 4.5 6.4b 13.1
Pueraria 40.6 12.3 12.2a 7.7 12.3ab 20.0
F (Fertil. level) 0.00 0.75 2.61 0.06 1.27 0.62
F (Species) 1.78 0.96 4.85c 2.60 5.28c 1.79
F (Fertil. × Species) 0.48 0.39 0.36 0.64 4.01b 2.53
a Sum of ammonium-N and nitrate-N.
b p < 0.05; values followed by similar letters in the same column are not significantly different at p < 0.05 (LSD test).
c p < 0.01; values followed by similar letters in the same column are not significantly different at p < 0.05 (LSD test).

any of the investigated soil layers, whereas the species

effect was significant at 30–50 cm and at 100–150 cm
depth, and the fertilizer-species interaction was signif-
icant at 100–150 cm depth. At the 50–100 cm depth,
the species effect was significant at p = 0.076 for both
nitrate and mineral N. At 30–50 cm depth, the soil un-
der the cover crop had significantly more mineral N
than the soil under the four tree crops, which did not
differ between each other. The mineral N concentra-
tion in the soil at 0–30 cm depth was also higher under
the cover crop than under the tree crops, especially at
the low fertilization level, although this was not sig-
nificant. At 100–150 cm depth, the soil under cupuaçu
with full fertilization had significantly more mineral
N than the soil at all other sampling positions of both
fertilization levels, including the cupuaçu in the low
fertilization plots. The soil under Pueraria in the full
fertilization treatment had significantly more mineral
N than the soil under peach palm and Brazil nut at the
same fertilization level. At 150–200 cm, the soil under
cupuaçu and Pueraria still had more mineral N than
the soil under the other tree crops in the full fertiliza-
tion plots, but the difference was not significant. These
differences resulted in significantly higher mineral N
stocks in the soil profile under cupuaçu (27.5 g m−2 )
and the cover crop (27.7 g m−2 ) than under the remain-
ing three tree crop species (14.5–15.4 g m−2 , Table 4).
The spatial patterns of the distribution of mineral N
within the polyculture plots were related to the fertil-
ization level of the respective tree species (Table 2).
Cupuaçu received the highest N fertilization in May
Fig. 3. Profiles of mineral N under peach palm monoculture and
Vismia fallow on a ferralitic upland soil in central Amazonia
(means and S.E.). The area between the y axis and the circles
represents the ammonium-N, and the area between the circles and
the squares represents the nitrate-N.
1997 and had also the highest accumulation of min-
eral N in the subsoil. Brazil nut received a low fer-
tilization with N and had the lowest accumulation of
mineral N in the subsoil. However, the difference in
subsoil N between the plots with full fertilization and
those with low fertilization minus N was less than what
might have been expected, probably because more N
had been lost below the sampling depth of 2 m in the
full fertilization plots.
3.3. Distribution of mineral N in the soil under
different land use types (systems comparison)
The profiles of mineral N for the peach palm mono-
culture and the fallow are given in Fig. 3. In compar-
ison to the sampling position under peach palm in the
polyculture plots (Fig. 2), the soil in the peach palm
monoculture tended to have less mineral N in the top-
soil. However, the nitrate concentration in the soil of
 G. Schroth et al. / Agriculture, Ecosystems and Environment 75 (1999) 109–120 117

Table 6
Mineral N forms in perennial polyculture systems at two fertiliza-
tion levels (full and low minus N), peach palm monoculture, fallow
and primary forest on a ferralitic upland soil in central Amazoniaa
Nitrate-N Ammonium-N Mineral N
(g m−2 )

Polyculture full 20.8a 1.2 22.0a

Polyculture low −N 16.7a 1.4 18.1a Fig. 4. Profiles of mineral N under two primary forest species
Monoculture 20.1a 0.9 21.1a (Eschweilera sp. and Oenocarpus bacaba) on a ferralitic upland
Fallow 4.2b 1.1 5.3b soil in central Amazonia (means and S.E.). The area between the
Forest (Eschweilera) 12.7 1.9 14.6 y-axis and the circles represents the ammonium-N, and the area
Forest (Oenocarpus) 9.8 2.4 12.2 between the circles and the squares represents the nitrate-N.

F 10.24b 2.54 10.62b

a 0–2 m soil depth.
b p < 0.01; values followed by similar letters in the same column
are not significantly different at p < 0.05 (LSD test). The two
forest tree species were not included in the statistical analysis.
the monoculture plots also showed a pronounced in-
crease below 1 m soil depth, and the total accumula-
tion of mineral N in 0–2 m soil depth under peach palm
was even slightly higher in the monoculture than in
the polyculture, although the difference was not sig-
nificant (Tables 4 and 6).
In the fallow, in contrast, there was only a very slight
increase in the nitrate concentration in the soil below
1 m depth (Fig. 3). The concentration of mineral N be-
low 10 cm soil depth in the fallow soil was lower than
in all sampling positions of the agricultural plots (in
0–10 cm, the monoculture soil had slightly less min-
eral N than the fallow, Fig. 3). This resulted in signifi-
cantly lower accumulations of nitrate and mineral N in
the fallow, compared with the agricultural plots, which
did not differ significantly from each other (Table 6).
Comparing the fallow with the individual sampling
positions in the agricultural plots, the accumulation
of mineral N in the top 2 m of soil was between 2.5
and 7 times lower in the fallow (5.3 g m−2 ) than un-
der the tree and cover crops (12.4–37.3 g m−2 , Fig. 2).
Ammonium N did not differ significantly between the
systems (Table 6).
3.4. Distribution of mineral N in the primary forest
soil
As in most other sampling positions, the mineral N
concentration in the forest soil decreased with increas-
ing soil depth near the soil surface. Below 1 m depth,
the nitrate concentration in the soil increased again,
and the nitrate accumulation in the subsoil was more
pronounced in the forest than in the fallow (Fig. 4).
Also, the concentration of mineral N in the topsoil was
higher in the forest than in the fallow. For the whole
profile (0–2 m), the average mineral N accumulation
in the soil (14.6 g m−2 for Eschweilera and 12.2 g m−2
for Oenocarpus, Table 6) was intermediate between
the fallow and the agricultural plots. In the topsoil un-
der Oenocarpus, the highest ammonium concentration
from all sampling positions was found (59% of the
total mineral N), possibly indicating that nitrification
was retarded in comparison to the agricultural plots in
the proximity of this palm species (Fig. 4).
4. Discussion
Substantial quantities of mineral N were found in
the subsoil of a ferralitic Amazonian upland soil un-
der perennial crops at the beginning of the rainy sea-
son, shortly before the fertilizer application. Subsoil
accumulations of mineral N have been reported from
different tropical regions under annual crops (Cahn
et al., 1993; Weier and Macrae, 1993; Hartemink et
al., 1996), but leaching losses from perennial crops
are often thought to be rather low (Santana and
Cabala-Rosand, 1982; Seyfried and Rao, 1991; Bab-
bar and Zak, 1995). However, the accumulation of
2200 kg of nitrate-N per hectare at 1–5 m soil depth
under a fertilized coffee plantation in Kenya has re-
cently been reported by Michori (1993, cit. in Buresh
and Tian, 1998).
The significantly higher accumulation of mineral N
in the fertilized plots, compared with the unfertilized
fallow, indicates that part of the subsoil N was de-
118 G. Schroth et al. / Agriculture, Ecosystems and Environment 75 (1999) 109–120
rived from mineral fertilizer. This also explains why
the tree species with the highest N fertilization within
the polyculture systems (cupuaçu) also had the high-
est N accumulation in the subsoil, and why the N ac-
cumulation under cupuaçu increased with the fertil-
ization level. The rapid nitrification of the ammonium
derived from the fertilizer facilitated N leaching into
the subsoil (Robertson, 1989) where the nitrate was
retained by anion sorption (Cahn et al., 1992).
The high mineral N accumulation, especially in
the topsoil, under the cover crop within the polycul-
ture plots cannot be explained by N fertilizer, because
these positions had never been fertilized with N. Pre-
sumably, this N originated from the mineralization of
leguminous biomass and soil organic matter. Part of
the N may have been derived from biological N fixa-
tion, although the nodulation of the Pueraria was rela-
tively weak. The same N sources may also explain the
mineral N accumulation in the low fertilization plots,
where no N fertilizer had been applied for about 16
months before the sampling (Table 2).
Due to the close spacing of the trees in the peach
palm monoculture and the fallow, these plots had no
leguminous, and little non-leguminous understorey
vegetation. This and the more efficient N uptake by
the closely spaced trees may explain the lower con-
centration of mineral N in the topsoil of the monocul-
ture compared with the polyculture plots (Figs. 2 and
3). However, even the dense palm stand of the mono-
culture could not absorb all the applied fertilizer and
mineralized soil N, as evidenced by the pronounced
nitrate accumulation in the subsoil.
The growth and yield data and the distribution of
mineral N in the soil indicated that the applied fer-
tilizer N was not efficiently taken up by the trees.
This may have been because the trees were not lim-
ited by N, or because the applied N was leached too
rapidly out of the topsoil due to the combination of
high nitrification rates, intensive rainfalls and perme-
able soils. This points to the need to develop scien-
tifically based fertilizer recommendations, which do
not exist at present, for the investigated (and many
other) fruit tree species for central Amazonian condi-
tions. These would probably include splitting of the
annual fertilizer dose into more than two applications
to keep the nutrient concentration in the soil solution
low (Alva et al., 1998), and improved synchronization
of the applications with growth and nutrient uptake dy-
namics of the crops. Moreover, hydrological measure-
ments in the experimental plots of this study indicated
that no soluble fertilizers should be applied close to
the stem of species with high stemflow because of the
increased water infiltration and consequently leaching
risk in these positions. Such species include Brazil
nut and peach palm for fruit (but not peach palm for
palmito; fruit palms were not included in the mineral
N study) (Schroth et al., 1999).
The mineral N accumulation under the unfertilized
cover crop in the polyculture system suggests that its
potential with respect to the N nutrition of the tree
crops was strongly underutilised. Better use could be
made of this surplus N through reduced spacing of the
trees, either permanently or temporarily, by including
semi-perennial crops. In addition, it is likely that a re-
duction of the N fertilization of the trees would not
only reduce directly the N leaching, but would also in-
crease the competition between the trees and the cover
crop for N in the intertree spaces. The advantages may
be severalfold: a reduced need for N fertilization of the
tree crops that may, in some cases, even become un-
necessary; reduced nitrate leaching (and consequently
soil acidification) under the cover crop; and eventu-
ally even an increased fixation of atmospheric N by
the cover crop when the availability of mineral N in
its root zone is reduced by tree root uptake (Giller and
Wilson, 1991).
The inspection of the N profiles in the soil under
the two rainforest trees shows that even the unfertil-
ized and undisturbed natural vegetation lost nitrate into
the subsoil (Fig. 4). Amazonian upland forests on fer-
ralitic soils are limited by P, Ca and Mg, rather than
by N (Cuevas and Medina, 1986; Cuevas and Medina,
1988; Vitousek and Matson, 1988), and this may ex-
plain why they have not evolved more efficient recy-
cling mechanisms for nitrate. Much lower subsoil ni-
trate contents than in the agricultural plots were found
only under the impoverished fallow vegetation (Fig.
3). This observation discourages expectations that a
vegetation with relatively high N availability (either
through fertilization, biological N fixation or, as in the
rainforest, rapid internal N cycling) could be made
completely tight for very mobile nutrients under con-
ditions of high rainfall and permeable soils. Under the
pedoclimatic conditions of this study, productive agri-
cultural systems will probably always lose nutrients
into the subsoil.
 G. Schroth et al. / Agriculture, Ecosystems and Environment 75 (1999) 109–120
It has been hypothesized that in agroforestry sys-
tems deep rooting trees could reduce leaching by form-
ing a ‘safety-net’ with their roots under shallower
rooted crops (van Noordwijk et al., 1996). From the
tree crops in this study, none was able to play this role,
as all of them were sources of subsoil nutrients, instead
of being sinks. This may have partly been an effect
of systems design and management. Trees which are
expected to intercept leached nutrients should proba-
bly not be fertilized with these nutrients. In addition,
a certain amount of root competition in the topsoil
through close spacing with other trees or certain cover
crop species has been found to increase tree root de-
velopment in the subsoil, and consequently the use of
subsoil resources (Atkinson et al., 1976; Comerford et
al., 1984; Eastham et al., 1990; Neves et al., 1998).
5. Conclusions
The finding of considerable amounts of mineral N
in the subsoil under perennial polyculture and mono-
culture systems on a ferralitic Amazonian upland soil
indicates that there is a potential for reducing fertil-
izer inputs through more efficient nutrient use, includ-
ing increased N uptake by tree crops from the sub-
soil and from the soil under the cover crop. How-
ever, even in well-designed perennial cropping sys-
tems, high rainfalls and permeable soils, which are
typical for large areas in the humid tropics, make it un-
likely that leaching losses of mobile nutrients can be
completely avoided. The fact that all tree species in this
study were sources, and not sinks, for mineral N in the
subsoil shows that not every tree species/management
combination is suitable for the reduction of nutrient
leaching and the recycling of subsoil nutrients. Not
only must the species possess suitable characteristics,
such as the tendency to form extensive and deep root
systems under the given pedoclimatic conditions, but
the management of the trees must also favour their
development.
In addition to the improvement of cropping sys-
tems for better nutrient use, there is a need to develop
site- and species-specific fertilizer recommendations
for Amazonian tree crops. Soil analyses as a basis for
N fertilization of tree crops should include the sub-
soil, as significant amounts of previously applied fer-
tilizer nutrients may still be within the reach of the
119
crop roots, especially in the case of older trees with
well-developed root systems.
Acknowledgements
This work was funded by the German Ministry
of Education, Science, Research and Technology
(BMBF) together with the Brazilian Conselho Na-
tional de Desenvolvimento Cientı́fico e Tecnológico
(CNPq) as part of the SHIFT program, project
0339641/ENV 45. José Pereira da Silva Junior made
helpful comments on a previous draft of the paper. We
thank Luiz Gonzaga for careful assistance in sample
collection and processing.
References
Alva, A.K., Paramasivam, S., Graham, W.D., 1998. Impact of
nitrogen management practices on nutritional status and yield
of Valencia orange trees and groundwater nitrate. J. Environ.
Qual. 27, 904–910.
Atkinson, D., Naylor, D., Coldrick, G.A., 1976. The effect of tree
spacing on the apple root system. Hortic. Res. 16, 89–105.
Babbar, L.I., Zak, D.R., 1995. Nitrogen loss from coffee
agroecosystems in Costa Rica: leaching and denitrification in
the presence and absence of shade trees. J. Environ. Qual. 24,
227–233.
Buresh, R.J., Tian, G., 1998. Soil improvement by trees in
sub-Saharan Africa. Agrofor. Syst. 38, 51–76.
Cahn, M.D., Bouldin, D.R., Cravo, M.S., 1992. Nitrate sorption
in the profile of an acid soil. Plant Soil 143, 179–183.
Cahn, M.D., Bouldin, D.R., Cravo, M.S., Bowen, W.T., 1993.
Cation and nitrate leaching in an oxisol of the Brazilian
Amazon. Agron. J. 85, 334–340.
Comerford, N.B., Kidder, G., Mollitor, A.V., 1984. Importance of
subsoil fertility to forest and non-forest plant nutrition. In: E.L.
Stone (Ed.), Forest soils and treatment impacts, 6th Proc. of
the North American Forest Soils Conference, Knoxville, June
1983. University of Tennessee, Knoxville, pp. 381–404.
Cravo, M.S., Smyth, T.J., 1997. Manejo sustentado da fertilidade
de um latossolo da Amazônia central sob cultivos sucessivos.
R. bras. Ci. Solo 21, 607–616.
Cuevas, E., Medina, E., 1986. Nutrient dynamics within amazonian
forest ecosystems I. Nutrient flux in fine litter fall and efficiency
of nutrient utilization. Oecologia 68, 466–472.
Cuevas, E., Medina, E., 1988. Nutrient dynamics within amazonian
forests II. Fine root growth, nutrient availability, nutrient
availability and leaf litter decomposition. Oecologia 76, 222–
235.
Eastham, J., Rose, C.W., Cameron, D.M., Rance, S.J., Talsma,
T., Charles Edwards, D.A., 1990. Tree/pasture interactions at a
range of tree densities in an agroforestry experiment II. Water
120 G. Schroth et al. / Agriculture, Ecosystems and Environment 75 (1999) 109–120
uptake in relation to rooting patterns. Aust. J. Agric. Res. 41,
697–707.
Ewel, J.J., 1986. Designing agricultural ecosystems for the humid
tropics. Ann. Rev. Ecol. Syst. 17, 245–271.
FAO/Unesco, 1990. Soil map of the world, revised legend. Rome,
pp. 119.
Giller, K.E., Wilson, K.J., 1991. Nitrogen Fixation in Tropical
Cropping Systems, CAB International, Wallingford, pp. 313.
Hartemink, A.E., Buresh, R.J., Jama, B., Janssen, B.H., 1996. Soil
nitrate and water dynamics in Sesbania fallows, weed fallows,
and maize. Soil Sci. Soc. Am. J. 60, 568–574.
Herrera, R., Mérida, T., Stark, N., Jordan, C.F., 1978.
Direct phosphorus transfer from leaf litter to roots.
Naturwissenschaften 65, 208–209.
Hurlbert, S.H., 1984. Pseudoreplication and the design of
ecological field experiments. Ecol. Monogr. 54, 187–211.
Little, T.M., Hills, F.J., 1978. Agricultural experimentation. John
Wiley, New York, pp. 350.
Neves, C.S.V.J., Dechen, A.R., Feller, C., Saab, O.J.G.A., Piedade,
S.M.S., 1998. Efeito do manejo do solo no sistema radicular
de tangerineira ‘Ponca’ enxertada sobre limoeiro ‘Cravo’ em
latossolo roxo. Rev. bras. Frut. 20, 246–253.
Robertson, G.P., 1989. Nitrification and denitrification in humid
tropical ecosystems: potential controls on nitrogen retention.
In: Proctor, J. (Ed.), Mineral Nutrients in Tropical Forest and
Savanna Ecosystems. Blackwell Scientific Publications, Oxford,
pp. 55–69.
Santana, M.B.M., Cabala-Rosand, P., 1982. Dynamics of nitrogen
in a shaded cacao plantation. Plant Soil 67, 271–281.
Schroth, G., da Silva, L.F., Wolf, M.A., Teixeira, W.G., Zech, W.,
1999. Distribution of throughfall and stemflow in multi-strata
agroforestry, perennial monoculture, fallow and primary forest
in central Amazonia, Brazil. Hydrol. Process., in press.
Seyfried, M.S., Rao, P.S.C., 1991. Nutrient leaching loss from two
contrasting cropping systems in the humid tropics. Trop. Agric.
(Trinidad) 68, 9–18.
Stark, N.M., Jordan, C.F., 1978. Nutrient retention by the root mat
of an Amazonian rain forest. Ecology 59, 434–437.
Szott, L.T., Palm, C.A., Sanchez, P.A., 1991. Agroforestry in acid
soils of the humid tropics. Adv. Agron. 45, 275–301.
Szott, L.T., Kass, D.C.L., 1993. Fertilizers in agroforestry systems.
Agrofor. Syst. 23, 157–176.
Tran, T.S., Simard, R.R., 1993. Mehlich III-extractable elements.
In: Carter, M.R. (Ed.), Soil Sampling and Methods of Analysis.
Lewis Publishers, Boca Raton, pp. 43–49.
van Noordwijk, M., Lawson, G., Soumaré, A., Groot, J.J.R.,
Hairiah, K., 1996. Root distribution of trees and crops:
competition and/or complementarity. In: C.K. Ong and P.
Huxley (Eds.), Tree-Crop Interactions. CAB International,
Wallingford, pp. 319–364.
van Wambeke, A., 1992. Soils of the tropics. McGraw-Hill, Inc.,
New York, 343 pp.
Vitousek, P.M., Matson, P.A., 1988. Nitrogen transformations in a
range of tropical forest soils. Soil Biol. Biochem. 20, 361–367.
Weier, K.L., Macrae, I.C., 1993. Net mineralization, net
nitrification, net nitrification and potentially available nitrogen
in the subsoil beneath a cultivated crop and a permanent pasture.
J. Soil Sci. 44, 451–458.