STAND ROTATION FREQUENCY
AS A DETERMINANT OF LEACHING
IN THE HUMID TROPICS
A DISSERTATION PRESENTED TO TH]E GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
The supervisor for my master's and doctoral programs, Jack Ewel, has had a large
influence on my formation as an ecologist. He provided the intellectual, financial, and
physical infrastructure for my research, has vigorously critiqued untold hundreds of pages
of my writing, and has been a worthy example of absolute commitment to science and
teaching. I am grateful. Tom Sinclair has patiently guided me over many years through the
intricacies of environmental plant physiology. I appreciate his way of treating me as an
equal and of giving my ideas a careful listening. Jack Putz, Kimberlyn Williams, and
Nicholas Comerford have been consistently helpful committee members who have given
me substantial aid with experimental design and analysis, editing, and other matters. I
particularly appreciate Kimberlyn Williams's help with equipment purchases. I also thank
David Eissenstat, an early committee member.
Earl Stone provided an exceptionally thoughtful critique of my research proposal.
Other scientists who have helped with discussion are Marco Bindi, P. S. C. Rao, Ram
Oren, Henry Gholz, Wendel Cropper, Peter Nkedi-Kizza, David Whitehead, Ian Calder,
John Grace, Steve Mulkey, Michael Keller, Edzo Veldkamp, Marco Gutierrez, Ankila
Hiremath, Hank Loescher, Ken Smith, and Matt Kelty. Jay Harrison spent many hours
assisting with statistics, and Michelle Curtis taught me how to use the computer programs
SAS and FoxPro. Becky Ostertag critiqued most of my dissertation and provided many
insightful suggestions on presentation of ideas.
The people involved with the Huertos project on sustainable use of tropical soils
were of inestimable help. My research built on experiments initiated and data collected by
Jack Ewel and collaborators Jeremy Haggar and Fabio Chaverri. Ankila Hiremath has
been a steadfast colleague throughout my doctoral program, and I am particularly grateful
for her help in collecting data after I returned from Costa Rica. Fabian Menalled helped
with data collection. The field crew--Silvino Villegas, Pablo Gonzalez, Gilberth Hurtado,
Roger Gomez, Denis Zufiiga, and Danis--maintained the experiments well and were a real
pleasure to interact with. Research assistants Olman Paniagua and Virgilio Alvarado did a
great job collecting data and kept me entertained with many a good "chile." Miguel
Cifuentes has responded effectively to innumerable requests for information.
The staff of the Glasnost project--Michael Keller, Patrick Crill, Ed Veldkamp,
Antje Weitz, Bil Grauel, Marvin Nufiez, Rodolfo Vargas, and others-provided
unpublished meteorological data and other resources. Antje Weitz shared her vast
knowledge of soil physics and field techniques, gave me access to unpublished soil
hydraulic conductivity and moisture retention data, and devoted several days to assisting
with experiments. Nalini Nadkarni graciously loaned meteorological equipment
Organization for Tropical Studies employees Gerardo Orosco and the workshop staff
helped with construction and trouble-shooting of field equipment. I also thank the station
directors, Bruce Young and Cynthia Echeverria for their assistance.
I appreciate the assistance with chemical analyses offered by the following: James
Bartos and the staff of the Institute of Food and Agricultural Sciences Analytic Research
Lab; Marian Sanchez and Victor Carmona at the Organization for Tropical Studies; Mary
Jeanne Sanchez and the staff at the lab at the International Institute for Tropical Forestry;
and, Eloy Molina and the staff of Centro de Investigaciones Agron6micas. I am grateful to
Michael Keller and Hank Loescher for transporting heavy samples across international
Financial assistance was provided by teaching assistantships from the University of
Florida Department of Botany, gifts from May and Dave Bigelow, and NSF grant DEB-
9318403 to Jack Ewel. I appreciate the secretarial and accounting help of Paula Rowe, Pat
Custer, Debbie Folks, and Corrine Arnold.
I thank my fellow graduate students for companionship and intellectual
stimulation. They include Jeff Hillard, Charlie Pedersen, Zuleika Pinz6n, Claudia Romero,
Carol Lippincott, Kim Heuberger, Lisa Zweede, Brett McMillan, Dan Wenny, Kevin
Baldwin, Ron Edwards, Garry Peterson, Deborah McGrath, Jose Escamilla, Patti
Anderson, Tracy Feldman, Hank Loescher, and Ken Smith. Doug Hornbeck and Sue
Mauk hosted an unforgettable graduation party. I deeply appreciate the friendship of
Jeremy Schwartz, David Dilcher, Becky Ostertag, and Ankila Hiremath. Jon and Meg
Blanchard, and Don Flickinger and Jennifer Silveira have been constant sources of support
and friendship over many years. Most of all, I thank my parents and the rest of my family
for their love and support.
TABLE OF CONTENTS
ACKNOW LEDGMENTS ............................................... ii
A B STRA CT ........................................................ vii
1 INTRODUCTION ................................................... 1
Study Site ..................................................... 2
Scope of D issertation ............................................. 4
2 STAND STRUCTURE, LEAF SIZE, AND STOMATAL PHYSIOLOGY EFFECTS
ON ESTIMATED ANNUAL EVAPOTRANSPIRATION FROM TROPICAL TREES 8
Introduction .................................................... 8
M ethods ...................................................... 10
R esults ....................................................... 20
D iscussion .................................................... 23
3 NUTRIENT CONCENTRATIONS IN SOIL WATER IN RESPONSE TO CUTTING
FREQUENCY OF TROPICAL TREE PLANTATIONS ..................... 36
Introduction ................................................... 36
M ethods ...................................................... 37
R esults ....................................................... 42
D iscussion .................................................... 44
4 NUTRIENT LEACHING IN SHORT-ROTATION TROPICAL FOREST STANDS 59
Introduction ................................................... 59
M ethods ...................................................... 61
R esults ....................................................... 69
D iscussion .................................................... 72
5 EFFECTS OF MINERALIZATION AND UPTAKE ON NITROGEN LEACHING
FROM STANDS OF THREE TROPICAL TREES .........................
M ethods .............................................
R esults ....... ......................................
D iscussion ...........................................
6 CONCLU SION S ..........................................
... .. .90
Species Effects on water Balance in a Tropical Tree Plantation ...........
Magnitude and Duration of Nutrient Loss After Cutting Plantations ........
Nutrient Limitation to Growth in Annually Cut Stands ..................
C onclusions ..................................................
LIST OF REFERENCES .............................................
BIOGRAPHICAL SKETCH ....................................
Abstract of A Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
STAND ROTATION FREQUENCY AS A DETERMINANT
OF LEACHING IN THE HUMID TROPICS
Seth Warren Bigelow
Chairperson: John J. Ewel
Major Department: Botany
In the humid tropics, large quantities of nutrients can be rapidly leached under
disturbance regimes that bare the soil frequently, such as short-rotation cropping. To
determine the impact of disturbance frequency on nutrient loss from a fertile soil in the
humid lowlands of Costa Rica, leaching was measured from plantations of three tree
species, each of which was subjected to three cutting cycles: annual, every four years, and
uncut. Leaching was determined by combining measurements of mineral nutrients in soil
water with estimates of water percolation rates. The latter were estimated using a water
balance model; inputs to the model were obtained from a study of evapotranspiration from
the stands, soil hydraulic conductivity and water retention measurements, and
micrometeorological measurements. Soil water for nutrient analysis was drawn from
porous ceramic cup samplers located 1 m below the soil surface. Uptake of nutrients into
plant tissues, nitrogen mineralization, and nitrification were measured to assess their
effects on leaching rates.
Nitrate leaching was much greater under the annual cutting cycle than in uncut
stands; there also were increases in potassium and other cations, but they were not as
large. Estimated annual evapotranspiration differed by as much as 10 per cent among the
three tree species, indicating that broad-leaved tree species can affect leaching through
their effects on stand hydrology. After trees in the 4-yr rotation plots were cut, nitrate
leaching increased substantially, but the duration of the increase was brief. The magnitude
of nitrate leaching immediately after cutting was similar in 1-yr and 4-yr rotations (range
29-41 kg/ha). Increased nitrate leaching was due to a combination of increased nitrogen
mineralization and nitrification in concert with decreased uptake. After several rotations,
trees in the 1-yr rotation stands began to show diminished growth, decreased leaf nitrogen
concentrations indicate that they were probably suffering from nitrogen limitation. In these
soils, short-rotation cropping has a deleterious impact on nitrogen retention leading to
nutrient limitations of productivity.
When lands in the humid tropics are converted from forest to agriculture, rapid
declines in plant productivity often ensue within just a few cropping rotations (Nye and
Greenland 1960, Sanchez et al. 1982, Dagar et al. 1995, Lal 1995). Reduced yields are
due to factors such as weed infestations, deterioration of soil physical properties, and
increased pressures from insects and pathogens (Sanchez 1976). Soil fertility depletion, a
particularly important factor, can occur swiftly, because leaching is facilitated by the high
rainfall and rapid decomposition rates in humid lowland tropics.
Leaching is the loss of a soil's soluble salts in percolating water. Plants influence
nutrient leaching by taking nutrients out of the soil solution, adding organic matter to the
soil, and, by transpiration, altering the rate at which water drains through the soil. When
vegetation is cut, all of these processes are affected: uptake and transpiration cease, and
organic matter enters the soil as a sudden pulse instead of a steady input. Nutrients such as
nitrogen, for which rates of supply are mediated biologically, are particularly affected,
because increased soil temperature and moisture after disturbance can lead to enhanced
mineralization and nitrification (Smethurst and Nambiar 1990), which in turn facilitate
nitrate leaching (Williard et al. 1997).
Nutrient uptake has been hypothesized to be the most important mechanism by
which plants affect leaching (Vitousek and Reiners 1975). A molecule that has been
incorporated into plant tissues is more resistant to leaching than one that is in the soil
solution (Bormann et al. 1960); therefore, the rate of uptake should exert some control
over the rate of leaching. Uptake rate, in turn, is related to species, growth rate, and stage
of a tree's life cycle (Miller 1995). In trees, the rate of accumulation of nutrients peaks
very early, while it is still building its crown. After crown closure, uptake is broadly
correlated with growth rate, but prior to that, species-specific differences related to leaf
longevity can play an important role in determining uptake rates (Wller 1995).
Cropping systems in which harvesting is frequent and involves substantial soil
disturbance (e.g., monocultures of annual crops and short-rotation monocultures of timber
trees) can have a serious impact on soil fertility in the humid tropics. The increasing
prevalence of these land-use systems in the humid tropics (Brewbaker 1985) demands a
closer examination of the relationship between harvest frequency and nutrient leaching. To
do this, I carried out a study whose objectives were to quantify nutrient leaching from
stands of varying rotation length and species and to quantify the relative contributions of
uptake and mineralization in determining the leaching of nitrate.
The study was carried out at the La Selva Biological Station, in the province of
Heredia, Costa Rica. The previous landowner, Dr. Leslie R_ Holdridge, established
multistoried agroforestry systems (Hunter and Camacho 1961, McDade and Hartshorn
1994), but little agroecological research was done in the years following the acquisition of
the land by the Organization for Tropical Studies in 1968. In the late 1980s, the site's
potential began to be realized with the initiation of investigations into soil properties
(Robertson and Sollins 1987) and plantations of native trees (Butterfield 1994) and has
since grown with the addition of projects on the effects of agricultural practices on
greenhouse gas emissions (Keller and Reiners 1994) and effects of plant life form diversity
and stand rotation length on sustainability, including maintenance (or loss) of soil fertility
(Haggar and Ewel 1994). My dissertation research was done under the auspices of this
The project was conceived as an outdoor laboratory, in which large-scale, well-
replicated experimental treatments could be used by a variety of investigators. It consists
of approximately 8 ha of experimental stands located on a floodplain terrace at the
confluence of the Sarapiqui and Puerto Viejo rivers (Fig. 1-1). There are two main
experimental treatments. In one, the effects of plant life form diversity are examined by
growing three native tree species in monocultures and in polycultures that incorporate
monocotyledonous plants. Research utilizing the monoculture-polyculture comparison has
been conducted on aspects of productivity (Haggar and Ewel 1997) and canopy
development (Menalled et al. in press). Research involving monocultures alone has been
conducted on competitive effects (Gerwing 1995), growth and allocation (Haggar and
Ewel 1995), and attack by a stem-boring moth larva (Rodgers et al. 1995). In the second
main experimental treatment, the effects of cutting cycle frequency are examined by
growing the same three trees in monocultures subjected to three rotation frequencies: 1 yr,
4 yr, and 16 yr. I utilized the latter experiment for my dissertation research and was able to
draw upon baseline data on plant growth, plant tissue nutrient concentrations, and soil
properties that had been collected since the establishment of the stands in 1991.
Scope of Dissertation
The dissertation unfolds in four main parts, in addition to the introduction and
conclusion. In the second chapter, evapotranspiration (ET) rates from stands of three tree
species, Hyeronima alchorneoides, Cedrela odorata, and Cordia alliodora, are
compared. ET is partitioned into three components: canopy transpiration, understory
transpiration, and evaporation from wet vegetation surfaces. Canopy transpiration was
estimated using the Penman-Monteith equation (Monteith and Unsworth 1990). Canopy
conductance, an important and hard-to-measure term in the Penman-Monteith equation, is
estimated using a combination of measurements of stomatal conductance, leaf area index,
and canopy gap fraction.
The third chapter explores the impact of species and cutting-cycle treatment on
nutrient concentrations in leaching water. A 3.5-yr time sequence of concentrations of the
plant macronutrients nitrate, calcium, potassium, and magnesium in leaching water is
presented. Leaching water was collected using porous cup samplers under weak vacuum,
at a depth of 1 m below the soil surface.
In the fourth chapter, the quantities of nutrients leached from the various cutting
frequency and species treatments are estimated by linking concentrations of nutrients in
soil water (presented in chapter 3) to the volume of water drainage. Drainage was
estimated from ET, precipitation, and soil moisture. ET was estimated by calculating
potential ET from daily temperature maxima and minima, then multiplying it by a crop
factor obtained for each tree species using ET estimates from the second chapter. Soil
moisture is estimated using the water-balance model SWAP (Soil-Water-Air-Plant) 1993
(van den Broek et al. 1994), which incorporates detailed information on soil physical
structure and water retention characteristics, measured on the study site soils in the study
area by a collaborator, Antje Weitz.
The fifth chapter compares the amount of nitrate leached (calculated in chapter 4)
in each treatment to the amount mineralized and the amount incorporated into tree tissues.
Nitrification and N mineralization were estimated using an in situ paired-core incubation
method (Pastor et al. 1984); N accumulation in trees was estimated by linking biomass of
leaves stems, and roots (estimated by harvest and allometry) to N concentrations in these
Short-rotation forest plantations are increasingly prevalent in the tropics
(Brewbaker 1985), including the Pacific islands (Phillips et al. 1995), South Asia (Toky
and Singh 1993), and South America (Spangenburg et at. 1996). In such plantations, a
high proportion of the rotation is spent in site-preparation and stand establishment,
exposing the system to nutrient losses (McColl and Powers 1984). Nevertheless, few data
on leaching losses in short-rotation forestry have been reported (see reviews by Raison
and Crane 1986, Heilman 1992). Examination of the link between rotation age and
nutrient retention will improve understanding and expand options for sustainable
management of tropical agroecosystems.
HYAL HYERONIMA ALCHORNEOIDES
COAL CORDIA ALLIODORA
Figure 1-1. Central America, with inset of study site at La Selva Biological Station. The 4-
letter acronyms denote tree species: HYAL is Hyeronima alchorneoides, CEOD is
Cedrela odorata, and COAL is Cordia alliodora. Dashed lines within boxes delineate
borders between uncut stands and 1- and 4-yr rotation stands. Numbers inside boxes
STAND STRUCTURE, LEAF SIZE, AND STOMATAL PHYSIOLOGY EFFECTS ON
ANNUAL EVAPOTRANSPIRATION ESTIMATED FOR TROPICAL TREES
Plantations of tropical trees are becoming a viable means of augmenting the
world's timber supply (Myers 1991), and an increasingly broad variety of trees is being
used. As such plantations increase in extent, so does the need for information on their
effects on ecosystem processes such as evapotranspiration (ET). ET is important in the
humid tropics for its role in the hydrologic balance: higher ET means decreased drainage
from soils, thus lower nutrient leaching and lower yields of water from watersheds (Lee
1980). Notwithstanding the many studies on physiological properties of tropical trees
(e.g., Fetcher 1979, Grace et al. 1982, Oberbauer et al. 1987, Roberts et al. 1990, Roy
and Salagher 1992, Meinzer et al. 1993, Granier et al. 1996), there is little information on
ET from single-species stands (but see Calder et al. 1991).
A traditional method of studying ET, the paired watershed experiment (Bosch and
Hewlett 1982), has a number of disadvantages, not least of which is difficulty in finding
sites that fit exacting requirements (e.g., an impermeable basement substrate). This
problem is avoided in an alternative approach: scaling up physiological and structural
measurements on a small number of individuals to the level of the stand. This approach is
particularly apt given the number of new species being used in tropical plantation forestry,
because it is mechanistic. It offers the possibility of predicting ET from species that are
relatively unknown to foresters, without the necessity of establishing extensive forest
stands for empirical measurements.
The Penman-Monteith equation provides a framework for estimation of ET from
scaled-up physiological and plant structural measurements coupled with environmental
measurements (Monteith and Unsworth 1990). It is based on physical principles of energy
balance, yet methods for obtaining two parameters that describe the plant's effect on ET
are still the subject of debate (Finnegan and Raupach 1987, Baldocchi et al. 1991, Norman
1993, Meinzer and Goldstein 1996). These parameters are vapor conductance of the plant
canopy and the atmosphere. Modeling the behavior of the canopy and surrounding air at
increasingly fine scales of description has led to models that are mathematically elegant,
yet are felt by some to be too complex for practical application (Raupach and Finnegan
1988, Baldocchi 1993). There is a need for simple, robust methods for estimating these
parameters for stands of different tree species.
The objective of this study was to adapt the Penman-Monteith equation to estimate
annual ET from even-aged stands of three tropical trees that contrast sharply in leaf size,
timing and duration of leaf drop, and canopy structure. Two of the species, Cedrela
odorata and Cordia alliodora, have been used extensively in tropical tree plantations,
whereas the third, Hyeronima alchorneoides, has only recently been used for reforestation
and has not yet been brought into widespread commercial production. It was expected
that ET would be lowest under Cedrela and Cordia, which are leafless for part of the year
and have rather sparse canopies. Hyeronima, in contrast, produces a dense canopy which
persists throughout the year, and it was expected that ET would be increased accordingly.
This did not prove to be true, though; ET from Hyeronima was lower than from the other
Site and species description
The study took place at La Selva Biological Station in the Atlantic lowlands of
Costa Rica. Stands of the three trees, all of which are native to the region, were
established in 1991 on the site of an abandoned cocoa plantation (Haggar and Ewel 1995).
The stands were strikingly different in appearance and phenology. Hyeronima
alchorneoides (Euphorbiaceae) is a massive canopy emergent with large evergreen leaves
(Table 2-1). It produced stands with even, dense, shallow canopies that were concentrated
within a small fraction of total tree height (Menalled 1996). Once complete canopy cover
was established, approximately 1.5 yr after planting, very few plants successfully colonized
in the understory. Cordia alliodora (Boraginaceae) is a slender canopy tree with simple
leaves which are much smaller than those of Hyeronima. Mature trees defoliate for several
months at the beginning of the wet season. In the experimental stands, Cordia crowns
were much sparser than those of Hyeronima, permitting growth of herbs and shrubs
below. This understory was cut approximately every 6 wk but grew back rapidly. The
third species, Cedrela odorata (Meliaceae), is a canopy tree that is leafless for several
months towards the beginning of the drier season, which starts in January. It has pinnately
compound leaves with leaflets that are smaller than the leaves of Hyeronima and Cordia.
Like Cordia, it formed stands with deep, sparse canopies and rapidly growing
Each tree species was established in thrice-replicated 50 x 40 m stands. At the
beginning of the study, stands of Hyeronima, Cordia, and Cedrela respectively averaged
2200, 2205, and 2444 stems/ha; a thinning in March 1995 reduced these densities by
roughly 101/o. Mean height of all species was between 10 and 15 m (Menalled 1996).
The area receives a long-term average annual precipitation of 3962 mm, most of
which falls in the May-December wet season (Sanford et al. 1994). Even the driest months
average more than 100 mm of rain. The soils are alluvial, from volcanic parent material,
and have a strongly aggregated structure that allows rapid drainage despite high clay
content and high water-holding capacity (average moisture in the top 0.1 m is 0.58 m3/m3;
Weitz et al. 1997) The combination of high water-holding capacity, good moisture-release
characteristics, and substantial rain even in the driest months probably prevents the
prolonged episodes of moisture stress that are common in more seasonal tropical climates.
The ET model
A slightly different form of the Penman-Monteith equation was used to express
each of three evaporative fluxes from the stands: tree transpiration, understory
transpiration, and evaporation from wet tree leaves during and immediately following rain.
Its general form is
AR +pcpDg,, f
(A + y (1 +ga/g ))fA
where E (minis) is evaporation and transpiration, R (J m"2 s-1) is net radiation flux density,
and D (kPa) is vapor saturation deficit in the air above the tree canopy (Monteith and
Unsworth 1990). The effects of the plant canopy are incorporated into the terms for
canopy (g) and aerodynamic (g) conductance (both in mol m-2 s'). The rate of increase in
saturated vapor pressure with temperature, A (kPa/K), is related to air temperature and
was calculated for each time increment of the model. Air density (p, 1.16 kg/n3), heat
capacity of air at constant pressure (c ; 1010 J kg-' K'), the latent heat of vaporization (A;
2.43 x 106 J/kg), and the psychrometer constant (r, 0.067 kPa/K) are all weak functions
of air temperature and were treated as constants by assuming a temperature of 30 *C.
Two conversion factors are included:f, (0.0245 m3/mol) converts conductance from units
of mol m2 s-' to rn/s assuming a temperature of 30 C and sea level atmospheric pressure
(Pearcy et al. 1989). The second factor, f2 (10' kg/mm3), converts ET from mass to depth
Meteorological data were obtained from instruments mounted on a portable
scaffolding tower. Net radiation flux density at 2 m above canopy was measured at 30 s
intervals with a net radiometer (model Q7, Radiation and Energy Balance Systems, Inc.,
Seattle, Washington USA). The net radiometer was initially mounted for 4 mo above a
stand of Cordia interspersed with palms and heliconias, which formed part of a different
study. During this time, comparisons were made with adjacent monocultures of all three
species by mounting another net radiometer above them on scaffolding for several days
each. After 4 mo, the net radiometer was placed in a Cedrela stand for 3 mo, and finally in
a Hyeronima stand for the remaining 5 mo.
There were 16 wk of missing net radiation data from December 1, 1994 to
November 30, 1995, which occurred due to instrument failures and movement of
equipment between stands. The gaps fell on these dates: December 28, 1994 to January 2,
1995; January 21 to January 23, 1995; March 25 to April 4, 1995; April 17 to April 24,
1995; June 7 to July 26, 1995; August 29 to September 15, 1995; and November 17 to
November 30, 1995. The gaps were filled by using photosynthetically active radiation
(PAR) data from two weather stations that were within 1 km of the study site. Equations
expressing net radiation as a function of PAR were made by linear regression; a separate
regression was made for the 10 d prior to each missing data segment, thus ensuring
appropriate calibration. All regressions had an r2 value > 0.93. Calibration of net
radiometers themselves was done every 3 mo by placing them alongside a net radiometer
that was recently factory-calibrated and had not been used in the field.
From December 1, 1994 to April 23, 1995 air temperature and relative humidity
were measured at a weather station adjacent to the study site using a solid-state humidity
sensor and thermistor (Campbell Scientific, Inc., Logan, Utah USA); after this, readings
were obtained from an aspirated thermocouple psychrometer mounted on the tower
alongside the net radiometer, except during times of instrument failure (listed in the
preceding paragraph) when data from the adjacent weather station were used. Vapor
pressure deficit (D) was obtained from the formula
D = e,(7) (1- h / 100)
where h is relative humidity (%) and e,(I) is saturated vapor pressure as a function of
temperature. The latter quantity was estimated using the Tetens (1930; also given in
Monteith and Unsworth 1990) formula to calculate saturated vapor pressure. It is
e,(T) = 0.611 exp (17.27 / (1 + 237 / ).
Precipitation was measured by a recording rain gauge. Instruments were connected to data
loggers (model CR10, Campbell Scientific, Inc.); averages or sums were recorded hourly
prior to 12 January 1995 and half-hourly thereafter.
Transpiration from the understory, E, was estimated using the equation
A R PC
Denmead 1984, Black and Kelliher 1989, Whitehead and Kelliher 1991). In this form of
the Penman-Monteith equation the terms that link transpiration to atmospheric vapor
content are omitted, so transpiration is determined entirely by radiant energy flux. The
understory was assumed to have insignificant water storage capacity and thus interception
by it was not modeled.
To estimate energy flux to the understory, above-canopy net radiation was
multiplied by the gap fraction of the tree canopy, PC. Canopy gap fraction was estimated
from measurements made in July 1994 and January 1995 with a plant canopy analyzer
(model LAI-2000, LiCor Co., Lincoln, Nebraska USA). This instrument measures
transmittance of light through the tree canopy at five angle ranges from 0' (zenith) to 690;
canopy gap fraction for each angle range is assuired to equal transmittance (Welles and
Norman 1991). In each of the three stands of the three tree species, one above-tree-
canopy and four below-tree-canopy readings were taken at nie locations. Canopy gap
fraction was averaged across the 5 angle ranges; this value was averaged across the three
replicate stands to arrive at a single canopy gap fraction for each species.
Transpiration from tree canopies
Terms for radiant energy input and aerodynamic (g) and canopy (g) conductance
were obtained to estimate transpiration from tree canopies. Radiant energy input to the
tree canopy was estimated by multiplying above-canopy net radiation by J-p, thus
producing the complement to understory energy flux.
Aerodynamic conductance (g) is typically estimated using a momentum-transfer
equation (Monteith and Unsworth 1990), but plot dimensions were too small to allow
development of the uniform boundary layer that is a requirement of this method (Munro
and Oke 1975). Instead, a published value, 10 mol m2 s' (Calder et al. 1986), was used.
This is at the high end of the range of aerodynamic conductances for tropical forests
(Table 2-2); it was assumed that aerodynamic conductances at the study site would be
high because the abrupt edges of the small stands should promote formation of large
Canopy conductance (g) is sometimes defined as the sum of stomatal
conductances above a representative unit area of ground (Roberts et al. 1980, Grace et al.
1982, Jarvis and McNaughton 1986, Baldocchi et al. 1991), but this definition excludes
boundary layer conductance. In practice, the leaf boundary layer is often incorporated into
forest canopy conductance: the Penman-Monteith equation is algebraically rearranged to
solve for canopy conductance, using independent estimates of aerodynamic conductance
and evaporation as inputs (Calder et al. 1986, Granier et al. 1996, Grace et al. 1995).
Because existing estimates of tropical forest canopy conductance incorporate leaf
boundary layer (Table 2-2), this was also done in the present study.
An electrical conductance analogy,
was used to scale stomatal and leaf boundary layer conductances to the level of the
canopy. Stomatal (g) and boundary layer (g.) conductances were assumed to be in series,
and the resulting conductances were summed in parallel over leaf area index (L). Stomatal
conductances were obtained using a null-balance porometer (model LI-1600, LiCor Co.,
Lincoln, Nebraska USA), whose calibration was checked each day before use. Four to five
days of porometer measurements, in August and September 1996, were made for each
species. A portable scaffolding was erected from which leaves of four to six trees could be
reached. Measurements were made throughout the canopy profile at approximately 2-hr
intervals; leaves were selected regardless of age or condition. Leaves were classified
according to apparent age (juvenile, mature, or senescent/diseased), and lighting (sunlit or
shaded). Vapor pressure deficit and incident photosynthetic photon flux density were also
Leaf boundary layer conductances were obtained for Hyeronima and Cedrela
using wetted filter-paper replicas of leaves (Jones 1992). At mid-canopy height on the
portable scaffolding, a wetted leaf replica was suspended from threads attached to a wire
frame. The areas of the Hyeronima and Cedrela leaf replicas were 318.5 and 30.7 cm2,
respectively. Evaporation rate was measured by weighing at 5-min intervals, and surface-
to-air vapor pressure deficit was determined from thermocouple measurements of leaf
replica temperature and psychrometer measurements of air temperature and humidity.
Boundary layer conductance was then calculated using the equation for latent heat flux
(Jones 1992). Measurements were made on a small number of replicates done at one
height on a single day, and thus do not incorporate variation due to canopy height or daily
Leaf area index was estimated with the plant canopy analyzer, using the
measurements made for canopy gap fraction determinations- The instrument uses gap
fraction to solve an exponential equation similar to Beer-Bouget-Lambert to arrive at leaf
area (Welles and Norman 1991).
Evaporation from the wet canopy
Evaporation of intercepted rainfall was modeled by setting the term g/g, in the
denominator of the Penman-Monteith equation to zero, because canopy conductance is
considered to be infinitely large when the canopy is wet (Shuttleworth 1988). The amount
of water stored on the canopy in thejth time interval, C, following rainfall P was
predicted using the equation
C. = C_ + ((I P) P E. H)At,
where C., is canopy water storage in the preceding time interval, Pt is free throughfall
(fraction of rain that falls to the ground without striking the canopy), and E, and H
respectively are evaporation and drainage from the canopy during the period At. It was
assumed that the tree canopy was either completely wet or dry at any given time (e.g.,
Calder et al. 1986, Lloyd et al. 1988). Although this is not the approach of the original
model (Rutter et al. 1971), the errors introduced by using this simplification are small
(Calder and Wright 1986).
It was assumed that no drainage occurred when canopy water storage was less
than maximal (Whitehead and Kelliher 1991); thus,
H=O when C
When canopy water storage exceeded the maximum, the rate of drainage was
controlled by the drainage functions
H = Ho e C-s) when C > S
in which H is drainage rate, H0 is drainage rate when canopy water depth is at maximum
capacity (S), and b is a drainage parameter representing the log-transformed rate of
drainage with respect to canopy water depth (Rutter et al. 1971).
Maximum canopy water storage, S, was approximated with a standard value for
tropical forests of 1 mm (Jetten 1996). Canopy drainage parameters were those used by
Lloyd and coworkers (1988) for tropical forest in central Amazonia; minimum drainage
rate, Ho, was taken to be 0.0014 mm/min, and the slope of the graph of log-transformed
drainage rate plotted against canopy water depth (b) was taken to be 5.25 ln(ml/min)/ml.
The free throughfall fraction (p) was estimated from the overhead (zenith angle 0'
to 70) gap fraction from the LAI-2000 canopy analyzer. Measurements were a subset of
those used for the canopy gap fraction determinations.
Seasonal variation in cumulative net radiation and mean vapor pressure deficits,
although present, was not pronounced (Fig. 2-1). As expected, these quantities tended to
be lower during months of high rainfall, although this pattern is attenuated because the
time of highest potential radiant energy influx, the summer months of June and July, falls
at the height of the local rainy season. Maximum daily temperature averaged over the
entire year was 31.3' 2.40 C, and maximum daily vapor pressure deficit varied between
0.01 and 2.5 kPa. Regressions of net radiation above the stands plotted against net
radiation measured simultaneously above a reference stand (which consisted of a Cordia
overstory with an understory of palms and heliconias) did not differ significantly from 1 (p
Canopy gap fraction
Hyeronima, whose high leaf area causes a darkened understory, had a canopy gap
fraction of only 0.03 m2/m2 Table (2-3). Cedrela and Cordia stands, which are much
sparser, had far larger gap fractions of 0.14 and 0.17 m2/m2, respectively. Once Cedrela
had shed its leaves during the drier season, its gap fraction increased to 0.62 m2/m2;
interception of light by branches, trunks, and leaves remaining on juvenile trees kept this
figure from being larger. Cordias gap fraction doubled to 0.38 m/m2 during the drier
season. These trends also held true for free throughfall fraction (Table 2-3).
The original intent was to weight the canopy conductance according to fraction of
canopy that was sunlit and shaded (Sinclair et al. 1976), but inspection of the data proved
this unnecessary. A three-way analysis of variance of stomatal conductance data showed
no effect due to sun or shade (p = 0.3), although species and time of day were highly
significant (p = 0.0001 and 0.0005, respectively). Daily trends of stomatal conductance
were similar among the three species, except that Hyeronima showed a pronounced
tendency towards midday stomatal closure (Fig. 2-2), which has also been noted by other
workers (S. Iriartea and R_ Montgomery, personal communication). With the exception of
midday Hyeronima readings, average stomatal conductances in all three species tended to
range between 0.3 and 0.4 mol m2 s1 for most of the day.
Leaf-boundary-layer conductance measurements made with filter-paper replicas
yielded much lower estimates for Hyeronima than for Cedrela (Table 2-1), which was
expected given the large size of the former's leaves. Measurements were not made on
Cordia, but its leaves are similar in shape and area to Cedrela leaflets and were therefore
assumed to have similar boundary-layer conductance.
Although the low stomatal and boundary layer conductances of Hyeronima
resulted in lower canopy conductances than for the other species, they were mitigated
somewhat by a high leaf area index (Fig. 2-3). Mid-morning canopy conductance of
Hyeronima was 0.33 mol m2 s-1, whereas for Cedrela and Cordia it was 0.45 mol n2 S.1
and 0.37 mol m2 s-', respectively. Lower dry-season leaf area index for Cedrela caused
mid-morning canopy conductance to diminish to 0.17 mol m2 s'. Although mature Cordia
trees become leafless during the beginning of the wet season, the trees in this study did not
show this pattern: instead, they had much lower leaf area during the dry season (Table 2-
3). Presumably as the trees age they will assume the typical adult phenological pattern of
early wet-season leaf loss (Greaves and McCarter 1990).
The simulations indicated substantial differences in annual ET among stands of the
three species: 1169 mm for Hyeronima, 1401 mm for Cedrela, and 1349 mm for Cordia
(Table 2-4). Tree transpiration in Hyeronima was 86% that of Cedrela, and 91% that of
Cordia. Understory transpiration (E.) in Hyeronima plots was negligible (Fig.2-4), in
contrast to Cedrela and Cordia, in which it constituted = 16% of total ET. Understory
transpiration in the latter species was substantial during the first 4 mo of the simulation
(December to April), when loss of leaves in the overstory increased light transmission to
Stomatal conductances in all three species tended to be high: even Hyeronimds
midday low of 0.12 mol m"2 s-1 exceeds values often found in woody tropical plants.
(Midday depression in stomatal conductance is by no means unusual in woody tropical
plants; Roy and Salagher 1992, Zotz and Winter 1996). Many such species have maximum
stomatal conductances of 0.1 to 0.2 mol m2 sI (Langenheim et al. 1984, Pearcy and
Calkin 1983, Oberbauer et al. 1987, Fanjul and Barradus 1985, Mulkey et al. 1996), in
contrast to the 0.3 to 0.4 mol m" s-' range of mean conductances reported herein. Other
woody tropical plants do have higher maximum stomatal conductances, ranging from 0.2
to 0.4 mol m"2 s" (Aylett 1985, Koch et al. 1994, Kelliher et al. 1995) to over 1 mol m" s-1
(Grace et al. 1982, Chiariello 1984). Good plant nutrition and hydration are strongly
correlated with high stomatal conductance (Kozlowski et al. 1991); thus, the values found
for the three species investigated at this site are attributable to high rainfall and deep,
Even at mid-day, leaf boundary-layer conductance in Hyeronima was lower than
stomatal conductance, thus at all times vapor phase conductance was dominated by the
former term. This boundary-layer conductance is lower than most reported values: in
central Amazonia, for example, single-sided boundary layer conductances ranged from
0.12 mol m2 s-' at the forest floor to 0.7 mol m-2 s-' in the upper canopy (Roberts et al.
1990), and values for woody species in gaps in Panama ranged from 0.08 to 0.33 mol m2
s"' (Meinzer and Goldstein 1996). Studies on tropical trees with large leaves comparable in
size to those of Hyeronima have shown boundary layer conductances of 0.4 mol m"2 s-' or
roughly 4 times that of Hyeronima (Meinzer et at. 1993). This discrepancy is explained in
part by the higher average wind speeds (1.5-2.0 m/s) in those studies. Wind speed in the
present study was measured at 2 m above the canopy and averaged 0.74 m/s from
November 1994-June 1995. This slower wind speed at the Costa Rica site would cause
increased boundary layer thickness and thus decreased conductance. Furthermore, in the
present study measurements were done in mid-canopy, where wind speeds would be
decreased still further, contributing to lower leaf-boundary-layer conductance.
Leaf boundary-layer conductances, five times higher in Cedrela than in
Hyeronima, were the most important factor in the latter's low estimated transpiration. The
influence of leaf boundary-layer on ET estimates can be appreciated by running a model
simulation using leaf area and stomatal conductance parameters derived for Hyeronima,
but increasing leaf boundary layer conductance to equal that of Cedrela (0.5 mol m2 s-').
This produces a 90% increase in estimated transpiration, from 648 mm to 1231 mm. Thus,
leaf-size effects on boundary layer conductances may be an important factor inducing
differences in transpiration rates among tropical trees.
There is no consensus on how to scale up leaf conductances to the level of a tree
canopy in part because the concept of canopy conductance itself is ill-defined (Finnegan
and Raupach 1987, Lhomme 1991, Baldocchi 1994). The method used herein, which
explicitly combined stomatal and leaf boundary layer conductances, yielded canopy
conductance estimates that were similar to those obtained by different methods for several
tropical forests (Table 2-1). When the leaf boundary layer was omitted from the canopy
conductance term and assigned instead to the aerodynamic conductance term, the resulting
estimated annual ET was improbably low. Future studies of evapotranspiration using the
Penman-Monteith approach should state where the boundary layer is assigned in the
system of canopy and aerodynamic conductances (Sinclair 1990).
Forest understories can contribute substantially to stand ET (Black and Kelliher
1989, Whitehead et al. 1994), in some cases causing ET from stands with sparse
overstories to equal that from denser stands (Roberts et al. 1982). In the present study,
understory transpiration contributed roughly 16% to total ET in both Cedrela and Cordia.
During the dry season, however, understory transpiration was the major evaporative flux
in Cedrela stands (Fig. 2-4). Indeed, simulated ET during this time from Cedrela stands,
which were virtually devoid of overstory leaves, was greater than that of Hyeronima,
which had a tree leaf area index of 4.5.
Lower simulated ET from Hyeronima than from Cordia or Cedrela was surprising
in light of the high leaf area and rapid growth (Haggar and Ewel 1995) of this species.
Leaf boundary layer and stomatal conductances were so low that they overwhelmed the
effects of leaf area on canopy conductance (Equation 1). It must be recognized, though,
that there is substantial uncertainty associated with the estimation of interception, because
general canopy drainage parameters were used, rather than ones specific to the three
species. Conceivably, actual evaporation of intercepted rain from Hyeronima stands might
be several hundred mm higher than that modeled with the present parameters, causing ET
to be equal between species. It is clear, though, that at least when the canopy is dry the
expectation of lower ET from stands of the partially deciduous species (Cordia and
Cedrela) is wrong, due to higher canopy conductances and large evaporative fluxes from
Annual ET estimated in this study, approximately 1450 mm/yr for the species with
highest ET, is somewhat lower than previous ET estimates for La Selva Biological
Station, which range from 1635 mm (estimate based on purely meteorological parameters;
Hancock and Hargreaves 1977) to over 2100 mm (based on detailed study of natural
forest; Luvall 1984, Sanford et al. 1994). All ET studies in tropical rain forest should be
interpreted with respect to cumulative annual net radiation, because latent heat loss (i.e.,
evaporation) should approximately equal the net input of radiant energy over long periods
in warm, wet climates (Priestly 1966). This theory has been confirmed in several studies:
89% of radiant energy was dissipated as latent heat over 3 yr in tropical rain forest in
central Brazil (Shuttleworth 1988); the comparable figure was 96% over 1 yr in a tropical
rain forest on Java (Calder et al. 1986). Cumulative annual net radiation in the present
study was 3.5 GJ/m2, which is equivalent to 1440 mm of evaporated water-quite close to
the 1451 mm estimated from Cedrela stands. In contrast, Luvall (1984) measured 4.1 GJ
m-2 y-', which is equivalent to 1640 mm of evaporated water. In light of this, the 2100 mm
of ET estimated by Luvall seems improbably high; this may in part be due to the canopy
conductance parameter he used in the Penman-Monteith equation, which at 1 mol m-2 s-1 is
higher than most published values for tropical forests (Table 2-1).
Results from this study indicate that there are substantial differences in ET among
broad-leaved tree species, mediated in large part by leaf boundary layer effects. These
differences are large enough to influence drainage rates and watershed yields: given the
3192 mm of rainfall over the year of this study, the increase in ET simulated from Cedrela
relative to Hyeronima would be enough to cause a 16% decrease in water yields and deep
drainage. Such differences are large enough to warrant notice from the standpoint of
watershed management. Observations on phenology, physiology, and plant structure can
be successfully scaled up to explore the effects of species on ecosystem processes, and
these approaches will become increasingly important as humans continue to cause large-
scale transformations of landscapes.
Table 2-1. Average leaf dimensions and one-sided leaf boundary-layer conductances for
three tropical tree species. Values are means (standard deviation; n)
Units Hyeronima Cedrela Cordia
leaf width mm 124 (14; 20) 44 (4; 20) 67 (8; 20)
leaf length mm 197 (25; 20) 132 (20; 20) 150 (17; 20)
boundary-layer mol m2 s 0.083 (0.029; 4) 0.478 (0.249; 6) no data
Table 2-2. Aerodynamic and canopy conductances estimated for several tropical forests.
Study Location Aerodynamic Canopy
mol m2 s1 mol m2 s-1
Calder et al. 1986 Java 4 10 0.342
Shuttleworth 1988 Amazonia not reported 0.32'
Grace et al. 1995 Amazonia 0 4 0.4- 1.04
Granier et al. 1996 French Guiana not reported 0.32'
Luvall 1984 Costa Rica 8.2 1.01
2mean over three weeks
3mean daily maximum
range of daily maxima
Table 2-3. Leaf area indexes and gap fractions for three-year-old (in June 1994) stands of
three tropical tree species. May is the onset of the wet season, and January is the onset of
the dry season. Values were determined with an LAI-2000 plant canopy analyzer. Values
are means and standard deviations from three plots; nine readings were taken in each plot,
each reading consisting of one above-canopy and four below-canopy readings.
Hyeronima Cedrela Cordia
6/94 1/95 6/94 1/95 6/94 1/95
leaf area 4.4 4.5 2.4 0.9 2.1 1.3
index (L) (0.3) (0.3) (0.2) (0.6) (0.2) (0.1)
free throughfall 0.03 0.02 0.21 0.62 0.22 0.44
fraction (p,) (0.02) (0.01) (0.06) (0.10) (0.05) (0.10)
canopy gap fraction 0.03 0.02 0.14 0.53 0.17 0.38
(PI) (0.01) (0.01) (0.03) (0.15) (0.02) (0.07)
Table 2-4. Components of modeled annual evaporation and transpiration from stands of
three tropical trees. Values are mm/yr.
Component Hyeronima Cedrela Cordia
understory transpiration (E,) 26 235 217
tree canopy transpiration (EC) 648 754 711
evaporation of intercepted rain (E.) 495 462 470
total 1169 1451 1398
DJ F MAMJ J A
1994 1995 MONTH
Figure 2-1. Climate data for La Selva Biological Station, Costa Rica, December 1994 to
November 1995: cumulative monthly net radiation (Rn), average daily vapor pressure
deficit (D; narrow vertical lines are standard deviations), and monthly rainfall.
J f B H H
J .. .. .. .
H J J f .....
a I J J r
,-. 0. 5
< 0 Hyeronima
oo 0.0 1 1
8-10 10-12 12-14 14-16 16-18
TIME OF DAY (hours)
Figure 2-2. Daily courses of stomatal conductances in 5-yr-old plantations of tropical
trees. Measurements were made throughout the canopy of 5-yr-old trees, over 4 to 5 d.
Vertical lines are standard deviations.
12-14 14-16 16-18
Figure 2-3. Daily courses of canopy conductance of three monospecific stands of tropical
trees estimated from stomatal conductance, boundary layer conductance, and leaf area
A ........ A...... & .. ..
TIME OF DAY (hours)
40 Eu + Ec + E
40-Eu + Ec .,.,,,
'E l 0 . . . . . . .. . . . . . .. . . .o. O O ~ .... . . . . . .. .
I --oCedrela odorata< 05-I/ .
4 0 -
U), 20 -
< 10 V
C- 50 Cordia alliodora
A0 V I.t
10 20 30 40 50
Figure 2-4. Simulated weekly ET from December 1994 to November 1995 in stands of
three tropical trees. ET is partitioned into transpiration from understory and canopy, and
evaporation of intercepted rain.
NUTRIENT CONCENTRATIONS IN SOIL WATER IN RESPONSE TO CUTTING
FREQUENCY OF TROPICAL TREE PLANTATIONS
Plants contribute to retention of mineral nutrients within ecosystems by
sequestering them in their tissues, fostering microbial growth and cation exchange capacity
in the soil through addition of organic matter, and decreasing soil water percolation by
augmenting evapotranspiration (Vitousek and Melillo 1979). Disruption to plant cover
affects these processes and can result in leaching of nutrients below the rooting zone,
where they are unavailable for plant uptake and may produce deleterious effects such as
contamination of drinking water and eutrophication of rivers and lakes (Bormann and
Given that loss of plant cover can result in increased leaching, it follows that
systems in which disturbance is frequent are likely to have high rates of nutrient leaching
from the soil. For example, in short-rotation forest plantations, a high proportion of time is
spent in site-preparation and the period before full site-occupancy, thus exposing the
system to nutrient losses (McColl and Powers 1984). Nevertheless, few data on leaching
losses in short-rotation forestry have been reported (reviewed by Raison and Crane 1986
and Heilman 1992). This issue is important given that plantation forestry is becoming an
increasingly important source of fiber. There are significant areas of short-rotation forest
plantations in tropical regions (Brewbaker 1985), including the Pacific islands (Phillips et
al. 1975), South Asia (Toky and Singh 1993), and South America (Spangenburg et at.
I designed an experiment for testing the effects of disturbance frequency on
leaching losses that eliminated the potentially confounding factors of soil type and species.
The approach was to impose different cutting-cycle lengths on single tree species: short
rotations of 1 yr, intermediate rotations of 4 yr, and an uncut control. I asked whether
shorter rotation times induce elevated nutrient concentrations in soil water at a 1 m depth
in the soil of four macronutrients: nitrate (N03*), calcium (Ca+2), magnesium (Mg+2'), and
potassium (K'). All experimental treatments were carried out on the same soil type. To
improve generality of results, the rotation treatments were applied to three fast-growing
tree species which differed substantially with respect to rooting pattern, leafing phenology,
crown architecture, and litter quality (Haggar and Ewel 1995, Menalled et al. in press).
The study took place at La Selva Biological Station in the Atlantic lowlands
(elevation 40 m) of Costa Rica. Mean annual rainfall and temperature are 3962 mm and
25.8 C, respectively (Sanford et al. 1994). There is a short dry season in February-April,
although mean rainfall exceeds 100 mm in all months.
The experimental site is a flat, recently formed alluvial terrace with deep, loamy
soil classified as a Eutric Hapludand (Weitz et al. 1997). Due to the volcanic origin of the
parent material, the soils have very low bulk density (about 0.7 g/cm3 in the upper 10 cm)
and are porous, permitting rapid drainage. This, combined with flat topography and
absence of micro-relief, contributes to the virtual absence of overland flow even in the face
of high rainfall. The A horizons, which extend to -0.2 m, are a highly aggregated clay
loam with many fine pores. The B horizons, extending to -0.6 m, are a sandy loam, also
with many fine pores. There is a probable buried A horizon at --0.7 m, followed by a C1
horizon of coarse sand with few fine pores (A. F. Bouwman, National Institute of Public
Health and Environmental Protection, The Netherlands, personal communication.)
The soil cation exchange complex is dominated by Ca2; extractable values in the
upper 0.1 m in 1993 (means and standard deviations of all treatments averaged together;
3 species 3 cutting cycles 3 blocks = 27 plots) were 15.8 cmolfkg (standard
deviation = 6.1), compared to 1.7 (0.3) cmoljkg K' and 3.1 (0.7) cmolkg Mg*2 (Haggar
and Ewel 1994). There was little exchangeable hydrogen or aluminum, thus base
saturation was effectively 100%. Effective cation exchange capacity was 10-25 cmoljkg.
Nitrogen retained on the exchange complexes was negligible compared to the base-
forming cations: 0.040 (0.008) cmoldg of ammonium, and 0.058 (0.010) cmol/kg of
The three tree species planted, all native to Central America, are strikingly
different in appearance and phenology. Hyeronima alchorneoides (Euphorbiaceae) is a
massive canopy emergent with evergreen leaves. Once complete canopy cover was
established, approximately 1.5 yr after planting, very few plants successfully colonized in
its understory. Cordia alliodora (Boraginaceae), a slender-crowned canopy tree, becomes
leafless for several months at the beginning of the wet season. In contrast to Hyeronima
stands, the herb layer grew well under Cordia, and was cut back approximately every 6
wk. The third species, Cedrela odorata (Meliaceae), is a canopy tree that becomes leafless
for several months towards the beginning of the drier season. (Henceforth, generic names
will be used.) Like Cordia, its crowns were deep and sparse, with a dense herbaceous
layer growing beneath them.
Root morphology also differs among the three species, although all of them have
tap roots. Hyeronima produces a dense, compact ball of roots, whereas Cordia produces
long roots near the surface that sometimes extend far beyond the edge of the crown.
Cedrela has fairly low rooting density, and lacks the long near-surface roots of Cordia.
In April 1991 the site was cleared of an abandoned cocoa plantation, merchantable
overstory trees (mostly Cordia alliodora) were harvested, and the slash was burned.
Immediately following manual clearing of charred logs, tree plantations were established
(2200 stems/ha) in three blocks. Each block comprised plots of the three tree species, each
of which contained smaller plots with the three cutting cycle treatments. The smallest
plots, 10 by 40 m, were used for annually cut stands; 20 by 40 m and 30 by 40 m plots
were used for the 4-yr cutting-cycle and uncut stands, respectively, because trees in these
would grow taller and become more sparse (through thinning and natural mortality) than
the annually cut stands. No biomass was removed after cutting because the objective of
the experiment was to determine the effect of cutting frequency on nutrient retention;
removal of biomass would have introduced a confounding effect of differential export of
nutrients in above-ground biomass. Sites were immediately replanted with the same
species at the same density. Details on seed sources, nursery treatments, and out-planting
are in Haggar and Ewel (1995).
To examine the distribution of roots, fine (< 2 mm diameter) root length density
was measured on soil cores (48-mm diameter) taken in 0.1 m increments. Roots were
sampled to a depth of 2 m in one block, and to 1 m in the other two blocks. Cores were
taken from eight randomly selected locations in each plot, then the eight cores from each
0.1 m depth increment were composited.
Soil water was collected from porous cup samplers starting 4 June 1992, 11 mo
after establishment of the plantations. For the first 3 yr of the experiment, samples were
collected roughly every 3 mo; samples were collected more frequently. Tension (-16 kPa)
was placed on the samplers after more than 20 mm of rain had fallen within 3 d; soil water
was collected 24 hr later. Each sampler consisted of a 48-mm diameter by 60-mm length,
high-flow, round-bottom ceramic cup (Soil Moisture Equipment Corp, Santa Barbara,
CA) glued to a 48-mm diameter by 1.1 5-m length PVC pipe that was sealed on the other
end with a rubber cork. The cork was perforated by a semirigid 7.7 mm diameter
polyethylene tube whose tip was inserted inside a soft neoprene tube; by clamping the tube
the entire assemblage could be sealed to hold a vacuum. The unit was inserted vertically to
a depth of 1 m. To collect samples, a -1.3 m length, flexible Teflon tube was passed
through the tube that penetrated the cork until it reached the cup at the bottom of the
sampler; a hand-pump was used to extract water. To inhibit N transformations, 0.05 ml of
chloroform was added to the samples then they were refrigerated.
NO,- was analyzed on-site with an automated analyzer (Technicon 1973b); some
Ca+E, Mg2, and K' analyses were done at Centro de Investigaciones Agron6micas (CIA)
in Costa Rica (using atomic absorption spectroscopy; Helmke and Sparks 1996, Suarez
1996), whereas others were done at University of Florida's Institute for Food and
Agricultural Sciences Analytic Research Lab using inductively coupled plasma
spectroscopy (Soltanpour et al 1996). Analytical problems with Ca'2, evidenced by
extremely low (< 2 mg/L) concentrations in a large number of samples submitted to CIA,
made it possible to utilize only the Ca+2 measurements made after the cutting of the 4-yr-
old trees. Measurements of electrical conductivity, pH, sodium, and chloride were made
on 200 (25% of the total) samples. Nutrient concentrations in soil water before and after
cutting of the 4-yr-old stands were summarized with time-weighted means: this was done
by measuring the area under the curve of nutrient concentration plotted against time, then
dividing by the time elapsed.
The data were analyzed as a complete factorial design with time as a repeated-
measure using Proc Mixed (SAS 1997). Species, cutting-cycle treatment, time, and their
interactions were treated as fixed effects; block and its interactions with species and time
were treated as random. Compound symmetry covariance structure was used, which
assumes that variance is constant over time. Model adequacy was tested by inspection of
residuals for obvious patterns. Concentrations before and after the cutting of the 4-yr-old
trees in mid-1995 were analyzed separately.
Soil water beneath stands subjected to an annual cutting cycle had persistently
elevated concentrations of NO3- and K' in comparison to that beneath uncut stands (Fig.
3-1a,b). The effect was most pronounced for NO3. There was substantial seasonal
variation in NO3" concentration in the annual rotation (Fig. 3-1 a), which was also in
synchrony with the cutting and replanting. Cutting cycle was a statistically significant
source of variance for NO3 and K' (Table 3-1). In contrast to NO3, there were inter-
annual fluctuations in K' concentrations that occurred in both cut and uncut stands (Fig.
3-1b), which caused time to be a significant source of variance (Table 3-1). Time was also
a significant source of variance for Mg2, but concentrations of this nutrient did not
increase under the annual cutting cycle (Fig. 3-1 c).
Cutting the 4-yr-old stands caused a strong response in soil water N03 (Fig. 3-1
a). For all three species, N03" concentrations peaked within 7 mo after cutting then
dropped back close to concentrations in uncut stands after 8 or 9 mo. The three species
showed distinct patterns: NO3 concentration under Hyeronima rose rapidly to a sharp
peak within 3 mo of cutting then declined gradually; under Cordia stands, concentrations
rose gradually, reached a peak roughly 6 months after cutting, then declined rapidly; and
under Cedrela stands, there were had several peaks. There was a statistically significant
effect of cutting cycle on both NO3" and K4, but for the latter element it was due to
persistently elevated concentrations in the annual rotations rather than an increase after
cutting the 4-yr-old trees. Concentrations of all nutrients showed significant trends over
time in the post-cutting period.
As expected, the top 1 m of soil had the greatest fine root density for all three
species (Fig. 3-2); Hyeronima had 85% of fine root length from 0-2 m soil depth
concentrated in the top 1 m, compared to 60% for Cedrela and 65% for Cordia. The tree
species differed in absolute values of root length density as well as distribution;
Hyeronima had approximately 1.5 times as much root length as Cordia, and 2.3 times that
Despite their differing phenologies, root architectures, and root length densities,
the three tree species did not differ significantly in their effects on the concentrations of
any of the nutrients measured (Table 3-1). The significant interaction for NO3- between
time, species, and cutting cycle after cutting the 4-yr-old trees (Table 3-2) probably
occurred because the three species were cut approximately 1 mo apart; Hyeronima, which
was cut first, showed the earliest peak, followed by Cedrela, which was cut second, and
then Cordia, which was cut last.
NO3 leaching has been implicated in the leaching of cations because nitrification
produces excess hydrogen ions, which displace cations from exchange sites; the N03" then
acts as a balancing anion, allowing the cations to leach. To explore this, samples for which
sodium and chloride (in addition to the other ions) concentrations were known were
ranked in order of increasing NO3" concentration (Fig. 3-3). On average, increases in
negative charge due to higher NO3 concentration were balanced by increases in positive
charge from cations, providing support for the hypothesis that an increase in NO3 (and
nitrification rate) enhances cation leaching. The sum of negative charge was consistently
less than the sum of positive charge, indicating that an anion that contributed substantially
to the negative charge, possibly bicarbonate or sulfate, went unmeasured.
Leachate concentrations after cutting tropical forests
Peak N03 concentrations were comparable to those obtained after cutting of
mature tropical forest in other locations (Poels 1987, Bruijnzeel 1990). The highest single
concentration observed was 13 mg/L; in comparison, Parker (1985), who conducted his
research near La Selva Biological Station, and Brouwer (1996) both observed peak N03"
concentrations after cutting of 11.3 mg/L. For all three species in the present study,
though, such elevations in concentration remained for only a short time, which presumably
limited post-harvest losses.
Nitrate had dropped to negligible levels by 1.5 yr after initial establishment of the
plots in 1991; other studies of tropical forests indicate similar lengths of time needed to
attain pre-harvest levels (Brouwer 1996). Potassium and Mg2 concentrations were similar
to those from other tropical soils (Russell and Ewel 1985, Brouwer 1996, Poels 1987),
but Ca2 concentrations were high: average Ca2 concentration was 9 mg/L, in comparison
to a maximum of 3 mg/L reported in the other studies.
Cutting frequency and soil water nutrient concentrations
The annual cutting cycle greatly increased the concentrations of NO3- and K'. Such
elevations in N03' are due both to cessation of uptake, and increased mineralization due to
increased temperature and moisture. Low rates of uptake are probably the main
contributing factor to high nutrient concentrations in soil water under the annual cutting
cycle. Imposing an annual cutting cycle on trees, which was done to avoid the
confounding effect of having different species for different age lengths, is a somewhat
artificial treatment, because the trees did not achieve full site occupancy and full use of
resources by 1 yr. It is probable, therefore, that elevated nutrient concentrations in soil
water under the annual rotations were as much due to low uptake rates between cutting
events as to the actual cutting.
The trees in the 4-yr-cutting cycle stands, in contrast, had reached the phase of
rapid nutrient uptake long before they were harvested (Haggar and Ewel 1995).
Consistent with this, nutrient concentrations in soil solution were low prior to cutting.
Given the simultaneous cessation of uptake when the trees were cut, and the liberation of
nutrients from plant biomass, it is perhaps surprising that the subsequent nutrient pulse
was not more prolonged. NOa concentrations under all three species peaked within 7 mo
of cutting, and by 8 to 9 mo were returning to pre-cutting levels.
The presence of large quantities of fine roots in the subsoil in the uncut plots
indicates that nutrient uptake occurred below I m, the level at which the soil water was
collected. In stands of 1-yr-old trees, in contrast, there were few roots at 1 m depth
(unpublished data), so there was little uptake below this level. This does not affect the
conclusions reached herein, because It can be assumed, therefore, that the concentrations
of nutrients at this level, which were already low, were minimal.
Studies of leaching from forests have focused on effects of cutting (e.g., Bormann
et al. 1968, Brown et al. 1973, Tamm et al. 1974, Poels 1987, Malmer and Grip 1993,
Pardo et al. 1995), fertilization (e.g., Cole and Gessel 1965), harvest intensity (e.g.,
Stevens et al. 1995, Mann et al. 1988), atmospheric deposition (e.g., Friesleben and
Rasmussen 1986, Van Grinsven et al. 1987, Stein and Van Breemen 1993), or
successional stage (e.g., Vitousek and Reiners 1975, Binkley et al. 1982). In contrast, few
studies explicitly compare effects of different tree species on soil water chemistry.
Nevertheless, these few reports indicate that species can differentially affect leaching. For
example, alder (Alnus rubra), an N-fixing tree, markedly accelerated NO3- leaching
compared to Douglas fir (Pseudotsuga menziesii; Van Miegrot and Cole 1989), and
leaching of Cae2 was much lower under mixed oak than under loblolly pine stands
(Johnson and Todd 1987).
The present study, in contrast, showed no significant differences in nutrient (NO3,
Ca+2, Mg 2, and K') concentrations in soil water among the three tree species in the uncut
stands. Differing leafing phenologies--evergreen, dry-season deciduous, and wet-season
deciduous-might have been expected to cause temporal differences in nutrient
concentrations in soil water, due to leaf drop at different times of year. The only evidence
for this occurred 44 mo after stand establishment: elevated NO3" concentrations in Cedrela
stands were detected after a single storm during the dry season. Presumably, the storm
caused a flush of mineralization from the leaf litter. This pattern was not detected, though,
in the subsequent year.
Differences in ions
The greater sensitivity of NO3- and K+ than of Ca+2 and Mg2 to the cutting cycle
treatments is consistent with the lyotropic sequence: ionic species with low charge density
tend to disappear more rapidly from the soil profile than more highly charged ions (Bohn
et al. 1985).
NOa" has frequently been found to be the ion that is most sensitive to forest
cutting, in both temperate (Vitousek and Melillo et al. 1979) and tropical systems (Poels
1987, Brouwer 1996). Thus, it was not surprising to find that the response of N03" to
cutting was much more pronounced than for the base-forming cations. Differences in
modes of supply are probably responsible because N03 comes into solution continually
through the mineralization of organic matter, and there is little possibility for storage on
the exchange complexes. The base-forming cations, in contrast, are present in large
quantities on the exchange complexes; although these ions are continually exchanging with
others in solution, there is no net exchange unless ions are being removed from the soil
solution (Bohn et al. 1985). Thus, when plant demand for nutrients ceases, as when plants
are harvested, the supplying of N3 to the soil solution continues unabated, whereas that
of the base-forming cations largely ceases, except as induced by nitrification.
Notwithstanding the primacy of NO3 in leaching, when NO3 concentrations in soil
solution were low there were still substantial amounts of base-forming cations present
(Fig. 3-3), thus being lost in drainage. Most of their charge was balanced by an
unmeasured anion, possibly bicarbonate (HCO3"). Bicarbonate is important as a balancing
anion in tropical soils with soil solution pH greater than 6.0 (Nye and Greenland 1960);
pH at the study site was typically between 6.0 and 7.0. A previous study at La Selva
Biological Station indicated that HC03" constitutes roughly 85% of the negative charge in
the A horizon, whereas sulfate and NO3 each contributed about 7% (Johnson et al. 1977).
Therefore, the role of NO3 production in causing cation leaching is uncertain; elevated
NO3 in response to the cutting cycle treatments is probably more consequential in terms
of its loss from the site than its effects on cation leaching.
Very significant increases in N03- and less significant increases in K' were
observed in the soil solution at 1 m depth after clear-cutting all three tree species.
Concentrations were highest after cutting the 4 yr rotations, although the sustained
elevated NO3"concentrations under the annual cutting cycle treatment indicate that
cumulative losses over 4 yr were highest under the annual cutting cycle. Thus rotation age,
or disturbance frequency, played an important role in determining nutrient retention,
whereas species did not.
Table 3-1. Results (probability values) of an analysis of variance of nutrient concentrations
in annual rotations and uncut stands from 1992 to 1995 (i.e., until cutting of the 4-yr-old
stands). A mixed statistical model containing random and fixed effects and incorporating
time as a repeated measure was used.
Variance Source NDF* DDF" NO3" K Mg+2
species 2 4 0.89 0.22 0.54
cutting cycle 1 96 / 75" 0.0001 0.003 0.40
species cutting cycle 2 96 /75 0.61 0.19 0.53
time 9 / 7 18 / 14 0.84 0.0004 0.0001
time species 18/14 96/75 0.11 0.32 0.36
time cutting cycle 9/7 96/75 0.79 0.79 0.18
time species cutting cycle 18 / 14 96/75 0.14 0.90 0.98
"numerator degrees of freedom
'denominator degrees of freedom
t number to left of solidus refers to N03", number on right, t
differ because NO3 was sampled on several additional dates.
o K' and Mg2. Numbers
Table 3-2. Results (probability values) of an analysis of variance of nutrient concentrations in annual and 4-yr rotations and
uncut stands for 7 mo (June 1995 February 1996) after cutting of the 4-yr and annual stands. A mixed statistical model
containing random and fixed effects and incorporating time as a repeated measure was used.
Variance source NDFt DDFt N03" Ca+2 K+ Mg2
species 2 4 0.89 0.54 0.38 0.59
cutting cycle 2 254 / 275t' 0.0001 0.16 0.002 0.15
species cutting cycle 4 254 / 275 0.74 0.31 0.39 0.97
time 16/17 32 /34 0.0053 0.002 0.0004 0.0001
time species 32/34 254/275 0.0001 0.19 0.55 0.19
time cutting cycle 32 / 34 254 / 275 0.0001 0.20 0.11 0.11
time species cutting cycle 59/63 254 / 275 0.0003 0.86 0.98 0.98
Tnumerator degrees of freedom
denominator degrees of freedom
-tnumbers to left of solidus refer to N03-, numbers to right refer to Ca+2, K, and Mg2. Numbers differ because there was an
additional sampling date for the cations.
Figure 3-1. A-D. Mineral nutrients in soil water draining from tropical tree stands
subjected to three cutting cycles: annual, every 4 yr, and uncut. Each datum was based on
a composite water sample collected from three to four porous cup samplers. Samples from
the 4-yr-cutting-cycle and uncut plots are averaged together for the first 4 yr (i.e., while
these treatments were identical), until the cutting of the 4-yr-old stands in 1995. Thin
arrows indicate approximate dates of cutting of annually cut stands; the thick arrow
indicates approximate date when 4-yr-old stands were cut in addition. The inserted graphs
show data from the 7 mo after cutting on an expanded axis. A. NOa" B. K+ C. Mg 2 D.
30 36 42
MONTHS SINCE ESTABLISHMENT
6- 00 8 4 -
E 7- 3
6 2 "
5 -1 I I I I I I
U) 4 "[ 50 51 52 53 4 .. 56 57 =
2 @ .. . ... .....
-. annual rotation A A CO
8 -- 4yr rotation 4 "
7 *--- uncut 3
0 I ..i i i i i I I I I I I ... I I 1.1 1 1 I
12 18 24 30 36 42 48
MONTHS SINCE ESTABLISHMENT
3- 50 51 52 53 54 55 56 57
D 4 -
U) 3 50 51 52 53 54 55 56 571
< I .
/ A A Cordia
6 --.--annual rotation 3 Z:Tr _--T- ,
5 ---- 4 yr rotation 2 -
4.. .... uncut 1 -1 ;
12 18 24 30 36 42 48 54
MONTHS SINCE ESTABLISHMENT
- annual rotation
- 4 yr rotation
... uncut T
50 51 52 53 54 55 56
MONTHS SINCE ESTABLISHMENT
0-1 m mean = 0.093 cm/cm3 (85%)
.... ... ... ] ... ... .... ... ... .......... . . . . . . .
1-2 m mean = 0.017 cm/cm3 (15%)
-0-1 m mean = 0.029 cm/cm3 (60%)
54,0-1 m mean = 0.040 cm/cm3 (53%)
........ ...................................... . . .........................
1-2 m mean = 0.035 cm/cm3 (47%)
I I !
ROOT LENGTH DENSITY (cm/cm3)
Figure 3-2. Fine (5 2 mm) root length densities of three tropical tree species at 4 yr of age.
Means and standard deviations from three blocks are reported for the uppermost 1.1 m of
soil; values for the lower 0.9 m came from one block and therefore do not have standard
deviations. Each datum was derived from a composite sample of eight cores.
1-2 m mean = 0.019 cm/cm3 (40%)
Mg+2 + K+ + Ca+2 + Na+
-Mg+2 + K+ + Ca+2
2.0 Mg+2 + K+
... .:A- .
-0.5 V_. NO3" -, ..
NO3- + CI"
-1.0 1 1 1
10 20 30 40 50 60 70 80
RANK ORDER (by NO3- concentration)
Figure 3-3. Electrical charge of ionic species in soil water samples collected at 1 m depth
from experimental plots of tropical trees at La Selva Biological Station, Costa Rica.
Eighty-one samples are shown, ordered from lowest to highest nitrate concentration.
NUTRIENT LEACHING IN SHORT-ROTATION TROPICAL FOREST STANDS
In moist climates, harvesting of crops generally induces an increase in rates of
leaching of mineral nutrients from upper into lower soil horizons. As a result, soil fertility
is diminished and the environment may be degraded (e.g., when nitrates leach into
waterways). Management of ecosystems, therefore, should aim to minimize leaching
losses, and one way to do this may be to use crops that do not require frequent harvesting.
But estimates of the magnitude of leaching losses are needed to know whether it is
worthwhile to sacrifice other objectives (such as production of particular crops) to achieve
the goal of minimizing leaching.
Reviews of nutrient dynamics of forest operations in the temperate zone have
suggested that post-harvest hydrological losses are generally small, particularly when
amortized over the length of the entire rotation (Stone 1985, Mann et al. 1988).
Nevertheless, the studies reviewed involved long-rotation stands; there is little information
available on hydrologic losses of nutrients in short-rotation forestry (McColl and Powers
1984, Heilman 1992). Post-harvest nutrient losses that might be considered negligible
when averaged over a 50-yr rotation would assume much more importance if they
occurred over a 10-yr rotation. Furthermore, findings from temperate forests may not be
readily extrapolated to the tropics, where high rates of rainfall and decomposition create
the potential for high rates of leaching (Sanchez 1976).
Soils in lowland tropical forest regions may be particularly susceptible to
debilitating leaching loss of nutrients in response to short rotation times; after tropical
forest is cleared and planted to annual crops, there is often crop failure within a very short
time (Nye and Greenland 1960, Sanchez et al. 1982, Dagar et al. 1995, Lal 1995, Juo et
al. 1995). This observation, among others, has given rise to the recommendation that trees
be incorporated into tropical cropping systems (e.g., Muller-Samann 1994, Woomer and
Swift 1994): trees conserve nutrients within ecosystems by adding organic matter to the
soil, diminishing erosion through mitigation of raindrop impact (Calder et al. 1993),
cycling nutrients from the subsoil via deep roots (Burnham 1989), and continually taking
up nutrients. Thus, some of the nutrient-conservation services that trees provide are
inextricably linked to their perennial habit.
These questions are important given the increased prevalence of short-rotation
cropping in the humid tropics. Forest plantations cover approximately one per cent of the
tropical land mass capable of supporting forest cover (Hammond 1994), and many
hundreds of thousands of hectares of these are managed on short rotations (e.g., Lamb
1973, Brewbaker 1985, Ogawa and Matsuhira 1996, Phillips et al. 1995, Singh and Toky
1995). Furthermore, there is a trend towards supplanting traditional cultivars of perennial
crops such as coffee with newer ones that are higher-yielding and shorter-lived, effectively
shortening the cropping cycle and increasing the frequency of soil disturbance. The
relationship between cropping cycle duration and leaching losses must be understood and
quantified if sustainable cropping systems are to be designed.
To address the relationship between cutting frequency and leaching losses, I
measured leaching after manipulating the cutting cycle in stands of three tropical tree
species in the Atlantic lowlands of Central America. Two rotation frequencies, 1 yr and 4
yr, were employed, and uncut stands were used as controls. To isolate as much as possible
the effect of cutting-cycle treatment on nutrient dynamics, slash was left in situ; this
prevented differential export of nutrients in plant biomass from becoming a confounding
factor among the treatments. Leaching losses were measured over a 4 yr period extending
until 7 mo after the first harvest of the 4-yr rotation.
The experiment took place at La Selva Biological Station in the Atlantic lowlands
of Costa Rica (100 20 N, 83'50 W, elevation 40 m). In 1991, monospecific stands of
three tree species (Hyeronima alchorneoides, Cedrela odorata, and Cordia alliodora)
were established on the site of an abandoned cocoa plantation. The vegetation was cut,
trunks of large trees (mostly Cordia alliodora) were removed, and the remaining slash
was burned. Trees were planted at a density of 2200 stems/ha. They were set out in three
blocks, each of which contained large plots of the three tree species that were further
divided according to cutting cycle: plots with trees on the annual cutting cycle measured
10 by 40 m; those on the 4-yr cutting cycle, 20 by 40 m; and the uncut stands, 30 by 40 m
(Haggar and Ewel 1995). Annual rotation stands were cut and replanted each year; the 4-
yr-rotation stands were cut and replanted in mid-1995.
The soils at the study site have been variously classified as Eutric Hapludands (i.e.,
andesitic soils of humid climates with minimal horizonal development and high base status;
Weitz et al. 1997) and Typic Dystrandepts (i.e., typical andesitic inceptisols with low base
status; Haggar and Ewel 1995). The differing assignments to soil order occurred because
Andisols (soils formed from volcanic ejecta and dominated by allophane or aluminum-
humic complexes) have only recently been recognized as a soil order (Soil Survey Staff
1990); previously they were classified as Andepts. The the conflict in classification
probably arose due to an early analytic error in calcium (Ca2) determination, which led to
erroneously low estimates of base status.
The difference between pH measured in KCI and in water, ApH, yields information
about the charge status of soils, and hence their ability to retain mineral nutrients (Sanchez
1976). A strongly negative ApH such as occurs in the soils at the study site (Table 4-1)
indicates that negative charge predominates (Uehara and Gillman 1981); therefore, they
have little ability to retain anions such as nitrate (NO3). A ApH less than -0.5 indicates
that the charge system is not predominantly made up of variable-charge clays but instead
consists of permanent charge; the ApH values at the study site, which were mainly less
than -1, are surprising given that Andisols are known for having a high amount of pH-
dependent charge because of high allophane and organic matter (Sanchez 1976). The lack
of dominance of pH-dependent charge may mean that the ability of the soil to retain
cations is relatively insensitive to changes in pH.
Assessment of leaching
Solute leaching was determined by coupling leachate nutrient concentrations to
estimated drainage volumes using the trapezoid method (Lord and Shepherd 1993). In this
method, nutrient concentrations in drainage water are averaged between two successive
sampling dates then multiplied by the volume of water calculated to have drained between
the dates. Leachate nutrients were measured at intervals ranging from 1 wk to 3 mo.
(Samples were taken most frequently immediately after harvesting of the 4-yr rotations, at
intervals of 1 wk.)
Porous ceramic cup samplers (Soil Moisture Equipment Corp., Goleta, California,
USA) were inserted vertically so that the top of the cup was 1 m below the soil surface.
This depth was chosen as being below most roots (Chapter 3), yet above the vadose zone.
Tension of-16 kPa was applied with a hand pump after rainfall of at least 20 mm occurred
within 3 d; water was collected 24 h after tension was applied. The suction of-16 kPa was
chosen because tensiometer measurements indicated that saturated soil drained to
approximately this tension at 1 m within 24 h (J. Haggar and A. Nobre, unpublished data).
The tension applied therefore probably extracted predominantly freely draining water,
rather than water that was tightly held by the soil. Cups measured 48 mm in diameter by
60 mm in length. Further details of ceramic cup sampler placement are provided in
There were three samplers in each of the nine (three of each species) annual-
rotation stands, and two samplers in each 4-yr-rotation and uncut stand until rnid-1995,
when the 4-yr rotation stands were cut. At that time, an additional sampler was installed in
each 4-yr-rotation and uncut stand. Samples from the two to three samplers in each plot
were combined prior to chemical analysis.
Concerns have been raised (Schaffer et al. 1979, Tyler and Thomas 1981)
regarding the inability of porous cup samplers to sample preferential flow, i.e., rapid flow
through a system of large pores that bypasses the soil matrix (Beven and Germann 1982).
Although preferential flow is an important phenomenon in tropical soils that can affect
leaching estimates (Russell and Ewel 1985, Sollins and Radulovich 1988), at the study site
there are few large pores that extend deep into the soil profile (Lex Boumann personal
communication), thus the likelihood of large volumes of water being channeled from the
surface past the 1 m sampling depth without interacting with the soil matrix is small.
Analysis of solutions for N03" was done with an automated analyzer at La Selva
Biological Station; analyses for Ca2, Mg2, and K+ were done at the University of Costa
Rica's Centro de Investigaciones Agron6micas (CIA) using atomic absorption
spectrophotometry (Helmke and Sparks 1996, Suarez 1996), and at the University of
Florida Institute for Food and Agricultural Sciences's Analytical Research Laboratory
(ARL) using inductively coupled plasma-mass spectroscopy (Soltanpour et al. 1996).
Calcium determinations made by CIA from June 1993 to November 1995 were much
lower than values obtained on similar samples by ARL and at the USDA Forest Service's
International Institute of Tropical Forestry in Rio Piedras, Puerto Rico, using plasma-mass
spectroscopy and therefore are not reported here.
Statistical analysis of cumulative leaching losses was done by an analysis of
variance in which measurements made on the same plots over time were treated as
repeated measures. Species and cutting-cycle treatments were treated as fixed factors, and
block was treated as a random factor. Analysis was done used Proc Mixed in SAS (1997).
Drainage rates were calculated using the model SWAP 1993 (van den Broek et al.
1994). The model incorporates Richards's (1931) equation for unsaturated flow to
calculate water movement through a soil profile. The soil profile was modeled in four
layers (0-0.1, 0.1-0.3, 0.3-0.5, and 0.5-1.0 m) which roughly corresponded to soil
horizons, van Genuchten parameters (van Genuchten and Nielsen 1985) were used to
describe soil hydraulic properties. These parameters (saturated and residual volumetric soil
moisture, saturated hydraulic conductivity, and four fitting parameters) were obtained by
Antje Weitz (International Institute for Tropical Forestry) using the suction crust method
(Booltink et al. 1991) at a secondary forest site located on the same soil mapping unit
about 100 m distant from the study plots.
Precipitation was measured daily by manual rain gauge. Potential
evapotranspiration (ETP) was estimated from a regression of cumulative daily net radiation
against the difference between maximum and minimum daily temperature (AT). This was
necessary because net radiation data were only available for 1 yr of the study. A net
radiometer (Q7; Radiation and Energy Balance Systems, Inc., Seattle, Washington, USA)
was mounted 2 m above a green, freely transpiring plant canopy. Temperature was
measured with a shielded, aspirated thermocouple mounted alongside the net radiometer;
both instruments were recorded by a data logger (CR10; Campbell Scientific Instruments,
Inc., Logan, Utah USA). Data were available for 154 d. Because cumulative daily net
radiation appeared to reach a plateau after AT exceeded 11 C, a spline curve with one
knot (i.e., one break-point between two lines) was fitted to the data. Proc Reg (SAS
199 1) was used to estimate approximate slopes for the two lines using a knot that was
estimated visually. The resulting values were used as starting points for an iterative
estimation using Proc Nlin, which provided precise estimates of slopes and knot.
Temperature values over the 4-yr study were obtained from several different
sources. From 1 March 1992 to 24 January 1993 were obtained from a thermistor (Vaisala
50Y temperature probe) connected to a CR10 data logger at a meteorological station
operated by the Organization for Tropical Studies approximately 1 km from the study site.
Temperature values from 25 January 1993 to 30 November 1994 were obtained from a
recording thermograph (Belfort Instrument Co., Baltimore, Maryland, USA) located at
1.5 m above the ground in the center of the study site. From 1 December 1994 to 15
November 1995, measurements were obtained mainly from an aspirated fine-wire
thermocouple mounted on a tower above the tree canopy at the study site, but several
lapses in the data record due to equipment failure and moving the tower between stands
were filled with data from a thermistor (Vaisala SOY temperature probe) approximately 2
m above the ground at a meteorological station within 200 m of the study site. Values
from the latter station were also used from 16 November 1995 to 29 January 1996.
Estimated Rn was converted to ETP by dividing by the latent heat of vaporization,
which was justified by the observation that, in wet tropical forests, the energy required to
fuel estimated annual ET is very close to the measured annual input of net radiation
(Calder et al. 1986, Shuttleworth 1988). A crop factor for each of the three tree species
was determined by regression of ET estimated from the Penman-Monteith equation
(Chapter 2) during dry periods against ETP. Crop factors for Hyeronima, Cedrela, and
Cordia were 0.556, 0.704, and 0.673, respectively. Interception was calculated using an
algorithm embedded in SWAP1993 (Feddes et al. 1978), in which interception
asymptotically approaches 2 mni/d as rainfall increases.
Potential ET was assumed to be identical in the annual rotations and the uncut
rotations. This assumption was bolstered by paired measurements of net radiation over
green canopies of tall trees and annual rotation stands, which showed that net radiation
between them was identical (Seth Bigelow unpublished data). It is likely, however, that
ET was less in the annual stands that in the uncut ones, and therefore drainage from
annual stands was underestimated to an unknown but probably minor extent.
After the 4-yr rotation was cut, leaf area plummeted, and the energy balance was
altered due to the slash on the ground. The reduction in R was estimated for each species
by placing one net radiometer over the slash and another over adjacent, green vegetation
and measuring at 30 s intervals, averaged over 0.5 h periods, for approximately 1 wk.
Potential ET after cutting the 4-yr rotation stands was corrected by multiplying by the
fractional reduction in R, over slash with respect to green vegetation. The length of time
that the ground remained covered by slash was estimated by periodically making 100
random vertical probes in each plot, and noting whether the probe first touched green leaf;
slash, or bare ground. The fractional reduction in Rn was no longer applied after leaf cover
Soil moisture measurements taken during 1995 were used to evaluate the water
balance model. Measurements were made using time domain reflectometry (Trase 11, Soil
Moisture Equipment Corp., Santa Barbara, CA). In each plot, pairs of 1 m long, stainless
steel guide rods were inserted vertically in three randomly selected spots. Each pair of
rods sampled a cylindrical volume of soil, extending from the surface to 1 m, with a cross
sectional diameter of 0.2 m (twice the distance between the rods). Because there is doubt
as to whether the factory-supplied algorithm reliably converts the apparent dielectric
constant of the soil into soil moisture in fight-textured, low-density soils (Gray and Spies
1995) such as the ones at the study site, the physically based, three-phase mixing model of
Roth et al. (1990) was used to convert apparent dielectric constants into volumetric soil
moisture. This model has been tested on the soils at the study site by Weitz et al. (1988)
and found to perform better than a variety of other approaches. I used parameters
measured by them on the study site soils (i.e., porosity = 0.711 m3/m3, and soil geometry
factor a = 0.47) as inputs to the model.
Measurements were made at intervals of 1 d to 5.5 wk; the average period
between readings was 1 wk. Readings from the three pairs of rods in each plot were
combined to give a plot average. The resulting data were subjected to a repeated-measures
analysis of variance, using Proc Mixed in SAS (1997), incorporating species, cutting cycle,
and time effects. Data from before and after the cutting of the 4-yr rotations were
The regression of cumulative measured daily net radiation (R,) against the
difference between maximum and minimum temperature (AT) yielded an r2 of 0.67 (Fig.
4-1). The resulting equation was
if AT 10.2 then 1 = -1.626 + 1.296 AT
if AT > 10.2 then R. = 6.285 + 0.559 AT,
where AT and R have units of C and MJ m2 d1, respectively.
Potential ET averaged 3.7 mm/d over the 4 yr studied. Cutting of the 4-yr
rotations affected energy balance for two of the three species. Net radiation measured 2 m
above slash was 88% that of green vegetation for Hyeronima (r2=-0.84) and 91% for
Cedrela (r2=0.88), whereas for Cordia the figure was 98% (r2=-0.93). This reduction of
net radiation in comparison to freely transpiring vegetation in the former two species is
probably due to higher surface temperatures, due to the restriction of ET from the slash.
Higher surface temperatures would increase upward emission of long-wave radiation, thus
diminishing R.. The reduction in R, probably did not last longer than it took for the slash
to become covered with vines and other weedy vegetation. Herbaceous cover was 95% in
Cedrela and Cordia stands within 2 mo of felling. The herb layer in felled Hyeronima
stands, however, grew much more slowly: leaf cover was only 35% by 3 mo after felling,
the remaining area consisting almost entirely of slash.
Soil moisture (0-1 m) was measured on 47 d during 1995 (Fig. 4-2). Soil moisture
during the wet season (averaged among all blocks, species, and cutting cycles: n = 27) was
approximately 0.44 m3/m3, and although soil moisture dropped during the main dry season
(January-May) and the short dry season (September), average soil moisture never dropped
below 0.35 m3/m3. Differences in leaf area among the experimental treatments were
expected to influence water balance, but this was not reflected in the soil moisture
measurements. Inspection of the data did not indicate consistent differences in soil
moisture among treatments, even immediately after the trees in the 4-yr rotation were cut.
Species and cutting-cycle did not significantly affect soil moisture either before (Table 4-2)
or after (Table 4-3) the cutting of the 4-yr rotations.
The SWAP 1993 hydrology model produced estimates of soil moisture that agreed
closely with measured soil moisture (Fig. 4-2). Simulations from 1992 to 1996 indicated
that soil moisture stayed close to field capacity for most of the year, although soil moisture
deficits were incurred in the drier seasons (Fig. 4-3). During 1992, 1993, and 1995 some
drainage occurred every month, but the 1994 dry season was more pronounced, and little
or no drainage occurred from February through April. The volume of water that drained
each year varied from 1513 mm to 3540 mm. Drainage volumes under Hyeronima, (Table
4-4) which had the smallest crop factor (i.e., the greatest fractional reduction in ETP),
were greater than under the other species. Drainage was least in years in which
precipitation was lowest, mainly because of low precipitation itself but also because ET
was highest during those years.
Nitrate was the only ion that displayed a significant response to any of the
experimental treatments (Tables 4-5 & 4-6). Concentrations of N03" in soil water
collected at 1 m were elevated in all treatments after site preparation and planting in 1991,
but within 18 mo they had dropped to low levels (z0. 1 mg/L) in the uncut stands, and
differences between annually cut and uncut stands became evident (Chapter 3). In the
annually cut stands, NO3" concentration tended to vary between 1 and 3 mg/L, oscillating
with season. Cutting the 4-yr rotations caused dramatic increases in NO3 levels, which
peaked as high as 13 mg/L, but they were short-lived, returning to background levels
within 6 mo.
The elevated NO3 concentrations translated into substantial nutrient losses. NO3"
losses from 1992 through mid-1995, averaged across species, were 6.9 times greater in
the annually cut stands than in the uncut ones. After the cutting of the 4-yr rotation stands,
in mid-1995, NO3 losses in those stands increased to 17 times those in uncut stands.
Cumulative losses over the 3.5 yr reported herein, though, were more than 3 times greater
in the annually cut plots than in the 4-yr cutting-cycle plots. Tree species did not influence
leaching losses in any of the treatments, and there were no significant interactions between
species and cutting cycle.
Concentrations of K' in soil water collected at 1 m in the annual rotations were
consistently slightly elevated above those in the uncut stands (Chapter 3), but cumulative
leaching losses were not significantly different among any of the cutting-cycle treatments
or species. This was surprising because the same data, analyzed simply as concentrations
(i.e., prior to being multiplied by drainage) using a repeated-measures analysis of variance,
were found to be significantly affected by treatment.
Concentrations of Ca2 in drainage water, which were mostly >10 mg/L, were far
greater than those of any of the other solutes measured, which is consistent with the large
proportion of the cation exchange capacity occupied by Ca2. These high concentrations
led in turn to high leaching losses of over 100 kg ha' yr1 (Table 4-5). Losses of Mg 2,
which was present at much lower concentrations in drainage water, tended to be one third
to one quarter those of Ca2 on a weight basis.
Lack of differences in soil moisture among the species and cutting-cycle treatments
was surprising given the differences in ET among the species, and, in particular, the
differences in energy balance after the 4-yr treatment was cut (i.e., a decrease in R
compared to uncut stands). In a Pirrus radiata plantation, in contrast, soil moisture
remained elevated for 1.5 yr after cutting (Smethurst and Nambiar 1995). The explanation
lies partly in the large, steady volume of rainfall at the study site, which tends to keep the
soil moisture near field capacity, except during the drier season (Fig. 4-2, and Weitz et al.
1997). Additionally, small differences in moisture in surface soils are unlikely to be
detected by the sampling system used, which integrates water through the top 1 m of the
soil profile. Finally, the presence of a dense herbaceous understory undoubtedly helps to
equalize evapotranspiration among stands of differing overstory leaf area. Forest
understories can contribute significantly to ET rates (Roberts 1983, Kelliher and Black
1986), in some cases causing ET from stands with sparse overstories to equal that in
denser stands (Roberts et al. 1980).
Losses of N03- after cutting the 4-yr rotation were greater than those observed
after cutting of mature trees in many other studies. For example, in the temperate zone
studies reviewed by Mann et al. (1988) there was a maximum loss of 11 kg/ha in the first
year after clearcutting, and most sites lost less than 3.5 kg/ha. In the present study, in
contrast, the stands lost at least 30 to 40 kg/ha NOa in the first 7 mo after harvesting. The
high leaching is probably caused by the high rainfall and rapid nitrogen mineralization at
the study site. With a climate that is classified as tropical superwet (Richards 1996), the
Atlantic coast of Costa Rica is far wetter than the temperate forests where most leaching
studies have been conducted. The 2 to 3 m of rain that percolates through the solum each
year at La Selva, coupled with rapid nitrogen mineralization rates (Robertson 1984,
Vitousek and Denslow 1986, Marrs et al. 1988), creates abundant potential for leaching
when nutrient cycles are disrupted.
Average yearly loss of N03 from the annual rotations was 42.6 kg/ha, compared to
5.9 kg/ha in the uncut stands (means of the three species). This loss is well in excess of
atmospheric inputs, which are likely to be less than 5 kg ha-' y-1 (Hendry et al. 1984, Clark
1994, Eklund et al. 1997). Undoubtedly, there are additional N inputs into the system
from N-fixing algae, heterotrophic bacteria, and leguminous herbs, but these potential
sources of inputs were not examined.
I initially expected that leaching losses of NO3- after the cutting of the 4-yr
rotations would be much larger than those occurring after each cutting of the 1-yr
rotations, because of potentially rapid decomposition of the biomass of the 4-yr rotations,
in which large quantities of N were sequestered. Surprisingly, though, quantities of NO3
lost, 30-40 kg ha1 yr-, were similar in both cutting treatments; enough microbial
immobilization may have occurred in the 4-yr rotations to prevent excess leaching of No3
when the stands were cut. Similarly, despite the large variation in N03" sequestered in the
tissues of the three species (unpublished data) there was no difference in N03" leaching
among them. Long-term NO3" leaching losses, then, appear to depend mainly on cutting
The choice of species for the annual rotation may, however, be significant; by
imposing an annual rotation on a potentially long-lived tree crop, the stand was effectively
maintained continually in the establishment phase, when losses tend to be highest because
of partial canopy development and modest rates of uptake. Thus, some of the nutrients
that leached might, under other circumstances, have been incorporated into the tissue of
the crop. The rapid growth of the herbaceous layer, which was cut approximately six times
per year, undoubtedly helped to staunch the flow of NO3" leaching.
Diminished cation uptake due to cutting of stands should cause increased leaching
(Johnson and Todd 1990), but such losses are generally modest. Mann et al. (1988), in a
review of 18 studies on clear-cutting in temperate forests, found that most increases in
hydrologic losses of Ca"2 were less than 10 kg/ha in the first year, and most increases in K'
losses in the first year were less than 3 kg/ha. Increases in losses relative to uncut systems
had mostly disappeared by 3 yr after harvesting, and the authors concluded that hydrologic
losses were insignificant compared to harvest export losses.
Given the above, and the large differences in NO" leaching between cut and uncut
treatments, it was surprisingly to find no difference in cation leaching among species or
cutting cycle treatments. Nitrification promotes cation leaching, both by acidifying the soil
solution (which can displace cations from exchange complexes), and by serving as a
charge-balancing anion to accompany leaching cations (Raney 1960). In the present study,
increases in negative charge from NO" in the soil solution were indeed accompanied by an
equivalent increase in positive charge which was partitioned among three cations (Chapter
3). Nevertheless, the increase may have been too small to detect when partitioned among
three cations. The high cation exchange capacity of the soil (z20 cmolkg in the upper 0.1
m), which helps to retain cations, may also have contributed to lack of difference among
It was troubling to find no differences in K4 leaching, in light of previous findings
(Chapter 3) of statistically significant elevations in leachate K' concentrations under
annually cut stands. The explanation lies both in the small magnitude of the increases in K'
(median K' concentration in the annually cut stands was 1.45 times that in the uncut
rotation) and in differing statistical approaches. K4 concentrations (Chapter 3) were
analyzed using a repeated-measures design, which increased statistical power due to the
many sampling dates. Absolute leaching losses of K', in contrast (this chapter), were
analyzed with a simple analysis of variance: the many sampling dates simply improved the
precision with which leaching was estimated, and provided no additional statistical power.
It is likely that increased losses of K' from annual compared to uncut rotations are
As expected given the relative quantities of extractable cations present in the soil
(Table 4-1), losses of Ca2 were much greater than losses of K4 and Mg2. Annual levels of
loss, averaged across all species and treatments, were 211 kg/ha Ca42, 56 kg/ha Mg2, and
42 kg/ha K Bruijnzeel (1990) reviewed cation outputs of seven undisturbed tropical
forests: annual Ca2 outputs were 651, 583, 163, 86, 25, 20, and 10 kg/ha; Mg2 losses
were 78, 51, 50, 44, 43, 17, and 10 kg/ha; and K4 losses were 76, 49, 22, 15, 12, 9, and 3
kg/ha. Losses from the soil in the present study, although high, are well within these
ranges, and are consistent with the high base status of the soils (Table 4-1).
In summary, the present study showed that hydrologic losses of N03" after cutting
of very young tree plantations on a tropical, volcanically derived soil with allophanic
mineralogy were high. These losses were several times higher than post-harvest N03"
losses recorded from temperate forests. In contrast, there was no augmented cation
leaching due to cutting. The magnitude of NO3 loss in the year subsequent to cutting was
roughly similar regardless of whether stands with large or small amounts of sequestered N
(i.e., 4-yr-old compared to 1-yr-old stands) were felled. Thus, losses should be inversely
proportional to the cutting frequency. To prevent large losses of nitrate with potentially
deleterious effects on productivity, short cropping rotations should be avoided on these
Table 4-1. Soil characteristics of the study site (a Eutric Hapludand) in 1993. Means and
standard deviations of 27 plots. Ca"E, Mg'2, K', A113, and CECot are in units of cmoljkg,
P is in mg/kg, and N and C are in per cent (g/g 100).
depth below soil surface (m)
t includes data from one block sampled from 0.70 to 1.55 m.
"effective cation exchange capacity (CEC.) is the sum of Ca, Mg, K,
t"tApH is the difference between pH measured in KCI and in H20.
Al, and Na (not
Table 4-2. Results of repeated-measures analysis of variance of volumetric soil moisture in
annually cut and uncut stands of three tropical tree species. Measurements were taken on
eight dates between December 1994 and May 1995. Bold values are significant at p _.05.
variance source NDF' DDFt Type I F Pr> F
species 2 4 1.73 0.28
cutting cycle 2 86 1.26 0.29
species cutting cycle 4 86 0.47 0.75
time 7 10 60.64 0.0001
time species 11 86 3.00 0.0020
time cutting cycle 12 86 2.98 0.0016
time species cutting cycle 20 86 1.72 0.044,8
Tnumerator degrees of freedom
denominator degrees of freedom
Table 4-3. Results of repeated-measures analysis of variance on volumetric soil moisture
in 1-yr and 4-yr rotations and uncut stands of Hyeronima, Cedrela, and Cordia trees after
cutting and replanting of the 4 yr rotations. Values were averaged by month to comply
with assumptions of the statistical analysis. Eight, six, and five mo of data (spanning
June 1995-February 1996) were available for Hyeronima, Cedrela, and Cordia,
respectivelv. Bold values indicate significance at the P 0.05 level.
Species Source NDFt DDFtt Type 111 F Pr> F
Hyeronima stand age 2 32 0.21 0.8084
time 7 14 14.37 0.0001
time stand age 14 32 0.2776 0.2776
Cedrela stand age 2 24 1.91 0.1705
time 5 10 17.79 0.0001
time stand age 10 24 1.07 0.4243
Cordia stand age 2 20 0.45 0.6442
time 4 8 18.33 0.0004
time stand age 8 20 0.90 0.5358
*numerator degrees of freedom
'denominator degrees of freedom
Table 4-4. Hydrology of plots of Hyeronima alchorneoides. All values are in mm. P is
precipitation, Ej is interception, E. is canopy transpiration, ET is total evapotranspiration
(i.e., Ej + E, + E), and D is drainage.
Period P Ei ES EC ET D
March 1992 -December 1992 3362 243 44 571 858 2548
1993 3659 280 51 664 995 2714
1994 4450 268 49 607 924 3541
1995 2870 267 58 751 1076 1728
January 1996 February 1996 352 34 6 78 117 211
Table 4-5. Cumulative leaching losses (means and standard deviations; g/m2) from stands of three tree species, before and after cutting
of the 4 yr rotations.
Before 4-yr cut After 4-yr cut Cumulative total
(1992-mid 1995) (mid 1995-early 1996) (1992-early 1996)
genus ion annual 4-yr uncut annual 4-yr uncut annual 4-yr uncut
NO3- Hyeronima 16.7(8.6) 1.8(0.7) 2.3(2.6) 3.3(5.1) 4.1(1.7) 0.13(0.12) 20.0(13.6) 5.8(1.0) 2.7(2.5)
Cedrela 12.8(0.6) 3.0(2.4) 2.3(1.8) 2.3(2.2) 2.9(2.5) 0.37(0.54) 15.1(2.7) 5.9(4.8) 2.7(2.3)
Cordia 13.6(6.7) 1.3(0.9) 1.3(0.8) 1.3(0.7) 3.8(1.0) 0.13(0.18) 15.0(6.4) 5.1(1.6) 1.4(1.0)
Ca+2 Hyeronima t t 12.1(1.5) 14.7(3.4) 15.4(4.9) t t
t t t t t t
Cedrela 13.1(2.2) 12.5(5.2) 11.1(1.1)
tt t t
Cordia 10.4(2.1) 12.2(2.8) 8.8(1.6)
Mg+2 Hyeronima 19.8(2.3) 11.6(0.8) 12.9(2.4) 4.7(2.9) 2.7(1.6) 2.6(1.7) 24.5(4.1) 14.3(2.2) 15.5(4.1)
Cedrela 22.9(2.5) 23.0(8.1) 16.9(4.2) 3.8(0.4) 3.7(1.5) 3.0(0.9) 26.8(2.9) 26.6(9.1) 19.9(4.5
Cordia 23.0(6.7) 20.1(0.7) 18.0(6.1) 3.7(1.2) 2.9(1.6) 2.4(0.4) 26.7(11.4) 23.8(12.5) 20.4(6.4)
K Hyeronima 13.5(2.2) 11.1(2.2) 12.1(2.5) 3.4(0.9) 4.2(0.1) 3.8(1.6) 16.9(2.9) 15.3(2.6) 15.9(4.1)
Cedrela 14.3(1.7) 16.3(6.1) 12.1(3.7) 3.6(0.5) 3.9(3.5) 2.9(0.7) 17.9(1.8) 20.2(9.1) 15.0(4.4)
Cordia 11.8(1.5) 16.2(8.3) 12.0(0.5) 2.6(0.6) 3.3(1.6) 2.2(0.4) 14.4(1.8) 19.6(9.9) 14.3(0.8)
t Ca+2 data prior to mid-1995 are not reported due to analytical problems.
Table 4-6. Probability values for analysis of variance of cumulative leaching losses of
nutrients from stands of three tropical trees subjected to three cutting cycle treatments.
Highlighted values are significant at alpha = 0.05 level of probability.
1992 mid 1995 mid 1995 early 1996
variance source NO3" Mg2 K NO3" Ca2 Mg2 K-
species 0.62 0.34 0.68 0.81 0.26 0.88 0.49
cutting cycle 0.009 0.11 0.40 0.04 0.41 0.08 0.34
species*cutting cycle .74 0.55 0.50 0.80 0.26 0.55 0.92
0 1 2 3 4 5 6 7 8
9 10 11 12 13 14 15 16 17 18 19 20
Figure 4-1. Cumulative daily net radiation above a tree canopy plotted against maximum
minus minimum daily temperature. Straight lines are a spline curve regression.
o Hyeronima alchorneoides
(1 0.38 __simulated soil moisture
0 measured soil moisture
0 100 200 300
Figure 4-2. Simulated and measured soil moisture in top 1 m of soil in experimental
plantations of Hyeronima alchorneoides at La Selva Biological Station, Costa Rica. Soil
moisture was measured with time-domain reflectometry in stands of three species planted
in three blocks (Hyeronima is shown as an example). Each data point is a mean of nine
measurements, one for each species and block; each measurement is a mean from three
sets of probes in each plot. Bars on circles represent standard errors.
Figure 4-3. Four yr of simulated soil moisture in upper 1 m of soil and monthly rainfall and drainage totals in uncut stands of
MA M J J AS ONDJ FM AM J J A SO NDJ F MA M J J A SOND J F MA M J J A SOND
1992 1993 1994 1995
EFFECTS OF NITRIFICATION AND UPTAKE ON NITROGEN
LEACHING FROM STANDS OF THREE TROPICAL TREE SPECIES
Nitrate (NO;3) leaching is a major pathway for nitrogen (N) loss in the humid
tropics, where precipitation can greatly exceed evapotranspiration (Young 1976). Such
losses can have important implications for soil fertility and water quality. The magnitude
of N03" leaching is determined by the balance between the process that supplies NO3', i.e.,
microbial oxidation of organic nitrogen, and processes that tend to retain N03' within
ecosystems, such as uptake by plants (Vitousek and Melillo 1979). Despite the recent
burst of research on N in terrestrial ecosystems, the relative contribution of these two
factors in determining leaching is not often measured in forested ecosystems (but see
Vitousek and Matson 1985, Dyck et al 1987, Matson et al. 1987, Smethurst and Nambiar
Although early studies discounted the role of variation in N mineralization and
nitrification rates as determinants of N leaching rates among forested ecosystems
(Vitousek and Reiners 1975), more recent research has highlighted their importance.
Nitrogen mineralization rates were the single most important factor in explaining the
variation in NO;" leaching from 17 temperate forest sites (Van Miegrot et al. 1992). In
another study, the proportion of mineralization due to nitrification, along with total N in
the upper 0.1 m of soil, explained 88% of variation in watershed N03" export in nine
forested watersheds in the mid-Appalachians of eastern North America (Williard et al.
Despite this variation, uptake and mineralization are often tightly coupled in
growing stands, and where this is true leaching losses are minimal. This relationship can
dissolve after large-scale disturbances. For example, after plant harvest, uptake ceases
suddenly, while mineralization continues unabated or may even increase due to increased
temperatures and soil moisture (Burger and Pritchett 1984, Vitousek and Matson 1985,
Matson et al. 1987, Smethurst and Nambiar 1990), resulting in sharply elevated levels of
N03" leaching (Bormann et al. 1968, Vitousek and Melillo 1979, Mann et al. 1988,
Stevens and Hornung 1988).
The relative importance of increased mineralization compared to increased N03"
retention in early stand establishment has been quantified only infrequently (Smethurst and
Nambiar 1995). Mechanisms limiting NO; losses from disturbed forests have been
reviewed by Vitousek and Melillo (1979). They stressed that a limited number of
mechanisms can prevent losses to groundwater of N mineralized in ecosystems. The ones
of potential importance in humid tropical ecosystems include plant uptake, immobilization,
volatilization, anion adsorption, and prevention of nitrification. Plant uptake is likely to be
the most important of these in tropical forest ecosystems, where regrowth after
disturbance is rapid. For example, in wet evergreen forest in West Africa, N accumulation
in vegetation was estimated as 114 kg ha-1 y-' during the first 5 yr of succession (Nye and
Greenland 1960). In comparison, denitrification in lowland tropical rainforest in early
succession has been measured at <5 kg ha-' y-1 (Robertson and Tiedje 1988, Keller and
Reiners 1994). Anion adsorption can also be an major nitrate retention mechanisms in
variable-charge clays in the tropics (Uehara and Gillman 1981, Matson et al. 1987).
A field experiment was utilized to explore the relative importance of mineralization
and uptake in determining N03 leaching in lowland wet tropical agroecosystems. Two
treatments were compared: stands that were established and then allowed to grow, and
stands that were cut and replanted annually. After cutting the annual stands the slash was
left on the soil. Thus, I compared growing stands, in which net uptake and sequestration
of N were potentially positive year after year, to repeatedly cut stands in which there was
no opportunity for net uptake of N from year to year. Stands of three tropical tree species
were examined: the species differed in leafing phenology, root distribution, leaf size, and
growth rate. I predicted that both the elevated mineralization rates and the low uptake
rates that occur after harvesting would lead to very high leaching losses of N from annual
rotations. The three species, which differed in N uptake rate, were further examined to see
whether higher uptake rates led to lower leaching losses
The study took place at La Selva Biological Station, in the Atlantic lowlands of
Costa Rica. Mean annual rainfall and temperature are 3962 mm and 25.8 'C, respectively
(Sanford et al. 1994). There is a short dry season in February-April, although mean rainfall
exceeds 100 mm in all months. The experimental site is a recently formed alluvial terrace
with a deep, loamy soil classified as a eutric Hapludand (Weitz et al. 1997). Due to the
volcanic origin of the parent material, the soils have very low bulk density (about 0.7
g/cm3 in the upper 10 cm) and are porous, permitting rapid drainage. Their highly
aggregated structure probably limits the contact that draining water has with the soil and
thus may serve to decrease nutrient leaching (Sollins et al. 1994).
The soil cation exchange complex is dominated by Ca: extractable values (means
of all species and cutting cycle treatments; upper 0.1 m of soil) from a 1993 sampling were
15.8 cmolkg Ca (standard deviation = 6.07), compared to 1.7 (0.3) cmoljIkg K and 3.1
(0.7) cmolkg Mg. There was little exchangeable hydrogen or aluminum; thus, base
saturation was effectively 100%. Effective cation exchange capacity was 21.1 cmoljkg.
Nitrogen retained on the exchange complexes was negligible compared to the base-
forming cations: there was 0.040 (0.008) cmol/kg of ammonium (NH4+), and 0.058
(0.0 10) cmolkg of N03.
The three tree species planted, all native to Central America, contrast strongly in
appearance and phenology. Hyeronima alchorneoides (Euphorbiaceae) is a massive
canopy emergent with evergreen leaves. Once complete canopy cover was established,
approximately 1.5 yr after planting, very few plants successfully colonized in its
understory. Cordia aiodora (Boraginaceae), a slender canopy tree, becomes leafless for
several months at the beginning of the wet season. In contrast to the Hyeronima stands, an
understory grew readily and was cut back approximately every 6 wk. The third species,
Cedrela odorata (Meliaceae), is a canopy tree that becomes leafless for several months
toward the beginning of the drier season. Like Cordia, its crowns were deep and sparse,
with a thick herbaceous layer growing beneath them.
In April 1991 the site was cleared of an abandoned cocoa plantation, merchantable
overstory trees (mostly Cordia alliodora) were harvested, and the slash was burned.
Immediately following manual clearing of charred logs, tree plantations were established in
three blocks. In each block, species were planted in two plots: the annual rotation was
planted in 10 by 40 m plots, and the uncut rotation was planted in 30 by 40 m plots.
Larger plots were used for the uncut treatments in anticipation of the trees in them
becoming much larger and fewer (due to natural mortality and thinning) over time.