Several workshops have been held to formulate research plans for the various components of the Large Scale Biosphere Atmosphere experiment in Amazonia (LBA). This draft document forms the next step to the implementation of a research programme for the Land Surface Hydrology component in Rondônia, Pará State, Brazil. The document should serve as a starting point to the formulation of a more detailed programme for meso- and micro-scale catchment studies, in which all collaborating research groups should specify their objectives and requirements relating to the forest and pasture hydrological work, such as to arrive at a well-balanced and efficient working plan. The document should be circulated among all partners to obtain their views such that suggested modifications to the preliminary programme described below can be incorporated.
The main research questions
for LBA have been formulated in the concise experimental plan
of the LBA Science Planning Group (1996), and are detailed below:
How does Amazonia currently
function as a regional entity?
How will changes in land
use and climate affect the biological, chemical and physical functions
of Amazonia, including the sustainability of development in the
region and the influence of Amazonia on the global climate?
With respect to these very
broad questions, the LBA land surface hydrology group, in collaboration
with research groups from related disciplines (e.g. physical climate,
ecology, geobiochemistry), aims to address the following issues,
which are more specifically directed towards understanding the
hydrological behaviour and water quality (dissolved and particulate
matter) in the Amazon basin and their sensitivity to changes in
land use and climate at the micro-scale, as well as at the meso-
and macro-scales. The guiding science question for hydrologist
in LBA is (Dunne et al., 1997):
What are the spatial and
temporal patterns of water storage, water flux, and the transport
of bio-active materials in the soils and river corridors of the
Amazon Basin, and how are they influenced by variations in climate
and land use?
This can again be subdivided
into research questions pertaining to micro-scale catchment (<10
questions pertaining to meso-scale
basin (102-104 km2) studies:
and those pertaining to macro-scale
basin (>105 km2) studies (Dunne et al.,
The land surface hydrology
is also involved in research on the transfer of nutrients, particulate
matter and organic matter from the land to the river, and subsequently
to the basin outlet. Research questions pertaining to this subject
are formulated below (Dunne et al., 1997):
The area selected for the land surface hydrology is in the Rondônia Province, where large-scale conversion of tropical rain forest to pasture and crop-land occurs. The European-Brazilian contribution to the Land Surface Hydrology component will consist of micro- and meso-scale field measurements of water stores, fluxes and nutrient transport, to be made over a period of several years to account for temporal variations in rainfall, combined with micro- and meso-scale model development and modelling. The field measurements, and micro- and meso-scale model development and testing will be integrated to facilitate the extrapolation of the results of the meso-scale modelling to other meso-scale basins in the Amazon. The subsequent coupling of the meso-scale model with a macro-scale model (American-Brazilian contribution) should produce a macro-scale model treating the Amazon basin as a whole. In this respect, the meso-scale modelling is seen as crucial to bridging the gap in scales between micro-scale hydrological modelling and the macro-scale data requirements for climate and land use impact modelling. The European contribution to the Brazilian LBA experiment is therefore mainly in the micro- and meso-scale data collection and model development, whereas the development, testing and running of a macro-scale model will be the American contribution.
For an accurate description
of the hydrology of the macro-scale Amazon Basin, the micro- and
meso-scale land surface hydrology programme should ideally cover
the full range of climate, geology, topography, soil and vegetation
types encountered at different scales in the Amazon Basin. This
is for obvious reasons not possible. As a compromise, the present
study will initially deal with the two most common land use types,
being primary (or old secondary) lowland 'terra firme' rain forest
and regularly burned pasture. Furthermore, the range of climatic,
topographic, geologic and soil factors is restricted by the location
of the research area in the Rondônia Province on the Brazilian
shield. Measurements made by other research groups in areas with
different environmental conditions may however be used in the
modelling to gain further insight into the hydrological cycle.
However, additional measurements elsewhere in the Amazon basin
(e.g. Central Amazon, Pará State, Andes region) may be
necessary in the future to improve our understanding of hydrological
processes active in the different parts of the basin.
2. Participating Groups
The following institutions and persons take part in the micro- and meso-scale hydrological study.
CENA, Brazil: R.L. Victoria
CPTEC/INPE, Brazil: J. Marengo, J. Tomasella
DLO - Winand Staring Centre, The Netherlands: P. Kabat, M.J. Waterloo
Institute of Hydrology - Wallingford, UK: M. Hodnett
LTHE, France: R. Haverkamp
ORSTOM, France: J.L. Guyot
PIK, Germany: A, Becker
UNESCO IHP, France: M. Bonell
University of Newcastle, UK: P.E. O'Connell, C.G. Kilsby
Vrije Universiteit Amsterdam, The Netherlands: L.A. Bruijnzeel
Wageningen Agricultural University, The Netherlands: R.A. Feddes
3. Inventory of existing data and publications
The first phase of the Land Surface Hydrological programme should involve the preparation of an inventory of existing data records, such as long-term rainfall and streamflow records for micro- and meso-scale basins, detailed topographical information, soil maps, geomorphological maps, satellite and radar images, aerial photographs, etc., and to make an assessment of their quality. This should be a priority and should preferably be finished before site selection occurs, such that sites with existing data can be taken into evaluation. For instance, it would be a strong advantage to work in meso-scale basins for which long-term rainfall - runoff records are available as the strategic modelling component must necessarily rely heavily on the existence of such data for initial test runs.
The following data sources
for rainfall and streamflow have been identified (Dunne et al.,
1997), and an assessment should be made of the usefulness of the
data to the present study:
ORSTOM (Brazil, Bolivia, Guyana¼?)
A concise overview of literature
relevant to the hydrology of the Amazon Basin has been provided
in Annex 1 to give an impression of the existing knowledge, and
of the gaps therein. General information on a range of biogeophysical
and socio-economical issues regarding the Amazon Basin has been
provided in books by Sioli (1984) and Hemming (1985a,b). The following
conclusions concerning the current status of hydrological research
in the Amazon basin can be drawn from the review presented in
4. Research approach
The paired catchment approach is the traditional, and most rigorous, way of determining the effects of changes in land use on the rainfall - runoff response on the micro-scale catchment level, and to a lesser extend on sediment concentrations and water chemistry. Such a study involves the simultaneous monitoring of two micro-scale catchments of similar size, topography, geology, soils, climate and land use over a number of years (calibration period) until the rainfall - runoff response has been sufficiently established for a wide range of climate (i.e. rainfall, evaporation) conditions. The calibration ensures that differences in the rainfall - runoff response, sediment concentrations and water chemistry between the catchments, due to differences other than in land use, can be identified. The land use in one of the catchments then becomes subjected to a change, whereas that in the other remains untouched to serve as a control. Any changes observed in the rainfall - runoff response, sediment concentration or water chemistry in the treated catchment as compared to those in the control, can then be fully attributed to the change in land use.
Due to the increase in the variability of factors governing runoff, erosion, etc., with increasing catchment size, such a paired catchment study is not feasible on meso- or macro-scale levels. The effects of land use conversions and changes in climate on the hydrology can then only be determined through the use of models in which the dominant hydrological processes are accurately described.
For a number of reasons (time, logistics), the LBA Land Surface Hydrology programme has opted for a direct comparison of two micro-scale catchments with different land use. This is, technically speaking, not a paired catchment study and it should be noted that the pre-existing conditions (e.g. differences in subsurface leakage, water quality, etc.) in the catchment that is used for treatment (pasture) remain unknown, and were not necessarily similar to those of the catchment being effectively used as a control (rain forest). This is all the more true when suitable micro-scale catchments cannot be located in the same area, as an increase in distance between the catchments will in most cases increase the likelihood of differences in geology, soils, climate, etc. However, it is assumed that these uncertainties can be overcome through the development, validation and use of physically-based, distributed-parameter hydrological micro- and meso-scale models. These models should incorporate the dominant hydrological processes and how they are influenced by land use changes, as determined from plot and transect studies. The biophysical information obtained in micro-scale catchment and transect studies in different land use, climate and vegetation zones will assist in the extrapolation of models to different biophysical environments in the Amazon Basin. A nested approach (HYNEST) has been proposed, which means that the micro-scale catchments should preferably be located in meso-scale research basins such that the importance of hydrological processes and river hydrographs can be traced from the micro-scale to the meso-scale.
For an accurate assessment of the impact of changes in land use and climate on the hydrology and nutrient cycling, and to improve the modelling thereof, the importance of the various pathways of water through the ecosystem at a range of scales needs to be identified, and their sensitivity to land use and climate change be assessed. To achieve this, detailed information should be obtained on the following issues for contrasting land use types:
In addition, information will
be needed on:
During the first phase of
the project, the understanding of key hydrological processes,
obtained from the micro-scale catchment field campaigns, will
be used to improve the physically-based parameterizations of the
hydrological modelling systems. In the second phase, the improved
parameterizations will be incorporated in the modelling systems
and these will subsequently be validated on rainfall - runoff
data obtained from other micro- and meso-scale catchments using
ordinary, less-detailed, geology, soil and vegetation maps (remotely
sensed data) as input. The validated model will then be used to
assess the impact of land use and climate changes on the hydrology
and to provide input to the macro-scale hydrological model.
The ideal model would be (Wageningen
Workshop Thematic Inventory, 1997):
The structure of the meso-scale
model should comply to the following conditions:
The classification scheme to be incorporated should be based on commonly available information for the whole Amazon Basin, such as on land use and cover, topography, channel length, geology, soil and climate.
After the initial parameterization, based on data from micro-scale rain forest and pasture catchments, the model will be applied to predict the rainfall - runoff response for other micro-scale catchments in the same area, of which the rainfall - runoff response has been measured. When these predictions are satisfactorily, the model will be used to predict the response for two meso-scale basins and the results will again be compared with measured streamflow data. When deforestation and climate change scenarios are available, the validated model can be used to predict their effects on the hydrology of other ungauged micro-scale as well as meso-scale basins.
The physically-based distributed-parameter TOPOG model, developed at CSIRO, Australia, in combination with the SVAT model VAMPS, developed at the Vrije Universiteit Amsterdam, will be used to model the hydrology of micro-scale catchments in Central Amazonia using existing data, supplemented by data collected during short campaigns.
The Topog Online web site
(http://www.clw.csiro.au/topog) provides the following description
of TOPOG in its introduction: TOPOG is a terrain analysis-based
hydrologic modelling package which can be used to:
TOPOG describes how water
moves through landscapes; over the land surface, into the soil,
through the soil and groundwater and back to the atmosphere via
evaporation. Conservative solute movement and sediment transport
are also simulated. The primary strength of TOPOG is that it is
based on a sophisticated digital terrain analysis model, which
accurately describes the topographic attributes of three-dimensional
landscapes. It is intended for application to small catchments
(up to 10 km2, and generally smaller than 1 km2).
TOPOG is referred to as a 'deterministic', 'distributed-parameter'
hydrologic modelling package. The term 'deterministic' is used
to emphasise the fact that the various water balance models within
TOPOG use physical reasoning to explain how the hydrologic system
behaves. The term 'distributed-parameter' means that the model
can account for spatial variability inherent in input parameters
such as soil type, vegetation and climate. TOPOG is intended as
a research tool and is relevant to a wide range of water and land
management issues, including:
A user guide and additional information (applications) is available at the TOPOG ONLINE web site.
The features presented above indicate that the model will be very suitable for micro-scale forest and pasture catchment modelling studies of LBA, also because the required input and validation data can be collected during the various field campaigns.
I therefore propose that the TOPOG model will be used for micro-scale modelling in Rondônia as well, which implies that the model will have to be installed at SC-DLO, IH Wallingford and various Brazilian Research Centres, such that a start can be made with strategic modelling soon after the forest and pasture research catchments have been selected and topographic data becomes available such that digital elevation models can be developed.
The TOPOG model has originally been compiled for IBM-AIX and Sun Solaris operating systems, and only recently for the LINUX (UNIX for PCs) operating system. The source code, however, is freely available through the web site of CSIRO after a user name and password have been obtained from CSIRO to provide access. The program can, and will have to be ported to other operating systems in use at the various institutes, such as DEC/ALPHA Unix (at SC-DLO). As the model will have to be implemented at a number of research institutes, and may also have to be adapted when specific information on hydrological processes in Amazonia becomes available through LBA research, I should like to recommend that a formal cooperation be established with the Cooperative Research Centre for Catchment Hydrology at CSIRO, Canberra (Dr. Rob Vertessy). It may also be necessary for the various researchers involved in micro-scale catchment modelling to follow some kind of basic training for modelling and model adaptation at CSIRO.
The strategic modelling exercise will provide an understanding of which processes should be better represented in the models, which have traditionally been developed for temperate climates, to deal with the humid tropical environment. The TOPOG model will also be a valuable tool in the identification of hydrological response units to be used in meso-scale modelling.
Another option would be to use the Limburg Soil Erosion Model (LISEM, De Roo et al., 1990)), developed at the University of Utrecht, The Netherlands, to model runoff, erosion and sediment yield from micro-scale catchments. The model runs on a PC. As the model still contains some bugs in the runoff and erosion routines and has only limited graphical possibilities, should prefer the use of TOPOG.
The UP model has been identified for the meso-scale modelling work(?). The model data requirements and the information needed to adapt the model to Amazonian conditions should be listed in this document. Additional information should be included.
As an alternative, the Large Scale CAtchment Model (LASCAM), developed at the Centre for Water Research (Woods, ) may be used.
5. Site selection
The research and modelling approach detailed above implies that at least two micro-scale catchments (for measurement and validation) and at least one meso-scale basin for each land use type be instrumented and monitored for a number of years. The micro-scale catchments should preferably be located within the meso-scale basin (nested approach, HYNEST).
Ideally, all research groups
should be working in the same catchment area, such that information
collected by the various disciplines can readily be combined and
used for modelling. However, site requirements may conflict and
it is therefore proposed that at least one member of each of the
research groups (e.g. Physical Climate, Ecology, Land Surface
Hydrology, Biogeochemistry, Carbon Exchange and Storage, Land
Use and Land Cover Change (?), Remote Sensing(?)) joins in the
actual process of site selection. The following is a tentative
list of site conditions which should preferably be met for the
hydrological study in micro-scale catchments.
The following conditions,
most of which are the same as for the micro-scale catchments,
should apply for the meso-scale basins
Additions to this list of site-selection criteria should yet be made for other research programmes.
A number of forest and pasture catchments have been identified during the LBA Reconnaissance flight in September 1996 (Kabat et al., 1997). The Reserva Jaru was recommended as the only site where the rain forest would presumably remain undisturbed for the duration of the LBA experiment, and where a micro-scale forested catchment could be instrumented. The Fazenda Triangulina (15 km2) and the Fazenda Agropecuaria Rio Machado (17+19 km2) were identified as suitable for the instrumentation of micro-scale pasture catchments. The area of the former ranch is also partly covered by selectively logged rain forest. These sites should again be visited to assess their suitability for hydrological research and the present (and future) condition of the vegetation. In addition, some possible sites (which could not be visited during the last reconnaissance mission should be visited during the next mission to assess their potential. If a suitable site is found, permission should be negotiated with the land owners to ensure their collaboration, while stressing that the vegetation in the selected forested areas should remain untouched during the length of the study. An accessible, forested meso-scale basin was difficult to find, and we may have to settle for a catchment at the lower end of the meso-scale range. The Rio Jaru and Rio Jamari were identified as the most likely candidates for the selection of a disturbed meso-scale basin. Additional visits and negotiations with landowners will have to be made before final site selection.
The micro- and meso-scale catchments should be selected as soon as possible, because the choice and density network of the instrumentation will depend for a large part on their size, accessibility, topography, channel morphology, soil, etc. An additional constraint in timing is that certain instruments (e.g. discharge measurement structures) can only be installed at the end of the dry season when streamflow levels are low. Strategic modelling also depends on the availability of detailed topographical information of the selected site, although some preparations can be done at present (choice and installation of model, training, compiling relevant data from literature).
The availability to all partners
playing a role in the selection of sites of the following material
on the research area in Rondônia, before actual site selection
would be of great advantage to the site-selection process:
A set of the data quoted above should become available to all groups before the site selection mission.
6. Meso-scale basin data collectionMeso-scale basin data collection
Because the exact locations of research sites are yet unknown, the following list necessarily provides a tentative approach towards carrying out the field measurement campaigns. A more definite programme can only be made after sites have been selected and their size, topography, vegetation distribution, etc. have been assessed. Meso-scale basin measurements basically consist of rainfall - runoff measurements and water quality measurements. Additional information for modelling should be obtained from the micro-scale catchment studies and from existing data sources such as maps, remote sensing images, etc. The list below does not take into account constraints related to funding, access, travel times or manpower!
The University of Newcastle (lead), together with CPTEC and SC-DLO are involved in measuring meso-scale rainfall. The spatial and temporal variation of rainfall may be considered high in a convective rainfall regime such as that in Rondônia. This should therefore be reflected in the rainfall measurements, and in particular in the number and areal distribution of gauges used to estimate the rainfall input over the basin area (one gauge per 10-100 km2?). This dense network of (automatic) rain gauges should be complemented with rainfall radar and TRMM data to obtain a fair estimate of the spatial rainfall distribution. The ground data can, in turn, be used for the calibration of the TRMM data. The location of the meso-scale basin should therefore be such to profit from the presence of the rainfall radar in the area, which may best be located at Ariquemes, Rondônia (Kabat et al., 1997). Stratified sampling of rainfall, based on the topography (and possibly other relevant features) of the basin, may be necessary to identify trends in the rainfall distribution (orographic, land use effects). (option?: to reduce hardware costs of the network, the number of automatic rain gauges could be reduced in favour of standard rain gauges, emptied on a daily basis)
The installation and maintenance of such a dense rainfall measurement network is laborious to say the least, taking into account the size of the meso scale basin, and the number of gauges which can be installed will therefore be largely determined by practical restrictions (e.g. all-year access, presumed difficulties in finding open sites where rainfall measurements can be made conforming to the WMO standard, maintenance costs).
To account for the temporal
variation, measurements should preferably be made over a number
of years to include very wet as well as very dry years. A list
of possible hardware is provided in Annex 2.
IH Wallingford and SC-DLO are the leading institutes for the monitoring of meso-scale basins under forest and pasture, respectively, but all other partners are involved in some or other way, with the exception of PIK. In large rivers, construction of a fixed discharge measurement structure is not economically feasible due to the high discharges associated with the size of the drainage area (meso-scale). Hence, a straight, stable section (preferably on bedrock) in the river should be identified, in which water level measurements should not suffer from backwater effects during wet periods.
At these sites, automatic water level recorders (pressure transducer type) or preferably acoustic flow measurement equipment should be installed and maintained, and a fixed staff gauge should be installed as a reference. Measurements may be made at 15-30 minute intervals, although a 5-minute interval may be preferred for consistency with the micro-scale measurements. To avoid gaps in the long-term data record due to equipment malfunctioning I should strongly suggest that a second system (water level recorder, perhaps of a different design (robust float type mechanical recorder?)), be installed to serve as a backup system.
When a water level logger
is used, a discharge rating curve has to be developed to translate
the observed water levels into discharges. This incorporates the
frequent use of a boat and current meter to gauge the discharge
in individual sections of the river channel profile. Simultaneous
measurements of water level and discharge should be made both
during low and high flow conditions (wet and dry season campaigns).
Measurements should be made throughout the study to test for the
stability of the river bed. When a more recently developed system
(by Ott), based on sound propagation, can be installed, the procedure
will be greatly simplified as this system provides direct average
streamflow velocity measurements (at one ore more depths), which
can be combined with the river cross-section area at the corresponding
height to provide discharge values. A list of related hardware
is provided in Annex 2.
Soil hydraulic properties
LTHE is the lead in the soil
physical component of LBA. Existing soil data (maps) and data
obtained in the micro-scale catchments will be used to create
input into the meso-scale hydrological model.
Streamflow chemistry and
Regular water samples should be collected at the outlet of the basin for the determination of nutrient and sediment concentrations (CENA?). Sampling on a weekly basis may be sufficient under baseflow conditions. However, an automatic water sampler (1 hour sampling interval) should be installed during the wet season to capture the dynamics of the streamflow. A more detailed sampling scheme can only be designed when information on the hydrologic response of the basin to rainfall has been assessed.
Electrical conductivity (EC),
water temperature (Tw) and pH should be measured in
the field. A EC/Tw/pH probe, connected to a datalogger
can provide detailed information, which may assist in the separation
of stormflow and baseflow components, and may be used for the
determination of nutrient exports when relations can be obtained
bteween these parameters and other chemical components. Samples
should be collected for cation (Na, K, Mg, Ca, Si, Al, Fe, ¼),
anion (HCO3, Cl, SO4, NOx, PO4,
TOC, TIC, BOD, Total P, Total N, etc. analysis. These should be
analysed shortly after sampling. Concentrations are presumably
very low and therefore prone to contamination or analytical errors.
It is therefore imperative that a qualified laboratory, which
can carry out such analysis with very low detection limits, will
Relevant information on the above-ground parts of the vegetation must be obtained from the remote sensing group (LAI, biomass, vegetation types), in combination with ground-based measurements in the micro-scale catchments (ecology, biogeochemistry components).
Root distribution with depth,
root biomass, etc. will be obtained from samples collected in
deep soil pits (down to bedrock) in micro-scale catchment studies.
Topography, channel morphology
The topography, channel width
and morphology, etc. can be determined from maps, satellite images
or aerial photographs (remote sensing group). Ground-based measurements
(river depth, channel roughness?) may be necessary to complement
Please provide us with additional data needed for meso-scale modelling.
7. Micro-scale catchment
Measurements in the micro-scale
basins will include catchment-wide measurements (e.g. runoff,
evaporation), as well as plot measurements. The leading institutes
for monitoring are again IH-Wallingford (forest) and SC-DLO (pasture)
in collaboration with the Brazilian institutes, but other groups
are expected to play a dominant role during the dry and wet season
campaigns. The forest and pasture measurements should preferably
be made in the same way (methods and instrumentation) to avoid
differences related to differences in equipment. Two complete
set of instruments to carry out the measurements detailed below
are therefore deemed necessary. Study plots in forest and pasture
should provide the following information, with which the water
storages, fluxes and related transfers of nutrients and sediment,
can be quantified:
The instrumentation of several
research plots on different terrain elements (e.g. catena
transects) will provide a good indication of the spatial variability
of these factors within the catchment as related to the topography.
A more detailed discussion of these measurements is given below.
The spatial variation of rainfall will be less in the micro-scale catchment than in the meso-scale catchment. Depending on the size (1-2 km2), shape and topography of the selected catchment, a minimum of 4-8 automatic rainfall recorders should be installed to obtain a fair estimate of the areal precipitation. Additional rain gauges may have to be installed when large differences are observed between the gauges. At least one special totalizing rain gauge should be installed in each catchment for sampling of rainwater for chemical analysis (nutrient input, bulk samples?).
The installation in pasture should be fairly straightforward due to the limited height of the vegetation. However, several towers may need to be installed in the forest catchment to provide above canopy access. Such towers may also be required by other groups (ecology, plant physiology, biogeochemistry?) for canopy level measurements. All rain gauges should form part of the dense rain gauge network to be set up in the meso-scale basin. A list of hardware is provided in Annex 2.
Rainfall interception (throughfall
Because of the low height of the pasture vegetation (< 60 cm) the traditional method using rain gauges cannot be applied. A more suitable method involves the installation of low troughs (height less than 20 cm, area 0.1 m2, three troughs in each plot) on the soil surface along the slopes, with their outlets connected to tipping bucket - datalogger systems and/or collecting vessels (nutrient flux). If feasible, the troughs should be repositioned within the study plot at regular time intervals (bi-weekly?) to reduce errors on account of the spatial variation of throughfall. Throughfall plots should be located in the vicinity of the automatic rain gauges. The proposed automated tipping bucket system allows for the extraction of single storm events, from which vegetation-specific parameters may be derived for use in interception modelling (Gash or Rutter models). Moisture leaving the tipping bucket system may be collected in appropriate collecting vessels for chemical analysis.
The traditional method using
30-50 roving rain gauges and 30 stemflow gauges to obtain above
and below canopy flux data can be used in the rain forest. In
addition, several trough-datalogger systems should be installed
to sample event data. The throughfall gauges could be distributed
over the rain gauge plots. Water samples can readily be obtained
from the throughfall gauges in the forest. Care should be taken
to avoid sources of biological contamination in the gauges (use
of dark flasks to avoid the growth of algae, frequent replacements
of collecting vessels).
Rainfall interception by
the litter layer
If a dense litter layer is
present, rainfall may be intercepted in it and this, combined
with the interception by the vegetation may constitute of up to
10% of incident rainfall for grass. However, direct measurements
of the litter percolate may be difficult to obtain and it is therefore
suggested to determine the evaporation from the litter layer from
repeated (daily-weekly) measurements of the litter layer mass/moisture
content during dry periods following rainstorms. A simple empirical
model was developed by Waterloo (1994) to use such data to model
rainfall interception by the litter layer from daily rainfall
data. To determine the flushing of nutrients from the litter layer,
several litter percolation gauges can be installed.
The research and modelling strategy demands for several micro-scale catchments for each land use type to be selected for rainfall - runoff measurements. One (or two) of these catchments will be selected for detailed studies, whereas at the other sites measurements will be limited to rainfall and spot measurements of discharge. This requires the installation of fixed staff gauges at the outlet of each catchment, as well as the development of discharge rating curves.
In the catchments selected for detailed studies, fixed discharge measurement structures (concrete flumes) should be build to last for at least several years in view of the possible extension of the study beyond 2003. The construction should preferably take place at the end of the dry season of 1998, when flow conditions are low. Water level recorders should be installed just upstream from each structure in stilling wells. Measurements should be made at a 5-minute interval to capture the stormflow hydrograph. The water level recorders should be complemented by fixed staff gauges, to be used as a reference. It is imperative that backup systems be in place.
In order to calibrate the
discharge measurement structure and to obtain discharge rating
curves for sites where no structures are constructed, current
meters should be available, as well as portable EC meters (salt
dilution measurements at low flows or in very turbulent shallow
Ground water discharge
Significant ground water discharge may occur when soils are deeply weathered as is the case in Rondônia. Ground water flow out of the catchment may be primarily through unconsolidated material in the valley. The quantity may be determined using a groundwater monitoring system (wells and water level logging equipment), providing information on the local gradient, in combination with measurements of the saturated hydraulic conductivity (pumping tests). The dip wells may also be used to obtain samples for the chemical analysis of groundwater. Such samples should be collected at bi-weekly to monthly intervals to capture the seasonal variation. Hydrochemistry (isotopes) may also be used to determine the groundwater contribution to the runoff, or tree transpiration.
The extent of the valley
fill and the depth of the bedrock should be determined by geophysical
measurements to provide the necessary information for quantifying
Lateral subsurface flow
Lateral subsurface flow can
be measured in soil pits using custom-made steel troughs connected
to tipping bucket datalogger systems or to large collecting vessels.
The water should be analysed to obtain the chemical characteristics
of this flow component.
Soil moisture content and
Soil moisture content and
tension measurements will be made at selected rainfall and throughfall
plots in the micro-scale catchments to assess the spatial and
temporal variability. Soil moisture depth profiles, preferably
down to the groundwater level or bedrock, should be measured one
or two times a week with a neutron probe in five access tubes
at each plot. To obtain information with a higher temporal resolution
(5-30 minutes interval), the neutron probe data should be complemented
with a number of TDR/FDR sets, with measurements made in deep
pits at a number of depths in the soil down to the rooting depth
(e.g. at 0.1 m, 0.2 m, 0.3 m, 0.5 m, 0.7 m, 1.0 m, 1.5 m, 2.0
m, 3.0 m, 4.0 m, 5.0 m, 6.0 m, 8.0 m and 10.0 m).
Chemical properties of
Vacuum type ceramic cup or
teflon samplers can be used to extract moisture available for
the plant root system from the micro-pore soil system during wet
periods. Samplers should be installed at different depths in the
soil down to he rooting depth. A fixed pressure (60-70 kPa) should
be applied to ensure that water is derived from the same minimum
pore size, such that seasonal variations may be studied. Samples
may be collected on a weekly basis.
Transpiration by trees and grass can be determined using heat pulse sapflow measurement equipment. Due to the large variation between and within species, at least 8 trees should be monitored. The saplow site should be close to the micrometeorological tower to relate measured evaporation rates with the sapflow rates.
Sapflow measurements can be made in pasture with the Dynamax system using micro-sondes (2-5 mm stem diameter). Again, as many plants as possible should be monitored. An estimate of the combined soil evaporation and transpiration flux can be obtained from soil moisture measurements using the zero flux plane method during dry periods. In addition, information on the stomatal conductivity, which controls transpiration, may be obtained from porometer measurements during the dry and wet seasons.
A different technique for the determination of forest transpiration uses a tracer (Deuterium), injected in the stem of the tree (Calder, 1992; Calder et al., 1992). Moisture evaporated from the leaves is then sampled in plastic bags over a period of time and analysed, until the tracer has been fully recovered. This method requires canopy access.
The depth from which moisture is extracted by roots can be determined by isotope sampling of the xylem sap, provided that isotope profiles in the soil are known (these can be obtained from moisture sampling in deep pits) and the isotope sampled shows a distinct variation with depth.
A porometer may be used to
determine the stomatal conductance of grass and forest leaves
to detect if the values vary in relation to seasonal changes in
soil moisture. However, such measurements are known to vary with
the position in the forest canopy as well.
Soil physical and chemical
A detailed map of relevant
soil properties should be prepared for use as input to the micro-scale
model and for the determination of the nutrient budget. Soil properties
to be determined may include:
The sampling and measurements
should be made on a grid basis. More details on the grid (spacing,
stratified sampling schemes, etc.) and depth of sampling can only
be given when the sites have been selected.
Streamflow chemistry and
Regular water samples may
be collected at the outlet of the basin for the determination
of nutrient concentrations and sediment concentrations. Due to
the short-lived stormflow events, it is necessary to install an
automatic water sampler to collect water samples for chemical
and sediment analysis).
Overland flow and Erosion
Erosion plots (e.g. Wischmeier plots, Gerlach troughs) can be installed to measure overland flow and associated particle transport. Wischmeier plots should contain a tipping bucket flux measurement system and large containers to quantify and collect the runoff. These containers should be emptied on a weekly basis. Gerlach troughs are smaller and should be installed along the slope as a cascade system. Some of these troughs should be connected to a tipping bucket flux measurement system to determine the quantity, as well as the timing of runoff. The rest of the troughs can be connected to totalizers. The measurements should again be made on a weekly basis, but may be made on an event basis (daily) during short campaigns.
Erosion pins can be installed
along a catena to assess the spatial variation of erosion and
deposition. These should be measured twice a year, preferably
at the start and end of the wet season. However, erosion may not
be high enough under forested conditions to be measured this way.
Soil samples for bulk density estimations should be collected
to transform the measurements to soil loss or deposition rates
in kg km2. Splash cups may be used for the assessment
of whether splash erosion is an important sediment delivery process.
Soil cohesion and aggregate stability can be assessed using laboratory
methods. Soil micro-topography and roughness affect ponding and
redistribution of water and should also be determined.
The forest and pasture vegetation will have to be mapped (LAI, ground cover, etc.) and the seasonal variation needs to be assessed to see if the vegetation responds to moisture stress (e.g. pasture) during the dry season.
Light interception techniques
may be used to estimate the LAI (sunfleck ceptometer, LAI-2000
plant canopy analyzer) in forest, whereas such light measurements
may be combined with destructive sampling in pasture. Measurements
should be made at least once a month at different locations to
estimate the spatial and temporal variation.
Annex 1. Literature review
The literature review presented
below has been based on publications obtained from internationally
available sources. Additional information may well be available
in the form of unpublished technical reports, papers, unpublished
data, etc. at various government departments, private companies
or universities. An attempt should be made to find out if such
documentation exist, particularly for the research area in Rondônia.
Catchment water balance
Few catchment water balance
studies have yet been carried out in lowland 'terra firme' rain
forest in the Amazon Basin (Franken, 1979, 1980; Franken and Leopoldo,
1984, 1987; Leopoldo et al., 1985, 1995; Richey et al., 1986)
and none of those in seasonally flooded forest, secondary regrowth,
man-made pasture or crop lands. Furthermore, the majority of the
studies were located near Manaus, in central Amazonia, and there
is little information for areas elsewhere in the Amazon Basin,
including the Rondônia Province. Recently, results have
been published of several catchment studies on the effect of clear-cutting
of rain forest and subsequent land use conversion (Fritsch, 1993),
and of selective logging followed by natural regrowth (Jetten,
1994; Brouwer, 1996), on streamflow and evaporation in Guyana.
A summary of the results of these studies in terms of magnitudes
of water balance components has been provided in Table 1.
Table 1. Summary of forest catchment water balance components as published in selected South American studies. Area is in km2, P is precipitation (mm), Qb is baseflow component (mm), Qq is quickflow component (mm) and 'Loss' represents the sum of evaporation, sub-surface leakage and moisture storage differences (mm).
|Reserva Ducke, Manaus||1.3||2209||650||66||1493+||Leopoldo et al., 1995|
|Reserva Ducke, Manaus||23.5||2089||541*||-||1548||Leopoldo et al., 1985|
|Lake Calado, Manaus||0.23||2870||1562||88||1120||Lesack, 1993|
|Serra do Mar, São Paulo A||0.56||2319||1368||269||682||Fujieda et al., 1997|
|Serra do Mar, São Paulo B||0.37||1941||1004||384||554||Fujieda et al., 1997|
|Obidos+Altamira basins||5.1·106||2153||1014*||-||1139||Matsuyama et al. 1993|
|Ibidem, selectively logged||0.062||2190||820||55||1133||Jetten, 1994|
|French Guyana, (I)||3.2||3676||2148*||-||1528||Roche, 1982|
|French Guyana, (II)||8.4||3697||2269*||-||1437||Roche, 1982|
|French Guyana, (III)||12.4||3751||2307*||-||1444||Roche, 1982|
|French Guyana, A||0.013||3423||15||650||2758||Fritsch, 1993|
|French Guyana, B||0.016||3267||20||595||2952||Fritsch, 1993|
|French Guyana, C||0.016||3265||92||239||2933||Fritsch, 1993|
|French Guyana, D||0.014||3257||31||480||2746||Fritsch, 1993|
|French Guyana, E||0.016||3350||8||426||2916||Fritsch, 1993|
|French Guyana, F||0.014||3102||435||1058||1609||Fritsch, 1993|
|French Guyana, G||0.015||3173||423||947||1803||Fritsch, 1993|
|French Guyana, H||0.010||3165||489||1088||1588||Fritsch, 1993|
|French Guyana, I||0.011||3285||96||364||2825||Fritsch, 1993|
|French Guyana J||0.014||3219||83||748||2388||Fritsch, 1993|
actual evaporation, *Total surface/subsurface runoff
Franken and Leopoldo (1995) and Leopoldo et al. (1985) published rainfall and streamflow totals from the Bacia Modelo (Taruma Açu) and Barro Branco catchments in the Reserva Ducke 'terra firme' rain forest near Manaus, Central Amazonia. The topography of these catchments is gently undulating and the soil is a deep Oxisol. Measurements were made during a rather dry period, with annual rainfalls of 2089 mm (1980) and 2312 mm, 2365 mm and 1949 mm (1981-1983), respectively, as compared to a long-term mean of 2410 mm y-1 (Lesack, 1993a). The rainfall regime is seasonal with a 2-3 month dry season during which rainfall is generally less than 100 mm. In spite of the dry conditions, evaporation rates as calculated using the water balance method were high in comparison to that obtained by Shuttleworth (1988) for the same area using two years (1983-1985, average rainfall 2636 mm y-1) of micro-meteorological measurements (ET= 1320 mm y-1). This lead Bruijnzeel (1990) to suggest that a significant part of the water may have left the catchment as ungauged sub-surface flow through the deep Oxisols. Lateral flow (overland flow) has rarely been observed on the hillslopes as the deep Oxisols are highly permeable (Nortcliff and Thornes, 1981). Baseflow is therefore the dominant streamflow component, accounting for over 90% of the runoff in the Barro Branco catchment (Leopoldo et al., 1995).
The most complete water balance study up to date is that by Lesack (1993a), who measured all components of the water balance, including sub-surface outflow, over a year in an undisturbed 'terra firme' rain forest catchment near Lake Calado, Central Amazonia (deep Oxisol). At this site, which experienced a 1 in 10 years wet year with 2870 mm of rainfall, total surface and sub-surface runoff amounted to 1650 mm and 42 mm, respectively, implying an annual evaporation total of 1120 mm. Stormflow runoff (173 storms) amounted to only 5% of the total runoff, and baseflow was therefore again the dominant streamflow component. The stormflow to rainfall ratio (which may serve as an estimate of the minimum contributing area) amounted to 3%, which suggests that most of the stormflow was generated in the channel floodplain, presumably through direct channel precipitation and saturation overland flow.
South of the Amazon Basin, in the humid subtropical area of Serra do Mar of the São Paulo State, Brazil, Fujieda et al. (1997) have studied rainfall, rainfall interception, hillslope processes and streamflow in two small catchments (A, B) covered with montane rain forest over a 10 year period. The bedrock consists of gneiss and crystalline schist and the rather shallow soils (depth = 1.5 m) are sandy clay Oxisols with a permeable A-horizon. Slopes were steep (32-46%). The climate is characterised by a short dry season with rainfall less than 100 mm month-1 in the period June - August. The water balance calculations yielded low mean annual evaporation totals, which were explained by the altitude of the sites (1000-1200 m a.s.l.) and the fact that the region is often covered by dense afternoon fog from the Atlantic ocean during the dry season. If cloud stripping occurs, the actual moisture input could well have been larger than that of rainfall alone, suggesting that evaporation rates may have been underestimated. Runoff occurred mainly as baseflow, with stormflow constituting 16% and 28% of the total runoff. Surface runoff on the hillslopes amounted to a low 0.6% of the annual rainfall. The dominant runoff producing mechanism was vertical infiltration, followed by lateral groundwater flow to the stream. As in Central Amazonia, stormflow was generated mainly by rainfall on the riparian zone, which covered 5% (A) to 10% (B) of the catchment area.
Jetten (1994) used runoff and soil data to model evaporation of a small catchment with sandy soils (Albic Arenosols) at Mabura Hill, Guyana. The climate was humid, with rainfall at the nearby Great Falls climate station exceeding 100 mm month-1 throughout the year. He obtained evaporation values similar to those of Lesack (1996) and Shuttleworth et al. (1988) and again observed that baseflow was the dominant streamflow component, with quickflow amounting to about 6% of the total flow.
In the coastal zone of French Guyana, Fritsch (1993) measured rainfall and runoff over a two year calibration period from 11 very small paired rain forest catchments on mica-schist bedrock, overlain by permeable ferralitic soils. Losses, as calculated form the water balance, varied between 1588 mm y-1 and 2952 mm y-1, which indicates that there is significant (ungauged) groundwater outflow in most of these catchments. A large part of the streamflow occurred as stormflow in all catchments. Two stormflow generating mechanisms were recognised. For catchments with highly permeable soils, vertical or lateral drainage over a less permeable B-horizon was dominant, and saturation overland flow was the main stormflow generating mechanism. In these catchments, water tables never reached the valley floor surface, and stormflow amounted to 10-26% of rainfall. In catchments G, F and H, where soil conditions were different, a rapid rise of the water table above the valley floor surface occurred during rainfall and the resulting overland flow contributed also to stormflow, resulting in higher percentages of rainfall leaving the catchment as stormflow (30-34%).
After the calibration period, the vegetation in six catchments was clear-cut, resulting in increases in runoff varying between 166 mm y-1 (H) and 299 mm y-1 (C) for 'bare soil' conditions. Subsequent conversion to pasture and grass (Digitaria swazilandensis) resulted in a runoff increase of 59% in the first year and 27% in the fifth year, as compared to the forested situation. The highest increase (catchment C, 73% in the first year, decreasing to 46% in the fifth year) was observed for conversion to a grapefruit orchard (Fritsch, 1993). Only slight increases in runoff were observed for traditional slash and burn and selective logging (no clear-cutting) followed by regrowth (26-30%).
Measurements of rainfall and runoff were made in the larger Gregoire I-III catchments (Ultisols) in high rainfall areas in French Guyana by Roche (1982). These catchments were presumably watertight (Bruijnzeel, 1990) and the evaporation estimates are at the higher end of the range obtained for Amazonian rain forest.
At Tonka, Surinam, Poels measured rainfall and discharge from a small rain forest catchment in slightly undulating terrain with sandy-loam soils over a 4 years and nine months period. The length of the dry season varied between two to five months a year. Assuming that sub-surface leakage from the catchment was negligible, Poels arrived at the rather high evaporation value given in Table 1, which amounted to 94% of the potential evaporation. Surface and lateral flow were only important in valleys and footslopes, and groundwater flow was again found to be the main contributor to the streamflow.
On the macro-scale, Matsuyama et al. (1993) estimated rainfall over the Amazon Basin from 34 precipitation stations and measured discharge at the Obidos (Amazonas river) and Altamira (Xingu river) stations, which combined were considered to give a good approximation for the discharge of the whole Amazon Basin. They obtained a macro-scale evaporation value which, considering the possible errors in rainfall and discharge on this scale, was similar to those obtained by Lesack (1993a) and Shuttleworth (1988).
Evaporation and rainfall
In addition to the 'losses' obtained from the catchment water balance studies presented in Table 1, which in most cases include evaporation as well as moisture storage changes and subsurface losses, estimates of evaporation for 'terra firme' rain forest and pasture have been obtained from micro-meteorological studies, soil moisture studies and through modelling.
Shuttleworth (1988) used micro-meteorological measurements collected for the ARME project in the Reserva Ducke rain forest in Central Amazonia over a 25 month period, in combination with modelling, to obtain an annual evaporation rate of 1320 mm, of which 328 mm was lost through rainfall interception and 992 mm through forest transpiration.
Evaporation by rain forest and pasture was modelled by Kabat et al. (1997) using data collected by the ARME and ABRACOS projects in the Reservas Jaru (West Amazonia), Ducke (Central Amazonia) and Vale (East Amazonia) rain forests and at the Nossa Senhora (Ji-Paraná, West Amazonia), Dimona (Manaus, Central Amazonia) and Boa Sorte (Marabá, East Amazonia) pasture ranches. In the pastures at Ji-Paraná and Manaus, the dominant grass species were Brachiaria humidicola and Brachiaria decumbens and there was little invasion by shrubs. The dominant grass species at the degraded pasture at Marabá was Panicum maximum, and there was some invasion of woody shrubs and small palms. Predicted annual rain forest evaporation ranged from 1155 mm at Reserva Ducke to 1372 mm and 1384 mm at Reserva Vale and Jaru, respectively. Annual evaporation from pasture was on average 323 mm lower at 915 mm (Dimona), 1004 mm (Boa Sorte) and 1024 mm (Nossa Senhora), which was attributed to the higher albedo of grass, as well as to the occurrence of water stress during the dry season. Modelled interception loss in rain forest amounted to 18-20% of rainfall. Interception losses in pasture were also modelled, but details were not reported. Water balance calculations using the evaporation totals given above indicate that runoff would increase 41-48% after conversion of these forests to pasture (assuming that there is no significant loss in infiltration capacity).
Based on soil moisture measurements down to a depth of 3.6 m and storage calculations, Hodnett et al. (1996) obtained minimum dry season evaporation rates for the same ABRACOS pasture sites in East, Central and West Amazonia. The seasonal rainfall distribution at the sites varies considerably. The Pará site has the longest dry season (5 months), followed by the Rondônia site (4 months) and the Manaus site, where the dry season is short with only two months with rainfall less than 100 mm mo-1. Minimum dry season evaporation ranged from a low 0.6 mm d-1 at Marabà (Panicum maximum) where the grass senesced almost completely, to 1.2 mm d-1 and 2.5 mm d-1 at Manaus and Ji-Paraná (Brachiaria sp.), respectively. The difference between the sites was attributed to differences in the plant available water capacity in the 0-2 m soil layer, as well as to possible differences in the response to drought between the grass species. No estimates could be obtained for adjacent forest sites as water was extracted by the trees from below the depth of the soil moisture measurements.
In the Pará State, Nepstad et al. (1994) obtained dry season evaporation rates of 3.6 mm d-1 for rain forest and 3.0 mm for a degraded, woody pasture from a soil water balance over a 5.5 month dry period with only 95 mm of rainfall. The relatively high pasture evaporation rates may be due to the presence of (deeper rooting) woody shrubs and treelets, which make up for 50% of the leaf area.
The results presented above were based on relatively short measurement periods (less than 3 years) and the inter-annual variation of rainfall could therefore not be taken into account. Although water stress has not been observed for the deep-rooting rain forest, it is yet unknown if such stress may occur in areas with shallow soils, during particularly long dry seasons, or in areas where the rooting depth is restricted for some reason or another.
A summary of rainfall interception and stemflow estimates is given in Table 2. Studies of rainfall interception in the Amazon basin have also been concentrated in the 'terra firme' rain forests in central Amazonia, although some work has recently been carried out in Rondônia (Ubarana, 1996). Franken et al. (1982a,b) estimated rain forest canopy interception losses at 18.2-22.0%. Lloyd and Marques Filho (1988) and Lloyd et al. (1988) obtained much lower values of 7-9% of rainfall for the Ducke rain forest reserve and argued that the difference was due to the sampling method used (fixed versus roving gauges). Using the same methodology as Lloyd and Marques (1988), Ubarana obtained somewhat higher values for forests in the Rondônia and Pará States.
In San Carlos de Rio Negro,
Venezuela, throughfall amounted to 87% of the rainfall over a
two year period, but the rainfall interception loss was low at
5% due to high stemflow (Jordan and Heuveldop, 1981). Stemflow
observations in rain forest suggest that this component amounts
generally to less than 2% of incident rainfall and the exceptionally
high value of 8% obtained by Jordan and Heuveldop (1981) was attributed
to the high proportion of small trees at their site. A very high
stemflow value of 39% was observed for a sugar cane crop. Measured
interception losses for Amazonian rain forest compare well with
those observed for humid tropical forests elsewhere in the world
Table 2. Summary of stemflow and crown interception losses expressed as a percentage of above-canopy rainfall for various South American studies. AM: Amazonia, MG: Mato Grosso, RJ: Rio de Janeiro, SP: São Paolo.
|Location||Vegetation type||Stemflow||Interception loss||Author (rainfall P)|
|Manaus (AM)||Rain forest||-||12%||Nortcliff & Thornes 1981b|
|Manaus (AM)||Rain forest||0.3%||22%||Franken et al., 1982a|
|Manaus (AM)||Rain forest||-||18%||Franken et al., 1982b|
|Manaus (AM)||Rain forest||-||26%||Leopoldo et al., 1985|
|Manaus (AM)||Rain forest||1.8%||7%||Lloyd & Marques, 1988|
|Manaus (AM)||Rain forest||1.9%||9%||Lloyd, 1988|
|Jaru (Rondônia)||Rain forest||1.4%||12%||Ubarana, 1996|
|Vale (Pará)||Rain forest||0.8%||13%||Ubarana, 1996|
|Andes, Peru||Rain forest||?||?||Elsenbeer et al., 1994|
|Viçosa (MG)||Natural forest||0.2%||12%||Castro et al., 1983|
|Rio de Janeiro (RJ)||Reforested area||-||17%||Coelho Netto et al., 1986|
|Cunha (SP)||Natural forest||1.1%||16%||Fujieda et al., 1997|
|S. Carlos, Venezuela||Rain forest||8%||5%||Jordan & Heuveldop, 1981|
|Andes, Colombia||Andean forest||-||11%*,1||Vis, 1986 (P= 2123 mm)|
|Andes, Colombia||Sub-Andean for.||-||15%*||Vis, 1986 (P= 2779 mm)|
|Andes, Colombia||Sub-Andean for.||-||25%*||Vis, 1986 (P= 3968 mm)|
|Andes, Colombia||Lowland forest||-||22%*||Vis, 1986 (P= 673 mm)|
|Mabura, Guyana||Rain forest||-||16%||Brouwer, 1996|
|Mabura, Guyana||Sel. Logged||-||15%||Brouwer, 1996|
|São Moronel (SP)||Cerrado||2.9%||17%||Leopoldo & Conte, 1985|
|Botucatu (SP)||Sugar cane||39%||4%||Leopoldo et al., 1981|
Stemflow not measured and assumed negligible. 1 Possible
moisture inputs from fog stripping not included.
Rainfall interception is
usually ignored in modelling of the hydrology of short crops and
pasture. No data have yet been published on the rainfall interception
of Amazonian pasture, nor for pasture elsewhere in the humid tropics.
The nearest is a study by Leopoldo (1981) on sugar cane, who arrived
at a interception loss of 4% of rainfall. A similar value was
calculated using Gash's analytical model for a tall natural grassland
site (Pennisetum polystachyon) in Fiji (4% in canopy, 7%
in litter layer; Waterloo, 1994). Rainfall interception may well
be lower in the much less dense grass vegetation of Amazonia.
However, because rainfall interception may be in the same order
as predicted decreases in rainfall following land use conversion,
I suggest that measurements of grass (and litter layer) interception
be made to see whether this component of the hydrological cycle
can really be ignored.
Hillslope processes and
Northcliff et al. (1979) and Nortcliff and Thornes (1981) concluded from a combination of soil moisture tension and hydraulic conductivity measurements on a hillslope in the Barro Branco catchment that lateral flow (overland flow and sub-surface flow) through the permeable Oxisol soil was limited under normal soil moisture conditions. The direct contribution of the hillslope was therefore mainly to the baseflow component of the stream hydrograph. Stormflow was generated through direct channel precipitation and saturation overland flow in the floodplain, where the shallow groundwater table would rise rapidly in response to rainfall.
Hodnett et al. (1997) observed a similar decoupling of the hillslope hydrological pathways and stormflow dynamics from soil moisture and groundwater table measurements at the Fazenda Dimona cattle ranch, about 100 km north of Manaus. The soils on the slopes were permeable Oxisols, with a high clay content, whereas a sandy soil was found in the well-defined floodplain area. Under normal rainfall conditions, a high water table in the valley was maintained by groundwater flow from the hillslope. Stormflow was generated in the valley through direct channel precipitation and saturated overland flow. During a dry year, rainfall was not sufficient to replenish the soil moisture storage on the hillslope and deep drainage to the valley became limited. This caused lower groundwater levels in the valley and a subsequent decrease in the contribution of stormflow to the total stream runoff (from 5% when the water table was near the surface to 2-3% when the groundwater table was deeper).
The lack of lateral surface or sub-surface flow on the hillslopes, the generation of stormflow in the riparian zone through direct precipitation and saturated overland flow, and the dominant role of baseflow, were also observed by Fujieda et al. (1997), Jetten (1994), Poels (1987) and Fritsch (1993) in the catchment studies quoted above, which were all made on undulating terrain with rather permeable soils.
Using small runoff plots, Thornes et al. (1990) determined the surface runoff component for top slope, mid slope and bottom slope positions under rain forest, partially cleared forest and bare soil conditions on Maracá Island in the Rio Uraricoera, Roraima State, North Brazil, over a 300-days period. The soil texture changed with depth from coarse sandy loam A-horizon to a dense (gravelly) clay B-horizon, which was underlain by a lateritic layer (at about 1 m depth) in the mid slope positions. Soil disturbance was kept to a minimum by the manual extraction of all vegetation above breast height in the partially cleared plots, and that of all vegetation and litter in the completely cleared plots. Measured surface runoff amounted to about 6% of rainfall in the forest and partially cleared plots, and to 16% of rainfall for bare soil conditions. In the partially and completely cleared plots, runoff coefficients were highest for the mid slope positions (23% and 12%) and lowest for the top slope positions (8% for both treatments), respectively. Under forest, runoff increased gradually with the position in a down slope direction. Based on the available evidence, Thornes at al. (1990) concluded that runoff was controlled more by the treatment than by the position on the hillslope.
In contrast to the studies presented above, Elsenbeer and Lack (1996) and Elsenbeer et al. (1995) observed that, for a much less permeable ultisol in a small Amazonian headwater catchment in the Peruvian Andes (La Cuenca, 0.01 km2), a significant proportion of the flow on the hillslope occurred as lateral flow, either as sub-surface flow or as overland/pipe flow. At this site, the contribution of stormflow, generated by overland flow on he hillslopes to the total runoff may therefore be considered of much more importance.
This stresses the importance
of variations in soil permeability on the hillslope runoff generation
processes and the streamflow dynamics in the Amazon basin. It
also shows that a full range of runoff generating mechanisms is
likely to occur in the Amazon basin and that models based on data
from Central Amazonia alone are not likely to predict the rainfall-runoff
response of catchments with less permeable or shallow soils and
steeper slopes accurately.
Spatial and temporal variations
in soil moisture
Nepstad et al. (1994) used a soil water balance approach to determine the hydrological role of deep penetrating roots in rain forest, degraded pasture (Panicum maximum, Brachiaria humidicola, and woody vegetation) and managed pasture in the Pará State in Eastern Brazil. The climate is strongly seasonal with a 5.5-months dry season, during which only 95 mm of rain was recorded during their study. Annual rainfall averaged 1750 mm y-1. Roots were found down to depths of 8 m in the managed pasture, 12 m in the degraded pasture and 18 m in the rain forest. The rain forest leaf canopy was reduced by 16% during the dry season, and transpiration was sustained by the extraction of soil moisture from depths of more than 8 m. In contrast, the degraded pasture leaf canopy was reduced by 68% during the dry season and that of the managed pasture by 100%. Plant available water in the upper 8 m of the soil declined by 510 mm for rain forest and by 410 mm for the degraded pasture, with the 2-8 m soil layer supplying over 75% of the moisture extracted from the soil in both ecosystems. The lower moisture depletion at the pasture site indicates that the ecosystem can store less rainfall than the rain forest ecosystem. Seepage to groundwater may therefore be higher, leading to higher dry season baseflow levels in the rivers.
As part of the ABRACOS project, Hodnett et al. (1995, 1996a) studied the long-term soil water storage behaviour and water uptake in the 0-2 m and 2-3.6 m soil layers under forest and pasture at the ABRACOS sites from weekly neutron probe soil moisture measurements over a two to three years period. Large seasonal soil water storage variations were observed between the sites, which could be related to soil properties, vegetation cover, climate and water table response. Seasonal changes of moisture content in the 0-2 m soil layer were smallest at Manaus, where frequent rainfall and a short dry season gave the vegetation less opportunity to extract all moisture before re-wetting of the soil occurred, and where the plant available water capacity of the soil was lowest. Here, the seasonal change in storage amounted to 200 mm for the 0-3.6 m layer under forest and there was a relatively small difference between the seasonal storage change under forest and pasture (154 mm versus 132 mm for the 0-2 m layer). The duration of wet soil conditions varied between 2 months for a dry year to 6-7 months for normal years. Seasonal soil water changes in the 0-2 m soil layer and differences between forest and pasture were much larger at the Marabá and Ji-Paraná sites, with maximum seasonal changes of 263 mm and 262 mm observed for pasture and 483 mm and 365 mm for forest, respectively. At these sites, the maximum seasonal variation in the 0-3.6 m soil layer varied from 376-450 mm for pasture and 724-701 mm for rain forest, respectively. The storage changes at the Ji-Paraná site were affected by groundwater level fluctuations and the values presented above include saturated drainage as well as evaporation. At the Marabá forest site, the profile did not wet up to the same moisture level each year, suggesting that during dry years the subsoil may not be wetted to the maximum moisture storage. At all sites, the same minimum water content was observed in the 0-2 m layer each dry season, indicating that all water was extracted from this layer by the vegetation. The extraction limit was not reached in the 2-3.6 m layers, particularly at the pasture sites. Based on these observations, and using an ET of 3.5 mm d-1 (no water stress assumed), Hodnett et al. (1996a) estimated that 38% to 62% of the water used to sustain dry season transpiration was extracted from levels below 3.6 m. These observations confirm the observation of Nepstad et al. (1994) that a significant proportion of the soil moisture is extracted from depths below 2 m.
In Maracá Island,
North Brazil, Nortcliff et al. (1990) observed that soil moisture
tension, measured in three plots with different hillslope positions
(top slope, mid slope and bottom slope) and with different treatments
(forest, partially and completely cleared conditions), was controlled
by hillslope position (top slope > mid slope > bottom slope),
rather than by treatment during wet periods. The effect of treatment
on the soil moisture tension status, however, was more important
during dry periods (forest > partially cleared > bare soil
conditions). Nortcliff and Thornes (1981) observed that the soils
in the Barro Branco catchment were close to saturation during
the wet season. During the dry season, the lowest tensions were
again observed in the floodplain, whereas the highest tensions
(>65 kPa at a depth of 125 cm) occurred in the mid slope position.
Soil hydraulic conductivity and moisture retention
Nortcliff and Thornes (1989) published high saturated hydraulic conductivities, determined from small core samples, of 6500-13400 mm h-1 for the top soil, decreasing to 100-4600 mm h-1 and 140-300 mm h-1 at 33 cm and 1 m depth, respectively.
Elsenbeer et al. (1992) obtained much lower values ranging from 32-686 mm h-1 (median values) for surface soils and 0.0004-1.8 mm h-1 for the soil at a depth of 0.4 m in a small headwater catchment in Peruvian Andes. They also observed a link between the hydraulic conductivity and the topographic position. The hydraulic conductivity of the surface soil was lowest on side-slopes with active surficial processes (decreasing from 46 mm h-1 to 0.1 mm h-1 over a depth of 0.4 m), whereas that of level areas was the same irrespective of the relative elevation above the valley floor (335-462 mm h-1, decreasing to 0.01-0.1 mm h-1). The decrease with depth was most pronounced on interfluves, which had the highest surface permeability (from 550 mm h-1 to 0.015 mm h-1).
Jetten (1994) determined the hydraulic conductivity and moisture retention characteristics of four different undisturbed and disturbed (skid trails) sandy to sandy loam surface soils (Albic Arenosols, Ferralic Arenosols, Haplic ferralsols and Haplic Acrisols) in Guyana. The hydraulic conductivity, as determined from small cores in the laboratory, ranged from 55-2580 mm h-1 for undisturbed Xanthic Ferralsols, 307-2950 mm h-1 for Albic Arenosols, and 5-231 mm h-1 for Haplic Ferralsols. Significant changes in the van Genuchten parameters were observed upon disturbance of the soil by skidders for log extraction.
Soil physical properties
for Oxisols under pasture in Central Amazonia have been determined
by Tomasella and Hodnett (1996) using a ring permeameter (area
314 cm2) down to depths of 1.35 m. The mean saturated
hydraulic conductivity of the upper 1 m of the profile was high
(17-66 mm h-1), due to the presence of macropores,
both in forest and pasture soils. Below this depth, the saturated
hydraulic conductivity becomes more dominated by the particle
size distribution as macro- and mesopore effects become negligible.
To measure unsaturated hydraulic conductivity, the Instantaneous
Profile Method (IPM, Watson, 1966) was used and the results were
used to optimise parameters for the van Genuchten equation. The
optimised parameters were different from those which would have
been selected on a soil textural basis from published parameters
for temperate soils. They concluded that for Oxisol soils in the
Amazon Basin, which have a high clay content but where the pore
size distribution is strongly affected by the forming of iron
aggregates, texture seems not to be a good indicator for selecting
van Genuchten parameters from temperate data sets. This implies
that an extended set of van Genuchten parameters needs to be developed
for different soils in the Amazon basin, preferably for depths
down to the rooting depth.
Erosion, sediment and nutrient
fluxes in water
The only erosion study available for 'terra firme' Amazonian rain forest and partially and completely cleared areas is that of Nortcliff et al. (1990) and Ross et al. (1990) in Maracá Island, North Brazil. They observed a great contrast between soil losses in rain forest and partially cleared plots on the one hand, and those in completely cleared plots on the other hand. Soil losses from the forested and partially cleared plots were much less than 1000 t km-2 y-1. The highest loss occurred on the bottom slopes. In the cleared area, soil losses on the mid and bottom slope positions were 7400 t km-2 y-1 and 4500 t km-2 y-1, respectively, whereas the loss from the top slope plot was only slightly higher than those measured in the forested and partially cleared plots at 900 t km-2 y-1. It should be stressed that these measurements were made on undisturbed soils and that the soil infiltration capacity was probably not affected by the treatment. Much higher soil losses may be expected when mechanical clearing is practised, during which roads and tracks are constructed where the soil is severely disturbed and compacted. The fact that there was little difference in erosion between the rain forest plots and the partially cleared plots, where the undergrowth and litter layer remained intact, indicates that the undergrowth and litter layer is very important in providing a level of protection to erosion which is not very different from that of undisturbed rain forest.
Richey et al. (1986) measured discharge and sediment concentrations at 18 locations along the Solimoes River, the Amazon main stem. Most of the suspended sediment was fine-grained material with a particle size less than 0.063 mm. Sediment concentrations ranged from a low 5-10 mg l-1 in the Rio Negro at discharges (Q) of 280 and 80 m3 s-1, respectively, to 211-513 mg l-1 in the head waters of the Solimoes at San Antonio do Ica (Q= 389-559 m3 s-1) and 93-385 mg l-1 at the downstream end of the river at Obidos near Belém (Q= 917-1770 m3 s-1).
Several studies of the water quality, nutrient inputs and outputs of forested ecosystems in humid tropical parts of South America have been carried out over the past decades. A general overview of the chemical composition of the water in lakes and large rivers in Central Amazonia has been presented by Furch (1984). Measured ion concentrations in Amazonian lakes and rivers were low as compared to the world averages. This was supported by the low electrical conductivity of the streams, which ranged from less than 10 S cm-1 for the Rio Negro, draining the northern peripheral region, and several forest streams in Central Amazonia, to up to 60 S cm-1 for the Solimoes River, draining the western peripheral region. The variation in the pH was considerable and could be related to the ion content of the waters. Waters with a low ion concentrations were acidic, with pH values ranging from 4.5 in the forest streams to 5.1 in the Rio Negro, whereas those waters with higher ion concentrations were near-neutral, as indicated by the pH of 6.9 observed for the Solimoes River. The HCO3- and Ca2+ ions were dominant in waters with relatively high ion concentrations and these were therefore classified as belonging to the carbonate water type. In contrast, alkali metal ions (Na, K) were dominant in the waters with low ion concentrations, reflecting the chemical composition of rainfall.
A summary of inputs and outputs
of nutrients in micro-scale catchments in humid tropical parts
of South America is given in Table 3. Brinkmann (1983, 1985) determined
the nutrient output of the Reserva Ducke, whereas Franken and
Leopoldo (1984) determined that of the Barro Branco catchment.
Although these nutrient fluxes were measured in the same area,
large differences were found in the inputs and outputs of Na,
K and Cl. Lesack and Melack (1991, 1993b) have rigourously studied
the rainfall chemistry (dry and wet deposition) and nutrient export
from a rain forest catchment near Lake Calado in Central Amazonia
and their input and output values are presumably the most reliable,
suggesting that rainfall and streamflow samples may have been
contaminated in the Reserva Ducke and Barro Branco studies.
Table 3. Nutrient in- and outputs from selected forested ecosystems in the Amazon Basin. All values in kg km-2.
|Reserva Ducke1, Brazil||Barro Branco2, Brazil||Lake Calado3, Brazil||Tonka, Surinam4||San Carlos, Venezuela5||Mabura Hill, Guyana6|
Brinkmann (1983, 1985), 2 Franken & Leopoldo (1984),
3 Lesack (1991, 1993), 4 Poels (1987), 5
Jordan (1982), 6 Brouwer (1996)
Nutrient fluxes in rainfall and streamflow were generally higher in the Venezuelan, Surinam and Guyanese studies, which may reflect the proximity to the coast, differences in soil and climate, as well as errors in the magnitude of the water balance components (e.g. runoff) and possibly errors in the sampling and analysis of these very dilute waters (Bruijnzeel, 1990).
Nortcliff and Thornes (1978,
1989) showed that the nutrient output in streamwater does not
necessarily reflect the soil nutrient status, because the supply
to groundwater and streamflow is mainly from drainage through
the macropore system, which has significantly lower concentrations
than held water in the micropore system. This supports the findings
of Gibbs (1967, 1972), who studied factors controlling the salinity,
concentration and composition of suspended solids in the Amazon
river system using data from 16 stations in the basin. He found
that relief accounted for 78% of the variance in salinity, with
the primary process in the lowlands being a dilution of river
water coming from the Andes, where both physical and chemical
surface weathering processes are active. At the mouth of the Amazon
river 86% of the dissolved solids and 82% of the suspended sediment
was derived from the mountainous Andes region, which covers only
12% of the basin area. This implies that chemical weathering processes
in the deep lowland soil and bedrock do not contribute much to
the salinity of the river system, partly because water is moving
so fast through the system (macro-pore flow) that there is little
time to react with the rock, and partly because of uptake by the
dense vegetation preventing the removal of weathering products
in river water (Drever, 1982).
Bruijnzeel, L.A., 1990. Hydrology of Moist Tropical Forests and Effects of Conversion: A State-of-Knowledge Review. UNESCO, Paris; Vrije Universiteit Amsterdam. 200 p.
Brouwer, L.C., 1996. Nutrient Cycling in Pristine and Logged Tropical Rain Forest. PhD thesis.University of Utrecht, The Netherlands. Tropenbos Guyana Series 1. 224 p.
Brinkmann, W.L.F., 1983. Nutrient bakance of a central Amazonian rainforest: comparison of natural and man-managed systems. IAHS Publ. 140: 153-163.
Brinkmann, W.L.F., 1985. Studies on the hydrobiochemistry of a tropical lowland forest system. Geojournal 11.1: 89-101.
Drever, J.I., 1982. The Geochemistry of Natural Waters. Prentice-Hall, Inc., USA. 388 p.
Calder, I.R., 1992. Development of the Deuterium tracer method for the estimation of transpiration rates and transpiration parameters of trees. In: Calder, I.R., Hall, R.L., and Adlard, P.G. (Eds.). Growth and water use of forest plantations. John Wiley and Sons. p.180-195.
Calder, I.R., Swaminath, M.H., Kariyappa, G.S., Srinivasalu, N.V., Srinivasa Murthy, K.V. and Mumtaz, J., 1992. Measurements of transpiration from Eucalyptus plantations, India, using Deuterium tracing. In: Calder, I.R., Hall, R.L., and Adlard, P.G. (Eds.). Growth and water use of forest plantations. John Wiley and Sons. p.196-215.
Dunne, T,, Vörösmarty, C.J., Richey, J.E., Becker, A., Bonell, M., Hodnett, M.G., Marengo, J.A., Martinelli, L.A., Mellack, J., O'Connell, P.E., Tomasella, J. and Victoria, R.L., 1997. Land Surface Hydrology and Water Chemistry. Chapter 5 in: The Large Scale Biosphere Atmosphere Experiment in Amazonia (LBA): Integrated Science Plan. (Draft Version).
Elsenbeer, H., Cassel, K. and Castro, J., 1992. Spatial Analysis of Soil Hydraulic Conductivity in a Tropical Rain Forest Catchment. Water Resources Research 28(12): 3201-3214.
Elsenbeer, H., Cassel, D.K. and Zuñiga, L., 1994. Throughfall in the terra firme forest of western Amazonia. Journal of Hydrology New Zealand 32: 30-40 or 430-440.
Elsenbeer, H. and Lack, A., 1996. Hydrological Pathways and Water Chemistry in Amazonian Rain Forests. In: Anderson, M.G. and Brooks, S.M. Advances in Hillslope Processes 2. Chapter 43. John Wiley and Sons Ltd. Chichester, England. Pp. 939-959.
Elsenbeer, H., Lack, A. and Cassel, D.K., 1996. The stormflow chemistry at La Cuenca, Western Amazonia. Interciencia 21: 1-7.
Gibbs, R.J., 1967. The geochemistry of the Amazon river system. 1. The factors that control the salinity and the composition and concentration of suspended solids. Geological Society of America Bulletin 78: 1203-1232.
Gibbs, R.J., 1972. Water chemistry of the Amazon River. Geochim. Cosmochim Acta 36: 1061-1066.
Franken , W., Leopoldo, P.R., Matsui, E. and Ribeirom, N.G., 1982a. Intercepção das precipitações em floresta amazônica de terra firme. Acta Amazônica 12, 15-22.
Franken , W., Leopoldo, P.R., Matsui, E. and Ribeirom, N.G., 1982b. Estudo de Intercepção da àqua de chuva em cobertura florestal Amazônica do typo terra firme. Acta Amazônica 12, 327-331.
Franken, W. and Leopoldo, P.R., 1984. Hydrology of catchment areas of Central-Amazonian Rain Forest Streams. In: H. Sioli (Editor). The Amazon: Limnology and Landscape Ecology of a Mighty Tropical River and its Basin. Junk. Pp. 501--519.
Fritsch, J.M., 1993. The hydrological effects of clearing tropical rainforest and the implementation of alternative land uses. IAHS Publication 216: 53-66.
Hemming, J. (Ed.), 1985a. Change in the Amazon Basin: Vol. 1. Man's Impact on Forests and Rivers. Manchester University Press. 222 p.
Hemming, J. (Ed.) 1985b. Change in the Amazon Basin: Vol. 2. The Frontier after a Decade of Colonisation. Manchester University Press. 295 p.
Hodnett, M.G., Vendrame, I., Oyama, O.D., Marques Filho, A. de O. and Tomasella, J., 1997. Soil Water Storage and Groundwater Behaviour in a Catenary Sequence beneath Forests in Central Amazonia. II. Floodplain Water Table Behaviour and Implications for Streamflow Generation. Submitted to ¼
Jetten, V., 1994. Modelling the Effects of Logging on the Water Balance of a Tropical Rain Forest. University of Utrecht PhD. Dissertation. 196 p.
Kabat, P., Dolman, A.J., Ashby, M., Gash, J.C., Wright, I., Culf, A., Calvet, J.C., Delire, C., Noilhan, J., Jochum, A., Silva Dias, M.A., Fisch, G.A., Santos Alvala, R.C., Nobre, A., Prince, S.D., and Steininger, M., 1997. The Large Scale Biosphere-Atmosphere Experiment in Amazonia (LBA). Use of Integrated Modelling for Experimental Design. Final Report CEC Environment & Climate Programme Project.
Kabat, P. et al., 1997. First LBA Reconnaissance flight, September 1996. SC-DLO report. Wageningen. 90 p.
LBA Science Planning Group (1996). The Large Scale Biosphere-Atmosphere Experiment in Amazonia: Concise Experimental Plan. INPE, Cachoeira Paulista, Brazil. 44 p.
Leopoldo, P.R., and Pàdua Sousa, A. de and Filho, S.T., 1981. Interceptação da àqua de chuva em cultura de cana-de-açùcar. Brasil Açareiro 6: 9-16.
Leopoldo P.R., Franken, W. and Matsui, E., 1985. Hydrological Aspects of the Tropical Rainforest in the central Amazon. In: J. Hemming (Ed.). Change in the Amazon Basin: Vol. 1. Man's Impact on Forests and Rivers. Manchester University Press, Chapter 7.
Leopoldo, P.R., Franken, W. and Villa Nova, N.A., 1995. Real Evapo-transpiration and Transpiration through a Tropical Rainforest in Central Amazonia as Estimated by the Water balance Method. Forest Ecology and Management 73: 185-195.
Lesack, L.F.W., 1993a. Water Balance and Hydrologic Characteristics of a Rain Forest Catchment in the Central Amazon Basin. Water Resources Research 29(3): 759-773.
Lesack, L.F.W., 1993b. Export of Nutrients and Major Ionic Solutes from a Rain Forest Catchment in the Central Amazon Basin. Water Resources Research 29(3): 743-758.
Lesack, L.F.W. and Melack, J.M., 1991. The Deposition, Composition, and Potential Sources of Major Ionic Solutes in Rain of the Central Amazon Basin. Water Resources Research 27(11): 2953-2977.
Matsuyama, H., Oki, T. and Masuda, K., 1993. The Water Budget in the Amazon River Basin during the FGCE Period. IAHS Publication 216: 35-42.
Nortcliff, S. and Thornes, J.B., 1978. Water and cation movement in a tropical rain forest environment. Acta Amazonica 8(2): 245-258.
Nortcliff, S., Thornes, J.B. and Waylen, M.J., 1979. Tropical Forest Systems: A Hydrological Approach. Amazoniana 6: 557-568.
Nortcliff, S. and Thornes, J.B., 1981. Seasonal Variations in the Hydrology of a Small Forested Catchment near Manaus, Amazonas, and the Implications for its Management. In: R. Lal and E.W. Russell. Tropical Agricultural Hydrology. John Wiley and Sons Ltd. New York. Pp. 37-57.
Nortcliff, S. and Thornes, J.B., 1984. Floodplain response of a small tropical stream. In: Burt, T.P. and Walling, D.E. (Eds.). Catchment experiments in fluvial geomorphology. Geobooks, Norwich.
Nortcliff, S. and Thornes, J.B., 1988. The dynamics of a tropical floodplain environment with reference to forest ecology. Journal of Biogeography 15: 49-59.
Nortcliff, S. and Thornes, J.B., 1989. Variations in soil nutrients in relation to soil moisture status in a tropical forest ecosystem. In: Proctor, J. (Ed.). Mineral nurients in tropical forest and savanna ecosystems. ES Special Publication 9. Blackwell Scientific Publishers, Oxford. p. 43-54.
Nortcliff, S., Ross, S.M. and Thornes, J.B., 1990. Soil moisture, Runoff and sediment Yield from Differentially Cleared Tropical Rainforest Plots. In: Thornes, J.B. (Ed.). Vegetation and Erosion. John Wiley and Sons Ltd., UK. p. 419-436.
Poels, R.L.H., 1987. Soils, Water and Nutrients in a Forest Ecosystem in Surinam. PhD thesis. Wageningen Agricultural University, Wageningen, The Netherlands. 253 p.
Roche, M.A., 1982. Evapotranspiration réelle de la forêt amazonienne en Guyane. Cahiers ORSTOM, Serie Hydrologie 19: 37-44.
Shuttleworth, W.J., 1988. Evaporation from Amazonian Rain Forest. Proceedings of the Royal Society London. Volume 223. Series B. Pp. 321-346.
Sioli, H. (Ed.), 1984. The Amazon: Limnology and Landscape Ecology of a Mighty Tropical River and its Basin. Monographiae Biologicae 56. Dr. W. Junk Publishers. Dordrecht, The Netherlands. 763 p.
Schmidt, G.W., 1972a. Chemical properties of some waters in the tropical rain forest region of Central Amazonia. Amazoniana 3: 199-207.
Schmidt, G.W., 1972b. Amounts of solids and dissolved substances in the middle reaches of the Amazon over the course of one year (August 1969 - July 1970). Amazoniana 3: 208-223.
Stallard, R.F. and Edmond, J.M., 1983. Geochemistry of the Amazon. 2. The influence of geology and weathering environment on the dissolved load. Journal of Geophysical Research 88: 9671-9688.
Tomasella, J. and Hodnett, M.G., 1996. Soil hydraulic properties and van Genuchten parameters for an Oxisol under pasture in Central Amazonia. In: Gash, J.H.C., Nobre, C.A., Roberts, J.M. and Victoria, R.L. (eds.). Amazonian Deforestation and Climate. John Wiley and Sons, Chichester, UK. pp. 151-162.
Ubarana, V.N., 1996. Observations and Modelling of Rainfall Interception at two Experimental sites in Amazonia. In: Gash, J.H.C., Nobre, C.A., Roberts, J.M. and Victoria, R.L. (eds.). Amazonian Deforestation and Climate. John Wiley and Sons, Chichester, UK. pp. 151-162.
Vis, M., 1986. Interception, drop size distribution and rainfall kinetic energy in four Colombian forest ecosystems. Earth Surface Processes and Landforms 11: 591-603.
Waterloo, M.J., 1994. Water
and nutrient dynamics of Pinus caribaea plantation forests
on former grassland soils in Southwest Viti Levu, Fiji. PhD thesis.
Vrije Universiteit Amsterdam, The Netherlands. 478 p.
Annex 2. List of hardware
I am still collecting information, will be included at a later stage.