HESSHydrology and Earth System SciencesHESSHydrol. Earth Syst. Sci.1607-7938Copernicus PublicationsGöttingen, Germany10.5194/hess-21-4629-2017Technical note: A hydrological routing scheme for the Ecosystem Demography
model (ED2+R) tested in the Tapajós River basin in the Brazilian
AmazonPereiraFabio F.FarinosiFabiofabio.farinosi@gmail.comhttps://orcid.org/0000-0002-4774-4854AriasMauricio E.LeeEunjeeBriscoeJohnMoorcroftPaul R.Sustainability Science Program, Kennedy School of
Government, Harvard University, Cambridge, MA 02138, USACa' Foscari University of Venice, Venice,
ItalyDepartment of Civil and Environmental Engineering,
University of South Florida, Tampa, FL 33620, USAnow at: Department of Renewable Energy Engineering,
Federal University of Alagoas, Maceió, AL, Brazilnow at: Goddard Earth Sciences Technology and
Research, Universities Space Research Association, Columbia, MD 21046, USAcurrent address: Global Modeling and Assimilation Office, NASA Goddard Space
Flight Center, Greenbelt, MD 22071, USAnow at: European Commission, DG Joint Research Centre, Ispra, Italydeceased, 12 November 2014Fabio Farinosi (fabio.farinosi@gmail.com)14September2017219462946488March20164May201620July20172August2017This work is licensed under the Creative Commons Attribution 3.0 Unported License. To view a copy of this licence, visit https://creativecommons.org/licenses/by/3.0/This article is available from https://hess.copernicus.org/articles/21/4629/2017/hess-21-4629-2017.htmlThe full text article is available as a PDF file from https://hess.copernicus.org/articles/21/4629/2017/hess-21-4629-2017.pdf
Land surface models are excellent tools for studying how
climate change and land use affect surface hydrology. However, in order to
assess the impacts of Earth processes on river flows, simulated changes in
runoff need to be routed through the landscape. In this technical note, we
describe the integration of the Ecosystem Demography (ED2) model with a
hydrological routing scheme. The purpose of the study was to create a tool
capable of incorporating to hydrological predictions the terrestrial
ecosystem responses to climate, carbon dioxide, and land-use change, as
simulated with terrestrial biosphere models. The resulting ED2+R model
calculates the lateral routing of surface and subsurface runoff resulting
from the terrestrial biosphere models' vertical water balance in order to
determine spatiotemporal patterns of river flows within the simulated
region. We evaluated the ED2+R model in the Tapajós, a 476 674 km2
river basin in the southeastern Amazon, Brazil. The results showed that the
integration of ED2 with the lateral routing scheme results in an adequate
representation (Nash–Sutcliffe efficiency up to 0.76, Kling–Gupta efficiency
up to 0.86, Pearson's R up to 0.88, and volume ratio up to 1.06) of daily to
decadal river flow dynamics in the Tapajós. These results are a
consistent step forward with respect to the “no river representation” common
among terrestrial biosphere models, such as the initial version of ED2.
Introduction
Understanding the impacts of deforestation (e.g., Lejeune et al., 2015;
Medvigy et al., 2011; Andréassian, 2004) and climate change (e.g.,
Jiménez-Cisneros et al., 2014) on the Earth's water cycle has been a
topic of substantial interest in recent years given its potential
implications for ecosystems and society (e.g., Wohl et al., 2012; Brown et
al., 2005). Analyses of climate change impacts on the Earth's water cycle
increasingly use terrestrial biosphere models, which are capable of
estimating changes in the vertical water balance as a function of climate
forcing and/or land-use-induced changes in canopy structure and composition
(Zulkafli et al., 2013). Terrestrial biosphere models actively used for
hydrological and Earth system sciences include the Joint UK Land
Environment Simulator (JULES) (Best et al., 2011; Clark et al., 2011), the
Community Land Model (CLM) (Lawrence et al., 2011; Oleson et al., 2010), the
Lund–Potsdam–Jena (LPJ) land model (Gerten et al., 2004; Sitch et al.,
2003), the Max Planck Institute MPI-JSBACH model (Vamborg et al., 2011; Raddatz et al.,
2007), and the Integrated Biosphere Simulator (IBIS) (Kucharik et al., 2000).
Initial formulations of the hydrological processes within terrestrial
biosphere models were based on simple “bucket” model formulations (Cox et
al., 1999 after Carson, 1982). Moisture within each climatological grid cell
of the domain was simulated in a single below-ground pool in which surface
temperature and specific soil moisture factors determined evaporation, while
runoff was equal to the bucket overflow (Cox et al., 1999; Carson, 1982).
Recently, the hydrologic schemes within terrestrial biosphere models have
become increasingly sophisticated. In the most recent generation of land
surface models, water fluxes in and out of the soil column are
vertically resolved and take into account feedback from the different
components, for instance, through an explicit formulation of the
soil–plant–atmosphere continuum. This enables the models to provide a
detailed representation of the interactions between evapotranspiration, soil
moisture, and runoff (Clark et al., 2015).
To couple the calculation of the one-dimensional water balance with the
estimation of daily river flows, it is necessary to simulate multiple
hydrological dynamics involved in the lateral flow propagation through the
landscape, ideally including the most complex hydraulic features of
floodplains, lakes, and wetlands (Yamazaki et al., 2011). The first step
towards representing the finer-scale hydrodynamic processes responsible for
patterns in river gauge observations is to consider the topographic and
geomorphological features that control water flow (Arora et al., 1999). The
coarse spatial resolution of regional land surface models, imposed by
computational constraints, does not allow for proper simulation of the
complex hydrological dynamics determined by fine-scale topography in river
channels and floodplains (Yamazaki et al., 2011; Kauffeldt et al., 2016).
However, the combination of the terrestrial models with routing schemes can
be used to simulate the implications of global and regional environmental
changes for flood/drought forecasting, water resources planning and
management, and infrastructure development (Andersson et al., 2015).
Consequently, several terrestrial biosphere models have been integrated with
routing schemes. For example, JULES has been integrated with the Total
Runoff Integrating Pathways (TRIP) to evaluate the accuracy of its estimates
of annual streamflow (Oki et al., 1999). This integrated model was used to
investigate the status of the global water budget (Oki et al., 2001).
Rost et al. (2008) also used a modeling framework composed of the global dynamic
vegetation model, LPJ, and a simple water balance model to quantify the
global consumption of water for rain-fed and irrigated agriculture. An
offline coupling of the dynamic vegetation model, ISIS, and HYDRA – which
simulates the lateral transport of water through rivers, lakes, and wetlands
– was proposed in Coe et al. (2008) with the purpose of reproducing
linkages between land use, hydrology, and climate. Moreover, Liang et al. (1994)
developed and tested the coupling of the well-known VIC model with a
general circulation model (GCM) to improve the GCM's ability to capture the
interactions between surface hydrology and atmosphere. For the same purpose,
the MPI hydrological discharge model was validated with NCEP reanalysis and
parameterized for simulating the river routing for climate analysis at global
scale (Hagemann and Gates, 2001; Hagemann and Dumenil, 1997). Several routing
schemes have been designed to date, including normal depth, modified pulse,
simple Muskingum, and Muskingum–Cunge (USACE, 1991). Most notably, the
semi-distributed kinematic wave-routing Muskingum–Cunge method has been
recognized for its stability over different spatial and temporal modeling
resolutions (USACE, 1991; Miller and Cunge, 1975; Cunge, 1969), and has been
adopted by the most widely used regional-scale hydrological models, such as
VIC, SWAT, and MGB-IPH.
Recent studies have investigated the influence of land use on regional
patterns of rainfall and biosphere temperature (Ostberg et al., 2015; Bahn et
al., 2014; Pearson et al., 2013). These studies tracked how the occurrence of
conversion of land from its natural state over the same time frame as
observed fluctuations of rainfall and air temperature occurred – aspects
fully analyzed by terrestrial biosphere models (Hurtt et al., 2006;
Goldewijk, 2001; Ramankutty and Foley, 1999). However, these models assumed
that global and regional changes in the biosphere were a result of dynamics
of vegetation in a collection of landscapes given by forests, deserts, and
farmland only. Inland surface waters (e.g., rivers, lakes, and wetlands) were
not considered as an interactive component of the biosphere and hence the
climate system (Cole et al., 2007).
Schematic of the enthalpy
fluxes (all arrows) and water fluxes (all but solid black arrows) that are
solved in ED2. The schematic is based on Walko et al. (2000) and Medvigy et
al. (2009). (Figure courtesy of Marcos Longo.)
The Ecosystem Demography (ED2) is a terrestrial biosphere model that
simulates the coupled water, carbon, and energy dynamics of terrestrial land
surfaces (Longo, 2014; Medvigy et al., 2009; Moorcroft et al., 2001) to
describe the coupled water, carbon, and energy dynamics of heterogeneous
landscapes (Hurtt et al., 2013; Medvigy et al., 2009; Moorcroft et al., 2001).
ED2's ability to incorporate sub-grid-scale ecosystem heterogeneity arising
from land-use change makes the model suited for investigating how the
combined impacts of changes in climate, atmospheric carbon dioxide
concentrations, and land cover affect terrestrial ecosystems. For example,
ED2 was successfully used to simulate the carbon flux dynamics in the North
American continent (Hurtt et al., 2002; Albani et al., 2006) and to assess
the impacts on Amazonian ecosystems of changes in climate, atmospheric
carbon dioxide, and land use (Zhang et al., 2015). Moreover, ED2, coupled with
a regional atmospheric circulation component, has also been successfully
applied to assess the impacts of deforestation on the Amazonian climate
(Knox et al., 2015; Swann et al., 2015). The aforementioned studies were not
aimed at assessing hydrological implications of changes in land use and
climate. These works demonstrated the validity of ED2 for assessing impacts
of global and regional changes on ecosystem function and built the
foundations for an integrated tool aimed at analyzing hydrological
implications.
In this technical note, we describe the integration of ED2 with a
hydrological routing scheme. The hydrological routing scheme chosen was
adapted from the MGB-IPH (Collischonn et al., 2007). This exercise aims to
calculate the lateral propagation and attenuation of the surface and
subsurface runoff resulting from the vertical balance calculations in order
to simulate daily river flows through a large river basin. The advantage of
the proposed model is its ability to predict the sensitivity of river flows
to global and regional environmental changes such as climate and land-use
changes. The new product combines the advantages of biosphere and
hydrological models, bringing together global-, regional-, and local-scale
hydrological dynamics in a single modeling framework. The resulting model is
intended to be used in future studies as a computational tool to explore a
variety of research questions. In particular, it could be used to analyze
how current and future climate and land cover affect water availability in
river systems; how land-use-driven changes can influence the water
availability for human activities (hydropower, food production, urban
supply); and what the implications of those changes are for water and land
resources management.
The identified research areas are in line with key problems raised in the
literature, focusing on the importance of large-scale modeling and remote
sensing to fill knowledge gaps in water resources and hydrological dynamics
(Alsdorf et al., 2007; Prigent et al., 2007). The product obtained from this
exercise was tested in the Tapajós Basin, a large river system in
the southeastern Amazon, Brazil.
Schematic representation of the connection between the terrestrial
biosphere model and the hydrological routing scheme. Calibrating parameters
are circled in red. The reservoirs are used to determine the contribution of
streamflow that comes from overland flow, interflow, and groundwater flow. The
daily sum of these three reservoirs is then moved from each grid cell into
the drainage network.
Ecosystem Demography (ED2) model
ED2 is a terrestrial biosphere simulation model capable of representing
biological and physical processes driving the dynamics of ecosystems as a
function of climate and soil properties. Rather than using a conventional
“ecosystem as big-leaf” assumption, ED2 is formulated at the scale of
functional and age groups of plants. Ecosystem-scale dynamics and fluxes are
calculated through a scaling procedure which reproduces the macroscopic
behavior of the ecosystem within each climatological grid cell. It simulates
ecosystem structure and dynamics as well as the corresponding carbon,
energy, and water fluxes (Fig. 1; Hurtt et al., 2013; Medvigy et al., 2009;
Moorcroft et al., 2001). ED2 simulates the dynamics of different plant
functional types subdivided into tiles with a homogeneous canopy (Swann et
al., 2015; Medvigy et al., 2009). The dynamic tiles represent the sub-grid-scale
heterogeneity in ecosystem composition within each cell. Grid cell
size is determined by the resolution of meteorological forcing and soil
characteristics data, typically from 1 to 0.001∘ (∼ 110
to 1 km). ED2 simulates biosphere dynamics by taking into consideration
natural disturbances, such as forest fires and plant mortality due to
changing environmental conditions, as well as human-caused disturbances,
such as deforestation and forest harvesting (Medvigy et al., 2009;
Albani et al., 2006). Disturbances are expressed in the model as annual transitions
between primary vegetation, secondary vegetation, and agriculture (cropland
and pasture) (Albani et al., 2006). Natural disturbance, such as wildfire, is
represented in the model by the transition from primary vegetation (forest
in the case of the Amazon) to grassland–shrubland, and subsequently to
secondary vegetation (forest regrowth); the abandonment of an agricultural
area is represented with the conversion from grassland to secondary
vegetation, while forest logging is represented by the transition from
primary or secondary vegetation to grassland. The model is composed of
several modules operating at multiple temporal and spatial scales, including
plant mortality, plant growth, phenology, biodiversity, soil
biogeochemistry, disturbance, and hydrology (Longo, 2014; Medvigy et al., 2009).
A selection of the main parameters and the input used for this study
are presented in Table 1, and for a more complete description of the model,
we refer the reader to the literature available (Zhang et al., 2015; Longo
2014; Kim et al., 2012; Medvigy et al., 2009; Moorcroft et al., 2001).
ED2+R calibrated parameters (based on Zhang et al., 2015; Longo, 2014; Knox, 2012). Additional information about ED2 parameter
calibration for the Amazon Basin are available in Zhang et al. (2015) and
Longo (2014).
InputSourceMeteorological forcingSheffield et al. (2006)Land useHurtt et al. (2006)Topography (DEM)SRTM, Shuttle Radar Topography Mission 90 m resolution (USGS, 2016)Soil dataQuesada et al. (2010) – IGBP-DIS global soil data (Global Soil Data Task, 2014)Geomorphological relationsCoe et al. (2008)Streamflow observationsHYBAM – ANA (ANA, 2016; Observation Service SO HYBAM, 2016)Carbon dioxide concentration378 ppmProcessMethodIntegration schemeFourth-order Runge–Kutta methodEnergy and water cyclesKnox (2012) and Longo (2014)Temperature-dependent function for photosynthesisQ10 functionCanopy radiation schemeTwo-stream modelAllometry for heightBased on Poorter et al. (2006)Allometry for above-ground biomassBased on Eq. (2) of Baker et al. (2004)Allometry for leaf biomassBased on Cole and Ewel (2006) and Calvo-Alvarado et al. (2008)ParameterValueUnitsBiophysics time step600sNumber of soil layers16–Depth of the deepest soil layer6mDepth of the shallowest soil layer0.02mCohort water holding capacity0.11kgwmleaf+wood-2Residual stomatal conductance10 000µmol m-2 s-1Leaf-level water stress parameter0.016molH2O molAir-1Oxygenase/carboxylase ratio at 15 ∘C4000–Power base for oxygenase/carboxylase ratio0.57–Power base for carboxylation rate2.4–Power base for dark respiration rate2.4–Environmentally determined parametersValueUnitsWeight factor for stress due to light1.0–Maximum environmentally determined mortality rate5.0year-1Steepness of logistic curve10.0–Band-dependent radiation parameters ∗ValueUnitsDry soil reflectance(0.20; 0.31; 0.02)–Wet soil reflectance(0.10; 0.20; 0.02)–Leaf transmittance(0.05; 0.20; 0.00)–Leaf reflectance (grasses)(0.10; 0.40; 0.04)–Leaf reflectance (trees)(0.10; 0.40; 0.05)–Wood transmittance(0.05; 0.20; 0.00)–Wood reflectance (trees)(0.05; 0.20; 0.10)–
∗ Radiation-dependent parameters are given in the format xPAR, xNIR, and xTIR
corresponding to values for photosynthetically active, near infrared, and thermal infrared,
respectively. ∗∗ PFT-dependent parameters are given in the
format xETR, xMTR, and xLTR corresponding to the values for early-,
mid-, and late-successional cohorts, respectively. ∗∗∗ The
residence time parameters are dimensionless and used to correct the
Kirpich formula for time of concentration as explained in Collischonn
et al. (2007). Their magnitude is influenced by the size of the grid cell and its topography.
ED2 hydrology module
The hydrological module of the ED2 model is derived from the Land
Ecosystem-Atmospheric Feedback model (LEAF-2) (Walko et al., 2000). The model
computes the water cycle through vegetation, air canopy space, and soils,
yielding daily estimates of subsurface and surface runoff from each grid
cell, isolated from the others in the domain. The number of soil layers and
their thickness influence the accuracy with which the model is able to
represent the gradients near the surface. Soil composition was derived from
Quesada et al. (2010) and from the IGBP-DIS global soil data (Global Soil
Data Task, 2014). As described in Zhang et al. (2015), the mean fraction
values of sand and clay were assigned to each grid cell at 1 km resolution
and then aggregated at 1∘ resolution. Due to limited data
availability, soils were assumed to be homogeneous for a depth of 6 m.
Hydraulic conductivity of the soil layers is a function of soil texture and
moisture (Longo, 2014). Groundwater exchange is a function of hydraulic
conductivity, soil temperature, and terrain topography. Water percolation is
limited to the bottom layer by the subsurface drainage, determining the
bottom boundary conditions. Vegetation historical records and land-use
transitions were derived from the Global Land Use Dataset (Hurtt et al., 2006).
A more detailed description of the hydrological subcomponent of the
ED2 model is available in Longo (2014).
ED2 runoff routing scheme (ED2+R)
River routing schemes are often used to compute the lateral movement of
water over land in hydrological models for large river basins. In this way,
the prediction performance of models can be evaluated using river discharge
measurements. The use of routing schemes was then extended to Earth system
models in order to capture the impacts of man-made structures (e.g., dams and
reservoirs) and floodplain wetlands on the climate system (Li et al., 2011;
Yamazaki et al., 2011).
Daily runoff estimates from ED2 were computed for specific grid cells
independently. A hydrological routing scheme was then linked to this model in
order to estimate flow attenuation and accumulation as water moved through
the landscape. The hydrological routing scheme chosen was adapted from the
original formulation of the MGB-IPH, a rainfall–runoff model that has been
used extensively in large river basins in South America (Collischonn et al.,
2007). This model was later developed using hydrodynamic solutions and
floodplain coupling (Pontes et al., 2015; Paiva et al., 2013b). Although the
later development increased the modeling capabilities of the MGB-IPH in
representing fine-scale dynamics, given the regional application of our tool,
for ED2+R we decided to use the typical application of the MGB-IPH
characterized by the Muskingum–Cunge approach. The original MGB-IPH model is
composed of four different sub-models: soil water balance,
evapotranspiration, intra-cell flow propagation, and inter-cell routing
through the river network. In the present study, only the catchment and river
routing methods were utilized. The resulting ED2+R model computes the daily
total volume of water passing through any given grid cell in the resulting
drainage network in two separate steps: first, ED2 estimates of daily surface
and subsurface runoff from each grid cell are divided into three linear
reservoirs with different residence times to represent overland flow,
interflow, and subsurface water flow (Fig. 2). The reservoirs are used to
determine the contribution and attenuation of river flow by different soil
layers, characterized by different routing times. The sum of overland flow,
interflow, and subsurface water flow is then moved from each grid cell into
the drainage network, designed in the preprocessing phase using data from a
digital elevation model (DEM) from the Shuttle Radar Topography Mission (SRTM
– USGS, 2016) at a 90 m resolution and the Cell Outlet Tracing with an Area
Threshold algorithm (COTAT) (Reed, 2003). Each DEM grid cell therefore
becomes part of a flow path, which then accumulates water to a final
downstream drainage network outlet. A complete description of the technique
for defining drainage networks from DEMs employed in this study can be found
in Paz et al. (2006). Once water reaches the drainage network, ED2+R adopts
the Muskingum–Cunge numerical scheme for the solution of the kinematic wave
equation, which also accounts for flow attenuation, using a finite-difference
method as a function of river length, width, depth and roughness, as well as
terrain elevation slope (Collischonn et al., 2007; Reed, 2003). Statistical
relationships for the river morphology were obtained as a function of the
drainage area based on geomorphic data collected by Brazil's National Water
Agency (ANA) and the Observation Service for the geodynamical, hydrological,
and biogeochemical control of erosion/alteration and material transport in
the Amazon Basin (HYBAM) at several gauging stations in the Amazon and
Tocantins basins as presented by Coe et al. (2008). Further studies
successfully derived geomorphological relations in order to estimate river
geometric parameters and carry out hydrodynamic simulations of the Amazon
River system using a similar approach (Paiva et al., 2011, 2013). Multiple
groups of grid cells with common hydrological features, or hydrological
response units, can be created in order to parameterize and calibrate
ED2+R. In our approach, hydrological traits associated with soil and land
cover are primarily computed in ED2; thus, we calibrated ED2+R at the
sub-basin level as delineated based on the DEM. Details about the calibration
procedure are provided in the next section.
The model's performance was calculated through the adoption of widely used
indicators:
Pearson's R correlation coefficient (Pearson, 1895), calculated as in
Eq. (1):R=∑sim⋅obs-∑sim∑obsn∑sim2-∑sim2n∑obs2-(∑obs)2n2,
where sim and obs are the simulated and
observed time series, while n is the number of time steps of the simulation
period;
volume ratio, calculated as ratio of the simulated (sim) and observed
(obs) total water volume in the simulation period without consideration
for the seasonal distribution of flow, as in Eq. (2):VR=Volsim/Volobs;
the Nash–Sutcliffe efficiency (NSE) coefficient (Nash and Sutcliffe, 1970),
calculated as in Eq. (3):NSE=1-∑1n|obsi-simi|2∑1n|obsi-obsi‾|2,
where obsi and simi are the observed and simulated data at
time i, obsi‾ is the mean of the observed data, and n
is number of time steps of the simulation period;
the Kling–Gupta efficiency (KGE) index, both 2009 and 2012 versions, calculated
as in Eq. (4):KGE=1-s1R-12+s2vr2009or2012-12+s3β-12,
where s values are scaling factors (set to 1 in this case), r is
Pearson's correlation coefficient, β is the ratio between the mean of
the observed values and the mean of the simulated values, and vr is the
variability ratio, defined as vr2009 (simulated vs. observed standard
deviation ratio; Eq. 5) for the 2009 method and vr2012
(ratio of coefficient of variation of simulated and coefficient of variation
of observed values; Eq. 6) for the 2012 method (Kling et al., 2012; Gupta et
al., 2009).
vr2009=σsim/σobsvr2012=CVsimCVobs=σsim/μsimσobs/μobs
The optimal value for the Pearson's R, VR, NSE, and KGE indexes is 1; the
closer the indexes are to this value, the more accurately the model
reproduces the observed values.
Missing observations in the river flow records (HYBAM and ANA) were filled
via linear spatial and temporal interpolation between the series in
neighboring gauge stations (Eq. 7):
Obsyt=K+β1⋅Obszt+β2⋅Obsqt+β3⋅Obsyt-365+β4⋅Obsy(t+365),
where z, y, and q are three gauge stations with time series highly
correlated (Pearson's R≥ 0.85), and t expresses time in days. The
estimated β coefficients in Eq. (7) were used for the estimation of
the missing observations in site y (Table 2). The interpolation of the
gauge historical records was necessary to have continuous time series with a
sufficient number of observations to calibrate and validate the ED2+R
application in the basin.
Statistics about the gauge information filling procedure
(correlation with the station to be filled, number of original observations,
and filled number of observations).
Sub-basin nameMain river gauge station – z in Eq. (7)Original number of daily gauge records (number of daily observations)Gap-filling station 1 – q in Eq. (7) (correlation with z)Gap-filling station 2 – y in Eq. (7) (correlation with z)Number of daily records after filling procedure (number of daily observations)JamanximJamanxim1928Jardim do Ouro (0.97)Novo Progresso (0.96)5382Upper Teles PiresCachoeirão10 356Teles Pires (0.91)Indeco (0.94)11 524Upper JuruenaFontanilhas10 469Foz do Juruena (0.94)Barra do São Manoel (0.89)11 688Lower Teles PiresTres Marias8682Barra do São Manoel (0.98)Santa Rosa (0.98)10 640Lower JuruenaFoz do Juruena2074Barra do São Manoel (0.98)Jatoba (0.97)11 447Upper TapajósJatoba10 218Fortaleza (0.99)Barra do São Manoel (0.98)11 517Lower TapajósItaituba5789Fortaleza (0.99)Jatoba (0.98)11 688
For the presentation of the results, in order to compare the simulated and
observed values, we also used flow duration curves (FDCs). FDCs are
cumulative frequency plots that show the percentage of simulations steps (days
in the case presented in this study) in which the discharge is likely to
equal or exceed a specific value, without taking into consideration the
sequence of the occurrence.
Average precipitation (a) and temperature (b) in
the Tapajós River basin (1986–2005). Re-drafted from Farinosi et
al. (2017). (c) Aerial imagery the Tapajós River basin
illustrating land-cover diversity in the catchment. Source: Google Earth
Pro.
(a) Organization of the Tapajós Basin into seven
sub-basins: Upper Juruena (UJ); Lower Juruena (LJ); Upper Teles Pires (UTP);
Lower Teles Pires (LTP); Jamanxim (JA); Upper Tapajós (UT); and Lower
Tapajós (LT). (b) ED2+R represents the domain in grid cells
with 0.5∘ resolution (∼ 55 km). The black segments indicate
the flow accumulation network.
Case study: Tapajós River basin
The ED2+R formulation was parameterized and evaluated for the Tapajós
River basin, the fifth largest tributary of the Amazon. This basin drains an
area of 476 674 km2 in the southeastern Amazon, within the Brazilian
states of Mato Grosso, Pará, and Amazonas. The main rivers in the basin
are the Tapajós (with a length greater than 1800 km and average
discharge of 11 800 m3-1), Juruena (length of approximately
1000 km and discharge of 4700 m3 s-1), and Teles Pires (also
known by the name São Manoel, about 1600 km long and average discharge
of 3700 m3 s-1). The river system flows northwards, with terrain
elevation ranging from about 800 m a.s.l. (above sea level) in the southern
region, to a few meters above sea level at its confluence with the Amazon
River (ANA, 2011). The basin ecosystems are mainly represented by tropical
evergreen rainforests in the northern region (in the states of Amazonas and
Pará), and Cerrado dry vegetation in the south (Mato Grosso).
Precipitation ranges from about 1500 mm year-1 in the headwaters
(southern region) to about 2900 mm year-1 towards the basin's outlet
(Fig. 3a, b). Rainfall temporal distribution is characterized by a clear
seasonal distinction; total precipitation in the wet season (September to
May) could be as high as 400 mm month-1 in the most tropical areas,
whereas in the dry season (June to August), precipitation is close to zero in
the Cerrado and as low as 50 mm month-1 in the wetter areas (Mohor et
al., 2015). As a result of the large rainfall seasonal variability, river
flows are also extremely variable: the mean monthly flow of the Tapajós
River ranges between about 2300 and 28 600 m3 s-1 according to
the historical records used for the calibration of the ED2+R model. Soils
vary from those typically seen in the Brazilian shield in the south of the
basin to alluvial sediments in the north. Land use, almost completely
represented by primary forest until the 1970s, was radically changed in
recent decades. As estimated from the land-use/land-cover dataset used in
this study (Hurtt et al., 2006), in the late 2000s only about 56 % of the
basin (270 000 km2) was covered by the original vegetation cover.
Large parts of the basin lying in the territory of Mato Grosso were cleared
to make room for agricultural and livestock production, while vast areas
around the border between the state of Pará and Mato Grosso were cleared
for cattle production. The northern portion of the basin is largely protected
by natural parks or indigenous lands, but significant deforestation hotspots
could be identified around the cities of Santarém and Itaituba and along
the main transportation routes (Fig. 3c). For a more detailed description of
the basin's physical characteristics and historical analysis of trends in
deforestation, precipitation, and discharge, we refer the reader to Arias et
al. (2017) and Farinosi et al. (2017).
For calibration purposes, the basin was divided into seven sub-basins, each
of them with a corresponding gauge for which historical daily river flow
observations were available (Fig. 4a). The domain was gridded with a
spatial resolution of 0.5∘ by 0.5∘, roughly
corresponding to 55 km by 55 km. Simulations were carried out for the period
1970–2008. The ED2 model was forced using reconstructed climate (Sheffield
et al., 2006) and land-use/land-cover data (Hurtt et al., 2006; Soares-Filho
et al., 2006) at 1∘ spatial resolution. The original meteorological
dataset has a 3 h temporal resolution, which was downscaled to an hourly
resolution, as described in Zhang et al. (2015). In this technical note, we
describe the calibration of the flow-routing component of ED2+R. The
parameterization of the ED2 terrestrial biosphere model was developed and
evaluated independently using eddy-flux tower observations of carbon, water,
and energy fluxes and forest inventory observations of above-ground biomass
dynamics. Further details are available in Zhang et al. (2015) and Longo (2014).
ED2+R model calibration
The ED2+R model was manually calibrated using gauge observations (ANA,
2016; Observation Service SO HYBAM, 2016) spanning a period of 17 years from
1976 to 1992 (the period 1970–1975 was not considered in order to avoid
simulation initiation effects) through a two-step procedure, as highlighted
in Fig. 2. The first step is partitioning the flows from the two reservoirs
(surface and subsurface) of the ED2 biosphere model into the three reservoirs
(surface, intermediate, base) of the ED2+R routed biosphere model
(parameters α and β in Fig. 2). In particular, α
(ranging from 0 to 1, or from 0 to 100 %) represents the share of ED2
surface runoff allocated to the ED2+R surface reservoir. The remaining part
(1-α) is allocated to the ED2+R intermediate reservoir. β
represents a similar partitioning coefficient for the ED2 subsurface
reservoir to the ED2+R intermediate and base reservoirs. The second step
relates to the adjustment of the residence times of the water flows in the
three reservoirs for each of the grid cells in each of the sub-basins
(overland, intermediate, and subsurface water flows – represented by the
adjustment parameters CS, CI, and CB in Fig. 2).
In the first step, following the methodology described by Anderson (2002),
the sensitivity of the α and β parameters was tested by
running the model multiple times (∼ 35). For each run, the
NSE (Nash and Sutcliffe, 1970) was quantified by
comparing the results of the simulation to historical flow observations. The
combinations of the α and β parameters characterized by the
largest NSE were selected. Parameters α and β were assumed to
be uniform for the whole basin. Figure 5 shows the different combinations of
the α and β parameters introduced in Fig. 2. The color bar
indicates the NSE resulting from the comparison between the simulated and
observed river flow values obtained using different combinations of the
parameters α (x axis) and β (y axis). The chosen combination
(indicated by an x in Fig. 5) lies in one of the optimal combination areas
(NSE ∼ 0.8).
Calibration of flow partitioning (α and β parameters in
Fig. 2) between the ED2 and the ED2+R reservoirs. The color bar indicates the
NSE values of the simulated vs. the observed river flow values (0: very
different; 1: very similar).
Calibration and validation results for NSE,
Kling–Gupta (2009 and 2012 methods), Pearson's R correlation, and volume
ratio. Optimal values are equal to 1 (statistics were calculated using
the R package hydroGOF; Zambrano-Bigiarini, 2014).
KGE-2009 method values are indicated with square brackets []; KGE-2012 method values are indicated with curly brackets {} (Eqs. 4, 5, and 6).
Calibration period (1976–1992) Validation period (1993–2008) Sub-basinNash–Sutcliffe Kling–Gupta Pearson's RVolume ratio Nash–Sutcliffe Kling–Gupta Pearson's RVolume ratio [2009 method] correlation vol sim/vol [2009 method] correlation vol sim/vol {2012 method} obs {2012 method} obs ED vs. OBSED2+R vs. obsED vs. obsED2+R vs. obsED vs. obsED2+R vs. obsED vs. obsED2+R vs. obsED vs. obsED2+R vs. obsED vs. obsED2+R vs. obsED vs. obsED2+R vs. obsED vs. obsED2+R vs. obsUpper Juruena-26.880.45[-3.60] {-5.75}[0.50] {0.51}0.610.680.720.98-27.470.29[-3.54] {-6.10}[0.39] {0.38}0.530.540.681.01Upper Teles Pires-3.350.37[-0.51] {-0.64}[0.61] {0.61}0.530.640.941.01-3.190.28[-0.51] {-0.59}[0.63] {0.63}0.570.630.961.03Lower Juruena-1.450.65[-0.23] {-0.18}[0.64] {0.67}0.770.821.020.94-2.170.63[-0.43] {-0.30}[0.72] {0.67}0.750.811.051.08Lower Teles Pires-0.200.71[0.25] {0.27}[0.68] {0.67}0.800.851.011.02-0.340.67[0.17] {0.34}[0.69] {0.60}0.820.851.111.17Jamanxim-0.740.67[0.01] {0.39}[0.79] {0.78}0.820.851.551.13-0.100.55[0.23] {0.52}[0.75] {0.73}0.830.771.431.09Upper Tapajós-1.010.77[-0.13] {0.21}[0.82] {0.83}0.840.881.200.99-1.230.75[-0.22] {0.16}[0.84] {0.81}0.840.881.211.08Lower Tapajós-0.400.76[-0.09] {0.28}[0.86] {0.83}0.840.881.111.06-0.500.68[0.09] {0.29}[0.80] {0.76}0.820.861.131.13
Model sensitivity to the initial conditions of baseflow (QB) for the
different ED2+R sub-basins in the domain: Upper Juruena (UJ); Upper Teles
Pires (UTP); Lower Juruena (LJ); Lower Teles Pires (LTP); Upper Tapajós
(UT); Jamanxim (JA); and Lower Tapajós (LT).
Calibration of the residence times (τ) of the flow within the
ED2+R reservoirs of different grid cells in the domain through the
adjustment of the non-dimensional C parameters. Overland, intermediate, and
subsurface water flows are calibrated, respectively, through the adjustment
parameters CS, CI, and CB (Fig. 2). The color bars refer to the model
performance (NSE) of the specific parameter
combination in the specific sub-basin. In red are the chosen combinations. At the
bottom of each graph, there is a table providing the corresponding approximate average
time of concentration (in days) for the cells in the
sub-basin. (a) Upper Juruena (UJ); (b) Upper Teles Pires
(UTP); (c) Lower Juruena (LJ); (d) Lower Teles Pires
(LTP); (e) Upper Tapajós (UT); (f) Jamanxim (JA);
and (g) Lower Tapajós (LT).
Calibration and validation of the river flow (m3 s-1) at
Itaituba (farthest downstream river gauge – Lower Tapajós sub-basin).
ED2 output (green line), ED2+R (red line), and observations (blue dotted
line). The dotted black line splits the calibration and validation periods.
Similar comparison for each of the seven sub-basins is available in Appendix A.
Flow duration curves (percentage of time that flow –
m3 s-1 – is likely to equal or exceed determined thresholds) of
observed values (blue), ED2 outputs (green), and ED2+R (red) at the outlet of
the seven sub-basins. (a) Upper Juruena (UJ); (b) Upper
Teles Pires (UTP); (c) Lower Juruena (LJ); (d) Lower Teles
Pires (LTP); (e) Upper Tapajós (UT); (f) Jamanxim (JA);
and (g) Lower Tapajós (LT).
In the second step, the residence times (τ) of flow within the ED2+R
reservoirs of each grid cell in the domain were calibrated through the
adjustment of the non-dimensional parameters (CS, CI, and CB in Fig. 2) used to
correct the Kirpich formula for time of concentration (as explained in
Collischonn et al., 2007). The calibration procedure characterizing the
second step is similar to the previous one but in this case the calibration
is repeated for each sub-basin sequentially. The calibration process was
conducted from the furthest upstream sub-basins – headwaters – to the
final outlet of the basin (Anderson, 2002). The model was run multiple times
(between 30 and 50 per sub-basin) with different combinations of the three
parameters (CS, CI, and CB in Fig. 2); for each run, the goodness of fit was
quantified. This allowed us to design a sensitivity curve of the model to
different combinations of the three parameters for each of the seven
sub-basins and to select the combination that best approaches the
historical observations. Figure 6 shows how the model is sensitive to
marginal variation in initial conditions of baseflow, particularly in the
upstream section (i.e., UTP – Upper Teles Pires, UJ – Upper Juruena, and LTP
– Lower Teles Pires). Changes in initial subsurface water were controlled
by the 5-year initialization period; thus, contributions to the downstream
part of the basin had minimal impact (i.e., UT and LT – Upper and Lower
Tapajós).
Figure 7 describes the calibration of the residence time adjustment
parameters for each of the sub-basins, as well as an approximate calculation
of the corresponding time of concentration for each of the reservoirs in the
cell. The different combinations of the values assigned to the parameters
CS, CI, and CB significantly affect the overall goodness of fit of the river flow
simulations (NSE indicator). The calibration process was conducted from the
furthest upstream sub-basins – headwaters – (UTP – Upper Teles Pires, UJ
– Upper Juruena, and JA – Jamanxim) to the final outlet of the basin (LT
– Lower Tapajós). The different combinations are marked with the
corresponding NSE value; the optimal combination is marked in red (Fig. 7).
The period 1993–2008 was used for model evaluation. Comparisons between
observations and simulated flows (goodness of fit) were carried out using
Pearson's R correlation coefficient (Pearson, 1895), volume ratio (VR), the
Nash–Sutcliffe efficiency (NSE) coefficient (Nash and Sutcliffe, 1970), and
the Kling–Gupta efficiency (KGE) index (Kling et al., 2012; Gupta et al., 2009) (Table 3).
Results
The integration of the routing scheme with ED2 increases the ability of the
model to reproduce the observed temporal variations in river flows at the
basin outlet (Fig. 8). This statement applies to all of the sub-basins, as
the application of the routing scheme improved the model's performance
between simulated and observed values with respect to all four measures
selected (NSE, KGE, Pearson's R correlation,
and volume ratio) (Table 3). Both routed (ED2+R) and non-routed (ED2)
simulation results manage to reproduce the observed water availability
(quantity of water available) in the basin in terms of volume. The volume
ratio at the furthest downstream sub-basin (Lower Tapajós), in fact,
ranges around the optimal value for both validation and calibration periods
(ED2: 1.11–1.13; ED2+R: 1.06–1.13). The routing scheme improves the ability
of the model to reproduce the spatiotemporal distribution of water flows
across the basin: both the NSE and the KGE indexes reached values ranging
between 0.76 and 0.86 in the calibration, and between 0.68 and 0.80 in the validation
period (Table 3). Also, the correlation values confirm the results of the
other indexes, reaching 0.88 for the calibration and 0.86 for the validation
period. The performance of the presented tool is evident also when analyzing FDCs
(Fig. 9a–g). The adoption of the river routing scheme allows a more
realistic representations of the high discharge values (flow equaled or
exceeded 0 to 20–30 % of the time), and low discharge values (flow equaled
or exceeded 60 to 100 % of the time) in all the sections of the basin
(Fig. 9). The model's performance in simulating river flows is generally
more robust in the downstream sub-basins (NSE: 0.68–0.77 and KGE: 0.76–0.84
in the Upper and Lower Tapajós) and poorer in the headwaters (NSE:
0.28–0.45 and KGE: 0.38–0.61 in the Upper Juruena and Upper Teles Pires). In
the Upper Teles Pires and Upper Juruena, the model achieved the lowest NSE
(0.28 and 0.29, respectively, in the calibration, and 0.37 and 0.45 in the
validation period), and KGE values (0.61 and 0.50 during calibration, and 0.63 and
0.38 during validation). Although water volumes are correctly reproduced in both
the sub-basins (VR between 1.01 and 0.98 in the calibration, and 1.03 and
1.01 in the validation period), the seasonal variability is less accurate
(correlation: 0.64–0.68 and 0.63–0.54). The KGE, NSE, and correlation indices
are closer to the optimal value in the central and lower parts of the basin,
particularly in the Lower Juruena (calibration – NSE: 0.65, KGE: 0.64,
and correlation: 0.82; validation – NSE: 0.63, KGE: 0.67, and correlation: 0.81), Lower
Teles Pires (calibration – NSE: 0.71, KGE: 0.67, and correlation: 0.85; validation
– NSE: 0.67, KGE: 0.60, and correlation: 0.85), Upper Tapajós (calibration –
NSE: 0.77, KGE: 0.82, and correlation: 0.88; validation – NSE: 0.75, KGE: 0.81,
and correlation: 0.88), and Lower Tapajós (calibration – NSE: 0.76, KGE: 0.83,
and correlation: 0.88; validation – NSE: 0.68, KGE: 0.76, and correlation: 0.82) (Table 3).
FDCs, representing the probability of the flow values to exceed a specific
discharge, highlight the positive effect of the application of the routing
scheme in ED2+R across the entire range of flow variability (Fig. 9).
The simulated FDCs follow the same shape of the observed ones in the
furthest upstream sub-basins, especially in the cases of the Upper Juruena
and Upper Teles Pires, implying that the routing scheme is effective in
maintaining the simulated discharge range (Upper Juruena:
1200–2480 m3 s-1, Upper Teles Pires: 393–4130 m3 s-1) in line with the
observations (1030–2400 and 302–2767 m3 s-1, respectively).
This is especially true for the lowest flows, where the error between
simulated and observed curves is lower than 15 % (Figs. 9a, b, A1).
Regarding the intermediate sub-basins, Lower Juruena and Lower Teles
Pires, flood duration curves show that the model overestimates the lowest
values of the distribution by approximately 30 % of the observed values
(flow equaled or exceeded 60 to 100 % of the time in Fig. 9c, d).
Similar overestimation of the model could be noticed in the furthest
downstream sub-basins, Upper and Lower Tapajós (Fig. 9e–g). The
overestimation of the lower discharge values highlighted in Fig. 9g is
also evident in the multiyear hydrograph (Fig. 8), which shows that the
ED2+R simulation results overestimate (by about 40 % on average in the
discharge values included in the range of 60 to 100 % in Fig. 9g) the
observations during the dry seasons of the period under consideration.
Discussion
As the results in Table 3 and Figs. 8–9 show, the one-way integration of
ED2 with a routing scheme improves the performance of simulated daily
discharges. Although this could appear obvious from a hydrological modeling
perspective, the significance of this study lies in the fact that
terrestrial biosphere models, which are widely applied to examine the
impacts of climate and land use on the hydrology of the land surface, are
typically “no-river representation” models. The incorporation of ecosystem
responses to climate, carbon dioxide, and land-use changes simulated by
terrestrial biosphere models with hydrological modeling improves the
representation of the hydrological characteristics of basins characterized
by large forest cover and/or large deforestation rates. In applications in
the tropics, the one-way integration of the terrestrial biosphere model and
the routing scheme (i.e., the two tools are not fully coupled) could lead to
a partially inaccurate representation of the seasonally flooded ecosystems,
a relevant aspect as documented in the literature (e.g., Cole et al., 2007).
As seen in Fig. 9, the performance of the model in simulating river flows
in the basin is generally higher in the downstream sub-basins and poorer in
the headwaters. Several factors are likely to cause this issue, both from
the simulation of the hydrological dynamics in ED2, the flow partitioning
(α and β parameters), and the basin hydraulic characteristics
in ED2+R. The accurate calibration of the biosphere model with flux tower
observations (Zhang et al., 2015; Longo, 2014) and the optimization of
the flow partitioning make us believe that this variation in performance is
due to the relatively coarse spatial resolution of the model in combination
with the limitations typical of most land surface models in capturing the
interactions with deep groundwater (Lobligeois et al., 2014; Zulkafli et al., 2013; Smith et al., 2004).
We believe that the error is arising from the
complexities associated with deep soils present in the headwaters of the
Tapajós Basin. In particular, in the model application developed, soil
layers are represented to a depth of 6 m (Table 1), which might be too
shallow to realistically represent the conditions in the headwaters of the
basin. The importance of groundwater is also evident from the calibration of
the residence time parameter of the subsurface water flow: as shown in Fig. 7,
in fact, especially in the headwaters, even small variations in
the CB parameter greatly affect the model performance (specifically
quantified with NSE in Fig. 7). The combined effect of groundwater
interactions and spatial resolution is more evident in the upstream
sub-basins because of the greater marginal contribution of baseflow in these
areas. Surface flow accumulation, in fact, is lower in the headwaters.
Therefore, in relative terms, the role of baseflow is more relevant in this
portion of any basin. Further downstream, the effect of groundwater
interactions and spatial resolution is, at least in part, masked by the
larger rainfall–runoff contribution and the overall flow accumulation from
the upstream sub-basins. Other recent hydrological simulations of the
Tapajós have obtained higher accuracy (e.g., Mohor et al., 2015;
Collischonn et al., 2008; Coe et al., 2008); however, these simulations were
set up discretizing the basin into a finer spatial resolution grid (9 to 20 km
vs. ∼ 55 km grid cells) and using hydrological tools
able to reproduce highly detailed hydrodynamic characteristics of complex
river systems (i.e., floodplain, lakes, wetlands, backwater effects) that are
out of the scope of the tool presented in this study. The advantage of the
ED2+R model is the ability to study the sensitivity of the river flows to
global and regional changes as computed by traditional terrestrial biosphere
models, but adding a more detailed hydrological feature with respect to a
very simplistic- or no-river representation. The coarse spatial resolution
of the global datasets used as input for ED2+R is, however, a limiting
factor. Higher-resolution climatological data, vegetation, and land-use
datasets, which would allow a finer resolution of the hydrological grid, are
expected to improve the performance of the model by providing more detailed
hydrological processes. On the other hand, a finer spatial resolution of the
hydrological grid would also require a more detailed representation of the
subsurface water in the model. In general, the tool can be used to study how
different hydrological systems are being affected by changes in climate
forcing and changes in ecosystem composition and structure arising from the
combination of changing climate, rising atmospheric carbon dioxide, and
land-use transformation. Additionally, ED2+R could potentially bridge one
of the missing gaps for diagnosing and assessing feedback between
atmosphere and biosphere with inland surface waters being represented as a
dynamic system.
Conclusion
In this technical note, we present the integration of the terrestrial
biosphere model Ecosystem Demography 2 (ED2) with the Muskingum–Cunge
routing scheme. We tested the integrated model (ED2+R) in the Tapajós
River basin, a large tributary of the Amazon in Brazil, for the period
1970–2008. The results showed that the integration of a biosphere model with
a routing scheme improves the ability of the land surface simulation to
reproduce the hydrological and river flow dynamics at the basin scale. The
main limitations highlighted in this case study were linked to the
relatively coarse spatial resolution of the model and the rough
representation of subsurface water flow typical of these kinds of models.
Moreover, the terrestrial biosphere model ED2 and the routing scheme are
presented here in a one-way integration. The full coupling of the routing
scheme and ED2 could further improve the tool's ability to reproduce the
water balance considering flooded ecosystems, a relevant feature in the
simulation of environments like the tropical forest, where local
evapotranspiration plays a primary role in the specific ecosystem's
dynamics. In this first integration, our goal was to give the terrestrial
biosphere model the ability to reproduce river flows through a routing
scheme. With a fully coupled (i.e., two-way) integration, the model would be
able to determine the grid cells that are likely to be saturated and use
this information for the modeling of the ecosystem's dynamics. For instance,
this could determine the increase of the mortality rate of plants that are
sensitive to inundation. An additional limitation of the model could be
identified in its inability to reproduce highly detailed hydrological
dynamics of complex river systems (as, for instance, floodplain hydraulic
features or backwater effects). However, such a detailed hydrological
complexity was out of the scope of this study. Future efforts will address
the highlighted limitations, while upcoming studies will use ED2+R to
understand historical changes and future projections of the impacts of
climate change and deforestation on the Amazon's water resources.
Meteorological forcing data are derived from Sheffield et
al. (2006) – 10.1175/JCLI3790.1; land use data are derived from Hurtt
et al. (2006) – 10.1111/j.1365-2486.2006.01150.x; topographic data are
derived from the Shuttle Radar Topography Mission (SRTM) 90 m resolution
(USGS, 2016); soil map is derived from Quesada et al. (2010) –
10.5194/bg-7-1515-2010 – and IGBP-DIS global soil data (Global Soil
Data Task, 2014) – 10.3334/ORNLDAAC/565; geomorphological relations
are obtained from Coe et al. (2008) – 10.1002/hyp.6850; streamflow
observations are obtained from Observation Service SO HYBAM (2016) and
ANA (2016).
Time series of river flow (m3 s-1) at the outlet of each sub-basins.
ED2 output (green line), ED2+R (red line), and observations (blue dotted
line). (a) Upper Juruena (UJ); (b) Upper Teles Pires
(UTP); (c) Lower Juruena (LJ); (d) Lower Teles Pires
(LTP); (e) Jamanxim (JA); (f) Upper Tapajós (UT);
and (g) Lower Tapajós (LT).
FP, PM, and JB designed the study; FP
developed the ED2+R model code; FF, MA, and EL
calibrated the model and carried out the analysis; FF, MA, and
PM wrote the paper.
The authors declare that they have no conflict of
interest.
Acknowledgements
This work was conducted while Fabio F. Pereira, Fabio Farinosi, Eunjee Lee, and
Mauricio E. Arias were Giorgio Ruffolo Fellows in the Sustainability Science Program at
Harvard University. Fabio Farinosi was also funded through a doctoral
scholarship by Ca' Foscari University of Venice. Support from Italy's
Ministry for Environment, Land and Sea is gratefully acknowledged. We would
like to thank Marcos Longo for letting us use one of his figures, and Angela Livino
for the useful comments. The authors would like to dedicate this
study to the late Professor John Briscoe (1948–2014), who envisioned and
co-led the Amazon Initiative of Harvard's Sustainability Science Program. We
are grateful to the editor, Graham Jewitt, and to Hartley Bulcock
and the three other anonymous referees for the valuable comments received
during the review process. Finally, the authors would like to thank Erin Ciccone
and Madeleine Marino for proofreading the manuscript.
Edited by: Graham Jewitt
Reviewed by: Hartley Bulcock and three anonymous referees
ReferencesAlbani, M., Medvigy, D., Hurtt, G. C., and Moorcroft, P. R.: The
contributions of land-use change, CO2 fertilization, and climate
variability to the Eastern US carbon sink, Glob. Change Biol., 12,
2370–2390, 10.1111/j.1365-2486.2006.01254.x, 2006.Alsdorf, D. E., Rodríguez, E., and Lettenmaier, D. P.: Measuring surface
water from space, Rev. Geophys., 45, RG2002, 10.1029/2006RG000197,
2007.ANA: Plano Estratégico de Recursos Hídricos da Bacia Amazônica
– Afluentes da Margem Direita, Brasilia, Brazil, Brazil, available at:
http://margemdireita.ana.gov.br/ (last access: 7 September 2017), 2011 (in Portuguese).ANA: Hidroweb –Sistema de informações hidrologicas, available from:
http://www.snirh.gov.br/hidroweb/, (last access: 7 September 2017),
2016.Anderson, E. A.: Calibration of Conceptual Models for Use in River
Forecasting, available at:
http://www.nws.noaa.gov/oh/hrl/calb/calibration1102/main.htm (last
access: 7 September 2017), 2002.
Andersson, J. C. M., Pechlivanidis, I. G., Gustafsson, D., Donnelly, C., and
Arheimer, B.: Key factors for improving large-scale hydrological model
performance, Eur. Water, 49, 77–88, 2015.Andréassian, V.: Waters and forests: from historical controversy to
scientific debate, J. Hydrol., 291, 1–27,
10.1016/j.jhydrol.2003.12.015, 2004.
Arias, M. E., Lee, E., Farinosi, F., Pereira, F. F., Moorcroft, P. R., and
Briscoe, J.: Decoupling the effects of deforestation and climate variability
in large tropical river basins, J. Hydrol., in review, 2017.Arora, V. K., Chiew, F. H. S., and Grayson, R. B.: A river flow routing
scheme for general circulation models, J. Geophys. Res., 104, 14347,
10.1029/1999JD900200, 1999.Bahn, M., Reichstein, M., Dukes, J. S., Smith, M. D., and McDowell, N. G.:
Climate-biosphere interactions in a more extreme world, New Phytol., 202,
356–359, 10.1111/nph.12662, 2014.Baker, T. R., Phillips, O. L., Malhi, Y., Almeida, S., Arroyo, L., Di Fiore,
A., Erwin, T., Killeen, T. J., Laurance, S. G., Laurance, W. F., Lewis, S.
L., Lloyd, J., Monteagudo, A., Neill, D. A., Patino, S., Pitman, N. C. A.,
Silva, J. N. M., and
Vasquez Martinez, R.: Variation in wood density determines spatial patterns
in Amazonian forest biomass, Glob. Change Biol., 10, 545–562,
10.1111/j.1365-2486.2004.00751.x, 2004.Best, M. J., Pryor, M., Clark, D. B., Rooney, G. G., Essery, R. L. H.,
Ménard, C. B., Edwards, J. M., Hendry, M. A., Porson, A., Gedney, N.,
Mercado, L. M., Sitch, S., Blyth, E., Boucher, O., Cox, P. M., Grimmond, C.
S. B., and Harding, R. J.: The Joint UK Land Environment Simulator (JULES),
model description – Part 1: Energy and water fluxes, Geosci. Model Dev., 4,
677–699, 10.5194/gmd-4-677-2011, 2011.Brown, A. E., Zhang, L., McMahon, T. A., Western, A. W., and Vertessy, R. A.:
A review of paired catchment studies for determining changes in water yield
resulting from alterations in vegetation, J. Hydrol., 310, 28–61,
10.1016/j.jhydrol.2004.12.010, 2005.
Calvo-Alvarado, J., McDowell, N., and Waring, R.: Allometric relationships
predicting foliar biomass and leaf area: sapwood area ratio from tree height
in five Costa Rican rain forest species, Tree Physiol., 11, 1601–1608,
2008.
Carson, D.: Current parametrisations of land-surface processes in atmospheric
general circulation models, in: Land surface processes in atmospheric general
circulation models, edited by: Eagleson, P., Cambridge University Press,
Cambridge, UK, 1982.Clark, D. B., Mercado, L. M., Sitch, S., Jones, C. D., Gedney, N., Best, M.
J., Pryor, M., Rooney, G. G., Essery, R. L. H., Blyth, E., Boucher, O.,
Harding, R. J., Huntingford, C., and Cox, P. M.: The Joint UK Land
Environment Simulator (JULES), model description – Part 2: Carbon fluxes and
vegetation dynamics, Geosci. Model Dev., 4, 701–722,
10.5194/gmd-4-701-2011, 2011.Clark, M. P., Fan, Y., Lawrence, D. M., Adam, J. C., Bolster, D., Gochis, D.
J., Hooper, R. P., Kumar, M., Leung, L. R., Mackay, D. S., Maxwell, R. M.,
Shen, C., Swenson, S. C., and Zeng, X.: Improving the representation of
hydrologic processes in Earth System Models, Water Resour. Res., 51,
5929–5956, 10.1002/2015WR017096, 2015.Coe, M. T., Costa, M. H., and Howard, E. A.: Simulating the surface waters of
the Amazon River basin: impacts of new river geomorphic and flow
parameterizations, Hydrol. Process., 22, 2542–2553, 10.1002/hyp.6850,
2008.Cole, J. J., Prairie, Y. T., Caraco, N. F., McDowell, W. H., Tranvik, L. J.,
Striegl, R. G., Duarte, C. M., Kortelainen, P., Downing, J. A., Middelburg,
J. J., and Melack, J.: Plumbing the Global Carbon Cycle: Integrating Inland
Waters into the Terrestrial Carbon Budget, Ecosystems, 10, 172–185,
10.1007/s10021-006-9013-8, 2007.Cole, T. G. and Ewel, J. J.: Allometric equations for four valuable tropical
tree species, For. Ecol. Manage., 229, 351–360,
10.1016/j.foreco.2006.04.017, 2006.Collischonn, B., Collischonn, W., and Tucci, C. E. M.: Daily hydrological
modeling in the Amazon basin using TRMM rainfall estimates, J. Hydrol., 360,
207–216, 10.1016/j.jhydrol.2008.07.032, 2008.Collischonn, W., Allasia, D., Da Silva, B. C., and Tucci, C. E. M.: The
MGB-IPH model for large-scale rainfall–runoff modelling, Hydrol. Sci. J.,
52, 878–895, 10.1623/hysj.52.5.878, 2007.Cox, P. M., Betts, R. A., Bunton, C. B., Essery, R. L. H., Rowntree, P. R.,
and Smith, J.: The impact of new land surface physics on the GCM simulation
of climate and climate sensitivity, Clim. Dynam., 15, 183–203,
10.1007/s003820050276, 1999.Cunge, J. A.: On The Subject Of A Flood Propagation Computation Method
(Musklngum Method), J. Hydraul. Res., 7, 205–230,
10.1080/00221686909500264, 1969.
Farinosi, F., Arias, M. E., Lee, E., Longo, M., Pereira, F. F., Livino, A.,
Moorcroft, P. R., and Briscoe, J.: Future climate and land use change impacts
on river flows in the Tapajós Basin in the Brazilian Amazon, Earth's
Future, in review, 2017.Gerten, D., Schaphoff, S., Haberlandt, U., Lucht, W., and Sitch, S.:
Terrestrial vegetation and water balance – hydrological evaluation of a
dynamic global vegetation model, J. Hydrol., 286, 249–270,
10.1016/j.jhydrol.2003.09.029, 2004.Global Soil Data Task: Global Soil Data Products CD-ROM Contents (IGBP-DIS),
Data Set, Oak Ridge National Laboratory Distributed Active Archive Center,
Oak Ridge, Tennessee, USA, 10.3334/ORNLDAAC/565, 2014.Goldewijk, K. K.: Estimating global land use change over the past 300 years:
The HYDE Database, Global Biogeochem. Cy., 15, 417–433,
10.1029/1999GB001232, 2001.Gupta, H. V., Kling, H., Yilmaz, K. K., and Martinez, G. F.: Decomposition of
the mean squared error and NSE performance criteria: Implications for
improving hydrological modelling, J. Hydrol., 377, 80–91,
10.1016/j.jhydrol.2009.08.003, 2009.Hagemann, S. and Dumenil, L.: A parametrization of the lateral waterflow for
the global scale, Clim. Dynam., 14, 17–31, 10.1007/s003820050205, 1997.Hagemann, S. and Gates, L. D.: Validation of the hydrological cycle of ECMWF
and NCEP reanalyses using the MPI hydrological discharge model, J. Geophys.
Res., 106, 1503, 10.1029/2000JD900568, 2001.Hurtt, G. C., Pacala, S. W., Moorcroft, P. R., Caspersen, J., Shevliakova,
E., Houghton, R. A., and Moore, B.: Projecting the future of the U.S. carbon
sink, P. Natl. Acad. Sci. USA, 99, 1389–1394, 10.1073/pnas.012249999,
2002.Hurtt, G. C., Frolking, S., Fearon, M. G., Moore, B., Shevliakova, E.,
Malyshev, S., Pacala, S. W., and Houghton, R. A.: The underpinnings of
land-use history: three centuries of global gridded land-use transitions,
wood-harvest activity, and resulting secondary lands, Glob. Chang. Biol., 12,
1208–1229, 10.1111/j.1365-2486.2006.01150.x, 2006.Hurtt, G. C., Moorcroft, P. R., and Pacala, S. W.: Ecosystem Demography
Model: Scaling Vegetation Dynamics Across South America, Ecosyst. Demogr.
Model Scaling Veg. Dyn. Across South Am. Model Prod., available at:
http://daac.ornl.gov/MODELS/guides/EDM_SA_Vegetation.html (last access:
7 September 2017), 2013.Jiménez-Cisneros, B. E., Oki, T., Arnell, N. W., Benito, G., Cogley, J.
G., Döll, P., Jiang, T., and Mwakalila, S. S.: Freshwater resources, in:
Climate Change 2014: Impacts, Adaptation, and Vulnerability. Part A: Global
and Sectoral Aspects. Contribution of Working Group II to the Fifth
Assessment Report of the Intergovernmental Panel on Climate Change, edited
by: Field, C. B., Barros, V. R., Dokken, D. J., Mach, K. J., Mastrandrea, M.
D., Bilir, T. E., Chatterjee, M., Ebi, K. L., Estrada, Y. O., Genova, R. C.,
Girma, B., Kissel, E. S., Levy, A. N., MacCracken, S., Mastrandrea, P. R.,
and White, L. L., 229–269, Cambridge University Press, Cambridge, United
Kingdom and New York, NY, USA, available at:
https://www.ipcc.ch/pdf/assessment-report/ar5/wg2/WGIIAR5-Chap3_FINAL.pdf
(last access: 7 September 2017),
2014.Kauffeldt, A., Wetterhall, F., Pappenberger, F., Salamon, P., and Thielen,
J.: Technical review of large-scale hydrological models for implementation
in operational flood forecasting schemes on continental level, Environ.
Model. Softw., 75, 68–76, 10.1016/j.envsoft.2015.09.009, 2016.Kim, Y., Knox, R. G., Longo, M., Medvigy, D., Hutyra, L. R., Pyle, E. H.,
Wofsy, S. C., Bras, R. L., and Moorcroft, P. R.: Seasonal carbon dynamics and
water fluxes in an Amazon rainforest, Glob. Change Biol., 18, 1322–1334,
10.1111/j.1365-2486.2011.02629.x, 2012.Kling, H., Fuchs, M., and Paulin, M.: Runoff conditions in the upper Danube
basin under an ensemble of climate change scenarios, J. Hydrol., 424–425,
264–277, 10.1016/j.jhydrol.2012.01.011, 2012.Knox, R. G.: Land Conversion in Amazonia and Northern South America:
Influences on Regional Hydrology and Ecosystem Response, PhD Thesis,
Massachusetts Institute of Technology, available at:
https://dspace.mit.edu/handle/1721.1/79489 (last access:
7 September 2017), 2012.Knox, R. G., Longo, M., Swann, A. L. S., Zhang, K., Levine, N. M., Moorcroft,
P. R., and Bras, R. L.: Hydrometeorological effects of historical
land-conversion in an ecosystem-atmosphere model of Northern South America,
Hydrol. Earth Syst. Sci., 19, 241–273,
10.5194/hess-19-241-2015, 2015.Kucharik, C. J., Foley, J. A., Delire, C., Fisher, V. A., Coe, M. T.,
Lenters, J. D., Young-Molling, C., Ramankutty, N., Norman, J. M., and Gower,
S. T.: Testing the performance of a dynamic global ecosystem model: Water
balance, carbon balance, and vegetation structure, Global Biogeochem. Cy.,
14, 795–825, 10.1029/1999GB001138, 2000.Lawrence, D. M., Oleson, K. W., Flanner, M. G., Thornton, P. E., Swenson, S.
C., Lawrence, P. J., Zeng, X., Yang, Z.-L., Levis, S., Sakaguchi, K., Bonan,
G. B., and Slater, A. G.: Parameterization improvements and functional and
structural advances in Version 4 of the Community Land Model, J. Adv. Model.
Earth Syst., 3, M03001, 10.1029/2011MS000045, 2011.Lejeune, Q., Davin, E. L., Guillod, B. P., and Seneviratne, S. I.: Influence
of Amazonian deforestation on the future evolution of regional surface
fluxes, circulation, surface temperature and precipitation, Clim. Dynam., 44,
2769–2786, 10.1007/s00382-014-2203-8, 2015.Li, R., Chen, Q., and Ye, F.: Modelling the impacts of reservoir operations
on the downstream riparian vegetation and fish habitats in the Lijiang River,
J. Hydroinformatics, 13, 229, 10.2166/hydro.2010.008, 2011.
Liang, X., Lettenmaier, D. P., Wood, E. F., and Burges, S. J.: A simple
hydrologically based model of land surface water and energy fluxes for
general circulation model, J. Geophys. Res., 99, 14415–14428, 1994.Lobligeois, F., Andréassian, V., Perrin, C., Tabary, P., and Loumagne,
C.: When does higher spatial resolution rainfall information improve
streamflow simulation? An evaluation using 3620 flood events, Hydrol. Earth
Syst. Sci., 18, 575–594, 10.5194/hess-18-575-2014, 2014.Longo, M.: Amazon Forest Response to Changes in Rainfall Regime: Results from
an Individual-Based Dynamic Vegetation Model, Harvard University, available
at: http://dash.harvard.edu/handle/1/11744438 (last access:
7 September 2017), 2014.Medvigy, D., Wofsy, S. C., Munger, J. W., Hollinger, D. Y., and Moorcroft, P.
R.: Mechanistic scaling of ecosystem function and dynamics in space and time:
Ecosystem Demography model version 2, J. Geophys. Res.-Biogeo., 114, G01002,
10.1029/2008JG000812, 2009.Medvigy, D., Walko, R. L., and Avissar, R.: Effects of Deforestation on
Spatiotemporal Distributions of Precipitation in South America, J. Climate,
24, 2147–2163, 10.1175/2010JCLI3882.1, 2011.
Miller, W. A. and Cunge, J. A.: Simplified equations of unsteady flow, in:
Unsteady Flow in Open Channels, edited by: Mahmood, K. and Yevjevich, V.,
Colorado State University, Water Resources Publication, Fort Collins, CO,
USA, 1975.Mohor, G. S., Rodriguez, D. A., Tomasella, J., and Siqueira Júnior, J.
L.: Exploratory analyses for the assessment of climate change impacts on the
energy production in an Amazon run-of-river hydropower plant, J. Hydrol.
Reg. Stud., 4, 41–59, 10.1016/j.ejrh.2015.04.003, 2015.Moorcroft, P. R., Hurtt, G. C., and Pacala, S. W.: A method for scaling
vegetation dynamics: The ecosystem demography model (ED), Ecol. Monogr.,
71, 557–586, 10.1890/0012-9615(2001)071[0557:AMFSVD]2.0.CO;2, 2001.
Nash, E. and Sutcliffe, V.: River flow forecasting Through conceptual models
PART I- A Discussion of principles, J. Hydrol., 10, 282–290, 1970.Observation Service SO HYBAM: SO HYBAM – Geodynamical, hydrological and
biogeochemical control of erosion/alteration and material transport in the
Amazon, Orinoco and Congo basins, available from:
http://www.ore-hybam.org/index.php/eng, (last access: 7 September
2017), 2016.
Oki, T., Nishimura, T., and Dirmeyer, P.: Assessment of Annual Runoff from
Land Surface Models Using Total Runoff Integrating Pathways (TRIP), J.
Meteorol. Soc. Japan, 77, 235–255, 1999.Oki, T., Agata, Y., Kanae, S., Saruhashi, T., Yang, D., and Musiake, K.:
Global assessment of current water resources using total runoff integrating
pathways, Hydrol. Sci. J., 46, 983–995, 10.1080/02626660109492890,
2001.Oleson, K. W., Lawrence, D. M., Bonan, G. B., Flanner, M. G., Kluzek, E.,
Lawrence, P. J., Levis, S., Swenson, S. C., and Thornton, P. E.: Technical
Description of version 4.0 of the Community Land Model (CLM), Boulder, CO,
USA, available at:
http://www.cesm.ucar.edu/models/cesm1.0/clm/CLM4_Tech_Note.pdf (last
access: 7 September 2017),
2010.Ostberg, S., Schaphoff, S., Lucht, W., and Gerten, D.: Three centuries of
dual pressure from land use and climate change on the biosphere, Environ.
Res. Lett., 10, 044011, 10.1088/1748-9326/10/4/044011, 2015.Paiva, R. C. D., Collischonn, W., and Tucci, C. E. M.: Large scale hydrologic
and hydrodynamic modeling using limited data and a GIS based approach, J.
Hydrol., 406, 170–181, 10.1016/j.jhydrol.2011.06.007, 2011.Paiva, R. C. D., Buarque, D. C., Collischonn, W., Bonnet, M. P., Frappart,
F., Calmant, S., and Bulhões Mendes, C. A.: Large-scale hydrologic and
hydrodynamic modeling of the Amazon River basin, Water Resour. Res., 49,
1226–1243, 10.1002/wrcr.20067, 2013a.Paiva, R. C. D., Collischonn, W., and Buarque, D. C.: Validation of a full
hydrodynamic model for large-scale hydrologic modelling in the Amazon,
Hydrol. Process., 27, 333–346, 10.1002/hyp.8425, 2013b.Paz, A. R., Collischonn, W., and Lopes da Silveira, A. L.: Improvements in
large-scale drainage networks derived from digital elevation models, Water
Resour. Res., 42, W08502, 10.1029/2005WR004544, 2006.Pearson, K.: Note on regression and inheritance in the case of two parents,
Proc. R. Soc. London, 58, 240–242, 10.1098/rspl.1895.0041, 1895.Pearson, R. G., Phillips, S. J., Loranty, M. M., Beck, P. S. A., Damoulas,
T., Knight, S. J., and Goetz, S. J.: Shifts in Arctic vegetation and
associated feedbacks under climate change, Nat. Clim. Chang., 3,
673–677, 10.1038/nclimate1858, 2013.
Pontes, P. R. M., Collischonn, W., Fan, F. M., Paiva, R. C. D., and Buarque,
D. C.: Modelagem hidrológica e hidráulica de grande escala com
propagação inercial de vazões, Rev. Bras. Recur. Hídricos,
20, 888–904, 2015.Poorter, L., Bongers, L., and Bongers, F.: Architecture of 54 moist-forest
tree species: traits, trade-offs, and functional groups, Ecology, 87,
1289–1301, 10.1890/0012-9658(2006)87[1289:AOMTST]2.0.CO;2, 2006.Prigent, C., Papa, F., Aires, F., Rossow, W. B., and Matthews, E.: Global
inundation dynamics inferred from multiple satellite observations,
1993–2000, J. Geophys. Res., 112, D12107, 10.1029/2006JD007847,
2007.Quesada, C. A., Lloyd, J., Schwarz, M., Patiño, S., Baker, T. R.,
Czimczik, C., Fyllas, N. M., Martinelli, L., Nardoto, G. B., Schmerler, J.,
Santos, A. J. B., Hodnett, M. G., Herrera, R., Luizão, F. J., Arneth, A.,
Lloyd, G., Dezzeo, N., Hilke, I., Kuhlmann, I., Raessler, M., Brand, W. A.,
Geilmann, H., Moraes Filho, J. O., Carvalho, F. P., Araujo Filho, R. N.,
Chaves, J. E., Cruz Junior, O. F., Pimentel, T. P., and Paiva, R.: Variations
in chemical and physical properties of Amazon forest soils in relation to
their genesis, Biogeosciences, 7, 1515–1541,
10.5194/bg-7-1515-2010, 2010.Raddatz, T. J., Reick, C. H., Knorr, W., Kattge, J., Roeckner, E., Schnur,
R., Schnitzler, K.-G., Wetzel, P., and Jungclaus, J.: Will the tropical land
biosphere dominate the climate–carbon cycle feedback during the twenty-first
century?, Clim. Dynam., 29, 565–574, 10.1007/s00382-007-0247-8, 2007.Ramankutty, N. and Foley, J. A.: Estimating historical changes in global land
cover: Croplands from 1700 to 1992, Global Biogeochem. Cy., 13, 997–1027,
10.1029/1999GB900046, 1999.Reed, S. M.: Deriving flow directions for coarse-resolution (1–4 km)
gridded hydrologic modeling, Water Resour. Res., 39, SWC 4,
10.1029/2003WR001989, 2003.Rost, S., Gerten, D., Bondeau, A., Lucht, W., Rohwer, J., and Schaphoff, S.:
Agricultural green and blue water consumption and its influence on the global
water system, Water Resour. Res., 44, W09405, 10.1029/2007WR006331,
2008.Sheffield, J., Goteti, G., and Wood, E. F.: Development of a 50-Year
High-Resolution Global Dataset of Meteorological Forcings for Land Surface
Modeling, J. Climate, 19, 3088–3111, 10.1175/JCLI3790.1, 2006.Sitch, S., Smith, B., Prentice, I. C., Arneth, A., Bondeau, A., Cramer, W.,
Kaplan, J. O., Levis, S., Lucht, W., Sykes, M. T., Thonicke, K., and
Venevsky, S.: Evaluation of ecosystem dynamics, plant geography and
terrestrial carbon cycling in the LPJ dynamic global vegetation model, Glob.
Change Biol., 9, 161–185, 10.1046/j.1365-2486.2003.00569.x, 2003.Smith, M. B., Koren, V. I., Zhang, Z., Reed, S. M., Pan, J.-J., and Moreda,
F.: Runoff response to spatial variability in precipitation: an analysis of
observed data, J. Hydrol., 298, 267–286, 10.1016/j.jhydrol.2004.03.039,
2004.Soares-Filho, B. S., Nepstad, D. C., Curran, L. M., Cerqueira, G. C., Garcia,
R. A., Ramos, C. A., Voll, E., McDonald, A., Lefebvre, P., and Schlesinger,
P.: Modelling conservation in the Amazon basin, Nature, 440, 520–523,
10.1038/nature04389, 2006.Swann, A. L. S., Longo, M., Knox, R. G., Lee, E., and Moorcroft, P. R.:
Future deforestation in the Amazon and consequences for South American
climate, Agric. For. Meteorol., 214–215, 12–24,
10.1016/j.agrformet.2015.07.006, 2015.USACE: A Muskingum-Cunge Channel Flow Routing Method for Drainage Networks,
available at:
http://www.hec.usace.army.mil/publications/TechnicalPapers/TP-135.pdf
(last access: 7 September 2017), 1991.USGS: Shuttle Radar Topography Mission (SRTM), available from:
https://lta.cr.usgs.gov/SRTM1Arc, (last access: 7 September 2017),
2016.
Vamborg, F. S. E., Brovkin, V., and Claussen, M.: The effect of a dynamic
background albedo scheme on Sahel/Sahara precipitation during the
mid-Holocene, Clim. Past, 7, 117–131, 10.5194/cp-7-117-2011,
2011.Walko, R. L., Band, L. E., Baron, J., Kittel, T. G. F., Lammers, R., Lee, T.
J., Ojima, D., Pielke, R. A., Taylor, C., Tague, C., Tremback, C. J., and
Vidale, P. L.: Coupled Atmosphere–Biophysics–Hydrology Models for
Environmental Modeling, J. Appl. Meteorol., 39, 931–944,
10.1175/1520-0450(2000)039<0931:CABHMF>2.0.CO;2, 2000.Wohl, E., Barros, A., Brunsell, N., Chappell, N. A., Coe, M., Giambelluca,
T., Goldsmith, S., Harmon, R., Hendrickx, J. M. H., Juvik, J., McDonnell, J.,
and Ogden, F.: The hydrology of the humid tropics, Nat. Clim. Chang., 2,
655–662, 10.1038/nclimate1556, 2012.Yamazaki, D., Kanae, S., Kim, H., and Oki, T.: A physically based description
of floodplain inundation dynamics in a global river routing model, Water
Resour. Res., 47, W04501, 10.1029/2010WR009726, 2011.Zambrano-Bigiarini, M.: hydroGOF: Goodness-of-fit functions for comparison of
simulated and observed hydrological time series. R package version 0.3–8,
available at: http://cran.r-project.org/package=hydroGOF
(last access: 7 September 2017), 2014.Zhang, K., de Almeida Castanho, A. D., Galbraith, D. R., Moghim, S., Levine,
N. M., Bras, R. L., Coe, M. T., Costa, M. H., Malhi, Y., Longo, M., Knox, R.
G., McKnight, S., Wang, J., and Moorcroft, P. R.: The fate of Amazonian
ecosystems over the coming century arising from changes in climate,
atmospheric CO2, and land use, Glob. Change Biol., 21, 2569–2587,
10.1111/gcb.12903, 2015.Zulkafli, Z., Buytaert, W., Onof, C., Lavado, W., and Guyot, J. L.: A
critical assessment of the JULES land surface model hydrology for humid
tropical environments, Hydrol. Earth Syst. Sci., 17, 1113–1132,
10.5194/hess-17-1113-2013, 2013.