You are on page 1of 12

Journal of Hydrology 406 (2011) 170181

Contents lists available at ScienceDirect

Journal of Hydrology
journal homepage: www.elsevier.com/locate/jhydrol

Large scale hydrologic and hydrodynamic modeling using limited data


and a GIS based approach
Rodrigo C.D. Paiva , Walter Collischonn, Carlos E.M. Tucci
Instituto de Pesquisas Hidrulicas IPH, Universidade Federal do Rio Grande do Sul UFRGS, Av. Bento Gonalves, 9500 Porto Alegre, 90050-260 RS, Brazil

a r t i c l e

i n f o

Article history:
Received 5 October 2010
Received in revised form 6 June 2011
Accepted 12 June 2011
Available online 25 June 2011
This manuscript was handled by K.
Georgakakos, Editor-in-Chief, with the
assistance of Enrique R. Vivoni, Associate
Editor
Keywords:
Large scale hydrodynamic model
Amazon
Flow routing
Flood inundation
STRM DEM
GIS algorithm

s u m m a r y
In this paper, we present a large-scale hydrologic model with a full one-dimensional hydrodynamic module to calculate ow propagation on a complex river network. The model uses the full SaintVenant equations and a simple oodplain storage model, and therefore is capable of simulating a wide range of uvial
processes such as ood wave delay and attenuation, backwater effects, ood inundation and its effects on
ood waves. We present the model basic equations and GIS algorithms to extract model parameters from
relatively limited data, which is globally available, such as the SRTM DEM. GIS based algorithms include
the estimation of river width and depth using geomorphological relations, river cross section bottom
level and oodplain geometry extracted from DEM, etc. We also show a case study on one of the major
tributaries of the Amazon, the Purus River basin. A model validation using discharge and water level data
shows that the model is capable of reproducing the main hydrological features of the Purus River basin.
Also, realistic oodplain inundation maps were derived from the results of the model. Our main conclusion is that it is possible to employ full hydrodynamic models within large-scale hydrological models
even using limited data for river geometry and oodplain characterization.
2011 Elsevier B.V. All rights reserved.

1. Introduction
Large-scale hydrological models have been used for the impact
assessment of land use change (e.g. Matheussen et al., 2000) and
climate change (e.g. Nijssen et al., 2001; Xu et al., 2005) on water
resources, for the study of hydrobiogeochemical processes (e.g. Foley et al., 1996; Zhang et al., 2002), as a basis for hydrological forecast systems (e.g. Wood et al., 2002; Collischonn et al., 2005;
Thielen et al., 2009) and also as the land surface component of climate models (Pitman, 2003; Wood et al., 1992; Liang et al., 1994).
These models are in constant development, partly as a consequence of a growing understanding of the hydrological processes,
partly due to increasing computational capacity, and also due to
the growing availability of remote sensing data sources (Schmugge
et al., 2002), such as energy uxes estimates (Bastiaanssen et al.,
1998) and river water level (Frappart et al., 2006; Getirana et al.,
2010).
Most of the large-scale hydrological models are conceptual
models having at least a small physical basis. Some of them try
to represent the role of different vegetation and soil types on the
streamow generation processes and on water and energy budgets
Corresponding author. Tel.: +55 5133087511.
E-mail addresses: rodrigocdpaiva@gmail.com (R.C.D. Paiva), collischonn@
iph.ufrgs.br (W. Collischonn), tucci@rhama.net (C.E.M. Tucci).
0022-1694/$ - see front matter 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.jhydrol.2011.06.007

of the basin. These features can be found in such models as VIC


(Liang et al., 1994), LASCAM (Viney et al., 2000), TOPKAPI model
(Liu and Todini, 2002), and IBISTHMB (Coe et al., 2007). However,
one key aspect in the behavior of large-scale river basins that has
been partially neglected in this generation of large-scale hydrological models is river hydrodynamics.
Large-scale hydrological models usually use simple ow routing
models that focus only on the ood wave delay and attenuation,
such as storage or linear reservoir based models (Coe, 2000,
1997; Arora, 2001; Liston et al., 1994; Decharme et al., 2008), linearized SaintVenant equations representing wave advection and
diffusion (Lohmann et al., 1996), kinematic wave approximations
(Liu and Todini, 2002; De Roo et al., 2000), or the MuskingumCunge model (Collischonn et al., 2007) and its variations (Beighley
et al., 2009; Todini, 2007; Koussis, 2009).
While these simplied ow routing methods can in principle
adequately represent ood wave delay and attenuation, they cannot deal with some of the hydrodynamic processes that are often
present in large-scale river systems, notably backwater and oodplain storage effects. This is the case in the Amazon basin, where
ow of several rivers is controlled by backwater effects, as reported
by Meade et al. (1991), Trigg et al. (2009) and Tomasella et al.
(2010), and the oodplains are a signicant component of the uvial system (Bonnet et al., 2008). Floodplains also play an important role in many other South American river basins (Hamilton

R.C.D. Paiva et al. / Journal of Hydrology 406 (2011) 170181

et al., 2002), in the transport of dissolved and particulate materials


(e.g. Bourgoin et al., 2007), in biogeochemical processes and in the
global carbon cycle (Richey et al., 2002; Zhang et al., 2002). Floodplains may also inuence feedback effects between land surface
and atmosphere (Mohamed et al., 2005; Collini et al., 2008; Betts
et al., 1996; Silva Dias et al., 2004). The explicit simulation of ood
inundation could provide the basis for the representation of all
these processes.
Hydrodynamic ow routing in large-scale hydrological models
could also provide results of other ow related variables, such as
river stage, velocity and slope. These models could also be used
as a basis for transport models by solving the advectiondiffusion
equation and sediment transport equations.
One dimensional (1D) hydrodynamic river models are well
developed since the late 1970s (Cunge et al., 1980), and there are
even well known packages for 1D hydrodynamic modeling (USACE,
2002; DHI, 2003). These models have been applied in relatively
large-scale problems (Paz et al., 2010; Remo and Pinter, 2007;
Biancamaria et al., 2009; Lian et al., 2007), however their use within large-scale distributed hydrological models, which also represent rainfall-runoff processes, is uncommon. There are also
approaches for ood inundation simulation ranging from simplied models such as a composite cross section (Cunge et al.,
1980) to 2D models (Galland et al., 1991; Bates and De Roo, 2000).
The reason why simple ow routing methods are still preferred
within large-scale hydrological models is probably related to the
lower input data needs and the lower computational burden they
impose, while still providing reasonable results in several cases.
Hydrodynamic river models need river cross section data that are
usually surveyed only for relatively short reaches. Hydrodynamic
models are also more computational intensive than simplied ow
routing methods, and require more detailed information on boundary and initial conditions.
However, computing capacity keeps growing, and also new and
higher accuracy remote sensing data such as the Digital Elevation
Model (DEM) from the Shuttle Radar Topography Mission (SRTM)
(Rabus et al., 2003; Farr et al., 2007) can be used to provide some
of the information that is needed to compute parameters for large
scale hydrodynamic models. In a recent example, Trigg et al. (2009)
showed that the use of a simplied bathymetry, i.e. a wide rectangular channel approximation and a simple bed slope, does not
introduce signicant errors in the hydrodynamic simulation results in part of the central Amazon.
In the present paper, we describe the development of a model
for large scale hydrologic and hydrodynamic modeling. The model
uses limited data that can be usually found in global scale for most
river basins, mainly SRTM DEM. GIS methods are used to extract
the needed information from SRTM DEM. The proposed model is
basically an improvement of the MGB-IPH large scale hydrological
model (Collischonn et al., 2007). First, a brief description of the
MGB-IPH hydrological model is presented (Section 2). Later, details
concerning the hydrodynamic model equations, then a representation of the oodplains and the numerical scheme are shown (Section 3). We also present the GIS based algorithms developed to
extract the hydrodynamic model parameters from the DEM (Section 4) and a simple approach for 2D ood inundation simulation
(Section 5). The model is tested in a case study of the river Purus,
one of the major tributaries of the Amazon (Section 6). Other results for an even larger region, and comparisons with simplied
ow routing methods are presented by Paiva et al. (in review).

2. The hydrological model


The MGB-IPH model (Modelo Hidrolgico de Grandes Bacias)
is a large scale distributed hydrological model. It is similar to other

171

large scale hydrological models such as LARSIM (Ludwig and


Bremicker, 2006) and VIC (Liang et al., 1994; Nijssem et al.,
1997). It is a process based model that uses physical and conceptual equations to simulate the terrestrial hydrological cycle: soil
water budget, energy budget and evapotranspiration, interception,
supercial, sub-supercial and groundwater ow generation and
routing and river ow routing. It uses a daily or hourly simulation
time step.
The early version of the MGB-IPH model was based on a square
cell discretization of the river basin (Collischonn et al., 2007), while
the version described in the present paper uses a division of the basin in small catchments using the ArcHydro methods (Maidment,
2002). Each catchment is subdivided into Hydrological Response
Units (HRUs) which are areas with similar hydrological behavior
and dened by a combination of soil and land cover maps (Beven,
2001; Kouwen et al., 1993). Vertical water and energy budgets are
computed independently for each HRU in each catchment. Canopy
interception is represented by a reservoir with maximum storage
as a function of vegetation leaf area index. Soil water budget is
computed simulating the soil as a single water reservoir. Soil inltration and runoff are computed based on the variable contributing
area concept of the ARNO model (Todini, 1996), which is also used
in the PDM (Moore and Clarke, 1981), VIC2L and LARSIM models.
Energy budget is computed using basic surface meteorological
variables and evapotranspiration from soil, vegetation and canopy
into the atmosphere is estimated based on the Penman Monteith
approach (Monteith, 1965; Shuttleworth, 1993; Allen et al.,
1998). Subsurface ow is computed using an equation similar to
the Brooks and Corey unsaturated hydraulic conductivity equation
(Rawls et al., 1993). Percolation from soil layer to groundwater is
calculated according to a simple linear relation between soil water
storage and maximum soil water storage. Then, the ow generated
within each catchment is routed to the stream network using three
linear reservoirs (base ow, subsurface ow and surface ow). A
detailed description of the model is given by Collischonn et al.
(2007), and recent applications of the model are presented by Getirana et al. (2010), Collischonn et al. (2008), and Collischonn et al.
(2005).
Until now, river ow routing within the MGB-IPH model used to
be calculated using the Muskingum Cunge method, but in this paper we describe a mixed method, which employs a hydrodynamic
model in at reaches of the main rivers and the MuskingumCunge
method in the headwater parts of the drainage network, where
slope is usually higher.
3. The hydrodynamic model
The hydrodynamic model is based on the IPH-IV model, rst
developed by Tucci (1978). The model solves the full Saint Venant
equations (Cunge et al., 1980):

@Q
@h
b
qcat  qfl
@x
@t


@Q
@Q
@h
@A
2v
gA  v 2 b
 v 2 
gAS0  Sf
@t
@x
@x
@x hcte

where the rst and second equations are the 1D channel mass and
momentum conservation laws, Q (m3 s1) is river discharge, t (s) is
time, x (m) is river longitudinal space coordinate, b (m) is river cross
section width at free surface elevation, qcat (m2 s1) is local catchment lateral inow (the sum of the surface, subsurface and base
ow from the catchment), q (m2 s1) is the river-oodplain ow
exchange, h (m) is water depth, v (m s1) is ow velocity averaged
over the cross section, g (m s2) is acceleration due to gravity, A
(m2) is the cross sectional ow area perpendicular to the ow direction and S0 (m m1) and Sf (m m1) are the bed slope and friction

172

R.C.D. Paiva et al. / Journal of Hydrology 406 (2011) 170181

slope in the x-direction. Friction slope is estimated using Mannings


equation. Flow at river conuences is modeled using a simple mass
continuity equation and the energy equation discarding energy
losses and the kinetic term (Cunge et al., 1980). It is also possible
to model hydraulic river singularities such as dams or other hydraulic structures using internal boundary conditions.
The river reaches are discretized into several river cross sections
(Fig. 1) where the hydraulic variables are computed. The model
also divides the catchments into oodplain units (Fig. 1). These
are areas between two river cross sections where the riveroodplain ow exchange and oodplain water storage are
computed.
Flood inundation is simulated using a simple storage model
(Cunge et al., 1980), which assumes that (i) the ow velocity parallel to the river direction is null on oodplains, (ii) the oodplain
units act only as storage areas and (iii) the oodplain water level
equals the water level at the main channel. Considering the model
basic assumptions and the mass conservation law, the
river-oodplain ow exchange q equals:

qfl

Aflz @h
@h
Lz
dx @t
@t

where A (m2) is the ooded area and L (m) is the oodplain equivalent width, measured for each oodplain unit, as described in
Section 4.5.
This is a simplistic approach that cannot fully represent all aspects of oodplain hydrodynamics such as oodplain water levels
that are not equal to the adjacent main channel, oodplain ows
subparallel to the main river channel, and bidirectional oodplain
ows, that where observed by Alsdorf et al. (2007) using remote
sensing data in the Amazon. Additionally, as observed by Alsdorf
et al. (2003), water level changes in the main channel may not instantly propagate across the entire oodplain, as is implicitly considered in the model described here. However, the proposed
model, which is relatively complex in terms of river hydraulics,
but somewhat simplied in terms of oodplain simulation, may
be appropriate because (i) this model is oriented to large scale
applications and aimed at providing discharge, water level and
ood extent results, by simulating translation and diffusion of
ood waves, backwater effects and the inuence of oodplain storage on ood waves, (ii) it is not oriented for studying smaller scale
oodplain hydraulics, (iii) it is compatible with large scale applica-

tions, different from complex oodplain models which require


higher computational resources and detailed oodplain bathymetry data, that cannot be provided by current global scale DEMs
and (iv) it is in accordance with Trigg et al. (2009), who stated that
it is important to get the hydraulics of the main channel right before tackling the more complex interactions with the oodplain.
The partial differential equations of the model are solved using
a linear and implicit nite difference numerical scheme developed
by Chen (1973), similar to the Preissman scheme (Cunge et al.,
1980). Since the model simulates a river network with lots of conuences, the set of discretized equations forms a non symmetric
sparse linear system. Then, for better computational efciency,
the matrix solver uses a modied Gauss elimination procedure
based on a skyline storage method, avoiding the storage of null elements of the linear system of equations. This method was developed by Tucci (1978) and is similar to methods used by HEC-RAS
(USACE, 2002). In addition, a modication to the method developed by Tucci (1978) is proposed. An additional pointer is set in
the beginning of the simulation to store only algebraic operations
that are actually needed for the Gauss elimination algorithm, i.e.
it avoids operations with null elements, and then computes only
those during the rest of the simulation. This improvement increased the speed of calculations but it will not be fully described
here.
4. GIS based algorithms
Any hydrodynamic model should be applied using data from
detailed surveys of river cross sections and, when necessary, oodplain extent. This kind of data is usually available only at small
scale, for relatively short river reaches. For large scale hydrologic
modeling the necessary information has to be derived from globally available data sets.
In our case, which is focused on applications of the model in
large river basins, most of the necessary data is extracted from Digital Elevation Models (DEM) (e.g. the SRTM DEM described in Rabus et al., 2003; Farr et al., 2007) and some GIS algorithms. The
algorithms use a raster representation of topography and all other
spatial variables. The spatial variables are described through matrices of r rows and c columns, which store gridded values in r  c
positions dened as pixels. The pixel hi, ji corresponds to the element of a matrix located at row i and column j. The Digital Elevation Model is denoted by Z hi, ji and provides the terrain elevation
in each point hi, ji. The algorithms for model parameter extraction
include, as presented in Fig. 2, (1) DEM preprocessing, (2) river network and catchment discretization, (3) river cross section discretization, location and topology, (4) river cross section geometry, (5)
river cross section bottom level, (6) oodplain geometry. These
procedures are presented in the following sections. Examples of results from these algorithms are presented in Section 6.3.
4.1. DEM preprocessing and catchment discretization
DEM preprocessing and the river and catchment discretization
algorithms presented in this section are computed using well
known algorithms as presented in Jenson and Domingue (1988)
and available in several GIS packages such as the ArcHydro tools
(Maidment, 2002).
The DEM preprocessing algorithms include rst the computation of the ow direction dhi, ji raster. Flow direction is a function
of DEM that returns the direction of the largest slope (N, NE, E,
SE, S, SW, W, NW or null). Given the ow direction, the coordinates
of the downstream pixel hydrologically connected to hi, ji are given
by the function F:

Fig. 1. Model discretization into catchments (grey lines), river reaches (black lines),
river cross sections (grey dots) and oodplain units (grey thin lines).

Fhi; ji hi; ji Ddhi; ji

DEM

Flow direcon
and DEM pit
removal

Cross secon
geometry

Flow
accumulaon

Drainage

Hydrological
model input

Cross secon
discresaon

River and
catchment
discresaon

River prole
and cross
secon
boom level

Hydrodynamic model input

DEM preprocessing

Input data

R.C.D. Paiva et al. / Journal of Hydrology 406 (2011) 170181

Floodplain
geometry

Fig. 2. Flow chart of GIS based algorithms.

considering D(dhi, ji) a function that returns the relative position


hDi, Dji of the downstream pixel (e.g. D(N) = h+1, +1i or
D(NE) = h+1, 1i). The ow direction algorithm also includes the removal of pits from the DEM to ensure hydrological continuity (Jenson and Domingue, 1988). The next step is the computation of the
ow accumulation raster A hi, ji, that returns for each pixel h i , j i,
the sum of the surface area of all upstream pixels (Burrough and
McDonnell, 1998).
Later, using the ow direction and ow accumulation results,
the river and catchment discretization is performed. The drainage
D hi, ji maps the river network on pixels with ow accumulation
A hi, ji larger than a threshold Amin (Burrough and McDonnel,
1998). Then, the river network is segmented into several river
reaches, dened by river segments between upstream and downstream conuences (or between a conuence and a river source)
and each river reach gets a different code. Finally, catchment discretization is performed by identifying pixels located upstream of
each river reach and the result is the catchment raster B hi, ji. For
each pixel hi, ji, the algorithms follow downstream the ow path
dened by the ow direction using the function F until it nds a
river reach and assigns B hi, ji equal to the river reach code (Maidment, 2002).
4.2. Cross section discretization
A river reach within a catchment is then subdivided in several
shorter sub-reaches, as shown in Fig. 1. These sub-reaches are limited by the positioning of model river cross-sections, where
hydraulic variables Q and h are computed. For a given river reach
p, the number of cross sections is dened as np = Lp/Dxp + 1, where
Dxp is the distance between two cross sections and L is the length
of the reach p. A predened value of the distance Dx between the
cross sections is dened considering the spatial scale of interest
and the performance of the numerical scheme of the hydrodynamic model (Cunge et al., 1980) and can be estimated a priori
considering recommendations such as Dx < cT/M as presented in
Castellarin et al. (2009), where. c = (gh)0.5 is the ood wave celerity,
T is the period of the ood wave and 30 < M < 50.
Then, the location of model cross sections is dened by dividing
the river reach in a number of sub-reaches that results in the subreach length (Dxp) as close as possible to a predened value. The

173

algorithm starts from any pixel of the reach p and then follows upstream the river ow path using the ow direction raster, the
drainage raster and the inverse of the function F (Eq. (4)) until it
nds the upstream extreme of the reach. The coordinates of the
rst cross section of the reach p equals this current position. Then
it sets the coordinates of the second cross section (m = 2) by
repeating the following steps: (1) follow the reach ow path downstream using Eq. (4) and compute the accumulated distance x from
the rst cross section; (2) when x P (m  1). Dxp for the rst time,
set the coordinates of the mth cross section equal to the current
position; (3) set m = m + 1 to compute the next cross section and
repeat steps 13.
4.3. Cross section geometry
River cross sections are represented by a rectangular shape,
with width B (m) and maximum depth H (m). Parameters B and
H are estimated trough geomorphologic equations of the type
H = a  Abd , where a and b are parameters and Ad is the upstream
drainage area. This kind of approach is used in several large scale
simplied ow routing algorithms, such as Coe et al. (2007), Arora
(2001) or Decharme et al. (2008). The parameters used in the case
study are presented in Section 6.3.
4.4. Cross section bottom level
This section describes the procedure for estimating cross section bottom levels. This procedure takes three steps: (i) rst a river
longitudinal prole is extracted from the DEM for each river reach.
Then, two corrections are applied to this longitudinal prole: (ii)
removal of effects of vegetation and water depth and (iii) removal
of random noise using an iterative moving average lter. The details of these procedures are described below.
A river longitudinal prole is extracted from the DEM for each
reach to estimate river bottom level of computational cross sections. The algorithm follows the reach path from upstream to
downstream as described in Section 4.2 and computes terrain elevation zDEM as a function of distance x from the beginning of the
reach:

zDEM x Zhi; ji

The elevations of the river longitudinal prole extracted from


SRTM DEM are not equal to river bottom level z0(x) (m). Systematic
errors related to vegetation, surface water effects and also random
error related to noise on DEM data (Rabus et al., 2003; Sun et al.,
2003; Farr et al., 2007; Kellndorfer et al., 2004; Lehner et al.,
2006) affect the SRTM DEM elevation, which we call zDEM(x). We
consider that the river bottom level z0(x) can be estimated using
the following equation:

z0 x zDEM x  Hx  Hv eg x  ex

where H (m) is the water depth, Hveg (m) is the effective vegetation
height (different from real vegetation height since SRTM C-band radar penetrates the canopy to some extent) and e (m) is a random
noise, as illustrated in Figs. 3 and 4.
The effective vegetation height Hveg is expected to depend on
the actual vegetation height, and on the ratio between river width
B (m) and DEM spatial resolution res (m). For different relations between river width and DEM resolution, DEM elevation may be
close to the actual water stage or more perturbed by vegetation
height, as shown in Fig. 4. In large rivers, the elevation obtained
from SRTM data is close to water level. In narrow rivers, zDEM is
close to the elevation of the vegetation canopy, which covers the
river. Considering this, a preliminary longitudinal prole of bottom
levels z0 is estimated by:

174

R.C.D. Paiva et al. / Journal of Hydrology 406 (2011) 170181

rivers. At the extreme of the reaches, the lter also takes a special
form: zi,t + 1 = 0.95zi,t + 0.05zi + 1,t. The main parameter of the lter
is the maximum number of iterations itmax. Its choice is subjective
and a verication of the results is needed. A large (or small) value
of itmax gives smooth (or noisy) longitudinal proles. This iterative
moving average lter approach was preferred under a simpler polynomial t for each river reach (e.g. LeFavour and Alsdorf, 2005) because the latter creates discontinuities in the conuences and
between neighbor river reaches.
4.5. Floodplain geometry

Fig. 3. River longitudinal prole extracted from DEM, showing original data (red)
ltered water line (grey) and estimated river bottom level (black). (For interpretation of the references to colour in this gure legend, the reader is referred to the
web version of this article.)

8
zDEM  H  Hv eg
if B 6 a  res
>
>
<


Bares
z0 zDEM  H  Hv eg  1  bres
if B 6 a b  res
>
>
:
zMDE  H
if B > a b  res

where H (m), B (m) are the cross section depth and width estimated
as described in Section 4.3 and a and b are parameters.
Finally, we use a heuristic approach to eliminate random noise
(e) from the longitudinal prole to avoid possible numerical instabilities on the numerical scheme of the hydrodynamic model. An
iterative moving average lter is applied on z0 data:

zi;t1 h  zi1;t 1  2h  zi;t h  zi1;t

where zi,t (m) is the bottom level of the pixel i in the tth iteration, h is
a parameter (h = 1/3, which makes the lter a simple moving average), i  1 and i + 1 are the indexes related to the upstream and
downstream pixels and t is the iteration. Eq. (8) is computed
itmax times. Bottom level discontinuities near the conuences are
avoided by solving Eq. (8) in all reaches at the same iteration. A particular form of the lter is applied in pixels located at conuences
and the parameter h is weighted by the upstream drainage area of
each pixel to avoid small tributaries to generate noise in larger

Floodplains are characterized by a function A (z,m) that relates


water level z with ooded area A for each sub-catchment connected to a sub-reach of the river. These sub-catchments, which
are called here oodplain units, are denoted by P hi, ji and relate
each pixel hi, ji to a cross section located immediately downstream.
These are the areas hydrologically connected to the river segment
between two cross sections, i.e. a set of pixels that are immediately
upstream of a cross section (Fig. 1). To compute P hi, ji, we use a
raster watershed algorithm similar to the one used for catchment
discretization (Burrough and McDonnel, 1998). The purpose of
dening those oodplain units is only to derive local functions
relating water stage and inundated area. Rainfall-runoff calculations are done at the catchment level only.
The function A is dened using the DEM and oodplain units
rasters. The total ooded area between the cross sections m and
m  1 when the average water level equals z is computed as the
sum of the surface area of the set of pixels inside the oodplain
unit m that are lower than z:

Aflz; m

dAhi; ji

hi;ji2S

where S = {hk, li|Phk, li = m, Z hk, li 6 z}, dA (km2) is the pixel surface


area (m2), Z hi, ji (m) is a corrected DEM. A simple approach to correct the DEM is to consider a constant value of vegetation height
Hveg for the whole DEM and compute Z hi, ji = Zhi, ji  Hveg, but
other approaches for DEM correction could also be used such as different Hveg values combined with land cover maps. The oodplain

AA

AA

z DEM

Hve
z DEM

H
z0

Fig. 4. Example of two drainage rasters and real river limits considering different ratio between river width and raster resolution and the errors on SRTM DEM data due to
vegetation and water level effects.

175

R.C.D. Paiva et al. / Journal of Hydrology 406 (2011) 170181

equivalent width curve is computed as L(z, m) = A(z, m)/Dxm and a


correction is made L = max[0, L  B) to remove the main river width.

5. Flood inundation model


In this section, we describe two procedures: (i) how the model
generates 2D water level results and (ii) how the model considers
the ood extent in the water and energy budget computations.
Considering the model hypothesis to simulate the oodplain
hydraulics, it is possible to generate inundation results in terms
of oodplain water level or depth. The oodplain water depth Yt
hi, ji in the time interval t and pixel hi, ji is computed using previous
results of 1D water levels in each cross section z, of oodplain units
raster P hi, ji and of corrected DEM Z hi, ji:

Y t hi; ji maxztPhi;ji  Z  hi; ji; 0

(a)

10

The ood extent is also considered in the surface water and energy budget of the model. To consider this, we use a methodology
called here as Dynamic HRU. Basically, the fraction area of the
Hydrological Response Units (HRUs) in each catchment is considered to be variable in time. The fraction area of the HRU representative of water surface equals to the total ooded area inside the
catchment. Also, the areas of the other HRUs are adjusted to keep
the sum of all HRUs equal the catchment area. Since we use a simplistic model for simulating oodplains, the Dynamic HRU approach may also have some limitations, although these might be
less severe than not considering the ood extent time variability
into water and energy balance computations.

6. The Purus River basin case study


As previously mentioned, the main objective of this work is to
develop methods for large scale hydrodynamic modeling considering limited data and GIS based algorithms for parameter extraction
from DEMs. In this section, we show an evaluation of the developed methods through an application of the model in the Purus
River basin. First, we give a brief description of the Purus River basin and a presentation of the main data used. Subsequently, the
evaluation of the developed methods is presented focusing on
GIS procedures, model operation and feasibility and also some
analyses on hydrological results. Hydrological results are not fully
explored here, since further conrmation of the developed model
and discussion about advantages of a full hydrodynamic model
for large scale hydrologic modeling is presented in Paiva et al. (in
review).

6.1. The Purus River basin


Purus River is one of the main tributaries of the Amazon River
(Fig. 5), its drainage area equals 370,000 km2 and its average discharge is 11,000 m3 s1. The Purus River basin has typical characteristics of the Amazon region. It is a very large river, the main
land cover of the basin is rain forest and both high rainfall rates
and seasonality play an important role in hydrology. Purus River
slopes are quite small (<5 cm/km) and there are large oodplains,
which become seasonally inundated (Hess et al., 2003). The oodplain width on the lower Purus River is of the order of 30 km,
which corresponds to approximately 30 times the main channel
width. Due to these features, ood waves travel slowly along the
river, taking 2 or 3 months from the headwaters to the river mouth
(Paiva et al., in review). Signicant backwater effects are also present on the main river (Meade et al., 1991) and its tributaries (as
suggested by Paiva et al., in review).

(b)
Fig. 5. The Purus River basin. (a) Location inside the Amazon and (b) model river
network, zoom area (black line), stream gauges used for model calibration (grey
squares) and stream gauges used for model validation (black squares).

6.2. Data
We used the SRTM DEM (Rabus et al., 2003; Farr et al., 2007) obtained from the Hydrosheds project data (Lehner et al., 2006). We
used the DEM15s product, which is a DEM based on SRTM data
with some corrections on missing data and 1500 resolution (approximately 500 m). An HRU map with 12 classes was developed using
soil and vegetation maps. The soil map used is a combination of the
Brazilian database RADAMBrasil (Projeto RADAMBRASIL, 1982)
and SOTERLAC/ISRIC (Dijkshoorn et al., 2005) for areas outside Brazil. Concerning the land cover, we used the Vegetation Map of
South America developed by Eva et al. (2002). Hydrologic data
(rainfall and river discharge and stage time series) was provided
by the Brazilian National Water Agency (ANA Agncia Nacional
de guas). Precipitation data from rain gauges was used as input
for the hydrological model. Discharge and water level data from
gauging stations were used for calibration and validation of the
model (see Fig. 5). We used four ANA stream gauges to validate
the model, coded 13710001 (1), 13750000 (2), 13880000 (3) and
13980000 (4).
River cross sections surveys from stream gauge locations were
used to develop geomorphologic equations relating river width
and depth with drainage area for the Amazon region. Other meteorological input variables of the model near surface air temperature, pressure, moisture, wind speed and incoming shortwave solar
radiation were obtained from NCEP reanalysis (Kalnay et al.,
1996).
6.3. GIS procedures results
In this section we show the main results from the GIS based
procedures for the large scale hydrodynamic model discretization
and parameter estimation described in Section 4. The aim is to offer a better understanding about the GIS based procedures and to
show its feasibility. For a better understanding and visualization,
the results are shown (Fig. 6) in the smaller area (165 
165 km) highlighted in Fig. 5. This area will be called here as the
zoom area. The DEM (gure 6a) shows the Purus River and its large
oodplain owing from the lower left corner of the gure to the
northeast. The drainage raster D hi, ji and catchments raster B hi, ji
(Fig. 6b and c) were created considering a threshold drainage area

176

R.C.D. Paiva et al. / Journal of Hydrology 406 (2011) 170181

Fig. 6. (a) Digital elevation model Z hi, ji, (b) drainage raster D hi, ji, (c) catchments raster B hi, ji, (d) location of computational cross sections and drainage network and (e)
oodplain units raster P hi, ji on the zoom area of the Purus River basin.

equal to 625 km2, giving rise to 345 catchments with an average


drainage area of 1000 km2. The river reaches discretization into
sub-reaches and cross sections (Fig. 6d) was performed considering

Dx 6 10 km. Fig. 6e shows results of oodplain units raster P hi, ji,


together with model river reaches cross sections. One of the oodplain units between two model cross sections is highlighted, and

177

R.C.D. Paiva et al. / Journal of Hydrology 406 (2011) 170181

60

250

50
40

z [km]

Afl [km2]

200
150
100

20
10

50
0
10

30

15

20

25

30

35

40

45

0
1300

will be used later to show some of the model simulation results. As


an example, Fig. 7 shows the curve of water level versus ooded
area (A) extracted from DEM (Eq. (9)) of the area highlighted in
Fig. 6e.
The parameters of the cross sections, i.e. width B (m) and maximum depth H (m), were estimated as functions of drainage area Ad
(km2), using the following geomorphologic equations:
B = 0.8054Ad0.5289 (R = 0.77) and H = 1.4351Ad0.1901 (R = 0.72).
These equations were developed for the Amazon River basin using
cross sections proles from 341 gauge stations located in the Brazilian part of the Amazon basin, including the rivers Araguaia and
Tocantins (Paiva, 2009). These geomorphologic equations provide
similar results as those previously developed by Coe et al. (2007)
and Beighley et al. (2009).
Fig. 8 shows the longitudinal prole of the Purus River on the
zoom area extracted from the DEM (black line) and after correction. The systematic errors related to vegetation and surface water
effects were removed (Eq. (7)) considering the following parameters: Hveg = 17 m, a = 0.8 and b = 1.0. A single value for Hveg was
adopted for the whole DEM, since rainforest land cover predominates in the Purus basin and its value was estimated through an
inspection analysis of the DEM at the border of forested and deforested areas. Considering that the number of iterations itmax of the
iterative moving average lter is a subjective parameter, different
values were tested. The ltered longitudinal prole is smoother
when larger values of itmax are used, however, when itmax tends
to innity, the longitudinal prole gets too smoothed, and information is lost. Therefore we used itmax = 5000, but this parameter
should be veried in future applications using a similar inspection.

6.4. Streamow and water level modeling results


In this section, Purus River simulation results are presented. The
MGB-IPH parameters related to soil water budget were calibrated,
using discharge data from the 1985 to 2005 period. Calibration was
performed using a combination of manual and automatic optimization, as described by Collischonn et al. (2007), using the MOCOM-UA algorithm (Yapo et al., 1998). During the calibration
phase only gauges located at the headwaters of the basin (shown
by the grey squares in Fig. 5) where considered for hydrograph
comparisons. The model was then validated using the downstream
gauges, identied by numbers 1 to 4 in Fig. 5. Gauges 13 provided
discharge data and gauges 24 water level data. The few hydrodynamic model and data preprocessing parameters were not calibrated and were set to: Mannings n = 0.030, Hveg = 17 m,
itmax = 5000, a = 0.8 and b = 1.0. Mannings value was adopted
based on values for natural rivers presented by Chow (1959) and
by other authors who applied models in the Amazon region, such
as LeFavour and Alsdorf (2005) and Trigg et al. (2009). Further

1500

1600

1700

1800

x [km]

z [m]
Fig. 7. Flooded area versus water level on the highlighted oodplain unit extracted
from de DEM.

1400

Fig. 8. Longitudinal prole of Purus River on the zoom area: DEM levels (zDEM) in
black line and estimated bottom levels (z0) with itmax = 50 (light grey line),
itmax = 5000 (dark grey line) and itmax = 50,000 (bold black line).

0
Sep Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep
Fig. 9. Flood wave delay and attenuation in Purus River showed by the ration
between daily and mean streamow in the 1990/1991 hydrological year in the
gauges 1 (grey line), 2 (black line) and 3 (dashed line) (see Fig. 5).

details on model calibration can be found in Paiva et al. (in review)


and Paiva (2009).
Fig. 9 shows observed daily stream ow from September 1990
to September 1991 at the stream gauges shown in Fig. 5. Hydrographs in Fig. 9 are arranged from upstream to downstream and
the ratio between daily stream ow and mean stream ow is
shown. Fig. 10 shows observed and simulated daily stream ow
in Purus River at the same stream gauges. In these gures, one
can see that while the ood peaks rise from more or less
6000 m3 s1 to almost 15000 m3 s1, the ood wave becomes
smoother. Seasonal precipitation in this region causes alternated
high and low water periods. Hydrographs of the upper part of
the basin are noisy, with several peaks related to intense rainfall
events. As the ood wave travels to the lower part of Purus River,
it is attenuated and delayed due to the storage of high volumes of
water on the oodplain. This process produces much smother
hydrographs and ood wave travel times of several months, as
can be seen in Fig. 9. These hydrological features were relatively
well represented by the model, as can be seen in Fig. 10. Also, error
statistics comparing observed and simulated stream ow during
the 19852005 period corroborate the good model performance,
although large errors are found in specic time intervals, such as
in high ows in Fig. 10c. Volume error, or bias, at the four gauge
stations was: 9%, 1% and 5%. NashSutcliffe efciency coefcient values for each of the gauge stations were: 0.84, 0.89 and
0.91.
Fig. 11 shows observed and simulated water levels from 1987 to
1991 at stream gauge stations presented in Fig. 5. The main hydrological features observed in the water levels of Purus River are well
represented by the model. The range of water levels is quite large
being 15, 15 and 11 m for the observed data in the three gauge stations. The simulated range of the water levels is very similar to the
observed one. Large errors were only found at gauge 2 (27%),

178

R.C.D. Paiva et al. / Journal of Hydrology 406 (2011) 170181

(a)

(a)

8000

y [m]

Q [m3/s]

6000

10

4000

2000
-5
0
1986

1987

1988

1989

1990

1991

(b) 12000

1986

1988

1989

1990

1991

1987

1988

1989

1990

1991

1987

1988

1989

1990

1991

(b)
5

10000
8000

y [m]

Q [m3/s]

1987

6000
4000

0
-5

2000
0
1986

1987

1988

1989

1990

1991

(c)

(c) 15000
10000

y [m]

Q [m3/s]

1986

5000

0
1986

-5

1987

1988

1989

1990

1991

1986

Fig. 10. Observed (grey line) and simulated (black line) daily streamow in the
gauges 1 (a), 2 (b) and 3 (c) of Purus River in 19861991 time period (see Fig. 5).

Fig. 11. Observed (grey line) and simulated (black line) daily water level in the
gauges 2 (a), 3 (b) and 4 (c) of the Purus River in 19861991 time period (see Fig. 5).

while at the other two gauges river stage amplitudes were correctly reproduced (errors of 6% and 1%, respectively at gauges
3 and 4). As in Fig. 10, water levels shown in Fig. 11 also illustrate
the attenuation and delay of the ood wave due to oodplain storage as it travels through the Purus River. Analysis of the hydrographs and relatively high NashSutcliffe efciency coefcient
values at those three gauges (0.84, 0.90 and 0.92) suggest that
the model is performing good and that the delay and smoothing
of the ood wave is being well represented by the propagation
module of the hydrological model.

several meters. It can be seen that the ooded areas expand from
the dry to the wet season, and that oodplain areas located in
the downstream area (right part of Fig. 12) are still inundated after
the passing of the ow peak.
Detailed results are given for one of the oodplain units, which
is highlighted in Fig. 12. For this oodplain unit we show time series of the ooded area A and exchange ux between river and
oodplain q, as calculated by the model (Fig. 13a). Fig. 13b shows
time series of the main river discharge Q, again compared to q. Finally, Fig. 13c shows water level z in the oodplain and its temporal derivative dz/dt.
The inundation dynamics are characterized by a strong seasonality driven by the main river discharge and water level variations.
According to model results, oodplain lateral exchange is driven by
the river water level variations. When the water level is rising
(lowering), water ows from the main river (oodplain) to the
oodplain (main river) and the ooded area also increases
(decreases).
Since one of the basic assumptions of the model is that there is
no water level difference between the main river and the connected oodplain, total ooded area exhibits a similar behavior
of water levels. Also, the model assumes that all marginal lakes
and other parts of the oodplain are fully connected to the river.
This can be usually considered valid, since normally these marginal
lakes are connected to the main river through small channels (e.g.
Bonnet et al., 2008).
Simulation results show that the majority of water in the oodplain comes from the main river, and that inow and outow

6.5. Simulated ood inundation dynamics


Based on model water level results maps of oodplain inundation can be obtained by comparing water levels calculated at
hydrodynamic cross sections with DEM elevations, according to
the procedure presented in Section 5. In this section, we present
some ood inundation results and discuss the potential of the
model in simulating different types of inundation dynamics.
Fig. 12 shows the simulated water depth on the oodplain at
three time intervals throughout the hydrological year 1995/1996
(Fig. 12ac). It also shows the computational river reaches and
cross sections of the model. The rst water depth map was derived
using water levels calculated at the beginning of the hydrological
year, at low water season (15-October-1995), the second is from
the high water season (22-April-1996) and the third is from the
declining water level period (31-July-1996). Areas appearing white
in Fig. 12 are not ooded, while dark areas are ooded to depths of

179

200

500

100

0
1994

x 10

z [m]

1997

1998

-500
1999

500

0
1994

(c)

1996

1995

1996

1997

1998

-500
1999

35

0.2

30

0.1

25

20

-0.1

15
1994

1995

1996

1997

1998

qfl [m3/s]

Q [m3/s]

(b)

1995

dz/dt [m/day]

Afl [km2]

(a)

qfl [m3/s]

R.C.D. Paiva et al. / Journal of Hydrology 406 (2011) 170181

-0.2
1999

Fig. 13. (a) Simulated daily ooded area A (grey line) and oodplain lateral
exchange q (black line), (b) simulated daily discharge Q (grey line) and oodplain
lateral exchange q (black line) and (c) simulated daily water level z (grey line) and
its temporal derivative dz/dt (black line) in a oodplain unit of Purus River for the
period of 19941999.

Fig. 12. Simulated oodplain water depth on the zoom area of Purus River on 15October-1995 (a), 22-April-1996 (b) and 31-July-1996 (c).

intensity is dened by variation in time of main river water level.


Both features were reported by Bonnet et al. (2008) who analyzed
inow and outow from a marginal lake of the Amazon River near

bidos using water level observations and modeling results. Similar main river driven oodplain inundation dynamics were also observed by Alsdorf et al. (2010) in the mainstem Amazon oodplain
using gravimetric and imaging satellite methods.
In addition, inundation results of the model were validated in
terms of ooded areas extent (Paiva et al., in review; Paiva, 2009)
and a good performance of the model was found.
It is worth noting that the model presented here is not able to
simulate a full 2D inundation dynamic, e.g. when two large rivers
can exchange water trough oodplain ow. This situation may occur in Amazon rivers, but it is not as common in this area as it is in
the Pantanal region (Paz et al., 2010), and probably it has relatively
low impact on large scale modeling results. To improve the representation of oodplain ow a 2D model should be more appropriate (Bates and De Roo, 2000; Bates et al., 2010). The model
presented here only considers the oodplain storage in the 1D
hydrodynamic model, taking into account its effects on ood wave
delay and attenuation; and then we use the 1D water levels calculated at cross section locations to generate 2D inundation maps.
Although it has some limitations, the approach presented here
can be considered sufcient for large scale hydrological modeling
and has much less computational demands.

180

R.C.D. Paiva et al. / Journal of Hydrology 406 (2011) 170181

7. Conclusions
We describe the development of a model for large scale hydrologic and hydrodynamic modeling. The proposed approach is suitable for large scale applications in regions with limited data, since
it uses regionalization methods and GIS algorithms that can be applied using globally available datasets.
The model and methods used to derive the necessary information were tested in the Purus River basin, which is a representative
part of the Amazon basin in terms of river hydraulics and oodplain abundance. The case study shows the feasibility of the GIS
based algorithms for parameter extraction. A comparison between
observed and simulated discharges and water levels at gauging stations shows that the model is capable of reproducing the main
hydrological features of the Purus River basin. Calibration of
parameters related to the hydrodynamic model was not necessary.
However, while our modeling approach for ow propagation in rivers is relatively complex and complete, our description of oodplain dynamics is very simple. Our approach does not fully
reproduce what is actually happening in the oodplains, nevertheless realistic oodplain inundation maps were derived from the results of the model. Our main concern is large scale applications and
on the effects that oodplain inundation and emptying has on variables measured at the river gauges.
Model errors may be related to input data, and limitations of the
model itself. In terms of the river ow propagation module, we believe that errors may be related to the input data uncertainty, e.g.
DEM precision and errors related to vegetation and cross section
geometry provided by geomorphologic relations. Nevertheless, results show that it is possible to employ full hydrodynamic models
within large-scale hydrological modeling even using limited data
for river geometry and oodplain characterization. In comparison
to other large scale ow routing algorithms, the model can simulate additional hydrological features such as backwater effects
and ood inundation dynamics and can provide other output variables as river water levels, total ooded area and 2D water depth as
well.
In a forthcoming work (Paiva et al., in review), we present a
model validation including an evaluation of model errors and its
sources. We evaluate its advantages if compared with simplied
ow routing algorithms and also relate the differences in results
from both techniques with the role played by the oodplains and
backwater effects.
Acknowledgements
The authors are grateful for the nancial and operational support from the Brazilian agencies FINEP and ANA as for constructive
comments of Marie-Paule Bonnet, Douglas Alsdorf, and anonymous reviewers.
References
Allen, R.G., Pereira, L.S., Raes, D., Smith, M., 1998. Crop Evapotranspiration:
Guidelines for Computing Crop Water Requirements. Irrigation and Drainage
Paper 56. FAO, Rome.
Alsdorf, D., Dunne, T., Melack, J., Smith, L., Hess, L., 2003. Diffusion modeling of
recessional ow on central Amazonian oodplains. Geophys. Res. Lett. 32,
L21405.
Alsdorf, D., Bates, P., Melack, J., Wilson, M., Dunne, T., 2007. The spatial and
temporal complexity of the Amazon ood measured from space. Geophys. Res.
Lett. 34, L08402.
Alsdorf, D., Han, S.-C., Bates, P., Melack, J., 2010. Seasonal water storage on the
Amazon oodplain measured from satellites. Remote Sens. Environ. 114, 2448
2456.
Arora, V.K., 2001. Streamow simulations for continental-scale river basins in a
global atmospheric general circulation model. Adv. Water Resour. 24, 775791.
Bastiaanssen, W.G.M., Menenti, M., Feddes, R.A., Holtslag, A.A.M., 1998. A remote
sensing surface energy balance algorithm for land (SEBAL) 1. Formulation. J.
Hydrol. 212213, 198212.

Bates, P.D., De Roo, A.P.J., 2000. A simple raster based model for ood inundation
simulation. J. Hydrol. 236, 5477.
Bates, P.D., Horrit, M.S., Fewtrell, T.J., 2010. A simple inertial formulation of the
shallow water equations for efcient two-dimensional ood inundation
modelling. J. Hydrol. 387, 3345.
Beighley, R.E., Eggert, K.G., Dunne, T., He, Y., Gummadi, V., Verdin, K.L., 2009.
Simulating hydrologic and hydraulic processes throughout the Amazon River
Basin. Hydrol. Process. 23 (8), 12211235.
Betts, A.K., Ball, J.H., Beljaars, A.C.M., Miller, M.J., Viterbo, P.A., 1996. The land
surfaceatmosphere interaction: a review based on observational and global
modeling perspectives. J. Geophys. Res. 101 (D3), 72097225.
Beven, K.J., 2001. Rainfall-runoff Modeling: The Primer. Wiley, 360p.
Biancamaria, S., Bates, P.D., Boone, A., Mognard, N.M., 2009. Large-scale coupled
hydrologic and hydraulic modelling of the Ob river in Siberia. J. Hydrol. 379,
136150.
Bonnet, M.P., Barroux, G., Martinez, J.M., Seyler, F., Turcq, P.M., Cochonneau, G.,
Melack, J.M., Boaventura, G., Bourgoin, L.M., Len, J.G., Roux, E., Calmant, S.,
Kosuth, P., Guyot, J.L., Seyler, F., 2008. Floodplain hydrology in an Amazon
oodplain lake (Lago Grande de Curua). J. Hydrol. 349, 1830.
Bourgoin, L.M., Bonnet, M.P., Martinez, J.M., Kosuth, P., Cochonneau, G., Turcq, P.M.,
Guyot, J.L., Vauchel, P., Filizola, N., Seyler, P., 2007. Temporal dynamics of water
and sediment exchanges between the Curua oodplain and the Amazon River,
Brazil. J. Hydrol. 335, 140156.
Burrough, P.A., McDonnel, R.A., 1998. Principles of Geographical Information
Systems: Spatial Information Systems and Geostatistics. Oxford University
Press, Oxford, 333p.
Castellarin, A., Di Baldassarre, G., Bates, P.D., Brath, A., 2009. Optimal cross-sectional
spacing in Preissmann scheme 1D hydrodynamic models. Journal of Hydraulic
Engineering 135 (2), 96105.
Chen, Y.H., 1973. Mathematical Modeling of Water and Sediment Routing in Natural
Channels. PhD Dissertation, Colorado State University, USA.
Chow, V.T., 1959. Open Channel Hydraulics. McGraw-Hill, 680p.
Coe, M.T., 1997. Simulating continental surface waters: an application to holocene
Northern Africa. J. Clim. 10, 16801689.
Coe, M.T., 2000. Modeling terrestrial hydrological systems at the continental scale:
testing the accuracy of an atmospheric GCM. J. Clim. 13, 686704.
Coe, M.T., Costa, M.H., Howard, E.A., 2007. Simulating the surface waters of the
Amazon River basin: impacts of new river geomorphic and ow
parameterizations. Hydrol. Process. 22 (14), 25422553.
Collini, E.A., Berbery, E.H., Barros, V., Pyle, M., 2008. How does soil moisture inuence
the early stages of the South American Monsoon? J. Clim. 21, 195213.
Collischonn, W., Haas, R., Andreolli, I., Tucci, C.E.M., 2005. Forecasting river Uruguay
ow using rainfall forecasts from a regional weather-prediction model. J.
Hydrol. 305, 8798.
Collischonn, W., Allasia, D.G., Silva, B.C., Tucci, C.E.M., 2007. The MGB-IPH model for
large-scale rainfall-runoff modeling. J. Hydrol. Sci. 52, 878895.
Collischonn, B., Collischonn, W., Tucci, C., 2008. Daily hydrological modeling in the
Amazon basin using TRMM rainfall estimates. J. Hydrol. 360 (14), 207216.
Cunge, J.A., Holly, F.M., Verwey, A., 1980. Practical Aspects of Computational River
Hydraulics. Pitman Advanced Publishing Program.
De Roo, A.P.J., Wesseling, C.G., Van Deursen, W.P.A., 2000. Physically based river
basin modeling within GIS: the LISFLOOD model. Hydrol. Process. 14, 1984
1992.
Decharme, B., Douville, H., Prigent, C., Papa, F., Aires, F., 2008. A new river ooding
scheme for global climate applications: off-line evaluation over South America.
J. Geophys. Res. 113.
DHI, 2003. MIKE 11 A Modelling System for Rivers and Channels Reference
Manual. DHI Software.
Dijkshoorn, J.A., Huting, J.R.M., Tempel, P., 2005. Update of the 1:5 million Soil and
Terrain Database for Latin America and the Caribbean (SOTERLAC, Version 2.0).
Report 2005/01, ISRIC World Soil Information, Wageningen.
Eva, H.D., De Miranda, E.E., Di Bella, C.M., Gond, V., 2002. A Vegetation Map of South
America. EUR 20159 EN, European Commission, Luxembourg.
Farr, T.G., Caro, E., Crippen, R., Duren, R., Hensley, S., Kobrick, M., Paller, M.,
Rodriguez, E., Rosen, P., Roth, L., Seal, D., Shaffer, S., Shimada, J., Umland, J.,
Werner, M., Burbank, D., Oskin, M., Alsdorf, D., 2007. The shuttle radar
topography mission. Rev. Geophys. 45 (2).
Foley, J.A., Prentice, I.C., Ramankutty, N., Levis, S., Pollard, D., Sitch, S., Haxeltine, A.,
1996. An integrated biosphere model of land surface processes, terrestrial
carbon balance, and vegetation dynamics. Global Biogeochem. Cycles 10 (4),
603628.
Frappart, F., Calmant, S., Cauhop, M., Seyler, F., Cazenave, A., 2006. Preliminary
results of ENVISAT RA-2-derived water levels validation over the Amazon basin.
Remote Sens. Environ. 100, 252264.
Galland, J.C., Goutal, N., Hervouet, J.M., 1991. TELEMAC a new numerical model for
solving shallow-water equations. Adv. Water Resour. 14 (3), 138148.
Getirana, A.C.V., Bonnet, M.-P., Rotunno Filho, O.C., Collischonn, W., Guyot, J.-L.,
Seyler, F., Mansur, W.J., 2010. Hydrological modelling and water balance of the
Negro River basin: evaluation based on in situ and spatial altimetry data.
Hydrol. Process. doi:10.1002/hyp.7747.
Hamilton, S.K., Sippel, S.J., Melack, J.M., 2002. Comparison of inundation patterns
among major South American oodplains. J. Geophys. Res., 107(D20), LBA-1
LBA-4.
Hess, L.L., Melack, J.M., Novo, E.M.L.M., Barbosa, C.C.F., Gastil, M., 2003. Dual-season
mapping of wetland inundation and vegetation for the central Amazon basin.
Remote Sens. Environ. 87, 404428.

R.C.D. Paiva et al. / Journal of Hydrology 406 (2011) 170181


Jenson, S.K., Domingue, J.O., 1988. Extracting topographic structure from digital
elevation data for geographic information system analysis. Photogrammetric
Eng. Remote Sens. 54 (11), 15931600.
Kalnay, E. et al., 1996. The NCEP/NCAR 40-years reanalysis project. Bull. Am.
Meteorol. Soc. 77, 437471.
Kellndorfer, J., Walker, W., Pierce, L., Dobson, C., Fites, J.A., Hunsaker, C., Vona, J.,
Clutter, M., 2004. Vegetation height estimation from shuttle radar topography
mission and national elevation datasets. Remote Sens. Environ. 93, 339358.
Koussis, A.D., 2009. Assessment and review of the hydraulics of storage ood
routing 70 years after the presentation of the Muskingum method. Hydrol. Sci. J.
54 (1).
Kouwen, N. et al., 1993. Grouping response units for distributed hydrologic
modelling. J. Water Resour. Manage. Planning 119 (3), 289305.
LeFavour, G., Alsdorf, D., 2005. Water slope and discharge in the Amazon River
estimated using the Shuttle Radar Topography Mission digital elevation model.
Geophys. Res. Lett. 32, L17404.
Lehner, B., Verdin, K., Jarvis, A., 2006. HydroSHEDS Technical Documentation. World
Wiuldlife Fund US, Washington, DC. <http://hydrosheds.cr.usgs.gov>.
Lian, Y., Chan, I.-C., Singh, J., Demissie, M., Knapp, V., Xie, H., 2007. Coupling of
hydrologic and hydraulic models for the Illinois River Basin. J. Hydrol. 344, 210
222.
Liang, X., Lettenmaier, D.P., Wood, E.F., Burges, S.J., 1994. A simple hydrologically
based model of land surface water and energy uxes for general circulation
models. J. Geophys. Res. 99, 1441514428.
Liston, G.E., Sud, Y.C., Wood, E.F., 1994. Evaluating GCM land surface hydrology
parameterizations by computing river discharges using a runoff routing model:
application to the Mississippi Basin. J. Appl. Meteorol. 33, 394405.
Liu, Z., Todini, E., 2002. Towards a comprehensive physically-based rainfall-runoff
model. Hydrol. Earth Syst. Sci. 6 (5), 859881.
Lohmann, D., Nolte-Holube, R., Raschke, E., 1996. A large-scale horizontal routing
model to be coupled to land surface parametrization schemes. Tellus Ser. A:
Dyn. Meteorol. Oceanogr. 48 (5), 708721.
Ludwig, K., Bremicker, M., 2006. The Water Balance Model LARSIM Design,
Content and Applications. Freiburger Schiriften zr Hydrologie, Band 22,
Institut fur Hydrologie der Universitt Freiburg.
Maidment, D., 2002. Arc Hydro GIS for Water Resources. ESRI Press, Redlands, CA.
Matheussen, B., Kirschbaum, R.L., Goodman, I.A., ODonnell, G.M., Lettenmaier, D.P.,
2000. Effects of land cover change on streamow in the interior Columbia River
basin (USA and Canada). Hydrol. Process. 14, 867885.
Meade, R.H., Rayol, J.M., Da Conceio, S.C., Natividade, J.R.G., 1991. Backwater
effects in the Amazon River basin of Brazil. Environ. Geol. Water Sci. 18 (2),
105114.
Mohamed, Y.A., van den Hurk, B.J.J.M., Savenije, H.H.G., Bastiaanssen, W.G.M., 2005.
Impact of the Sudd wetland on the Nile hydroclimatology. Water Resour. Res.
41, W08420.
Monteith, J.L., 1965. Evaporation and environment. In: Proc. 19th Symp. Soc. Exp.
Biol. Swansea (1964).
Moore, R.J., Clarke, R.T., 1981. A distribution function approach to rainfall-runoff
modeling. Water Resour. Res. 17 (5), 13671382.
Nijssem, B., Lettenmaier, D.P., Liang, X., Wetzel, S.W., Wood, E.F., 1997. Streamow
simulation for continental-scale river basins. Water Resour. Res. 33 (4), 711
724.
Nijssen, B., ODonnell, G.M., Hamlet, A.F.E., Lettenmaier, D.P., 2001. Hydrologic
sensitivity of global rivers to climate change. Clim. Change 50 (12), 143175.
Paiva, R.C.D., 2009. Large Scale Hydrological and Hydrodynamic Modeling: the Rio
Solimes Case Study (in Portuguese). M.Sc. Dissertation, Instituto de Pesquisas
Hidrulicas, Universidade Federal do Rio Grande do Sul. <http://hdl.handle.net/
10183/18927>.
Paiva, R.C.D., Collischonn, W., Buarque, D.C., in review. Validation of a full
hydrodynamic model for large scale hydrologic modelling in the Amazon.
Hydrol. Process.

181

Paz, A.R., Bravo, J.M., Allasia, D., Collischonn, W., Tucci, C.E.M., 2010. Large-scale
hydrodynamic modeling of a complex river network and oodplains. J. Hydrol.
Eng. 15 (2), 152165.
Pitman, A.J., 2003. The evolution of, and revolution in, land surface schemes
designed for climate models. Int. J. Climatol. 23, 479510.
Rabus, B., Eineder, M., Roth, A., Bamler, R., 2003. The shuttle radar topography
missiona new class of digital elevation models acquired by spaceborne radar.
J. Photogrammetry Remote Sens. 57, 241262.
RADAMBRASIL. 1982. Programa de Integrao Nacional, Levantamento de Recursos
Naturais. Ministrio das Minas e Energia, Secretaria-Geral.
Rawls, W.J., Ahuja, L.R., Brakensiek, D.L., Shirmohammadi, A., 1993. Inltration and
soil water movement. In: Maidment, D.R. (Ed.), Handbook of Hydrology.
McGraw-Hill, New York.
Remo, J.W.F., Pinter, N., 2007. Retro-modeling the Middle Mississippi River. J.
Hydrol. 337, 421435.
Richey, J.E., Melack, J.M., Aufdenkampe, A.K., Ballester, V.M., Hess, L.L., 2002.
Outgassing from Amazonian rivers and wetlands as a large tropical source of
atmospheric CO2. Nature 416, 617620.
Schmugge, T.J., Kustas, W.P., Ritchie, J.C., Jackson, T.J., Rango, A., 2002. Remote
sensing in hydrology. Adv. Water Resour. 25, 13671385.
Shuttleworth, W.J., 1993. Evaporation. In: Maidment, D.R. (Ed.), Handbook of
Hydrology. McGraw-Hill, New York.
Silva Dias, M.A.F., Silva Dias, P.L., Longo, M., Fitzjarrald, D.R., Denning, A.S., 2004.
River breeze circulation in eastern Amazonia: observations and modeling
results. Theor. Appl. Climatol. 78, 111122.
Sun, G., Ranson, K.J., Kharuk, V.I., Kovacs, K., 2003. Validation of surface height from
shuttle radar topography mission using shuttle laser altimeter. Remote Sens.
Environ. 88, 401411.
Thielen, J., Bartholmes, J., Ramos, M.-H., de Roo, A., 2009. The European ood alert
system Part 1: concept and development. Hydrol. Earth Syst. Sci. 13, 125140.
Todini, E., 1996. The ARNO rainfall-runoff model. J. Hydrol. 175, 339382.
Todini, E., 2007. A mass conservative and water storage consistent variable
parameter Muskingum-Cunge approach. Hydrol. Earth Syst. Sci. 11 (5), 1645
1659.
Tomasella, J., Borma, L.S., Marengo, J.A., Rodriguez, D.A., Cuartas, L.A., Nobre, C.A.,
Prado, M.C.R., 2010. The Droughts of 19961997 and 20042005 in Amazonia:
Hydrological Response in the River Main-stem. Hydrol. Process.
Trigg, M.A., Wilson, M.D., Bates, P.D., Horritt, M.S., Alsdorf, D.E., Forsberg, B.R., Vega,
M.C., 2009. Amazon ood wave hydraulics. J. Hydrol. 374, 92105.
Tucci, C.E.M., 1978. Hydraulic and Water Quality Model for a River Network. Ph.D.
Dissertation, Colorado State University, Fort Collins, USA.
USACE, 2002. HEC-RAS River Analysis System: Hydraulic Reference Manual, Version
3.1. US Army Corps of Engineers, Hydrologic Engineering Center. <http://
www.hec.usace.mil>.
Viney, N.R., Sivapalan, M., Deeley, D., 2000. A conceptual model of nutrient
mobilisation and transport applicable at large catchment scales. J. Hydrol. 240
(12), 2344.
Wood, E.F., Lettenmaier, D.P., Zartarian, V.G., 1992. A land-surface hydrology
parameterization with subgrid variability for general circulation models. J.
Geophys. Res. 97, 27172728.
Wood, A.W., Maurer, E., Kumar, A., Lettenmaier, D.P., 2002. Long-range
experimental hydrologic forecasting for the eastern United States. J. Geophys.
Res. 107, D20.
Xu, C.-Y., Widn, E., Halldin, S., 2005. Modelling hydrological consequences of
climate change progress and challenges. Adv. Atmos. Sci. 22 (6), 789797.
Yapo, P.O., Gupta, H.V., Sorooshian, S., 1998. Multi-objective global optimization for
hydrologic models. J. Hydrol. 204, 8397.
Zhang, Y., Li, C., Trettin, C.C., Li, H., Sun, G., 2002. An integrated model of soil,
hydrology, and vegetation for carbon dynamics in wetland ecosystems. Global
Biogeochem. Cycles 16 (4), 1061.

You might also like