Académique Documents
Professionnel Documents
Culture Documents
Universidad Nacional Agraria La Molina, Av. La Universidad s/n, La Molina, Apartado 12-056, Lima 12, Peru
CNRS, Universit Rennes 1, Gosciences Rennes, Campus de Beaulieu, 35042 Rennes cdex, France
c
IRD, CP 7091 Lago Sul, 71619-970 Brasilia DF, Brazil
d
CESTU, Universidade do Estado do Amazonas, UEA. Av. Manaus, Amazonas, CEP 69058807, Brazil
b
a r t i c l e i n f o
a b s t r a c t
Article history:
Received 15 December 2011
Accepted 24 October 2012
Hydrodynamic modelling of Amazonian rivers is still a difcult task. Access difculties reduce the
possibilities to acquire sufcient good data for the model calibration and validation. Current satellite
radar technology allows measuring the altitude of water levels throughout the Amazon basin. In this
study, we explore the potential usefulness of these data for hydrodynamic modelling of the Amazon and
Napo Rivers in Peru. Simulations with a 1-D hydrodynamic model show that radar altimetry can
constrain properly the calibration and the validation of the model if the river width is larger than
2500 m. However, sensitivity test of the model show that information about geometry of the river
channel and about the water velocity are more relevant for hydrodynamic modelling. These two types of
data that are still not easily available in the Amazon context.
2012 Elsevier Ltd. All rights reserved.
Keywords:
Hydrodynamic modelling
Amazon River
Radar altimetry
Model sensitivity
1. Introduction
The Amazon River is the largest in the world with a basin area of
7.0 106 km2 and an average ow at its mouth to 206,000 m3/s
(Callde et al., 2010). Crossing eight countries, this huge river is the
main channel of communication from the Andes to the Atlantic.
Therefore, understanding and modelling the hydrodynamic of the
specic Amazon context is of great interest for environment,
economic and social processes. Since the end of the 1980s, extreme
hydrological events have been increasing in the River Amazon
(Espinoza et al., 2009, 2011). These extreme events caused inundations, as in 1999, 2006 and 2009, or very low water stages, as in
1998, 2005 and 2010, which are harmful to people living nearby the
watercourse and damaging for agriculture and ecosystems (e.g.
Saleska et al., 2007; Phillips et al., 2009; Asner and Alencar, 2010;
Lewis et al., 2011; Xu et al., 2011).The impacts that may cause the
increased frequency of extreme hydrological events in the Amazon
put at risk their vast amount of natural resources and a population
of more than 38 million people. Predicting the impact of climate on
* Corresponding author.
E-mail
addresses:
echavarri@lamolina.edu.pe,
echavarriv@gmail.com
(E. Chvarri), alain.crave@univ-rennes1.fr (A. Crave), marie-paule.bonnet@ird.fr
(M.-P. Bonnet), jsdsilva@uea.edu.br (J. Santos Da Silva), jean-loup.guyot@ird.fr
(J.L. Guyot).
0895-9811/$ e see front matter 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.jsames.2012.10.010
Please cite this article in press as: Chvarri, E., et al., Hydrodynamic modelling of the Amazon River: Factors of uncertainty, Journal of South
American Earth Sciences (2012), http://dx.doi.org/10.1016/j.jsames.2012.10.010
E. Chvarri et al. / Journal of South American Earth Sciences xxx (2012) 1e10
vy 1 vQ
q
vt w vx
wvx
(1)
vQ
v
Q2
vy
b
f
gA gASf bqvL
vt
vx
vx
A
(2)
2. Methodology
This study is divided in two steps. First, we use 1-D hydrodynamic model to quantied the sensitivity of the variables: water
depth (y), longitudinal streamow (Q), bankfull width (w) and
velocity (v) according to the variability of input parameters: the
cross section geometry, Manning roughness coefcient (n) and
longitudinal slope of the river (s). This shows how the hydrodynamic model response is related to the level of uncertainty of input
parameters and how they rank in terms of model sensitivity. In
other words, we evaluate theoretical impacts of uncertainties on
natural data on simulation of y, v and Q. Second, we compare
uncertainties of y simulation to radar altimetry accuracy applying
the same 1-D hydrodynamic model to the Amazon and Napo Rivers.
The streamow model is an original 1-D hydrodynamic model to
simulate unsteady streamow in anabranching river form such as
the Amazon River and Napo River. In the following text, we present
the main equations and hypothesis relative to this numerical model.
where Sf is the energy line slope (friction slope), g is the acceleration due to gravity [LT2], A is cross-sectional area of the streamow [L2], vL is the velocity of the lateral streamow, that is, in the
same direction as the principal streamow of the river, f is the
Local partial inertial factor (Fread et al., 1986), b is the Boussinesq
coefcient.
Equations (1) and (2) are solved under the Preissmann numerical scheme. It offers the advantage that a variable spatial grid may
be used; steep wave fronts may be properly simulated by varying
the weighting coefcient of the time interval q and weighting
coefcient of the space interval f, and the scheme yields an exact
solution of the linearized form of the governing equation for
a particular value of Dx and Dt.
The model has the ability to simulate sections with islands,
creating a new hypothesis in the study. To solve this problem, there
exist diverse alternatives such as assuming that the water depths
around the islands are the same. However, in this study we considered
Please cite this article in press as: Chvarri, E., et al., Hydrodynamic modelling of the Amazon River: Factors of uncertainty, Journal of South
American Earth Sciences (2012), http://dx.doi.org/10.1016/j.jsames.2012.10.010
E. Chvarri et al. / Journal of South American Earth Sciences xxx (2012) 1e10
yi1
1
Q 2
1
Q 2
fi1 i1 yi
fi i
2g
2g
Ai1
Ai
(3)
yi2
1
Q 2
1
Q 2
fi2 i2 yi
fi i
2g
2g
Ai2
Ai
(4)
where: Qi1 and Qi2: streamows in both side of the island, yi1 and
yi2: water depth in both side of the island, Qi: streamow in the
conuence. yi: water depth in the conuence.
The spatial and temporal resolution of the model takes into
account the Courant condition. For one-dimensional equations, the
Courant number, or also known as the Courant, Friedrichs and Lewy
(CFL) number, is dened as Abbott (1979).
ju cjDt
1
Dx
(5)
Dx
Dt
(6)
Dw
Dy
(7)
Fig. 1. (a) Amazon basin, (b) the simulated stretch between Nuevo Rocafuerte station and Tempestad island e Napo river, (c) The simulated stretch between the Tamshiyacu and
Tabatinga stations e Amazonas River. Figures include radar altimetry paths and nodes simulation.
Please cite this article in press as: Chvarri, E., et al., Hydrodynamic modelling of the Amazon River: Factors of uncertainty, Journal of South
American Earth Sciences (2012), http://dx.doi.org/10.1016/j.jsames.2012.10.010
E. Chvarri et al. / Journal of South American Earth Sciences xxx (2012) 1e10
4
3
2
1
0
River Napo
(Bellavista station)
MarcheMay
AugusteOctober
34,815
1.69
1213
0.08e0.35
23.5
726,403
MayeJuly
JanuaryeFebruary
5838
1.18
1312
0.1e1.8
12.77
100,518
800
1000
1200
800
1000
1200
90
Elevation (a.m.s.l)
86
84
82
80
200
400
600
Days
Fig. 2. Calibration results: For (a) Streamow and (b) Water elevation. We presented
the simulated and recorded hydrographs in Tabatinga station for the rst 1100 days (01
September 2002 to 04 September 2005).
x 10
5
4
3
2
1
1000
1200
1400
1600
1800
2000
2200
Days
90
88
Elevation (a.m.s.l)
River Amazon
(Tamshiyacu station)
600
Days
76
0
(8)
General characteristic
400
78
Table 1
General characteristics of Amazon and Napo Rivers in Peru (Hybam project
database).
200
88
Flow (m3/s)
The sensitivity of the model is analysed by looking at the variability of y, Q, w and v at each section of the river, in response to the
variability of a, n and s.
According to the Hybam database, a varies in the range between
20 and 300 following a normal distribution with an average of 106
(Appendix A, Fig A.1). One value of a is dened for each section used
to dene the channel geometry for each simulation. Thus, there are
many values of a and sections used to describe the channel. The set
of a is chosen within a normal distribution by random drawing.
Therefore, there are two parameters to dene the geometry of the
channel, the average value and the standard deviations s of the
normal distribution of a. The sensitivity of the model to a is estimated doing averages on output variable variability on 10 simulations applying the same normal distribution of a values.
Manning roughness coefcient values are chosen in the range for
natural rivers, meaning between 0.025 s/m1/3 and 0.045 s/m1/3.
Values of longitudinal slope are chosen in the range of slope values
observed for the Amazon and Napo Rivers (Bourrel et al., 2009),
meaning between 0.08 m/km and 0.35 m/km for the Amazon plain
and between 0.1 m/km and 1.8 m/km for the Andeans foothills.
In order to have a baseline of simulation to quantify the sensitivity of the model, a reference theoretical case is chosen which
corresponds to the Napo River dataset information. The geometry of
this theoretical channel is xed with a a set values by random
drawing within a normal distribution with an average of 106 and a s
of 52. These two values correspond to the analysis performed on the
dataset of the Napo River (Appendix A). Manning coefcient and
longitudinal slope of this theoretical case are xed to 0.035 s/m1/3
and 0.07 m/km respectively.
To quantify the sensitivity of the 1-D hydrodynamic model to
the input parameters, we calculate differences between the daily
evolutions of each output variables corresponding to a specic
input parameter set of values and the theoretical reference case at
the last downstream station. The criterion of sensitivity is obtained
by percentage of variability using Equation (8).
r
2
PM PD
1
1 Voutput Vreference
100
%Variability P P
M
D
1
1 Voutput Vreference
x 10
Flow (m3/s)
86
84
82
80
78
76
1000
1200
1400
1600
1800
2000
2200
Days
Fig. 3. Validation results: For (a) Streamow and (b) Water elevation. We presented
the simulated and recorded hydrographs in Tabatinga station for the last 1100 days
(05 September 2005 to 28 October 2008).
Please cite this article in press as: Chvarri, E., et al., Hydrodynamic modelling of the Amazon River: Factors of uncertainty, Journal of South
American Earth Sciences (2012), http://dx.doi.org/10.1016/j.jsames.2012.10.010
E. Chvarri et al. / Journal of South American Earth Sciences xxx (2012) 1e10
Table 2
General characteristics of radar altimetry information.
River
Latitude
( ) WGS84
Longitude
( ) WGS84
Time period
Width
approx. (m.)
Observation
Napo
Amazon
966
164
837
794
74.86
70.4
71.6
72.5
1.29
3.79
3.77
3.52
29
06
24
01
77.0
321.2
163.9
31.5
580.0/618.0
5800.0
3260.0
490.0/2110.0
Island
Reach
Reach
Island
Fig. 5. The sensitivity of model variables according to: (a) the Manning roughness
coefcient and (b) Longitudinal slope.
Please cite this article in press as: Chvarri, E., et al., Hydrodynamic modelling of the Amazon River: Factors of uncertainty, Journal of South
American Earth Sciences (2012), http://dx.doi.org/10.1016/j.jsames.2012.10.010
E. Chvarri et al. / Journal of South American Earth Sciences xxx (2012) 1e10
PN
E 1
Pi 2
2
Oi O
1 Oi
PN
1
(9)
Fig. 6. Radar altimetry values follow the temporal variation of water level simulated. (a) Path 966 e Napo river, (b) Path 794 e Amazon river, (c) Path 164 e Amazon river, (d) Path
837 e Amazon river. Linear correlation between radar altimetry value and water level simulated. (e) Path 966, (f) Path 794, (g) Path 164, (h) Path 837.
Please cite this article in press as: Chvarri, E., et al., Hydrodynamic modelling of the Amazon River: Factors of uncertainty, Journal of South
American Earth Sciences (2012), http://dx.doi.org/10.1016/j.jsames.2012.10.010
E. Chvarri et al. / Journal of South American Earth Sciences xxx (2012) 1e10
solution (Tapley et al., 2004). This study analyses four Envisat satellite
tracks that cross the Amazon and Napo Rivers and have been providing
data since 2002 (Table 2). These tracks have been chosen in a manner to
explore the inuence of the river width on hr uncertainties and cross
the rivers at sections with and without islands. Extreme values of river
width are 102.0 m and 1486.0 m.
3. Results
3.1. Model sensitivity to input parameters
According to the results of the sensitivity tests, w is the most
sensitive output variable to either input upstream ow or either to the
variability of all input parameters. The y, the Q and the v are almost
constant when input parameters change (Fig. 4a and b). The model
regulates all input variations tuning the value of w at each node. The
sensitivity of w depends linearly to the average and standard deviation of the normal distribution of a values set. On the other hand,
there is no sensitivity of w to n and s. The % of variability comes from
the distribution of a and do not show any specic trend when n and s
change (Fig. 5a and b). Channel geometry parameterization is then the
rst parameter which controls the model response to upstream ow.
3.2. Relevance of radar altimetry data
yr values follow the temporal variation of y values (Fig. 6aed) for
whichever site where yr values are available. For these sites y versus
yr shows a signicant linear trend with linear correlation coefcients
between 0.64 and 0.93 (Fig. 6eeh). The coefcient of the linear
trends varies from 0.6 to 1.1. Note that a slope coefcient between y
versus yr is smaller than 1 which means that the amplitude of the yr
variation is higher than the amplitude of the y values and, therefore,
marks a greater sensitivity of yr. This sensitivity of yr decreases
linearly when the size of the channel section increases (Fig. 7). For
river sections larger than 2500 m variations of y and yr are similar.
For the Amazon River at TAB, yr values are denitely less relevant than y values because the latter has been validated with in-situ
measurements. Therefore, for this case, yr uncertainty is larger than
y uncertainty without any doubt. For the Napo River case at TEMP,
yr uncertainty looks also larger than y uncertainty. Without in-situ
validation, we cannot reject that y value may have a systematic bias.
However, seven yr values out of eleven are in the range of y
uncertainty and are dispersed throughout the range of y values.
This observation supports that y values are not biased and de facto
that yr uncertainty is larger than y uncertainty for the Napo River.
4. Discussion and conclusion
Our 1-D hydrodynamic model applied to the Amazon basin
adjusts the variation of input parameters changing mainly the wet
cross section area and to a lesser extent the water velocity through
the section. The relationship between the wetted area and the
water depth is then one of the most important input data of our 1-D
hydrodynamic model. To simulate this relationship, we propose
using the coefcient a, of the presupposed linear relationship
between the variation rate of the water level and the variation rate
of the cross section width. An analysis of a bathymetric dataset of
52 cross sections of the Napo River supports this assumption. a is
a local parameter and takes a wide range of values. It controls the
level of uncertainties on simulated water velocity and above all
section widths. Our results show a linear relationship between the
level of uncertainty of these two variables and the uncertainty on
the channel geometry. On the other hand, the water depth does not
vary regardless of the range of values for each input parameter.
These results suggest possible errors in the model assumptions,
their mathematical formulation or numerical resolution. Nevertheless, the validation with physical data is therefore fundamental to
test the model. Two types of validation data have been used to
validate the model. First, we used in-situ measurements of water
level and streamow at the Tabatinga gauging station on the
Amazon River using a half of the data to calibrate the model. Despite
few non-negligible discrepancies, the simulated water level and
streamow variations t the in-situ measurement variations for a 3
year period. This suggests that our model is appropriate to simulate
water level and streamow simulation. Second, we compare simulated water levels with radar altimetric data at four sections with
different widths on the Amazon and Napo Rivers. Simulated water
level t correctly again with the radar altimetry measurements only
for sections with widths larger than 2500 m. Indeed, satellite radar
altimetry accuracy depends strongly on the size of water surface
spot on which the altitude is calculated (Santos da Silva et al., 2010).
If the section width is too small, the radar spot contains points on the
river banks and vegetation. This work shows that the promising
satellite radar technology has currently limited application for the
calibration and validation of hydrodynamic models.
If we take into account the sensitivity of a standard 1-D
hydrodynamic model to channel geometry description, Manning
coefcient or channel longitudinal slope, the validation procedure
should be focused on the channel width and water velocity variations tting. Those variables show the greatest sensitivity to input
parameters for Amazonian conditions where topographic roughness is very low. However, reducing uncertainties on a values or
simply acquiring daily in-situ velocity measurements at numerous
section along Amazonian rivers is still a challenge, specically in
areas with sparse population.
Acknowledgements
This study was sponsored by the Environmental Research
Observatory (ORE) HYBAM (Geodynamical, hydrological and
biogeochemical control of erosion/alteration and material transport
in the Amazon basin) which has been operates in Peru since 2003. A
scientic cooperation agreement between the LInstitut de Recherche pour le Dveloppement (IRD-France) and the Universidad
Nacional Agraria La Molina (UNALM-Peru) as of 2005, allowing the
participation of master and doctorate programs for students in
project ORE-HYBAM. ORE-HYBAM was proposed by team of LMTG
scholars (Laboratoire des Mcanismes de Transferts en GologieUMR 5563 CNRS-UPS-IRD) which has been conducting research
projects in hydro geodynamics in the Amazon basin since 1995.
Appendices
A. Methodology for calculating the variability of the a parameter
Fig. 7. Linear coefcient between the simulated elevation and radar altimetry
elevation versus width river.
Please cite this article in press as: Chvarri, E., et al., Hydrodynamic modelling of the Amazon River: Factors of uncertainty, Journal of South
American Earth Sciences (2012), http://dx.doi.org/10.1016/j.jsames.2012.10.010
E. Chvarri et al. / Journal of South American Earth Sciences xxx (2012) 1e10
Fig. A.1. Relationship of width versus depth ow for all proles between Rocafuerte
Station and Tempestad station and the coefcients of correlation R2 calculated.
Table B.1
Amazon River geometry used for the hydrodynamic model.
Nodes (52)
Downstream
reach length (m)
Width (m)
Elevation (m)
13,035.2
21,309.1
21,309.1
10,830.8
7577.1
7577.1
12,664.9
30,320.2
30,320.2
5411.5
4921.5
4921.5
29,529
8991.4
8991.4
10,925.7
12,639.7
12,639.7
12,086.7
15,349
15,349
2913.4
4733.5
4733.5
5780.1
3948.4
3948.4
17,309.1
15,447.8
15,447.8
20,556.1
10,764.7
10,764.7
7593.6
7601.1
7601.1
11,790.1
9653.9
9653.9
10,080
10,661.4
9919.4
9919.4
11,024.1
11,024.1
11,152
14,167.9
14,167.9
7415.8
7415.8
7415.8
0
342.3
223
290.1
474.3
224.8
203.8
366.9
491.4
218.2
398.2
122.2
150.8
438.4
139.5
225.1
320.5
206.7
167.7
246.7
231.2
138.5
292.9
158.4
195.7
912.2
543.8
182.7
807.6
326.6
79.3
296.8
243.7
139.1
283.4
538.1
142.8
305.4
123.5
344.9
495.1
285.9
339.6
240.5
136.8
278.5
261.6
360.3
309.5
314
237.6
267.7
485.6
88
87.6
87.6
86.6
86
86
85.6
85.4
85.4
84.5
82.7
82.7
82.1
79.8
79.8
76.5
75.9
75.9
74.1
73.5
73.5
71.5
71.3
71.3
71.2
71
71
70.9
70.8
70.8
69.9
69.7
69.7
69
68.2
68.2
67.5
66.8
66.8
66.1
65.8
65.8
65
64.6
64.6
64.4
64
64
63.4
63
63
62.5
Fig. A.2. Normal variability of a from upstream to downstream. (a) Nuevo Rocafuerte
station, (b) Tempestad located, (c) Santa Clotilde station and (d) Bellavista Mazn station.
Please cite this article in press as: Chvarri, E., et al., Hydrodynamic modelling of the Amazon River: Factors of uncertainty, Journal of South
American Earth Sciences (2012), http://dx.doi.org/10.1016/j.jsames.2012.10.010
E. Chvarri et al. / Journal of South American Earth Sciences xxx (2012) 1e10
Table B.2
ENVISAT radar altimetry information (geoid EGM2008).
Day
Elevation
(a.m.s.l)
Day
Elevation
(a.m.s.l)
Day
Elevation
(a.m.s.l)
Day
Elevation
(a.m.s.l)
02
06
10
21
26
30
04
08
13
17
22
26
02
06
82.10
82.55
81.62
83.44
84.81
83.48
78.09
77.32
78.27
80.24
82.12
81.39
80.21
82.41
11
15
20
24
02
07
10
14
21
25
30
17
21
26
82.49
82.33
81.48
77.06
79.83
82.16
81.44
81.19
82.43
83.75
80.67
75.52
79.78
79.72
30
06
10
24
28
02
06
11
15
19
26
20
04
09
82.03
82.46
84.46
82.00
76.80
75.95
77.55
81.97
83.24
82.33
82.80
83.94
82.64
79.92
13
17
22
26
31
04
11
15
20
24
29
02
07
76.10
76.51
76.06
81.84
82.00
82.43
83.52
83.59
82.80
81.06
79.38
80.29
80.07
December 2002
January 2003
February 2003
April 2003
May 2003
June 2003
August 2003
September 2003
October 2003
November 2003
December 2003
January 2004
March 2004
April 2004
May 2004
June 2004
July 2004
August 2004
November 2004
December 2004
January 2005
February 2005
March 2005
April 2005
May 2005
October 2005
November 2005
December 2005
January 2006
March 2006
April 2006
July 2006
August 2006
October 2006
November 2006
December 2006
January 2007
February 2007
March 2007
April 2007
June 2007
July 2007
August 2007
September 2007
October 2007
November 2007
December 2007
February 2008
March 2008
April 2008
May 2008
June 2008
July 2008
September 2008
October 2008
Source: Joecila Santos da Silva, CESTU, Universidade do Estado do Amazonas, UEA e Brasil.
Downstream
reach length (m)
Width (m)
Elevation (m)
3501.4
3629
3629
5233.5
4138.1
4138.1
1471.1
2565.3
2565.3
1884.9
2432.4
2432.4
1992.1
1115.1
1115.1
2172.8
1440.7
1440.7
2113.6
4871.4
4871.4
3233.8
2727.8
2727.8
1986.4
1055.9
1055.9
3564.3
3564.3
3564.3
3564.3
1156.3
321
562.6
1000
590
657.7
800
506.3
450
840
600
609.3
1300
626.9
300
1220.7
496.2
324.3
980
551.8
415.4
730
592.9
460
500
390.7
500
750
344.2
403.2
1000
165.5
165
165
164.5
164
164
163.7
163.5
163.5
163.4
163.3
163.3
163.1
163
163
162.5
162
162
161.5
161
161
160.5
160
160
159.5
159
159
158
157
157
156.5
References
Abbott, M.B., 1979. Computational Hydraulics e Elements of the Theory of Free
Surface Flows. Pitman Publishing Limited, London.
Asner, G.P., Alencar, N., 2010. Drought impacts on the Amazon forest: the
remote sensing perspective. New Phytologist. http://dx.doi.org/10.1111/j.14698137.2010.03310.x.
Beighley, R.E., et al., 2009. Simulating hydrologic and hydraulic
processes throughout the Amazon River Basin. Hydrological Processes 23 (8),
1221e1235.
Bourrel, L., et al., 2009. Estudio de la relacin entre la pendiente de los rios
obtenidas a partir de mediciones DGPS y la distribucin de la granulometra por
tres tributarios andinos del ro Amazonas: el caso de los rios Beni (Bolivia),
Napo (Ecuador-Per) y Maraon (Per). Tercera reunin cientca del
Please cite this article in press as: Chvarri, E., et al., Hydrodynamic modelling of the Amazon River: Factors of uncertainty, Journal of South
American Earth Sciences (2012), http://dx.doi.org/10.1016/j.jsames.2012.10.010
10
E. Chvarri et al. / Journal of South American Earth Sciences xxx (2012) 1e10
Tapley, B., et al., 2004. The gravity recovery and climate experiment: mission
overview and early results. Geophysical Research Letters 31, L09607. http://
dx.doi.org/10.1029/2004GL019920.
Trigg, M., et al., 2009. Amazon ood wave hydraulics. Journal of Hydrology 374, 92e105.
Xu, et al., 2011. Widespread decline in greenness of Amazonian Vegetation due to
the 2010 drought. Geophysical Research Letters 38 (L07402), 4. http://dx.doi.
org/10.1029/2011GL046824.
Please cite this article in press as: Chvarri, E., et al., Hydrodynamic modelling of the Amazon River: Factors of uncertainty, Journal of South
American Earth Sciences (2012), http://dx.doi.org/10.1016/j.jsames.2012.10.010