You are on page 1of 9

International conference on innovation advances and implementation of flood forecasting technology

AUTOMATIC CALIBRATION OF INITIAL STATE


VARIABLES FOR FLOOD FORECASTING USING A
DISTRIBUTED MODEL

Jorge Julián Vélez (1) and Félix Francés (1)


(1) Universidad Politécnica de Valencia, Instituto de Ingeniería del Agua y Medio Ambiente IIAMA,
Camino de Vera s/n, 46022. Valencia, Spain

Abstract
The hydrological conceptual distributed model, called TETIS, divides river basins into regular cells,
all of which are connected according to the network drainage. The rainfall-runoff process is modelled
using linked tanks with different outflow relationships. The vertical movement is based on soil
properties. Routing along the channel network has been coupled using geomorphologic basin
characteristics and the kinematic wave procedure. The sensitivity analysis carried out using the
Generalised Likelihood Uncertainty Estimation (GLUE) methodology demonstrates the importance of
some of the initial state variables during flood events in the TETIS model. Therefore, these variables
must be estimated. For a real-time automatic calibration of these initial values, the shuffled complex
evolution (SCE-UA) algorithm was selected. The flood forecasting has been divided in two phases:
the “warming period” which focuses on a period of time previous to forecasting using real-time data,
and the “forecasting” phase. The main goal during the warming period is to estimate some initial state
variables of the TETIS model using automatic calibration. In this way, it is possible to get a better
accordance among observed and simulated discharges during the warming period and to increase the
forecasting reliability. The results indicated that SCE-UA is robust and efficient. It was highlighted
that calibrating initial state variables allowed adjusting properly the observed and simulated discharges
during warming period. This warming period strategy can be coupled to a Quantitative Precipitation
Forecast (QPF) scheme in order to produce better results and reduce the uncertainty.

Keywords: automatic calibration; distributed modelling, flood forecasting; initial state variables

INTRODUCTION
There are many methodologies which are able to explain flood processes in natural river basins and
provide a reliable estimation of timing, shape and volumes during floods depends on the
characteristics of the selected model (Wood and O’Connell, 1985). These conceptual models can be
considered as an excellent tool to simulate rainfall-runoff processes during flood events because they
usually have few parameters and are easier to implement than physically based models if the adequate
scale is selected.

Distributed models take into account the spatial and temporal variability included in different
hydrological processes. The hydrological cycle is heterogeneous in space and time, even assuming
climatic and basin stationarity. The inputs (precipitation and potential evapotranspiration (PET))
change in space and time and the watershed characteristics, represented by the parameters in any
model, change in space (Beven, 2000).

In this paper the conceptual distributed model and their automatic calibration procedure are briefly
described. The forecasting strategy is proposed and applications in three different basins are used to
show the advantages of distributed modelling. Finally, some conclusions are presented.

17 to 19 October 2005, Tromsø, Norway 1 ACTIF/FloodMan/FloodRelief


International conference on innovation advances and implementation of flood forecasting technology

TETIS MODEL DESCRIPTION


The TETIS model is a distributed hydrological conceptual model, which is able to simulate
continuously the main components of the hydrological cycle. The model has been tested satisfactorily
in different climatic scenarios with a wide range of basin areas in Spain and France (Vélez, 2001;
Francés et al., 2002; Vélez et al., 2002a, 2002b and 2002c, Vaskova et al., 2004).

Hydrological behaviour at cell scale

The model is distributed in space using a regular grid. The rainfall-runoff process is modelled using
five vertical tanks as shown in Figure 1, where each tank represents different water storage in an
“extended soil column”: snow pack, static, surface, gravitational, and aquifer. The relationship among
the tanks is based on soil properties such as field capacity and the hydraulic saturated conductivities.
These soil characteristics were previously estimated at each cell using land use, lithology, edaphology,
geomorphologic characteristics and other available information (Puricelli and Francés, 2002).
SNOWFALL
T0 SNOW RAINFALL
MELTING X0
H0
Snowpack Y0

PRECIPITATION
X1
EVAPOTRANSPIRATION
D1
Y1

EXCEDENCE T1
H u*
X2 H1
Static storage
D2

INFILTRATION T2 Y2
X3 H2 DIRECT
RUNOFF
Surface
D3 storage

PERCOLATION T3 Y3
X4 H3 INTERFLOW
Gravitational
D4 storage
GROUNDWATER
OUTFLOW T4
Y4
X5
H4 BASE
FLOW
Aquifer
T5

H5
Channel

Figure 1 General description of hydrologic behaviour of TETIS model at cell scale

The vertical connections among tanks describe the precipitation (rainfall or snow), snow melting,
evapotranspiration, infiltration and percolation processes. The horizontal connections describe the
overland flow, interflow and base flow. Finally, a groundwater outflow is considered in the lowest
tank to close the water balance.

The horizontal flow movement is defined by a three dimensional mesh of connected tanks which drain
toward the corresponding tank in the downstream cell, following the flow directions until channel
network is reached. The routing along channel network to the outlet has been coupled to the kinematic
wave procedure using geomorphologic basin characteristics. A detailed description of the model can
be found in Vélez (2001), Francés et al (2002) and Vélez et al. (2002a, 2002 b, 2002c).

17 to 19 October 2005, Tromsø, Norway 2 ACTIF/FloodMan/FloodRelief


International conference on innovation advances and implementation of flood forecasting technology

Initial state variables

The initial state variables included in TETIS model are: the static storage, H1, which is a percentage
of the maximum capacity (where 100% means that maximum storage capacity has been reached), the
surface water depth, H2 given in mm, the gravitational storage, H3 given in mm, the aquifer, H4
given in mm and the channel, H5 given as a percentage or the bankfull discharge (values over 100 %
indicate the flow is flooding over cell’s bank). Snow pack is a state variable which was not included
in this study.

The main goal of the proposed procedure is to guarantee the best initial state variables estimated
before to launch the forecasting strategy. In this way, it has been selected a period of time previous to
forecasting long enough to calibrate automatically the initial state variables. The number of
parameters to calibrate is important because if a higher number of parameters are selected the
computation time will be longer. In this way, it is recommended to optimize the minimum number of
initial state variables, the maximum is five, because there are five tanks on which it is necessary to
estimate their initial state. The initial state is assumed the same for all cells.

Automatic calibration

Traditionally, the manual calibration has been used to calibrate hydrologic models, but it requires an
expert hydrologist to obtain reliable results and usually it is high time consuming. In the last decade, it
different procedures have been developed to calibrate automatically some parameters of hydrological
conceptual models. These procedures are based on global search, multiple start and evolution
techniques.

In this study, the selected optimization algorithm for automatic calibration was the Shuffled Complex
Evolution – University of Arizona (SCE-UA) proposed by Duan et al. (1992). This algorithm has
been tested with satisfactory results in different conceptual models at different basins (Duan et al.,
1992; Sorooshian et al., 1993; Duan et al., 1994; Yapo et al., 1996; Gan and Biftu, 1996; Thyer et al.,
1999; Lidén and Harlin, 2000; Seranath et al., 2000; Madsen et al., 2002).

Other automatic calibration procedures such as Genetic Algorithms or Simulated Annealing have been
detailed in scientific literature with good results, but when they are compared to SCE-UA no
significant advantages were detected (Thyer et al., 1999; Madsen, 2000; Madsen et al., 2002).

Objective function

The TETIS model can estimate different objective functions as: Root Mean Square Error (RMSE),
Nash and Sutcliffe efficiency index (R²), Heteroescedastic Maximum Likelihood Estimator (HMLE)
and Autocorrelated Maximum Likelihood Estimator, among others. According to Sorooshian et al.
(1983), Gan and Burges (1990), Brath and Rosso (1996) and Vélez y Francés (2004a) it is not clear
which is the best performance estimator. The selected objective function was the RMSE.

Initial values and search range

The SCE-UA automatic optimization procedure requires an initial value and a search range for each
state variable. The search range must be supplied as a minimum and maximum value but within
feasible values. These values must be previously defined by the modellers according to their
experience. Table 1 shows the recommended values according to the experience obtained in previous
research (Vélez et al. 2002a; Vélez, 2003; Vélez and Francés, 2004).

This selection of initial values and search ranges gives a subjective character to the calibration process
(Beven, 2000). Vélez (2003) and Vélez and Francés (2004) noted that final result are sensitive to
search range and initial values.

17 to 19 October 2005, Tromsø, Norway 3 ACTIF/FloodMan/FloodRelief


International conference on innovation advances and implementation of flood forecasting technology

Table 1 Recommended initial values and search ranges during SCE-UA process

Initial state variable Minimum Maximum Initial value


search value search value
Static storage – H1 (%) 0 100 50
Surface water depth - H2 (mm) 0 50 0
Gravitational storage - H3 (mm) 0 100 0
Aquifer - H4 (mm) 0 1000 10
Channel - H5 (%) 0 200 20

Uncertainty analysis

The Generalized Likelihood Uncertainty Analysis (GLUE) proposed by Beven and Binley (1992)
must be carried out in all studied cases in order to perform the sensitivity analysis. This analysis is
useful because can show the predictive capacity of TETIS model and supply a qualitative estimation
of the model performance, specially highlighting the most sensitive initial state variables at each case.

FORECASTING STRATEGY
It is important to guarantee that the TETIS model was satisfactorily calibrated, where calibration
factors have been estimated properly and represent the rainfall-runoff process in the basin adequately,
and only initial state variables can be modified to simulate properly floods. The future rainfall
estimation has not been included in the analysis; in this study real events have been used in order to
observe the effect on initial state variables.

The main goal is to estimate the best possible initial state variables using TETIS model with real-time
data. In this way, it is possible to get a better accordance among observed simulated discharges during
a period of time previous to forecasting and then to launch forecasting. The proposed strategy is
divided in two phases:

The warming period

The warming period is the period of time immediately before the forecast is executed. During this
period the TETIS model receives real-time information about gauge stations and rain gauges.
Therefore, the warming period is dedicated to calibrating automatically the initial state variables using
the SCE-UA procedure in order to obtain the best adjust possible among observed and simulated
discharges. It is necessary to define the length of the period of time previous to forecasting according
to the basin size and the temporal scale of data.

The forecast

The second phase is the forecast. If a good fit among observed and simulated discharges is obtained
during the warming period, then it is possible to launch TETIS model with some confidence. In the
forecast phase the TETIS model uses the initial state variables obtained during warming period. The
model is run during at least one day ahead using heuristic or other approaches as future rainfall.
Uncertainty will decay if a good approach for future rainfall is performed.

APPLICATIONS
Three basins have been selected in order to show the advantage of using distributed conceptual
modelling and considering the initial state variables during flood forecasting. In all cases, “real
events” were processed and considered as “real-time” events, in this way the real effect of initial state
variables on distributed conceptual model can be studied without the influence and uncertainty
induced by future rainfall. The precipitation of each event was aggregated and used as future rainfall.

17 to 19 October 2005, Tromsø, Norway 4 ACTIF/FloodMan/FloodRelief


International conference on innovation advances and implementation of flood forecasting technology

“Rambla del Poyo” Basin

The “Rambla del Poyo” Basin (located at the East coast of Spain) was used to test different values for
initial state variable H1(%). This basin is characterized by the presence of convective storms; it is a
typical Mediterranean basin where flash floods are quite frequent. Real-time data are collected with a
temporal resolution of 15 minutes, which is aggregated to 1 hour. The cell size used in this case is
500 m. There is one gauging station in the basin with an associated area of 193 km² and there are
seven rain gauge stations. However, during the calibration process only one station was considered
because this is the same station is used during forecasting. This application is based on two real
events dated on 10 December 1988 and 24 May 1989. Finally, a forecasting strategy was applied to
this basin considering a warming period of six hours and the TETIS model was executed to predict
one day ahead as shown in Figure 2.
0 0
Ppt (mm)

Ppt (mm)
5 Future rainfall
4
10
200 8
a) d)
160 Observed
6
Discharge (m³/s)

Discharge (m³/s)
Predicted

120
4
80
2
40

0
0 24/05/1989 6:00
12/10/1988 18:55 13/10/1988 6:55 24/05/1989 18:00
Date
Date 8
b) 200
e)

160 6
Discharge (m³/s)

Discharge (m³/s)

120
4

80
2
40
0
0 24/05/1989 6:00 24/05/1989 18:00
12/10/1988 18:55 13/10/1988 6:55
Date
Date
8
c) 250 f)
200 6
Discharge (m³/s)

Discharge (m³/s)

150
4

100
2
50

0
0
12/10/1988 18:55 13/10/1988 6:55 24/05/1989 6:00 24/05/1989 18:00
Date Date
Figure 2 Results during flood event dated on 12/10/1988 and 05/24/1989 at “Rambla del Poyo”
basin. The upper graphs correspond to hyetographs and three graphs below point out
the percentage of initial static storage H1, as follow: a) 10 %, b) 50 %, c) 70 %, d) 0
%, e) 50 % and f) 90 %

The Figure 2 shows different initial values for static storage H1, where it can be observed how
important the selection of this value is. For instance, during the flood in 1988 the maximum predicted

17 to 19 October 2005, Tromsø, Norway 5 ACTIF/FloodMan/FloodRelief


International conference on innovation advances and implementation of flood forecasting technology

values varies from 23.5 m³/s to 69.3 m³/s to 212.9 m3/s considering initial static storage values of 10%,
50 % and 70 %, respectively. The maximum values of 70% and 90% were stated because higher
values in static storage generate hydrological responses during warming period, which is not realistic.

Real Collobrier basin

The “Real Collobrier” basin is an experimental basin located in the south of France with a
Mediterranean climate. This basin has been studied extensively during last years. The temporal
resolution of available data is one hour and cell size used is 500 m. There are three gauge stations:
“Maliere” and “Collobrieres” with areas of 12 km² and 29 km², respectively, and “Pont de Fer” located
downstream with an area of 70 km². There are nine rain gauge stations inside the basin and their
information is used during forecasting.

The SCE-UA methodology was used during a warming period of 24 hours to estimate three initial
state variables: the static storage H1, the aquifer H4 and the channel H5. Finally, forecasting strategy
was applied to this basin to predict 1 day ahead in three episodes as shown in Figure 3. The automatic
calibration was carried out only at “Pont de Fer” station during the warming period.
0 0
0
Ppt (mm)

Ppt (mm)
2
Ppt (mm)

10 Future rainfall
4
Observed precipitation Episode 1 4
20 6 Episode 3
30
8 Episode 2 8

7
Pont de Fer
25 Observed 6 Pont de Fer
40 Pont de Fer
Discharge (m³/s)

Predicted
20 5
Previously estimated Discharge (m³/s)
Discharge (m³/s)

30 4
15
3
10 20
2
5
10 1
0
29/01/1976 20:00 30/01/1976 20:00 31/01/1976 20:00 0
0 12/10/1972 2:00 13/10/1972 2:00 14/10/1972 2:00
Date 29/01/1974 0:00 30/01/1974 0:00 31/01/1974 0:00 Date
3.0 Date

2.5
Maliere 9 1.6
Maliere
8
Maliere
Discharge (m³/s)

7
Discharge (m³/s)

1.2
Discharge (m³/s)

2.0
6
1.5 5
0.8
4
1.0
3 0.4
0.5 2
1 0.0
0.0
29/01/1976 20:00 30/01/1976 20:00 31/01/1976 20:00 0 12/10/1972 2:00 13/10/1972 2:00 14/10/1972 2:00
Date 29/01/1974 0:00 30/01/1974 0:00 31/01/1974 0:00 Date
Date 7
12 30

Collobrieres
6 Collobrieres
10 25
Collobrieres
5
Discharge (m³/s)

Discharge (m³/s)

Discharge (m³/s)

8 20 4

6 15 3

4 10 2

2 5 1

0
0 0 12/10/1972 2:00 13/10/1972 2:00 14/10/1972 2:00
29/01/1976 20:00 30/01/1976 20:00 31/01/1976 20:00 29/01/1974 0:00 30/01/1974 0:00 31/01/1974 0:00
Date Date Date

Figure 3 Results during flood episodes dated on 29/01/1976, 29/01/1974 and 12/10/1972 at
“Rambla del Poyo” basin. The upper graphs are the hyetographs and three graphs
below correspond to hydrographs at “Pont de Fer”, “Maliere” and “Collobrieres”,
respectively

17 to 19 October 2005, Tromsø, Norway 6 ACTIF/FloodMan/FloodRelief


International conference on innovation advances and implementation of flood forecasting technology

In this case, the main advantage of distributed models is reflected because performing the real-time
forecasting at “Pont de Fer”, the results at upstream catchments “Maliere” and “Collobrieres” can be
observed immediately.

Alagon River Basin

The Alagon River Basin is located in western Spain; this is a continental basin with an area of
1857 km² draining to the “Gabriel y Galan” gauging station, located few kilometres upstream of
reservoir with same name. There are eight gauging stations with drainage areas ranging from 70 km²
to 1857 km². The temporal resolution is 15 minutes, which has been aggregated to 1 hour in order to
reduce the computation time. The spatial resolution used is 500 m. In Figure 4 the results at “Gabriel
y Galan” and “Garcibuey” gauging stations are shown for two events. In this case the forecast is
performed with differences of six hour, as if real-time forecasting would be started at six hour interval.
The warming period used in this case was 12 hours and four state variables were automatically
calibrated. These were: the static storage H1, the gravitational storage H3, the aquifer H4 and the
channel H5. In both cases, the warming period allows an adjustment to be made between observed
and simulated discharges and consequently there will be more confidence in the forecast. The
forecasting strategy was applied to predict an event one day in advance.
0
Ppt (mm)
a) d)
Ppt (mm)

Future rainfall 10
Observed precipitation
10
700

1600 600
Observed
Discharge (m³/s)

Discharge (m³/s)

Predicted 500
1200 Previously estimated
400

800 300

200
400
100

0
0
03/03/2001 12:00 04/03/2001 0:00 04/03/2001 12:00 03/03/2001 12:00 04/03/2001 0:00 04/03/2001 12:00
Date
Date 0
Ppt (mm)

b) e)
Ppt (mm)

10

10
700

1600 600
Discharge (m³/s)

Discharge (m³/s)

500
1200
400

300
800

200

400 100

0
0
03/03/2001 12:00 04/03/2001 0:00 04/03/2001 12:00
03/03/2001 12:00 04/03/2001 0:00 04/03/2001 12:00
Date Date
0
Ppt (mm)

f)
Ppt (mm)

0
c)
10

10
700

1600 600
Discharge (m³/s)

Discharge (m³/s)

500
1200
400

800 300

200
400
100

0
0
03/03/2001 12:00 04/03/2001 0:00 04/03/2001 12:00
03/03/2001 12:00 04/03/2001 0:00 04/03/2001 12:00
Date
Date

Figure 4 Results of TETIS model at “Gabriel y Galan” and “Garcibuey” stations located at
Alagon River Basin. The upper graph corresponds to hyetograph and three graphs
below point out the forecasting beginning at: a) 04/03/2001 0:00, b) 04/03/2001 6:00,
c) 04/03/2001 0:00, d) 04/03/2001 0:00, e) 04/03/2001 6:00 and f) 04/03/2001 0:00

17 to 19 October 2005, Tromsø, Norway 7 ACTIF/FloodMan/FloodRelief


International conference on innovation advances and implementation of flood forecasting technology

CONCLUSIONS
The use of conceptual distributed models in flood forecasting has been tested and validated. It has
been highlighted that the TETIS model is a robust tool to simulate flood and can be used satisfactorily
during forecasting. The SCE-UA procedure showed excellent and efficient performance and the
results were obtained easily and automatically. Therefore, the initial state variables can be calibrated
satisfactorily during warming period using SCE-UA.

The length of the warming period must be in accordance with the basin area and the temporal scale, in
order to guarantee the success of the automatic calibration procedure. In the studied cases, which vary
from small to large basins, an hourly temporal scale was proposed. This was enough to consider at
least 6 to 24 hour warming period for a one day in advance forecasting.

The selected applications showed the relevant role of the initial state variables and the automatic
calibration during the flood forecasting. However, flash floods in ephemeral rivers cannot be easily
predicted as shown in “Rambla del Poyo” application. It is not possible to use automatic calibration
because during the warming period because the observed hydrograph is zero, and the result strongly
depends on initial state variable, H1. Therefore, it is required a previous knowledge of soil moisture
before to launch the forecasting process.

REFERENCES
Beven, K., Binley, A. (1992), The future of distributed models: Model calibration and uncertainty
prediction. Hydrological Processes, 6, 279-298.
Beven, K.J. (2000), Parameter Estimation and Predictive Uncertainty. In: Beven, K.J., (Ed.), Rainfall-
Runoff Modelling. The Primer. John Wiley & Sons Ltd. Chichester. U.K. (Chapter 7).
Brath, A., Rosso, R. (1996), Adaptive calibration of a conceptual model for flash flood forecasting.
Water Resources Research, 29(8), 2561-2572.
Duan, Q., Sorooshian, S., Gupta, V.K. (1992), Effective and efficient global optimization for
conceptual rainfall-runoff models. Water Resources Research, 24(7), 1163-1173.
Duan, Q., Sorooshian, S., Gupta, V.K. (1994), Optimal use of the SCE-UA global optimization method
for calibrating watershed models. Journal of Hydrology, 158, 265-284.
Francés, F., Vélez, J.J., Vélez, J.I., Puricelli, M. (2002), Distributed modelling of large basins for a
real time flood forecasting system in Spain. Proceedings Second Federal Interagency Hydrologic
Modelling Conference. Las Vegas, USA. July. CD Format.
Gan, T.Y., Biftu, G.F. (1996), Automatic calibration of conceptual rainfall-runoff models:
Optimization algorithms, catchment conditions, and model structure. Water Resources Research,
32(12), 3513-3524.
Gan, T.Y., Burges, S.J. (1990), An assessment of a conceptual rainfall-runoff model's ability to
represent the dynamics of small hypothetical catchments 1. Models, model properties, and
experimental design. Water Resources Research, 26(7), 1595-1604.
Lidén, R., Harlin, J. (2000), Analysis of conceptual rainfall-runoff modelling performance in different
climates. Journal of Hydrology, 238, 231-247.
Madsen, H. (2000), Automatic calibration of a conceptual rainfall-runoff model using multiple
objectives. Journal of Hydrology, 235, 276-288.
Madsen, H., Wilson, G., Ammentorp, H.C. (2002), Comparison of different automated strategies for
calibration of rainfall-runoff models. Journal of Hydrology, 261, 48-59.
Puricelli, M., Francés, F. (2002), Estimación de parámetros hidrológicos del terreno en la modelación
hidrológica distribuida en grandes cuencas. Proceedings 3ª Asamblea Hispano-Portuguesa de
Geodesia y Geofísica. Ed. UPV. Valencia, Spain. February 7-8, 1688-1692.
Senarath, S., Ogden, F.L., Downer, C.W., Sharif, H.O. (2000), On the calibration and verification of
two-dimensional distributed, Hortonian, continuous watershed models. Water Resources Research,
36(6), 1495-1510

17 to 19 October 2005, Tromsø, Norway 8 ACTIF/FloodMan/FloodRelief


International conference on innovation advances and implementation of flood forecasting technology

Sorooshian, S., Gupta, V.K., Fulton, J.L. (1983), Evaluation of maximum likelihood parameter
estimation techniques for conceptual rainfall-runoff models: Influence of calibration data variability
and length on model credibility. Water Resources Research, 19(1), 251-259.
Sorooshian, S., Duan, Q., Gupta, V.K. (1993), Stochastic parameter estimation procedures for
hydrologic rainfall-runoff models: Correlated and Heteroescedastic error cases. Water Resources
Research, 29(4), 1185-1194.
Thyer, M., Kuczera, G., Bates, B.C. (1999), Probabilistic optimization for conceptual rainfall-runoff
models: A comparison of the Shuffled complex evolution and simulated annealing algorithms. Water
Resources Research, 35(3), 767-773.
Vašková, I., Francés, F., Vélez, J.J. (2004), Empleo de la modelación distribuida en el estudio de los
recursos hídricos del País Vasco. Proceedings 4ª Asamblea Hispano-Portuguesa de Geodesia y
Geofísica, Figueira da Foz, Portugal. February 3-7.
Vélez, J.I. (2001), Desarrollo de un modelo hidrológico conceptual y distribuido orientado a la
simulación de las crecidas. Doctoral Thesis. Universidad Politécnica de Valencia, Departamento de
Ingeniería Hidráulica y Medio Ambiente. 266 pp.
Vélez, J.J., Vélez, J.I., Francés, F. (2002a), Simulación hidrológica de crecidas en grandes cuencas
mediante el uso de la modelación distribuida. Proceedings 3ª Asamblea Hispano-Portuguesa de
Geodesia y Geofísica. Ed. UPV. Valencia, Spain. February 7-8, 1682-1687.
Vélez, J.I., Vélez, J.J., Francés, F. (2002b). Modelo distribuido para la simulación hidrológica de
crecidas en grandes cuencas. Proceedings XX Congreso Latinoamericano de Hidráulica. La Habana,
Cuba. October.
Vélez, J.J., Vélez, J.I., Puricelli, M., Francés, F. (2002c), Hydrological simulation of flood events at
large basins using distributed modelling. Proceedings XXVII General Assembly European
Geophysical Society. Nice, France. April 21-26. Ed. Geophysical Research Abstracts. POSTER.
Vélez, J.J., Francés, F. (2004), Sensibility analysis of a conceptual distributed model during flood
events. In Proceedings I General Assembly European Geosciences Union. Nice France. April 26-30.
Ed. Geophysical Research Abstracts. CD format.
Wood, E. F., O’Connell, P. E. (1985), Real-time forecasting. In: Hydrological forecasting. (Eds.
Anderson, M. G. and Burt, P. T.). John Wiley and Sons Ltd., Chapter 15, 59-84.
Yapo, P.O., Gupta, H.V., Sorooshian, S. (1996), Automatic calibration of conceptual rainfall-runoff
models: sensitivity to calibration data. Journal of Hydrology, 181, 23-48.

17 to 19 October 2005, Tromsø, Norway 9 ACTIF/FloodMan/FloodRelief

You might also like