Sei sulla pagina 1di 8

bs_bs_banner

Water and Environment Journal. Print ISSN 1747-6585

Contrast and transferability of parameters of lumped water


balance models in the Segura River Basin (Spain)
Francisco Pellicer-Martinez1 & Jos M. Martnez-Paz1,2
1
2

Water and Environment Institute (INUAMA), University of Murcia, Murcia, Spain and
Applied Economics, University of Murcia, Murcia, Spain

Keywords
hydrological model contrast; parameter
transferability; Segura River Basin; snowmelt;
water balance.
Correspondence
Francisco Pellicer-Martinez, Water and
Environment Institute (INUAMA), University of
Murcia, Campus de Espinardo; s/n; Edicio D,
30100 Murcia, Spain. Email:
francisco.pellicer@um.es;
pellicer.martinez@gmail.com
doi:10.1111/wej.12091

Abstract
Knowing the water resources in a river basin is a key aspect to appropriate water
planning, and they can be estimated by lumped water balance models. The possibility to transfer the calibrated parameters of a contrasted model to non-gauged
ones improves this knowledge with no need for direct measurements. This work
studies the transferability of the parameters of the lumped models in ve catchments in the southeast of Spain, testing the inuence of the snow and the effect of
serial autocorrelation in the goodness of t. The results are also compared with
those of the distributed model of reference in Spain (SIMPA). As principal results, the
abcd model is the selected one, the snow effect is negligible, the transferability of
parameters for this case is not totally possible, and serial autocorrelation must be
included for this kind of modelling.

Introduction
The assessment of water resources in a basin is a necessary
step before hydrological planning, and it is crucial in waterscarce regions where water resources are key for the development of the basic activities carried out by the inhabitants
of a basin for the distribution and allocation of volumes for
different economic and environmental uses (Gleick 1986). In
the Segura River Basin (Spain) the natural water resources are
not able to supply all the water demands, thus needing external resources (Perni et al. 2012) and becoming, undoubtedly,
one of the most water-scarce basins in Europe (Martnez-Paz
et al. 2014).
One of the recent advances in hydrological sciences is to
transfer the parameters of a contrasted hydrological model
calibrated at a catchment outlet to others. The result of estimating stream ows in ungauged catchments using the
parameters previously calibrated in gauged one increases
stream ow information and would gain in economy and save
time compared to directly measuring stream ows. Due to
this advantage, much research has been conducted on this
subject (van der Linden & Woo 2003) in which lumped
monthly water balance models are generally used because it
is then possible to focus the study on the behaviour of the
parameters (Mouelhi et al. 2006). The critical issue of this
process is to select an appropriate model in order to accurately reproduce the stream ows generated in the basin
under study (Hughes & Metzler 1998).

Water and Environment Journal 29 (2015) 4350 2014 CIWEM.

This paper is mainly aimed at nding a lumped water


balance model which provides a proper performance in the
upper Segura River Basin. For this purpose, four lumped
water balance models are applied and contrasted, along with
the testing of a snowmelt module. Once a proper model has
been selected, the transferability of the parameters is analysed. This is the rst time that this kind of study has been
performed in the area. The effects of serial autocorrelation
obtained in the goodness of the t are also studied.
The secondary objective of this paper is to compare the
results of the selected lumped water balance with the results
of the only hydrological model applied in this basin, the
so-called SIMPA model (Estrela & Quintas 1996). This distributed model, which is used by public authorities in water management to assess all the water resources in Spain, has two
drawbacks, namely (i) it is not specic to every Spanish river
basin, and (ii) it does not consider the snow as a hydrological
process that can be signicant to the headwater basins in the
Mediterranean region (Karpouzos et al. 2011). This comparison identies any possible differences between the two
models, as the distributed model requires a larger amount of
input information and a greater computing capacity than the
lumped water balance model.
This paper is structured as follows: the methodology is
included in the second section; the third section deals with
the main features of the studied area and the information
used; the next section shows the analysis and discussion of
results; and, nally, the conclusions are drawn.
43

Transferability of the parameters of the lumped models

F. Pellicer-Martinez and J. M. Martnez-Paz

Stage 1 - Hydrological Models

Stage 2 Snowmelt module

Stage 3 - Transferability

Four lumped water balance


models are applied.

A snowmelt module is added


to the hydrological model
selected in Stage 1.

The transferability study is


structured following the
proxy-basin criteria.

Calibration:
Split-sample test.

Series are compared using


linear regressions estimated by
feasible generalized least
squares.

Calibration:
Split-sample test.

Transferability analysis
At the end of this stage, a
hydrological model is selected.

Identify the relevance of the


snowfalls in the area under
study.

Contrast of hydrological models

Comparison of the results with


SIMPA model.

Analysis of the transferability


of the parameters

Fig. 1. Methodological framework.

Methodology
The complete methodology used in this work is divided
into three stages (Fig. 1). The rst two stages are aimed at
nding a hydrological model which provides a proper performance of the studied area. The objective of the third stage is
to analyse the transferability of the parameters and to
compare the selected hydrological model with the SIMPA
model.

Stage 1 Hydrological models


The four lumped water balance models applied are: T-
model (Alley 1984), GR2M model (Makhlouf & Michel 1994),
abcd model (Thomas 1981) and WASMOD (Vandewiele &
Win 1993). All of them are based on the continuity equation
[Eq. (1)] where V is the variation in the volume stored in
the catchment; P(t) is the precipitation acting as the input
into the catchment; and the outputs are simulated stream
ows, Q(t), together with actual evapotranspiration, ET(t)
(Zhang et al. 2008). Nevertheless, in the algorithm of all
these mathematical models, the input data in the form of
temporal series are precipitation, P(t), and the potential
evapotranspiration, EP(t).
44

V = P (t ) ET (t ) Q (t )

(1)

The equations that govern the ows between the storages of


the model (Fig. 2) are monitored by three or four parameters.
The T- model has three parameters; the rst one determines
the surface ow, Qs(t), the second limits the value of soil
moisture, S(t), in the upper storage, and the last one establishes the discharge, Qg(t), which depends on the stored
volume in the lower storage, G(t). The GR2M model has three
parameters; the rst one establishes the actual evapotranspiration, ET(t), and the soil moisture, the second one determines
the percolation and the surface ow, and the last one evaluates the discharge. The abcd model has four parameters; two
of them establish the actual evapotranspiration and the soil
moisture, another one divides the inltration by the percolation, G(t), and surface ow, whereas the last one determines
the discharge. The WASMOD model has four parameters; one
monitors the actual evapotranspiration, another one establishes the surface ow, and two more parameters evaluate the
discharge. All these parameters have to be calibrated by
means of naturalized stream ows (Wurbs 2005).
Figure 2 illustrates all the models: a) the T- model and
GR2M model; b) the abcd model; and c) the WASMOD model.

Water and Environment Journal 29 (2015) 4350 2014 CIWEM.

F. Pellicer-Martinez and J. M. Martnez-Paz

Transferability of the parameters of the lumped models

ET(t)

P(t)

P(t)

S(t)

ET(t)

ET(t)

P(t)

S(t)

G(t

S(t)

G(t
Qs(t)

G(t)

Qg(t)

G(t)
Q(t)

Qs(t)

Qg(t)

Qs(t)

Q(t)

Qg(t)

Q(t)

a)

b)

c)

Fig. 2. Frame of the lumped models.

Calibration

Stage 2 Snowmelt module

The calibration of parameters of the hydrological models,


used also at stage 2, is carried out by minimizing the sum of
the square of deviations, SSQ, [Eq. (2)]. These deviations are
the difference between the stream ow simulated by the
model, q (t ) , and the naturalized streamow, q(t), that have
previously been transformed by square root transformation.
This transformation provides a set of deviations that is
assumed normally distributed with zero expectation and stochastically independent.

Snowfalls can be relevant since they occur in other basins


located in the Mediterranean area with similar climatic characteristics (Karpouzos et al. 2011). Therefore, in order to
identify whether this hydrological process is relevant to this
area or not, a snowmelt module is included in the proper
water balance. In this case, the snowmelt module proposed
by Xu et al. (1996), which incorporates two new parameters
in the model that also have to be calibrated, is used to represent the following hydrological processes: snowfall, snow
accumulation on the peaks, and snowmelt.
If the incorporation of the snowmelt module improves the
NS values obtained by the selected lumped water balance,
then the snow will be considered as a relevant hydrological
process which has to be taken into account in the analysis of
the transferability of the parameters.

SSQ = (q (t ) q (t ))

(2)

t =1

The models are calibrated using the split-sample test dened


by Klemes (1986). The rst years of the stream ow data is
used in the take-off of the model, and the rest of the data is
divided into two periods. The rst period is used in the calibration, and the second one is used for the validation. Once
the hydrological models have been calibrated, the efciency
criterion NS (Nash & Sutcliffe 1970) is used to measure the
quality of the calibration [Eq. (3)], where, q (t ) and q(t) are the
same variables as used previously and, q (t ) , is the mean of
the naturalized stream ow series.
n

NS = 1

(q (t ) q (t ))
t =1
n

(q (t ) q )

t =1

Water and Environment Journal 29 (2015) 4350 2014 CIWEM.

(3)

Stage 3 Transferability of models


Parameter transferability is conducted using proxy-basin criterion (Klemes 1986). This criterion applies the calibrated
parameters into a catchment in another close catchment and
compares the results by means of simple linear regressions
(Zhang et al. 2008). These regressions do not have a constant
term, and the linear coefcient is estimated by ordinary least
squares (OLS). However, since the monthly stream ow series
usually have a meaningful serial autocorrelation, the estimation by OLS has the problem of residual correlation, with
inefcient estimators and an overestimated coefcient of
45

Transferability of the parameters of the lumped models

F. Pellicer-Martinez and J. M. Martnez-Paz

Fig. 3. Segura River Basin District and catchments analysed.

Table 1 Catchments analysed


Name

Anchuricas

La Novia

Argamasa

Tobarico

Lietor

Code
Area (km2)
River name
Average annual temperature (C)
Average annual precipitation (mm)
Average annual potential evapotranspiration (mm)
Average annual run-off (mm)
Run-off coefcient
Average elevation (m)

A
235
Segura
10.7
830
875
336
0.40
1419

B
273
Zumeta
10.6
805
908
145
0.18
1557

C
185
Tus
13.7
753
866
195
0.26
1203

D
179
Taibilla
11.3
655
938
85
0.13
1457

E
576
Mundo
12.2
685
817
197
0.29
1135

determination R2 (Kroll & Stedinger 1998). To solve this


problem, in this paper the estimation is carried out by feasible generalized least squares (FGLS) using the HildrethLu
model (Hildreth & Lu 1960), the formula of which is shown in
[Eq. (4)] below:

Qt = Q t 1 + ut

ut = ut 1 + t

t iid (0, 2 )

(4)

where Qt is the naturalized stream ow, Q t is the ow


simulated by the model, 1 is the linear coefcient in the
regression, t is the error series, 2 is the variance of the
error series, is the coefcient that monitors the residual
correlation, and ut is the corrected error series. When
autocorrelation does not exist, is not meaningfully distinct
to zero, and both methods (OLS and FGLS) converge (Dios
1998).
46

Study area
The methodology was applied to ve catchments located in
South-Eastern Spain, which belong to the Segura River Basin
District (Fig. 3).
Table 1 shows the catchment names, the codes used in
this paper, the name of the river and the most important
physical and climatic features.
Data are taken from ofcial Spanish sources. The digital
elevation model used is provided by the National Centre for
Geographic Information (http://www.cnig.es), the gauging
stations and the sets of gauged stream ows that have been
naturalized are those of the Centre for Hydrographical
Studies (http://hercules.cedex.es/anuarioaforos/default.asp),
and the climate series used in the models are obtained by the
treatment of the information on the monthly raster maps for

Water and Environment Journal 29 (2015) 4350 2014 CIWEM.

F. Pellicer-Martinez and J. M. Martnez-Paz

Transferability of the parameters of the lumped models

Table 2 Periods used in the optimization process


Catchment

Take-off

Calibration

Validation

Anchuricas(A)
La Novia(B)
Argamasa(C)
Tobarico(D)
Lietor(E)

12/196511/1970
12/196511/1970
10/194010/1945
10/194010/1945
10/197010/1975

12/197010/1989
12/197010/1989
11/194505/1954
11/194511/1949
11/197502/1984

11/198909/2008
11/198909/2008
06/195401/1963

03/198407/1992

all the Spanish areas that were created as baseline information in the SIMPA distributed model (http://servicios2
.magrama.es/sia/visualizacion/descargas/). The potential evapotranspiration and precipitation series are areal values that
have been estimated as the average of the raster variable
values for each catchment.

Results and discussion


Stage 1 Hydrological models
The four models have been applied in the study area. The rst
5 years in the series of naturalized stream ows have been
used to estimate the initial conditions in the storages of each
hydrological model. If the rest of the series are over 10 years,
they are divided into two parts each with the same amount of
years: the rst part is used for the calibration, and the second
one for the validation of the models. Since the period of naturalized stream ows is different in every gauging station,
Table 2 gathers the time intervals used in these processes.
The minimization of the sum of the square of deviations
have been carried out by using a non-linear optimization
algorithm known as GRG2 that nds the extreme values
of the functions by the generalized reduced gradient
algorithm (Lasdon et al. 1978). Once the SSQ functions for
each model were optimized, the efciency criterion NS was
calculated.
The NS values obtained by all the models show a heterogeneous performance in the area under study (Table 3). The
better performance is always achieved in Catchment A,
where the abcd model reached NS values of 0.88 and 0.84 for
the calibration and validation periods, respectively. On the
contrary, the lowest results are found in Catchment E, lower
than 0.50 in all models. Moreover, all the catchments for all
the models experiment a decrease in the NS value in the
validation period. Catchment C is the one that yields the
highest decrease in goodness of t for all models, where
some NS values become negative in the validation. This
decrease may be due to changes in vegetation cover in the
catchment area, or because the length of the naturalized
stream ow series is shorter. This highlights the climatic
changes within their own data series.
The abcd model is selected to study the inuence of the
snowfalls and the transferability of the parameters since it

Water and Environment Journal 29 (2015) 4350 2014 CIWEM.

almost always provides the highest NS values both in the


calibration and validation processes. This model achieves
values greater than 0.60 in three out of the ve catchments,
and the maximum (0.88) is found here. Although none of the
models systematically provides the best performance for all
catchments, these results can be considered as good as
those obtained by Mouelhi et al. (2006) and Wang et al.
(2011) since the NS values reproduce the same cumulative
frequency.
Figure 4 shows the representation of the results obtained
by all four models in Catchment A. It conrms, graphically,
the good performance obtained for all the models in this
catchment.

Stage 2 Implementation of the


snowmelt module
The abcd model has been applied again together with the
snowmelt module, and the calibration process carried out
again with the same structure as in the previous stage.
However, by including the snowmelt module in this Stage 2,
the results in NS values and simulated stream ows are identical to those in Stage 1, as illustrated in Table 4.
Therefore, the effect of the snow is negligible in this area
when the analysis is carried out on a monthly water balance,
so the abcd model without the snowmelt module is selected
for the analysis of the transferability of the parameters.

Stage 3 Transferability of model parameters


In the transferability analysis, the stream ow of each catchment area is simulated by the abcd model together with the
parameters calibrated in all other catchments. Next, these
series are compared to the entire naturalized stream ows
(Table 2). In this case, the transferability of the parameters is
analysed as a whole, instead of analysing the possibility of
transferring a single parameter (Heuvelmans et al. 2004).
Thus, the possibility of directly using a set of parameters in
neighbouring catchments is studied.
Each pair of series is compared by using FGLS to obtain the
coefcient of each simple linear regression. Next, the coefcient 1 as well as the R2 of these estimations are examined,
where a perfect transferability means an adjustment in which
1 = 1 and R2 = 1. In addition, to contrast the coefcients
obtained, the same series were compared by applying the
OLS method. Thus, it is possible to measure the overestimation related to R2 that is provided by this method.
Table 5 shows the results for the 25 adjustments of each
method. The nomenclature QM(X)-QY points out that the
results belong to the contrast between the naturalized
stream ows gauged in basin Y and the stream ows simulated for the same basin by using the parameters adjusted in
basin X. The elements in the diagonal represented in bold
47

Transferability of the parameters of the lumped models

F. Pellicer-Martinez and J. M. Martnez-Paz

Table 3 Efciency criterion NS calculated


T- model

GR2M model

abcd model

WASMOD

Code

Calibration

Validation

Calibration

Validation

Calibration

Validation

Calibration

Validation

A
B
C
D
E

0.76
0.46
0.44
0.52
0.38

0.65
0.17
0.02

0.16

0.85
0.67
0.58
0.62
0.25

0.79
0.50
0.87

0.04

0.88
0.69
0.63
0.52
0.36

0.84
0.38
0.18

0.20

0.82
0.68
0.35
0.43
0.42

0.73
0.54
0.26

0.32

Fig. 4. Naturalized streamow (Q) and simulated streamow in Catchment A.

type are the comparison between the naturalized stream


ows gauged in the basin and those simulated from a calibrated model of the same basin. The last row, whose values
are represented in italics, contains the comparison between
48

the naturalized stream ows, and the stream ows simulated


by the SIMPA model.
These results show that the majority of the values of 1
and R2 are very different to 1, so a systematic transferability in

Water and Environment Journal 29 (2015) 4350 2014 CIWEM.

F. Pellicer-Martinez and J. M. Martnez-Paz

Transferability of the parameters of the lumped models

Table 4 Efciency criterion NS calculated applying abcd model with the


snowmelt module
abcd model with snowmelt module
Catchment

NS

0.88
0.84

0.69
0.38

0.63
0.18

0.52

0.36
0.20

Validation
Calibration

the area is not possible, conrming the heterogeneous performance that the abcd model has provided. In addition, the
values of R2 obtained by using OLS are greater than those
using FLGS due to the fact that the serial autocorrelation is
meaningful. Consequently, this analysis conrms that using
OLS to contrast the results overestimates the value of R2 (31%
on average and a maximum of 109% as can be deduced from
the results in Table 5) and provides inefcient estimators of
the parameters.
Finally, the results obtained also reveal the superiority of
the abcd model over the SIMPA model, since the R2 coefcients of the abcd model using its own parameters are
greater than those of the SIMPA model. Besides, this model
tends to underestimate the stream ows in this area since all
the estimated values of 1 are less than 1, whereas the abcd
model provides values of 1 closer to 1.

Conclusions
(1) The NS values obtained by the application of these four
lumped water balance models are similar to those obtained
by other authors (Mouelhi et al. 2006; Karpouzos et al. 2011;
Wang et al. 2011). Furthermore, the abcd model provides the
highest values of NS, reaching values close to 0.90 so its use
is recommended in catchments with similar characteristics to
those in this study.
(2) This is the rst time that a monthly snowmelt module in
this area has been applied, and it does not improve the
results of the abcd model. Thus, these hydrological pro-

cesses are not relevant to the ow forecast in the Segura


River Basin.
(3) The inclusion of serial autocorrelation of the monthly
stream ow series in the estimation demonstrates that the
usual OLS method overestimates the R2 coefcient and the
goodness of t of the transferability of the parameters. In
addition, this study reveals that the transferability of the
parameters of the abcd model is not possible in the headwater of the Segura River Basin, thus reecting empirical evidence of the complex hydrology present in the upper Segura
River Basin (Lpez-Bermdez 1973).
(4) The lumped abcd model provides better stream ow forecast results than the SIMPA distributed model. However, the
lumped model requires less baseline information and the calibration is easier than in the distributed model.
To submit a comment on this article please go to
http://mc.manuscriptcentral.com/wej. For further information please
see the Author Guidelines at wileyonlinelibrary.com

References
Alley, W.M. (1984) On the Treatment of Evapotranspiration, SoilMoisture Accounting, and Aquifer Recharge in Monthly WaterBalance Models. Water Resour. Res., 20 (8), 11371149.
Dios, R. (1998) The Linear Model without a Constant Term and
the Coefcient of Determination: A Monte Carlo Study.
Questii: Quaderns dEstadstica, Sistemes, Informatica i
Investigaci Operativa, 22 (1), 334.
Estrela, T. and Quintas, L. (1996) A distributed hydrological model
for water resources assessment in large basins. Rivertech 96
-1st International Conference on New/Emerging Concepts for
Rivers, Proceedings, Vols 12: 861868.
Gleick, P.H. (1986) Methods for Evaluating the Regional Hydrologic Impacts of Global Climatic Changes. J. Hydrol., 88 (12),
97116.
Heuvelmans, G., Muys, B. and Feyen, J. (2004) Evaluation of
Hydrological Model Parameter Transferability for Simulating
the Impact of Land Use on Catchment Hydrology. Phys. Chem.
Earth, 29 (1112), 739747.

Table 5 Coefcients 1 and R2 of the simple linear regressions


1

R2

OLS/FGLS

QM(A)
QM(B)
QM(C)
QM(D)
QM(E)
QM(S)

OLS/FGLS

QA

QB

QC

QD

QE

QA

QB

QC

QD

QE

1.08/1.09
2.20/2.40
0.89/2.94
1.34/5.51
1.33/1.40
0.90/0.82

0.54/0.54
1.08/1.11
0.41/0.42
0.60/0.63
0.67/0.68
0.74/0.74

0.58/0.32
1.57/0.72
0.97/0.97
1.76/1.83
0.87/0.48
0.42/0.19

0.46/0.56
1.28/1.55
0.63/0.73
1.02/1.19
0.78/0.85
0.48/1.01

1.54/0.66
1.59/1.58
0.60/0.73
0.92/1.23
1.02/0.97
0.99/0.44

0.89/0.80
0.83/0.76
0.66/0.62
0.58/0.55
0.84/0.78
0.81/0.68

0.75/0.60
0.69/0.54
0.53/0.33
0.46/0.22
0.71/0.60
0.74/0.58

0.74/0.69
0.72/0.69
0.88/0.77
0.87/0.77
0.82/0.67
0.70/0.72

0.67/0.51
0.64/0.56
0.66/0.47
0.63/0.45
0.66/0.43
0.72/0.70

0.85/0.54
0.82/0.53
0.80/0.51
0.75/0.49
0.86/0.53
0.80/0.46

The elements in the diagonal represented in bold type are the comparison between the naturalized streamows gauged in the basin and those
simulated from a calibrated model of the same basin.
FGLS, feasible generalized least squares; OLS, ordinary least squares.

Water and Environment Journal 29 (2015) 4350 2014 CIWEM.

49

Transferability of the parameters of the lumped models

Hildreth, C. and Lu, J.Y. (1960) Demand Relations with


Autocorrelated Disturbances, Michigan State University Agricultural Experiment Station Technical Bulletin 276, East
Lansing, MI.
Hughes, D.A. and Metzler, W. (1998) Assessment of Three
Monthly Rainfall-Runoff Models for Estimating the Water
Resource Yield of Semiarid Catchments in Namibia. Hydrol.
Sci. J., 43 (2), 283297.
Karpouzos, D.K., Baltas, E.A., Kavalieratou, S. and
Babajimopoulos, C. (2011) A Hydrological Investigation Using a
Lumped Water Balance Model: The Aison River Basin Case
(Greece). Water Environ. J., 25 (3), 297307.
Klemes, V. (1986) Operational Testing of Hydrological SimulationModels. Hydrolog. Sci. J., 31 (1), 1324.
Kroll, C.N. and Stedinger, J.R. (1998) Regional Hydrologic Analysis: Ordinary and Generalized Least Squares Revisited. Water
Resour. Res., 34 (1), 121128.
Lasdon, L.S., Waren, A.D., Jain, A. and Ratner, M. (1978) Design
and Testing of a Generalized Reduced Gradient Code for
Nonlinear Programming. ACM Trans. Math. Softw., 4 (1),
3450.
van der Linden, S. and Woo, M.K. (2003) Transferability of
Hydrological Model Parameters between Basins in DataSparse Areas, Subarctic Canada. J. Hydrol., 270 (34),
182194.
Lpez-Bermdez, F. (1973) The upper area in Segura River basin.
Climate hydrology and geomorphology (in Spanish). PhD Dissertation, Department of Geography, University of Murcia, 288
pp.
Makhlouf, Z. and Michel, C. (1994) A 2-Parameter Monthly WaterBalance Model for French Watersheds. J. Hydrol., 162 (34),
299318.
Martnez-Paz, J., Pellicer-Martnez, F. and Colino, J. (2014) A
Probabilistic Approach for the Socioeconmic Assessment of

50

F. Pellicer-Martinez and J. M. Martnez-Paz

Urban River Rehabilitation Projects. Land Use Policy, 36, 468


477.
Mouelhi, S., Michel, C., Perrin, C. and Andreassian, V. (2006)
Stepwise Development of a Two-Parameter Monthly Water
Balance Model. J. Hydrol., 318 (14), 200214.
Nash, J.E. and Sutcliffe, J.V. (1970) River Flow Forecasting through
Conceptual Models Part I A Discussion of Principles. J.
Hydrol., 10 (3), 282290.
Perni, A., Martinez-Paz, J. and Martinez-Carrasco, F. (2012) Social
Preferences and Economic Valuation for Water Quality and
River Restoration: The Segura River, Spain. Water Environ. J.,
26, 274284.
Thomas, H.A. (1981) Improved Methods for National Water
Assessment. Report, contract WR 15249270. U.S. Water
Resources Council, Washington, DC.
Vandewiele, G.L. and Win, N.L. (1993) Monthly Water and Snow
Balance Models on Basin Scale. In Banasik, K. and Zbikowski,
A. (eds). Runoff and Sediment Yield Modelling, Warsaw, pp.
8388. Warsaw Agricultural University Press, Warsaw.
Wang, Q.J., Pagano, T.C., Zhou, S.L., Hapuarachchi, H.A.P., Zhang,
L. and Robertson, D.E. (2011) Monthly versus Daily Water
Balance Models in Simulating Monthly Runoff. J. Hydrol., 404
(34), 166175.
Wurbs, R.A. (2005) Modeling River/Reservoir System Management, Water Allocation, and Supply Reliability. J. Hydrol., 300
(14), 100113.
Xu, C.Y., Seibert, J. and Halldin, S. (1996) Regional Water Balance
Modelling in the NOPEX Area: Development and Application of
Monthly Water Balance Models. J. Hydrol., 180 (14), 211
236.
Zhang, L., Potter, N., Hickel, K., Zhang, Y. and Shao, Q. (2008)
Water Balance Modeling over Variable Time Scales Based on
the Budyko Framework Model Development and Testing. J.
Hydrol., 360 (14), 117131.

Water and Environment Journal 29 (2015) 4350 2014 CIWEM.

Potrebbero piacerti anche