You are on page 1of 13

WATER RESOURCES RESEARCH, VOL. 45, W07412, doi:10.

1029/2008WR007504, 2009

Relative merits of different methods for runoff predictions


in ungauged catchments
Yongqiang Zhang1 and Francis H. S. Chiew1
Received 5 October 2008; revised 13 February 2009; accepted 8 May 2009; published 14 July 2009.
[1] There have been numerous regionalization studies on runoff prediction in ungauged
catchments. This study evaluates the relative benefits of different methods using two
conceptual daily rainfall-runoff models, Xinanjiang and SIMHYD, on 210 relatively
unimpacted catchments in southeast Australia. The results show that runoff predictions in
ungauged catchments can benefit from a smart selection of donor catchments whose
optimized parameter values are used to model runoff in the target ungauged catchment,
output averaging of results from multiple-donor catchments and incorporating leaf
area index data into the rainfall-runoff models. The biggest benefit comes from an
educated selection of donor catchments (compared to a random selection of donor
catchments) and output averaging of results from multiple-donor catchments. The
difference between the three commonly used approaches for selecting donor catchments is
relatively small. The spatial proximity approach (where the geographically closest
catchment is used as the donor catchment) performs slightly better than the physical
similarity approach (where the catchment with the most similar attributes is used as the
donor catchment), and the integrated similarity approach, which combines the spatial
proximity and physical similarity approaches, performs only very marginally better than
the spatial proximity approach. The incorporation of leaf area index data into the
rainfall-runoff models shows marginal improvements to the modeling results, although a
more appropriate integration of vegetation and other remotely sensed data may further
improve the results.
Citation: Zhang, Y., and F. H. S. Chiew (2009), Relative merits of different methods for runoff predictions in ungauged catchments,
Water Resour. Res., 45, W07412, doi:10.1029/2008WR007504.

1. Introduction catchments and catchment descriptors or attributes (climatic


[2] Predictions in Ungauged Basins or Catchments (PUB) and physical), and then the parameter values for the unga-
are regarded as one of the most challenging tasks in surface uged catchments are estimated from its attributes and the
hydrology. The International Association of Hydrological established relationship. The spatial proximity approach
Sciences (IAHS) launched an initiative, the IAHS Decade uses the parameter values from the geographically closest
on PUB (2003 –2012) [PUB_Initiative, available at http:// gauged catchment hypothesizing that neighboring catch-
www.cig.ensmp.fr/~iahs/], focusing on ‘‘formulating and ments should behave similarly owing to similar physical
implementing appropriate science programmes to engage and climatic characteristics. The physical similarity approach
and energize the scientific community, in a coordinated transfers the entire set of parameter values from a physically
manner, towards achieving major advances in the capacity similar catchment whose attributes (climatic and physical)
to make reliable predictions in ungauged basins’’ [Sivapalan are similar to those of the target ungauged one.
et al., 2003]. [4] The regression approach is popular in regionalization
[3] Regionalization is typically used for water quantity studies [Young, 2006] but it has been strongly criticized
studies in PUB, which is referred as the process of trans- [Bardossy, 2007; McIntyre et al., 2005; Oudin et al., 2008b;
ferring parameter values from a gauged catchment to the Parajka et al., 2007]. This is because the cross-correlation
target ungauged catchment [Bloschl and Sivapalan, 1995]. between parameters are seldom taken into account and
Three regionalization approaches have been widely used to because model calibrations can produce vastly different sets
choose the donor gauged catchment whose optimized of parameter values that give similar model performance
parameter values are used to model runoff for the target (equifinality problem [Beven and Freer, 2001]). The spatial
ungauged catchment: regression; spatial proximity; and proximity and physical similarity approaches are very
physical similarity. The regression approach establishes a common in many recent regionalization studies [Bardossy,
relationship between parameter values calibrated on gauged 2007; McIntyre et al., 2005; Merz and Bloschl, 2004; Oudin
et al., 2008b; Parajka et al., 2005]. Merz and Bloschl
1 [2004] and Parajka et al. [2005] compare the three region-
CSIRO Water for a Healthy Country National Research Flagship,
CSIRO Land and Water, Canberra, ACT, Australia. alization approaches in over 300 Austrian catchments using
an 11-parameter HBV model and show that the spatial
Copyright 2009 by the American Geophysical Union. proximity approach performs best followed by the physical
0043-1397/09/2008WR007504

W07412 1 of 13
W07412 ZHANG AND CHIEW: RUNOFF PREDICTIONS IN UNGAUGED CATCHMENT W07412

the polar-orbiting Terra satellite, Leaf Area Index) data to


estimate surface conductance into a lumped rainfall-runoff
model and show that the revised model structure signifi-
cantly improved monthly and daily runoff estimates in
ungauged catchments in southeastern Australia. The use
of vegetation data in rainfall-runoff modeling will also be
explored in this study.
[7] The main focus of this paper is to quantify the relative
merits of output averaging, different regionalization
approaches and the use of MODIS-LAI data to predict daily
runoff in ungauged catchments. The specific objectives
include: (1) to evaluate the contribution of output averaging
from multiple-donor catchments, (2) to investigate four
regionalization approaches for selecting the donor catch-
ments: random selection and three educated approaches
incorporating spatial proximity, physical similarity and
integrated similarity (combination of spatial proximity and
physical similarity) and (3) to examine the benefits of using
MODIS-LAI time series data in rainfall-runoff modeling.
Two lumped conceptual daily rainfall-runoff models,
Xinanjiang and SIMHYD, and their revised versions to
incorporate the use of MODIS-LAI, are used in this study.
The models are applied to 210 relatively unregulated catch-
ments in southeast Australia using data from 1994 to 2006
(Figure 1).
[8] The large-scale regionalization study here with
Australian data is also unique as most regionalization
studies reported in the literature are for European countries
[Bardossy, 2007; Goswami et al., 2007; McIntyre et al.,
2005; Merz and Bloschl, 2004; Oudin et al., 2008b; Parajka
et al., 2005, 2007; Young, 2006]. A regionalization study in
Australia is likely to give poorer results compared to the
Figure 1. Locations of 210 catchments used in this study. European studies because Australia covers a much larger
region than most of the European studies resulting in a
larger variation of climatic and catchment physical condi-
similarity approach with the regression approach performing tions, and compared to similar regions in Europe, rainfall
worst. Oudin et al. [2008b] draw the same conclusion using and streamflow in Australia are more variable due to the
two rainfall-runoff models, GR4J and TOPMO, in 913 function of El Niño-Southern Oscillation and Inter-decade
French catchments. In their study, they argue that as the Pacific Oscillation [Chiew et al., 1998; Chiew and
spatial proximity approach does not systematically outper- McMahon, 2002; Piechota et al., 1998; Post et al., 1998;
form the physical similarity approach, combining the two to Power et al., 1999; Verdon et al., 2004], and the density of
select a donor catchment may improve the modeling results, the stream gauging network is also much lower in Australia
and this is one of the research questions that will be than in Europe.
explored in this paper.
[5] Various studies, particularly those by McIntyre et al.
[2005] and Oudin et al. [2008b] have shown that output
2. Models and Data
averaging can reduce the uncertainty in runoff predictions in
ungauged catchments. In output averaging, the target catch- 2.1. Rainfall-Runoff Models
ment is modeled using parameter values from many donor 2.1.1. Xinanjiang and SIMHYD Models
catchments, rather than one donor catchment, and results [9] Xinanjiang and SIMHYD are two lumped conceptual
from the modeling using the different sets of parameter daily rainfall-runoff models. The inputs into the models are
values from the different donor catchments, are averaged to daily rainfall and daily potential ET (ETp), and the models
provide the runoff estimate for the target catchment. The use estimate daily runoff.
of output averaging will also be explored in this paper. [10] The 14-parameter Xinanjiang model has been widely
[6] The use of more information, such as remotely sensed used, particularly in humid and semi-humid regions in
(RS) vegetation data, in rainfall-runoff modeling can improve China [Cheng et al., 2002; Gan et al., 1997; Jayawardena
runoff estimates in ungauged catchments (Y. Q. Zhang and and Zhou, 2000; Zhao, 1992; R. J. Zhao et al., The
F. H. S. Chiew, Can remote sensing data improve short-term Xinanjiang model, paper presented at Hydrological Fore-
rainfall-runoff simulation?, paper presented at Down Under casting Proceedings Oxford Symposium, IAHS, July,
2008, Engineers, Adelaide, Australia, April, 2008). Zhang 1980]. The model structure of the Xinanjiang model and
et al. [2008] integrate the Penman-Monteith evapotranspi- the model parameters are shown in Figure 2. The Xinan-
ration (PM-ET) equation which uses MODIS-LAI (the jiang model has three submodels, ET submodel (3-par: Um,
MODerate resolution Imaging Spectrometer mounted on Lm and C), runoff generating submodel (3-par: Dm, B and
2 of 13
W07412 ZHANG AND CHIEW: RUNOFF PREDICTIONS IN UNGAUGED CATCHMENT W07412

Figure 2. Model structures of Xinanjiang and SIMHYD and their revised versions (the dash-dots show
the evapotranspiration submodel, which is modified).

Im) and routing submodel (8-par: Sm, Ex, Kg, Ki, Cg, Ci, Cs, (assumed to be zero here), Ga is the aerodynamic
L). conductance and Gs is the surface conductance.
[11] The version of the SIMHYD model used here has [14] The terms Ae, D, g, ra and D in equation (1) can be
9 parameters. The structure of the SIMHYD model and the calculated from the basic daily meteorological time series
model parameters and algorithms that describe water move- and the term Ga in equation (1) can be calculated from land
ment into and out of the storages are shown in Figure 2. cover data [Zhang et al., 2008].
SIMHYD has been extensively used for various applica- [15] The surface conductance, Gs, is the only physiolog-
tions across Australia [Chiew et al., 2002; Siriwardena et ical variable in the PM equation. It is calculated using the
al., 2006; Zhang et al., 2008; N. Viney et al., Region- algebraic, biophysical two-parameter surface conductance
alisation of runoff generation across the Murray-Darling model [Leuning et al., 2008; Zhang et al., 2008]. The daily
Basin using an ensemble of two rainfall-runoff models, input data required for the model are LAI and basic
paper presented at Water Down Under 2008, Engineers, meteorological variables. The model has two parameters,
Adelaide, Australia, 2008]. the maximum stomatal conductance gsx and the fraction of
equilibrium evaporation at the soil surface f. The soil
2.1.2. Revised Versions of the Two Models evaporation factor f is directly dependent on moisture status,
[12] To use RS-LAI data in the two models, the ET sub- and the soil wetness modeled by the Xinanjiang and
models are replaced with the Penman-Monteith (PM-ET) SIMHYD models are used as the estimate for f. The gsx
equation to calculate actual ET directly (see Figure 2). term is considered as a parameter that is optimized together
[13] The PM-ET equation can be written as: with the other Xinanjiang and SIMHYD model parameters.
[16] The revised Xinanjiang model has 12 parameters.
1 DAe þ ra Cp DGa The three-layer ET submodel is replaced with a one-layer
ET ¼ ð1Þ
l D þ g ð1 þ Ga =Gs Þ ET submodel, removing the three parameters, Um, Lm and C
(see Figure 2). The parameter, f, in the PM model is
where l is the latent heat of vaporization, D = de*/dTa is the estimated as the soil wetness W/WM. The parameter gsx is
slope of the curve relating saturation water vapor pressure to treated as an additional model parameter and optimized
temperature, D = e*(Ta)  ea is the vapor pressure deficit together with the remaining eleven parameters.
of the air, e*(Ta) is the saturation vapor pressure at air [17] The revised SIMHYD model has 10 parameters (one
temperature, ea is the actual vapor pressure, g is the additional parameter, gsx). The ET in the evapotranspiration
psychrometric constant, ra is the air density, Cp is the submodel is calculated using the PM model. The parameter,
specific heat capacity of air, Ae is the available energy, f, in the PM equation is calculated as the soil wetness,
the difference of the net radiation to the soil heat flux SMS/SMSC.
3 of 13
W07412 ZHANG AND CHIEW: RUNOFF PREDICTIONS IN UNGAUGED CATCHMENT W07412

[18] To distinguish between the revised and original models, averaged to obtain aggregate daily data series for model
the revised models will be referred to as ‘‘Xinanjiang-ET’’ inputs.
and ‘‘SIMHYD-ET’’.
2.3. Model Calibration and Verification
2.2. Data [25] The Particle Swarm Optimization (PSO) toolbox in
MATLAB is used to optimize the parameters of the two
2.2.1. Runoff Timeseries Data
rainfall-runoff models. The PSO method is firstly presented
[19] Daily runoff data from 210 relatively unimpacted
by Eberhart and Kennedy, inspired from the behavior of
catchments (50 to 2000 km2) in southeastern Australia
schools of fish or flocks of birds [Eberhart and Kennedy,
(Figure 1) are used. The region includes the most populated
1995]. It originates from the swarm paradigm, called
and important agricultural areas of Australia. Data from
particle swarm, and is expected to provide the so-called
1994 to 2006 are used in this study, with the 2000 –2006
global or near-global optimum. The PSO method has been
data used for model calibration and regionalization assess-
successfully used in several rainfall-runoff model parameter
ment of all the four models and the 1994 – 2000 data used for
optimizations [Chau, 2006; Gill et al., 2006].
model verification and regionalization assessment for the
[26] The rainfall-runoff models are calibrated to maximize
original Xinanjiang and SIMHYD models. The Xinanjiang-
the Nash-Sutcliffe Efficiency (NSE) [Nash and Sutcliffe,
ET and SIMHYD-ET models are not assessed for 1994 –
1970] of daily runoff which is defined as:
2000 because the MODIS-LAI data required to drive these
revised model is only available from 2000.
X
N  2
2.2.2. Meteorological Timeseries Data Qobs;i  Qsim;i
[20] Daily time series of maximum temperature, minimum i¼1
NSE ¼ 1  ð2Þ
temperature, incoming solar radiation, actual vapor pressure XN  2
and precipitation from 1994 to 2006 at 0.05°  0.05° Qobs;i  Qobs
i¼1
(5 km  5 km) grid cells from the SILO Data Drill of the
Queensland Department of Natural Resources and Water
(www.nrw.gov.au/silo) [Jeffrey et al., 2001] are used. The where Qsim and Qobs are the simulated daily runoff and
SILO Data Drill provides surfaces of daily rainfall and other observed daily runoff, respectively, Qobs is the arithmetic
climate data interpolated from point measurements made by mean of the observed runoff, i is the ith day and N is the
the Australian Bureau of Meteorology. The rainfall data is total days sampled.
required as input to the rainfall-runoff models. The other [27] All four models are calibrated against runoff data
meteorological data are used to calculate potential ET (ETp) from 2001 to 2006, with the 2000 data used for model warm
using the Priestley-Taylor model [Priestley and Taylor, up.
1972] and to calculate PM-ET for the revised rainfall-runoff [28] Two criteria are used for the model assessment: NSE
models. of daily runoff described by equation (2) and absolute Water
2.2.3. Remote Sensing Data Balance Error percentage (WBE) which is defined as:
[21] The RS-LAI data required to calculate PM-ET in the  N 
X XN 
revised rainfall-runoff models come from the 8-day com-  Qsim;i  Qobs;i 

posite 1-km resolution MODIS-LAI products (MOD15A2,  
collection 4) which are obtained from the Land Processes WBE ¼ 100 i¼1 i¼1 
 ð3Þ
 XN

Distributed Active Archive Centre (LPDAAC) (http://lpdaac.  Qobs;i 
 
usgs.gov/dataproducts.asp) for the period 2000 – 2006. i¼1
Before the application, the MODIS-LAI data are quality
controlled and interpolated to a daily time step [Zhang and [29] The NSE and WBE results are presented for all four
Wegehenkel, 2006]. However, there still exist the impacts of models for the calibration and regionalization assessment
cloud contamination and atmospheric variability on daily against 2001– 2006 runoff data and for the original two
LAI data. Thus the interpolated daily LAI data are then models for the verification and regionalization assessment
smoothed by the Savitzky-Golay filtering method, a widely against 1995– 2000 runoff data (1994 data are used for
used method for obtaining high-quality vegetation index model warm up).
time series [Chen et al., 2004; Fang et al., 2008; Ruffin et [30] The model calibration and verification results are
al., 2008]. summarized in Figure 3 and Tables 2 and 3. The model
[22] Land cover data required to estimate Ga in equation calibration results for all four models are very similar. The
(1) are obtained from the MODIS land cover product, the calibrations are generally satisfactory with NSE values
yearly Land Cover classification product (MOD12Q1) greater than 0.6 in 80 percent of the 210 catchments
(http://edcdaac.usgs.gov/modis/mod12q1v4.asp). The data (median value of about 0.8) and WBE of less than 30 percent
set has 17 vegetation classes defined according to the in 80 percent of the catchments. The calibration results
International Geosphere-Biosphere Programme. are similar to most rainfall-runoff modeling studies in
[23] The albedo data required to calculate Ae in equation Australian catchments (Viney et al., presented paper, 2008;
(1) are obtained from an annual average albedo product at a W. Boughton and F. Chiew, Estimating runoff in ungauged
5-km resolution for Australia [Dilley et al., 2000; Schaaf et catchments from rainfall, PET and the AWBM model, paper
al., 2002]. presented at International Symposium on Environment
[24] All the remote sensing and meteorological data are Software System, Harrisonburg, Virginia, 18-21 May,
reprojected and re-sampled to obtain 1-km gridded data. 2004). The verification NSE results for the original models
The gridded data in each catchment are then cut out and are slightly poorer than the calibration results with the
4 of 13
W07412 ZHANG AND CHIEW: RUNOFF PREDICTIONS IN UNGAUGED CATCHMENT W07412

Figure 3. Summary of calibrated and verified Nash-Sutcliffe Efficiency (NSE) and absolute Water
Balance Error percentage (WBE) values for Xinanjiang and SIMHYD models of the 210 catchments. A
good model performance is indicated by large NSE values and small WBE values.

averaged NSE values in the model verification generally series for the ungauged catchment, which is called as output
being 0.1 to 0.2 lower than those in the model calibration. averaging method as described by McIntyre et al. [2005]
These results will be further discussed in section 5.1. and Oudin et al. [2008b].
3.2. Selection of Donor Catchment
3. Modeling Runoff in Ungauged Catchments [34] To model runoff in an ungauged catchment, param-
[31] To assess the model predictions of daily runoff in eter values from a donor catchment are used. Four
‘‘ungauged’’ catchments, each of the 210 catchments is left approaches for choosing the donor catchment are investi-
out in turn and considered as an ‘‘ungauged’’ catchment, gated here: random selection; spatial proximity; physical
and the entire set(s) of parameter values from the donor similarity; and integrated similarity.
catchment(s) are used to model runoff in the ‘‘ungauged’’ 3.2.1. Random Selection Approach
catchments. [35] In the random selection approach, a donor catchment
is chosen randomly, using a random integer generator. This
3.1. Single-Donor Catchment Versus Multiple-Donor approach is taken as a benchmark to evaluate the following
Catchments three educated or informed approaches.
[32] In the single donor catchment approach, parameter 3.2.2. Spatial Proximity Approach
values from a single donor catchment are used to model [36] In the spatial proximity approach, the geographically
runoff in the ungauged catchment. closest gauged catchment is chosen as the donor catchment.
[33] In the multiple-donor catchments approach, multiple- The distance between catchment centroids, D, is used as the
donor catchments are used. Parameter values from each distance measure.
donor catchment are used to independently model runoff in 3.2.3. Physical Similarity Approach
the ungauged catchment. Each of the daily runoff time [37] In the physical similarity approach, the gauged
series modeled using parameter sets from each of the donor catchment with the closest physical similarity to the target
catchments are then averaged to obtain the daily runoff time catchment is chosen as the donor catchment.
5 of 13
W07412 ZHANG AND CHIEW: RUNOFF PREDICTIONS IN UNGAUGED CATCHMENT W07412

Table 1. Summary of Catchment Attributes in the 210 Catchments


Catchment Characteristics Notation Min 25% Median 75% Max
2
Area A (km ) 51 160 333 633 2000
Aridity index (ETp/P) AI(-) 0.76 1.22 1.55 1.89 2.98
Mean elevation E (m) 57 307 519 814 1445
Mean slope in degree S (°) 0.42 2.75 4.55 7.78 13.85
Stream length SL (km) 27 121 246 475 1753
Mean Solum thick ST (mm) 0.44 0.86 0.96 1.20 2.00
Plant available water holding capacity PAWC (mm) 50.0 82.3 110.8 158.3 265.8
Mean woody vegetation fraction WF (-) 0.00 0.23 0.52 0.83 1.00

[38] Eight catchment attributes (see Table 1) are consid- therefore there is little meaning in comparing relative nega-
ered in this study: (1) catchment Area (A) in km2, (2) Aridity tive NSE values), negative NSE values are considered as zero
Index (AI): calculated as the ratio of mean annual ETp to in calculating the average), and the median WBE values.
mean annual P, (3) mean catchment Elevation (E) in m: Table 3 summarizes the regionalization results for the two
derived from the 9 second DEM for Australia (custodian: original models in the 1995 – 2000 verification period.
Geoscience Australia), (4) catchment Slope (S): derived Table 4 compares the NSE values between the three
from the 9 second DEM for Australia (custodian: Geosci- educated regionalization approaches, between multiple
ence Australia), (5) Stream Length (SL): derived from donors and single donor, and between the integrated simi-
watercourse lines in the Topo250k Series 3 Hydrography larity approach and random selection for the four models in
data set (custodian: Geoscience Australia), (6) median Solum the calibration period and verification period. Figures 4 to 7
Thickness in m(ST): derived from the Atlas of Australian present the distributions of NSE and WBE values respec-
Soils, (7) Plant Available Water holding Capacity in solum in tively obtained for the 210 catchments as whisker plots.
mm (PAWC): derived from the Atlas of Australian Soils [42] The use of one to one hundred donor catchments is
[McKenzie et al., 2000], and (8) mean Woody vegetation explored in this study, and the results show that the
Fraction (WF): derived from the National Carbon Account- optimum number of donor catchments that generally gives
ing System (NCAS) 2005 Forest Extent data set (custodian: the highest NSE and lowest WBE values for the different
Australian Greenhouse Office). modeling experiments is about eight to ten (Figure 8). For
[39] The rank-accumulated similarity is used to select the this reason, the results presented throughout this paper are
donor catchment [Oudin et al., 2008b]. For each attribute, based on the output averaging of results from eight donor
the catchment with the most similar attribute to the target catchments.
catchment is considered rank one, the catchment with the
second most similar attribute is considered rank 2, and so
on. Where several attributes are used for regionalization, the
rank numbers for each of the attributes are added. The
catchment with the smallest total rank is chosen as the target
catchment. Each attribute used for regionalization is given
equal weight in the ranking system [Oudin et al., 2008b].
3.2.4. Integrated Similarity Approach
[40] In the integrated similarity approach, the spatial
proximity and physical similarity measures are considered
together. The geographic distance, D, is taken as an attribute
together with the other catchment attributes. Like the
physical similarity approach, this approach also uses the
rank-accumulated similarity to select the donor catchment.

4. Regionalization Results
[41] The model performances, applied to the ungauged
catchments as described in section 3, are summarized as the
distributions of NSE and WBE values obtained for the 210
catchments. Table 2 summarizes for the four models (two Figure 4. Summary of Nash-Sutcliffe Efficiency (NSE)
original rainfall-runoff models and two revised models to values for the four regionalization approaches in the
integrate MODIS-LAI) for the various methods used to calibration period. All the NSE values are from 8-donor
select the donor catchments in the 2001 – 2006 calibration catchments except IS1, which is for integrated similarity for
period, the 25th percentile, median and 75th percentile NSE one-donor catchment. Whisker plots show 10th, 25th, 50th
values, the NSE values averaged over the 210 catchments (median), 75th and 90th results for the 210 catchment. PS,
(to avoid the influence of very high negative NSE values physical similarity; SP, spatial proximity, IS, integrated
(a negative NSE value indicates that the model performs similarity. The number below each plot means the number
poorer than a mean estimate for every single day and of catchments with negative NSE values.

6 of 13
W07412 ZHANG AND CHIEW: RUNOFF PREDICTIONS IN UNGAUGED CATCHMENT W07412

Figure 7. Summary of absolute Water Balance Error


percentage (WBE) values for the four regionalization
approaches in the verification period (see details in Figure 4).
Figure 5. Summary of absolute Water Balance Error
percentage (WBE) values for the four regionalization
approaches in the calibration period (see explanation in
Figure 4). 0.65 to 0.85 in the 2001 – 2006 calibration; Tables 2 and 3)
and about 0.05 to 0.10 higher than the NSE values in the
2001 – 2006 regionalization.
4.2. Single-Donor Catchment Versus Multiple-Donor
4.1. Regionalization Results Versus Calibration Catchments
and Verification Results [44] The output averaging of results from 8-donor catch-
[43] The model regionalization (prediction in ungauged ments are considerably better than the results from a single
catchments) results are poorer than the model calibration donor catchment (Tables 2– 4 and Figures 4 – 7). For the
results, with the median NSE values from the best region- integrated similarity approach, which is the best regional-
alization results (integrated similarity) being about 0.25 to ization approach, the 25th percentile, median and average
0.30 lower than the calibration results (Table 2). However, of the NSE values from the 210 catchments from the use of
the model regionalization results are similar to the model 8-donor catchments compared to the use of a single donor
verification results in the 1995 –2000 verification period, catchment are generally higher by about 0.10. The 75th
where the model regionalization and model verification percentile NSE values (lower quartile NSE value showing
NSE values are similar for the SIMHYD model, and the results from the poorer modeled catchments) for 8-donor
model regionalization NSE values only about 0.05 lower catchments are 0.15 and 0.23 higher than those for a single
than the model verification NSE values for the Xinanjiang donor catchment for the Xinanjiang model and the SIMHYD
model (Table 3). The NSE values for the 1995 – 2000 model, respectively in the calibration period (Table 2 and
regionalization are about 0.15 to 0.20 lower than the NSE Figure 4), and 0.14 and 0.15 higher in the verification
values in the 2001– 2006 calibration (25th percentile to 75th period (Table 3 and Figure 6). The median WBE values
percentile range of 0.45 to 0.70 in 1995– 2000 compared to from the 210 catchments from the use of 8-donor catch-
ments are almost same as those from use of a single donor
catchment in the calibration period (Figure 5), but are
better than those obtained from a single donor catchment
in the verification period (Figure 7). The 8-donor catch-
ment NSE values are more than 0.02 higher than the
single donor catchment NSE values in 120– 140 catchments,
and the single donor catchment NSE values are more than
0.02 higher than the 8-donor catchments NSE values in
30 –50 catchments (IS8-IS1 in Table 4). The comparisons
between single donor catchment versus multiple-donor
catchments for the random selection, physical similarity
and spatial proximity approaches also showed similar
results and are not shown here.
4.3. Approaches for Selecting Donor Catchments
[45] Tables 2 – 4 and Figures 4 – 7 compare the random
selection approach and the three educated approaches used
Figure 6. Summary of Nash-Sutcliffe Efficiency (NSE) to select donor catchments. As expected, the educated or
values for the four regionalization approaches in the informed selection of donor catchments based on spatial
verification period (see details in Figure 4). proximity, physical similarity and integrated similarity give
7 of 13
W07412 ZHANG AND CHIEW: RUNOFF PREDICTIONS IN UNGAUGED CATCHMENT W07412

Table 2. Calibration and Regionalization Results for the Four Regionalization Approaches and for the Four Rainfall-Runoff Models in
the Calibration Period (2001 – 2006)
Model Indicator Calibration Random8 PS8 SP8 IS8 IS1 Model Indicator Calibration Random8 PS8 SP8 IS8 IS1

Xinanjiang 25th NSE 0.86 0.59 0.65 0.71 0.71 0.64 SIMHYD 25th NSE 0.85 0.63 0.64 0.66 0.68 0.57
Median NSE 0.78 0.38 0.50 0.51 0.51 0.43 Median NSE 0.79 0.33 0.47 0.48 0.51 0.36
75th NSE 0.65 0.16 0.17 0.25 0.27 0.12 75th NSE 0.67 -0.19 0.16 0.19 0.24 0.01
Averaged NSEa 0.72 0.34 0.42 0.45 0.47 0.40 Averaged NSEa 0.73 0.34 0.42 0.43 0.46 0.35
Median WBE 14 47 34 32 30 34 Median WBE 15 54 32 37 33 34
Xinanjiang-ET 25th NSE 0.86 0.6 0.68 0.71 0.71 0.63 SIMHYD-ET 25th NSE 0.86 0.58 0.65 0.67 0.68 0.58
Median NSE 0.79 0.37 0.52 0.52 0.52 0.43 Median NSE 0.78 0.32 0.46 0.50 0.49 0.34
75th NSE 0.68 0.49 0.22 0.27 0.33 0.10 75th NSE 0.64 -0.45 0.08 0.22 0.28 0.02
Averaged NSEa 0.74 0.33 0.45 0.47 0.49 0.39 Averaged NSEa 0.73 0.32 0.40 0.44 0.46 0.34
Median WBE 15 53 42 42 38 36 Median WBE 12 48 37 34 33 32
a
PS, physical similarity; SP, spatial proximity; IS, integrated similarity. All the regionalization results are obtained from eight-donor output averaging
except IS1, which is integrated similarity for one-donor catchment.

results that are better than donor catchments selected obtained using various combinations of the catchment
randomly. The median and average of the NSE values from attributes.
the 210 catchments for the spatial proximity, physical [48] The spatial proximity approach generally performs
similarity and integrated similarity approaches are generally better than the use of a single physically similar catchment
more than 0.1 higher than the NSE values from the random attribute, but only marginally better when all eight catch-
selection approach (Tables 2 and 3). The median WBE ment attributes are used. The use of an integrated similarity
values from the 210 catchments for the spatial proximity, approach that combines both the spatial proximity and
physical similarity and integrated similarity approaches are catchment attributes may lead to a better choice of donor
generally more than 10 percent lower than those from the catchments. Because spatial proximity is generally the best
random selection approach in the calibration period single factor for selecting the donor catchments, it is used
(Table 2), but are similar to those from the random selection together with several other catchment attributes in the
approach in the verification period (Table 3). The integrated integrated similarity approach to select the donor catch-
similarity approach gives better results in 120– 140 of the ments. In general, the use of spatial proximity or geographic
210 catchments but the random selection approach gives distance (D) together with the aridity index (AI) and the
better results in 30– 50 catchments (IS8-random8 in Table 4). mean woody vegetation fraction (WF) gives the best results
[46] Although expected, the results here are presented to compared to other combinations, and the integrated simi-
quantify the benefit of an educated selection of donor larity results presented here are based D, AI and WF.
catchment, with the results indicating that the relative [49] On average, the integrated similarity approach per-
improvement in the modeling results from an educated forms marginally better than the spatial proximity approach,
selection of donor catchment versus a random selection is which in turn performs better than the physical similarity
similar to the relative improvement of output averaging of approach (Tables 2 – 4 and Figures 4 –7). The averages of
results from multiple-donor catchments versus using a the NSE values from the 210 catchments are 0.47, 0.45 and
single donor catchment. The results for the random selection 0.42 respectively for the integrated similarity, spatial prox-
approach may be dependent on the catchment that happened imity and physical similarity approaches for the Xinanjiang
to be selected randomly. To overcome this, the analysis for model and 0.46, 0.43 and 0.42 respectively for the SIM-
the random selection approach is repeated for several times HYD model in the calibration period (Table 2), and 0.55,
to obtain alternative distributions of results presented in 0.54 and 0.53 for the three approaches respectively for the
Tables 2 and 3 and Figures 4 – 7. The repeated analysis Xinanjiang model and 0.53, 0.52 and 0.52 for the XIMHYD
showed very similar results to those presented here because model in the verification period (Table 4). This difference
of the large number of catchments used here. between the three approaches are relatively small compared
[47] Tables 2 – 4 and Figures 4 –7 also compare the results to the difference between the educated versus random
from the physical similarity, spatial proximity and integrated selections of donor catchments and between the use of
similarity approaches. For the physical similarity approach, multiple-donor catchments versus one donor catchment
the results shown are for the use of all eight catchment presented earlier.
attributes because the use of all eight catchment attributes [50] Table 4 shows that the relative improvement of the
generally gives the best or very similar to the best results spatial proximity approach over the physical similarity

Table 3. Verification and Regionalization Results for the Four Regionalization Approaches and the Two Original Rainfall-Runoff
Models in the Verification Period 1995 – 2000 (See the Meanings of Abbreviations in Table 2)
Model Indicator Verification Random8 PS8 SP8 IS8 IS1 Model Indicator Verification Random8 PS8 SP8 IS8 IS1

Xinanjiang 25th NSE 0.74 0.63 0.70 0.71 0.71 0.63 SIMHYD 25th NSE 0.70 0.63 0.71 0.70 0.70 0.62
Median NSE 0.63 0.48 0.58 0.56 0.57 0.49 Median NSE 0.56 0.49 0.58 0.56 0.58 0.45
75th NSE 0.47 0.35 0.43 0.41 0.45 0.31 75th NSE 0.40 0.28 0.41 0.41 0.42 0.27
Averaged NSE 0.59 0.46 0.54 0.53 0.55 0.45 Averaged NSE 0.52 0.44 0.52 0.52 0.53 0.42
Median WBE 22 29 30 25 26 31 Median WBE 17 27 23 25 23 29

8 of 13
W07412 ZHANG AND CHIEW: RUNOFF PREDICTIONS IN UNGAUGED CATCHMENT W07412

<=0.02
Table 4. Comparison Between the Three Educated Regionalization Approaches, Between Multiple Donors and One Donor and Between The Integrated Similarity Approach and Random
approach is more significant than the relative improvement

46
33
36
42
34

47
of the integrated similarity approach over the spatial prox-
imity approach in the calibration period. The spatial prox-

IS8-Random8
imity approach outperforms the physical similarity approach

<0.02 &
>0.02
in 90– 100 catchments compared to the physical similarity

36
33
37
49
34

43
approach outperforming the spatial proximity approach in
40 –50 catchments (SP8-PS8 in Table 4). The number of
catchments where the integrated similarity approach outper-

>=0.02

128
144
137

142

120
119
forms the spatial proximity approach is only slightly more
than the number of catchments where the spatial proximity
<=0.02 approach outperforms the integrated similarity approach
(IS8-SP8 in Table 4).

46
46
45
53
42

50
[51] The relative difference between the three regional-
ization approaches is more significant in the poorer modeled
catchments. The 75th percentile of the NSE values from the
IS8-IS1

<0.02 &
>0.02

37 210 catchments are 0.27, 0.25 and 0.17 for the Xinanjiang
39
46
29
41

25 model and 0.24, 0.19, 0.16 for the SIMHYD model (Table 2),
and the number of catchments with negative NSE values
>=0.02

are 27, 34 and 47 for the Xinanjiang model and 31, 36 and
127
125

128
127

135
119

42 for the SIMHYD model (Figure 4).


4.4. Revised Models With RS-LAI
<=0.02

[52] The revised models with new ET algorithms using


76
65
64
67
58

83

remotely sensed LAI time series data generally perform


better than the original rainfall-runoff models. However, the
SP8-PS8

improvements in the performance of the revised models are


<0.02 &
>0.02

marginal, where the average of the NSE values from the 210
47
55
56
51
48

55

catchments is less than 0.02 higher in the revised-ET models


compared to the original models (Table 2 and Figure 4). The
>=0.02

WBE values for the revised-ET and original models are also
104
87
90
90
92

72

very similar (Figure 6). However, the revised models give


significantly better results compared to the original models
in the poorer modeled catchments. The 75th percentile NSE
<=0.02

values for the integrated similarity, spatial proximity and


75
64
69
69
61

86

physical similarity approaches respectively are 0.33, 0.27 and


0.22 (and 22, 32 and 38 catchments with NSE less than
IS8-PS8

zero) for the Xinanjiang-ET model compared to 0.27, 0.25


<0.02 &
>0.02

43
44
45
49
42

40

and 0.17 (27, 34 and 47) for the original Xinanjiang model.
The 75th percentile NSE values for the integrated similarity,
spatial proximity and physical similarity approaches respec-
>=0.02

tively are 0.28, 0.22 and 0.08 (and 26, 32 and 48 catchments
102

107
92
96
92

84

with NSE less than zero) for the SIMHYD-ET model


compared to 0.24, 0.19 and 0.16 (31, 36 and 42) for the
original SIMHYD model.
<=0.02
Selection (See the Meanings of Abbreviations in Table 2)

70
76
80
75
73

78

5. Discussion
IS8-SP8

<0.02 &

5.1. Regionalization Results in the Calibration


>0.02

64
52
47
52
48

58

and Verification Periods


[53] As expected, the regionalization results are signifi-
cantly poorer than the calibration results over 2001 – 2006.
>=0.02
a

The relative difference between the model calibration and


76
83
83
89

74
82

model regionalization results is slightly greater than that


reported in similar studies [Merz and Bloschl, 2004; Oudin
Verification
Calibration

et al., 2008b]. First, the calibration period, 2001– 2006, is a


Period

very dry period. For the 210 catchments used here, runoff
Catchment number.

coefficient (Q/P) varies in 0.01 – 0.76 with a median only


0.10 and aridity index (ETp/P) varies in 0.76 – 2.92 with a
median about 1.55 in the calibration period. Compared to
Xinanjiang-ET

SIMHYD-ET

the 210 catchments, the 913 French catchments [Oudin et


Xinanjiang

Xinanjiang
Model

SIMHYD

SIMHYD

al., 2008b] are much wetter, showing that runoff coefficient


varies in 0.03– 4.24 with a median 0.34 and aridity index
a

varies in 0.23 – 1.20 with a median only 0.68. The UK and


9 of 13
W07412 ZHANG AND CHIEW: RUNOFF PREDICTIONS IN UNGAUGED CATCHMENT W07412

Figure 8. Averaged regionalization Nash-Sutcliffe Efficiency (NSE) and median absolute Water
Balance Error percentage (WBE) for the 210 catchments versus number of donor catchments used for
regionalization.

Austrian catchments are also wetter than the catchments in 2001 – 2006 calibration, while the 1995– 2000 regionaliza-
this study [Merz and Bloschl, 2009; Oudin et al., 2008a]. tion results are compared to the 1995 – 2000 verification
Second, the catchments here spread across a large area results where parameter values used to model runoff come
covering a large range of climates compared to the catch- from the model calibration against 2001 – 2006 data and not
ment used in European studies [Bardossy, 2007; Oudin et the 1995– 2000 data. The second reason is that the 1995 –
al., 2008b; Parajka et al., 2007; Young, 2006]. 2000 period is significantly wetter than the 2001 – 2006
[54] On the other hand, the 1995 – 2000 regionalization period (90 percent of the catchments are wetter in the 1995 –
results (model parameters calibrated using 2001 – 2006 data 2000 period, and averaged across the 210 catchments the
from donor catchments used to model 1995– 2000 data for 1995 – 2000 period is 30 percent wetter than the 2001 – 2006
the target catchment) are only slightly poorer than the period) and modeling results are generally better for wet
1995 – 2000 verification results (model parameters calibrated catchments compared to dry catchments (Figure 9).
using 2001– 2006 data used to model runoff over 1995 –
2000 for the same catchment). The first reason for this 5.2. Output Averaging of Results From Multiple-Donor
observation is that the 2001 – 2006 regionalization results Catchments
are compared directly to the best possible results from the [55] The output averaging of results on average gives
considerably better daily runoff simulations than the use of

Figure 9. Regionalization Nash-Sutcliffe Efficiency (NSE) against observed catchment mean annual
rainfall for the Xinanjiang model in the 2001 – 2006 calibration period.
10 of 13
W07412 ZHANG AND CHIEW: RUNOFF PREDICTIONS IN UNGAUGED CATCHMENT W07412

Figure 10. The Nash-Sutcliffe Efficiency (NSE) difference between spatial proximity and physical
similarity and between integrated similarity and spatial proximity against mean annual rainfall, the
Xinanjiang model in the 2001 – 2006 calibration period.

a single donor catchment because it will average out the closest catchments are chosen as the donor catchments,
effect of choosing a poor donor catchment. performs reasonably because neighboring catchments are
[56] The optimum number of donor catchments is likely more likely to have similar hydrological characteristics. The
to be different for different models and modeling physical similarity approach also attempts to select donor
approaches and considerations (Figure 8), with most studies catchments with similar climatic and physical character-
using five to ten donor catchments [McIntyre et al., 2005; istics, but it is difficult to define meaningful lumped
Oudin et al., 2008b; Viney et al., presented paper, 2008]. It catchment-average attributes and derive these attributes
is not the focus of this paper to consider the optimum accurately.
number of donor catchments, but Figure 8 suggests that [59] The integrated similarity approach, which combines
while eight donor catchments is likely to be close to the the spatial proximity and physical similarity approaches, only
optimum number of donor catchments for the spatial performs very marginally better than the spatial proximity
proximity and physical similarity approaches, the integrated approach. The relative difference between the three region-
similarity approach may show better modeling results if alization approaches is more significant in the poorer mod-
more donor catchments are used. eled catchments, where the integrated similarity approach
outperforms the spatial proximity approach which in turn
5.3. Regionalization Approaches outperforms the physical similarity approach. However, there
5.3.1. Educated Approaches Versus Random Selection is no clear indication of the integrated similarity approach
[57] The daily runoff modeling results from an educated consistently outperforming the spatial proximity approach or
selection of donor catchments (spatial proximity, physical the spatial proximity approach consistently outperforming
similarity and integrated similarity approaches) are consid- the physical similarity approach in drier/wetter catchments or
erably better than the use of randomly selected donor in specific spatial locations (Figures 10 and 11).
catchments. The improved modeling results from an edu-
cated selection of donor catchments are, on average, similar 5.4. Use of RS-LAI in Rainfall-Runoff Modeling
to the improved modeling from output averaging of results [60] The revised rainfall-runoff models with new evapo-
from multiple-donor catchments versus the use of a single transpiration algorithms using remotely sensed LAI time
donor catchment. This observation is important but seldom series data performs better than the original models. How-
reported in the literature, where many studies focus on ever, like the above differences between the different
improving and assessing regionalization results using a approaches for selecting donor catchments, this improvement
single donor catchment while the use of multiple-donor is marginal in the better modeled ungauged catchments, and
catchments versus a single donor catchment can improve more significant in the poorer modeled ungauged catch-
the modeling results as much as a smart selection of donor ments. This slight improvement is likely due to the use of
catchments. additional data (in this case, remotely sensed LAI) to con-
5.3.2. Comparisons Between the Three Educated strain the model calibration and because of the important
Regionalization Approaches role that vegetation processes play in controlling runoff.
[58] The difference between the three regionalization [Siriwardena et al., 2006; Yildiz and Barros, 2007; Zhang
approaches considered here, and used in many other studies, et al., 2009].
is small compared to the improved modeling results from [61] However, unlike the regionalization approaches
the use of multiple-donor catchments versus a single donor which have been widely explored in numerous other stud-
catchment and an educated selection of donor catchments ies, this is one of few studies that consider the use of leaf
versus a random selection of donor catchments. The spatial area index in conceptual rainfall-runoff models for predic-
proximity approach generally gives better modeling results tion in ungauged catchments. It is likely that a better
than the physical similarity approach, which is also reported integration of vegetation and other remotely sensed data
in other studies [Oudin et al., 2008b; Parajka et al., 2005]. may further improve the modeling results. These may
The spatial proximity approach, where the geographically include calibrating the rainfall-runoff models against both
11 of 13
W07412 ZHANG AND CHIEW: RUNOFF PREDICTIONS IN UNGAUGED CATCHMENT W07412

Figure 11. Comparisons of Nash-Sutcliffe Efficiency (NSE) performance between spatial proximity
and physical similarity and between integrated similarity and spatial proximity, the Xinanjiang model in
the 2001– 2006 calibration period.

runoff and actual evapotranspiration and/or soil moisture which have been widely explored in numerous other stud-
[Zhang et al., 2009] from remote sensing and improving ies, this is one of few studies that consider the use of LAI in
model structure to better incorporate remotely sensed conceptual rainfall-runoff models for prediction in unga-
information. uged catchments and a better integration of vegetation and
other remotely sensed data may further improve the mod-
eling results.
6. Conclusions
[62] This study evaluates the relative benefits of different [65] Acknowledgments. The study was supported by the Science
Leadership Scheme and the Water Resources Assessment and Accounting
regionalization methods for modeling runoff in ungauged Project (WIRADA) in Australian Commonwealth Scientific and Industrial
catchments, using two conceptual daily rainfall-runoff mod- Research Organization (CSIRO). We thank Jenet Austin for calculating the
els, Xinanjiang and SIMHYD, on 210 relatively unimpacted catchment attributes used in this study and Hongxia Li, David Post, and Jai
catchments in southeast Australia. The results show that the Vaze for helpful discussions. We also thank the three Water Resources
reviewers and the Associate Editor whose detailed comments led to
biggest benefit comes from an educated selection of donor significant improvement of the article.
catchments and output averaging of results from multiple-
donor catchments. The benefit from an educated selection of References
donor catchment versus a random selection of donor catch- Bardossy, A. (2007), Calibration of hydrological model parameters for
ment is similar to the benefit of using multiple-donor ungauged catchments, Hydrol. Earth Syst. Sci., 11, 703 – 710.
catchments versus a single donor catchment. Beven, K., and J. Freer (2001), Equifinality, data assimilation, and uncer-
[63] The difference between the three commonly used tainty estimation in mechanistic modelling of complex environmental
systems using the GLUE methodology, J. Hydrol., 249, 11 – 29.
approaches for selecting donor catchments is relatively Bloschl, G., and M. Sivapalan (1995), Scale issues in hydrological model-
small. The spatial proximity approach (where the geograph- ing—A review, Hydrol. Processes, 9, 251 – 290.
ically closest catchment is used as the donor catchment) Chau, K. W. (2006), Particle swarm optimization training algorithm for
performs slightly better than the physical similarity approach ANNs in stage prediction of Shing Mun River, J. Hydrol., 329, 363 – 367.
Chen, J., P. Jonsson, M. Tamura, Z. H. Gu, B. Matsushita, and L. Eklundh
(where the catchment with the most similar attributes is used (2004), A simple method for reconstructing a high-quality NDVI time-
as the donor catchment). The integrated similarity approach series data set based on the Savitzky-Golay filter, Remote Sens. Environ.,
which combines the spatial proximity and physical similar- 91, 332 – 344.
ity approaches performs only very marginally better than the Cheng, C. T., C. P. Ou, and K. W. Chau (2002), Combining a fuzzy optimal
spatial proximity approach. The relative difference between model with a genetic algorithm to solve multi-objective rainfall-runoff
model calibration, J. Hydrol., 268, 72 – 86.
the three approaches are more significant in the poorer Chiew, F. H. S., and T. A. McMahon (2002), Global ENSO-streamflow
modeled catchments where the integrated similarity approach teleconnection, streamflow forecasting and interannual variability, Hydrol.
performs better than the spatial proximity approach which in Sci. J., 47, 505 – 522.
turn performs better than the physical similarity approach. Chiew, F. H. A., T. C. Piechota, J. A. Dracup, and T. A. McMahon (1998),
El Niño Southern Oscillation and Australian rainfall, streamflow and
[64] The incorporation of LAI data into the rainfall-runoff drought: Links and potential for forecasting, J. Hydrol., 204, 138 – 149.
models only show marginal improvement to the modeling Chiew, F. H. S., M. C. Peel, and A. W. Western (2002), Application and
results. However, unlike the regionalization approaches testing of the simple rainfall-runoff model SIMHYD, in Mathematical

12 of 13
W07412 ZHANG AND CHIEW: RUNOFF PREDICTIONS IN UNGAUGED CATCHMENT W07412

Models of Small Watershed Hydrology and Applications, edited by V. P. Parajka, J., R. Merz, and G. Bloschl (2005), A comparison of regionalisa-
Singh and D. K. Frevert, pp. 335 – 367, Water Resources, Littleton, Colo. tion methods for catchment model parameters, Hydrol. Earth Syst. Sci.,
Dilley, A. C., M. Edwards, D. M. O’Brien, and R. M. Mitchell (2000), 9, 157 – 171.
Operational AVHRR processing modules: Atmospheric correction, Parajka, J., G. Bloschl, and R. Merz (2007), Regional calibration of catch-
cloud masking and BRDF compensation, CSIRO Atmospheric Research ment models: Potential for ungauged catchments, Water Resour. Res., 43,
Paper, pp. 14 – 24. (Available at http://www.eoc.csiro.au/tasks/task4_2/ W06406, doi:10.1029/2006WR005271.
AVHRRFinal2.pdf) Piechota, T. C., F. H. S. Chiew, J. A. Dracup, and T. A. McMahon (1998),
Eberhart, R. C., and J. Kennedy (1995), A new optimizer using particle Seasonal streamflow forecasting in eastern Australia and the El Niño
swarm theory, in Sixth International Symposium on Micro Machine and Southern Oscillation, Water Resour. Res., 34, 3035 – 3044.
Human Science, Nagoya Municipal Industrial Research Institute, pp. 39 – Post, D. A., J. A. Jones, and G. E. Grant (1998), An improved methodology
43, doi:10.1109/MHS.1995.494215, IEEE Press, Piscataway, N. J. for predicting the daily hydrologic response of ungauged catchments,
Fang, H. L., S. L. Liang, J. R. Townshend, and R. E. Dickinson (2008), Environ. Modell. Software, 13, 395 – 403.
Spatially and temporally continuous LAI data sets based on an integrated Power, S., T. Casey, C. Folland, A. Colman, and V. Mehta (1999), Inter-
filtering method: Examples from North America, Remote Sens. Environ., decadal modulation of the impact of ENSO on Australia, Clim. Dyn., 15,
112, 75 – 93. 319 – 324.
Gan, T. Y., E. M. Dlamini, and G. F. Biftu (1997), Effects of model com- Priestley, C. H. B., and R. J. Taylor (1972), On the assessment of surface
plexity and structure, data quality, and objective functions on hydrologic heat flux and evaporation using large-scale parameters, Mon. Weather
modeling, J. Hydrol., 192, 81 – 103. Rev., 100, 81 – 92.
Gill, M. K., Y. H. Kaheil, A. Khalil, M. McKee, and L. Bastidas (2006), Ruffin, C., R. L. King, and N. H. Younani (2008), A combined derivative
Multiobjective particle swarm optimization for parameter estimation in spectroscopy and Savitzky-Golay filtering method for the analysis of
hydrology, Water Resour. Res., 42, W07417, doi:10.1029/2005WR004528. hyperspectral data, Geosci. Remote Sens., 45, 1 – 15.
Goswami, M., K. M. O’Connor, and K. P. Bhattarai (2007), Development Schaaf, C. B., et al. (2002), First operational BRDF, albedo nadir reflec-
of regionalisation procedures using a multi-model approach for flow tance products from MODIS, Remote Sens. Environ., 83, 135 – 148.
simulation in an ungauged catchment, J. Hydrol., 333, 517 – 531. Siriwardena, L., B. L. Finlayson, and T. A. McMahon (2006), The impact
Jayawardena, A. W., and M. C. Zhou (2000), A modified spatial soil of land use change on catchment hydrology in large catchments: The
moisture storage capacity distribution curve for the Xinanjiang model, Comet River, Central Queensland, Australia, J. Hydrol., 326, 199 – 214.
J. Hydrol., 227, 93 – 113. Sivapalan, M., et al. (2003), IAHS decade on Predictions in Ungauged
Jeffrey, S. J., J. O. Carter, K. B. Moodie, and A. R. Beswick (2001), Using Basins (PUB), 2003 – 2012: Shaping an exciting future for the hydrolo-
spatial interpolation to construct a comprehensive archive of Australian gical sciences, Hydrol. Sci. J., 48, 857 – 880.
climate data, Environ. Modell. Software, 16, 309 – 330. Verdon, D. C., A. M. Wyatt, A. S. Kiem, and S. W. Franks (2004), Multi-
Leuning, R., Y. Q. Zhang, A. Rajaud, H. A. Cleugh and K. Tu (2008), A decadal variability of rainfall and streamflow: Eastern Australia, Water
simple surface conductance model to estimate evaporation using MODIS Resour. Res., 40, W10201, doi:10.1029/2004WR003234.
leaf area index and the Penman-Monteith equation, Water Resour. Res., Yildiz, O., and A. P. Barros (2007), Elucidating vegetation controls on the
44, W10419, doi:10.1029/2007WR006562. hydroclimatology of a mid-latitude basin, J. Hydrol., 333, 431 – 448.
McIntyre, N., H. Lee, H. Wheater, A. Young, and T. Wagener (2005), Young, A. R. (2006), Stream flow simulation within UK ungauged catch-
Ensemble predictions of runoff in ungauged catchments, Water Resour. ments using a daily rainfall-runoff model, J. Hydrol., 320, 155 – 172.
Res., 41, W12434, doi:10.1029/2005WR004289. Zhang, Y. Q., and M. Wegehenkel (2006), Integration of MODIS data into a
McKenzie, N. J., D. W. Jacquier, L. J. Ashton, and H. P. Cresswell (2000), simple model for the spatial distributed simulation of soil water content
Estimation of soil properties using the Atlas of Australian Soils, Tech. and evapotranspiration, Remote Sens. Environ., 104, 393 – 408.
Rep. 11/00, CSIRO Land and Water, Canberra, ACT, Australia. Zhang, Y. Q., F. H. S. Chiew, L. Zhang, R. Leuning, and H. A. Cleugh
Merz, R., and G. Bloschl (2004), Regionalisation of catchment model (2008), Estimating catchment evaporation and runoff using MODIS leaf
parameters, J. Hydrol., 287, 95 – 123. area index and the Penman-Monteith Equation, Water Resour. Res., 44,
Merz, R., and G. Bloschl (2009), A regional analysis of event runoff coef- W10420, doi:10.1029/2007WR006563.
ficients with respect to climate and catchment characteristics in Austria, Zhang, Y. Q., F. H. S. Chiew, L. Zhang, and H. X. Li (2009), Use of
Water Resour. Res., 45, W01405, doi:10.1029/2008WR007163. remotely sensed actual evapotranspiration to improve rainfall-runoff
Nash, J. E., and J. V. Sutcliffe (1970), River forcasting using conceptual modelling in southeast Australia, J. Hydrometeorol., 10, doi:10.1175/
models. 1: A discussion of principles, J. Hydrol., 10, 280 – 290. 2009JHM1061.1171.
Oudin, L., V. Andreassian, J. Lerat, and C. Michel (2008a), Has land cover Zhao, R. J. (1992), The Xinanjiang model applied in China, J. Hydrol., 135,
a significant impact on mean annual streamflow? An international assess- 371 – 381.
ment using 1508 catchments, J. Hydrol., 357, 303 – 316.
Oudin, L., V. C. Andréassian, C. Perrin, C. Michel, and N. Le Moine
(2008b), Spatial proximity, physical similarity, regression and ungauged 

catchments: A comparison of regionalization approaches based on 913 F. H. S. Chiew and Y. Zhang, CSIRO Water for a Healthy Country
French catchments, Water Resour. Res., 44, W03413, doi:10.1029/ National Research Flagship, CSIRO Land and Water, P.O. Box 1666,
2007WR006240. Clunies Ross Street, Canberra, ACT 2601, Australia. (yongqiang.zhang@
csiro.au)

13 of 13

You might also like