Professional Documents
Culture Documents
Korbinian Breinl
To cite this article: Korbinian Breinl (2016) Driving a lumped hydrological model with precipitation
output from weather generators of different complexity, Hydrological Sciences Journal, 61:8,
1395-1414, DOI: 10.1080/02626667.2015.1036755
(1981). Daily precipitation occurrences were simulated days). Class one contains the lowest, class five the
with a Markov process where the transition probabil- highest observed precipitation amount. The six
ities depend on m previous days: intensity classes were estimated by k-means cluster-
ing (for details about k-means clustering see
PRfXtþ1 jXt ;Xt1 ;Xt2 ;...;X1 g Hartigan 1975 or Hartigan and Wong 1979). The
¼ PRfXtþ1 jXt ;Xt1 ;...;Xtm g with m < t 1 (1) squared Euclidean distance was used to build the
classes. Once the occurrences have been generated,
precipitation amounts are first resampled from the
A two-state Markov chain was used for simulating
observations so that a randomly drawn observed day
dry and wet days. Precipitation amounts on wet days
of a specific class and month is assigned a synthetic
were sampled from one of the four distribution func-
day that matches both class and month. Following
tions (separate distribution parameters for each
this, parametric precipitation amounts are generated
month) as described in Section 2.2. Before fitting
for each rainy day and reshuffled so that the ranks of
the parametric distributions, the daily amounts of
the resampled and parametric precipitation amounts
all raingauges were weighted into lumped precipita-
match. The reshuffling is applied over all simulated
tion time series by Thiessen polygons. Other meth-
years but separately for each month to make the
ods including inverse distance weighting (IDW),
resampled and synthetic (i.e. parametric) time series
ordinary kriging (OK) and kriging with external
as similar as possible. Related reshuffling algorithms
drift (KED) considering elevation were tested using
have been described by Clark et al. (2004) and Breinl
the “R” software package “hydroTSM” (Zambrano-
et al. (2013). The choice of the number of precipita-
Bigiarini 2012). The computed distribution para-
tion intensity classes is subjective. There is a trade-
meters did not differ significantly between methods,
off between the number of classes and the number of
except for KED in the Ubaye catchment. However,
observations in each class. In this study, the number
for Ubaye, KED was considered less reliable as it
varies between 109 (highest intensity) and 1656
produced almost 5% negative precipitation values, a
observations (lowest intensity) in the Salzach catch-
known issue related to this method (Haberlandt
ment, and 61 and 3240 observations in the Ubaye
2007, Ly et al. 2011). Moreover, using the question-
catchment. As in Model A, the daily amounts of all
able KED output for hydrological model calibration
raingauges were averaged/weighted into lumped pre-
likewise resulted in considerably lower Nash-Sutcliffe
cipitation time series by Thiessen polygons before
efficiencies (Nash and Sutcliffe 1970) in both the
fitting the precipitation distributions. Temperature
calibration and validation periods (−10.8%/–7.2%).
was simulated in the same way as for Model A.
Mean daily temperature was simulated independently
The principle of Model A+ is schematized in Fig. 1.
by autoregressive moving average processes (ARMA).
Fourier time series were first fitted to represent the
2.1.3 Model B
annual cycles of temperature and standard deviation,
The multi-site Model B was likewise set up on a monthly
and the remaining residuals were simulated using the
basis and is described in detail in Breinl et al. (2015).
ARMA process. See Breinl et al. (2015) for details.
They presented a model based on a precipitation algo-
The principle of Model A is schematized in Fig. 1.
rithm by Breinl et al. (2013). In this algorithm, daily
snapshots of precipitation fields (i.e. catchment-wide
2.1.2 Model A+ precipitation patterns) called “amount vectors” are first
Model A+ is an adapted version of the univariate clustered into similar classes representing similar types
Richardson type model as described in Section 2.1.1, of patterns, which are then simulated using a Markov
and has been developed for this study to reduce the process. The number of classes has direct impact on the
underestimation of the autocorrelation of precipita- duplication of observations in the simulation output.
tion amounts in Model A. The latter is caused by the The number of classes was empirically chosen so that
random sampling of precipitation amounts on any the duplication of observations in the simulation was
rainy day, ignoring the serial correlation of precipi- below 1%. Once time series of classes have been simu-
tation amounts in the observations. Instead of simu- lated, each single class is replaced by a randomly drawn
lating precipitation occurrences with a two-state observed “amount vector” that matches the simulated
Markov chain for dry and wet days, in Model A+, class and month. In the last two steps, precipitation
the precipitation amounts are simulated with a six- amounts are sampled from parametric distributions (at
state Markov chain, representing dry days and five each site separately) and reshuffled according to ranks of
different precipitation intensity classes (i.e. rainy the observed amount vectors to maintain the temporal
1398 K. BREINL
Figure 1. Schematics of the three weather generators, Model A, Model A+ and Model B. Model B is shown for an exemplary number
of three sites and 10 clusters.
and spatial correlation structures. Breinl et al. (2015) consisting of the empirical distribution for lower and
called this model the reduced-state-space model (RSS). the GPD for higher precipitation amounts.
The principle is schematized in Fig. 1 (right). The reader The probability density function (pdf) of the
is referred to Breinl et al. (2015) for more details on the Exponential distribution is given by:
algorithm. In Model B, the precipitation amounts from
all simulated sites were subsequently weighted by fE ðxÞ ¼ α expðαxÞ; α > 0 (2)
Thiessen polygons into lumped time series. As in the
cases of models A and A+, mean daily temperature was
simulated independently using an ARMA process. The
principles of the simulation procedures with Model A, where α is the single parameter of the distribution.
Model A+ and Model B are shown in Fig. 2. The orders The pdf of the Gamma distribution is given by:
of the Markov processes in models A/A+ and B were
βα xα1 x
estimated using the Akaike information criterion, AIC fG ðxÞ ¼ exp ; α > 0; β > 0 (3)
(Katz 1981) and the Bayesian information criterion, BIC ΓðαÞ β
(Schwarz 1978). For all three models, first-order pro-
cesses are appropriate for all months in both study areas.
Daily mean temperature was simulated using time series where α is the shape and β is the scale parameter.
extracted from the centroid of an E-OBS grid cell in the The pdf of the Weibull distribution is given by:
Ubaye catchment and from three sites (15, 16 and 19) in
α
the Salzach catchment (Fig. 3 and Section 2.4). In the α x α1 x
Salzach catchment, the arithmetic mean of the three fW ð x Þ ¼ exp ;α > 0;β > 0 (4)
β β β
temperature sites was used as input for the hydrological
model. The four parametric precipitation distributions
described in Section 2.2 were tested for all models.
where α is the shape and β is the scale parameter.
The pdf of the Hybrid distribution is defined by:
Figure 3. Overview of the two study areas including location of the rain and river gauges. In the Ubaye study area, temperature
records were taken from grid-based observation data (E-OBS). In the Salzach study area, temperature records were available at sites
15, 16 and 19. Precipitation records of site 19 were not considered.
1
σ 1 ðx uÞ 1κ 2009), or through peak-over-threshold (POT) analy-
fGP ðxÞ ¼ ¼ 1 þκ ; σ > 0;u 0 (6) sis (e.g. Furrer and Katz 2008). Dynamic merging of
κ σ σ
the distributions types is numerically unstable, data
1400 K. BREINL
sensitive and computationally expensive (Singh and quadratically-convergent method (Press et al.
Mishra 2011). However, the POT method can be 2002). As different parameter sets in conceptual
subjective and (for large observation networks) models can lead to equally good simulation results
very time-consuming. In this paper, thresholds (“model equifinality”, see Beven 2001), 100 different
were visually estimated with “mean residual life but suitable parameter sets were calibrated for each
plots” and “parameter stability plots” using the sta- catchment. In the simulations, the model was run
tistical software ExtRemes in “R” (see Gilleland and separately for each of the 100 parameter sets and
Katz 2011). the output was subsequently averaged. A 10-year
calibration and validation period in both catch-
ments resulted in an average Nash-Sutcliffe effi-
2.3 Hydrological model ciency (Nash and Sutcliffe 1970) of 0.87/0.82 for
Salzach and 0.82/0.74 for Ubaye.
The present study requires a large number of mod-
elling runs for probabilistic analyses and to reliably
estimate uncertainty. Unlike conceptual hydrologi-
2.4 Study areas and data
cal models, fully-distributed models can simulate
physical processes but they likewise come with Two catchments were chosen for this research: the
numerous challenges (Beven 1989, Refsgaard et al. Salzach catchment in Germany and Austria down to
2010). Most importantly, though, they are compu- the city of Salzburg (4637 km2), and the Ubaye
tationally expensive and data demanding (Arnold Catchment in France down to the village of
et al. 1998, Romanowicz et al. 2005, Shrestha et al. Barcelonnette (548 km2) (Fig. 3). The discharge of
2006). In addition, a higher-resolution model does the Salzach River is dominated by snowmelt.
not always improve the model performance (Das Extreme discharges and floods usually occur in
et al. 2008). The hydrological model HBV is a con- summer as a combination of snowmelt and long-
ceptual model that has been applied in numerous lasting rainfall, or rainfall caused by Vb weather
studies (e.g. Uhlenbrook et al. 1999, Bergström situations (Mudelsee et al. 2004). Extreme dis-
et al. 2001b, Booij 2005, Hagg et al. 2006, charges in the Ubaye River are generally associated
Bárdossy and Das 2008, Das et al. 2008, Steele- with snowmelt or high-intensity summer storms
Dunne et al. 2008, Gao et al. 2012, Beck et al. (Turkington et al. 2014). For Salzach, daily precipi-
2013). In this study, the HBV-light model was tation records were obtained for 1987–2010 at 18
used in a lumped set-up with 10 different elevation sites. Temperature records were available at three
zones for the distributed snow routine (to improve sites (15, 16, 19). For Ubaye, precipitation records
the simulation of snow retention), and with three were available from 1971–2004 at four sites. As
groundwater boxes (Seibert 1997, 1999, 2000, gauge station based temperature records in the
Seibert and Vis 2012). The 10 elevation zones were area had multiple missing years and major gaps,
computed in a GIS using a natural breaks classifica- gridded observed temperature data from the
tion according to Jenks (1967). HBV-light generates ENSEMBLES project (E-OBS dataset, 0.25 degree
daily mean discharge using daily precipitation, daily regular grid) (Haylock et al. 2008) were used.
mean temperature and potential monthly evapora- E-OBS data have been successfully used in other
tion. The model essentially consists of four main flood hazard related studies (e.g. Photiadou et al.
components: a snow routine, a soil moisture rou- 2011, Schädler et al. 2012, Freudiger et al. 2014).
tine, a response routine and a routing routine. The E-OBS temperature data show few inhomogeneities
15 model parameters are calibrated to discharge in southeastern France (Hofstra et al. 2009).
observations. The model was calibrated using a Temperature time series were extracted from the
GAP routine (Seibert 2000), consisting of a genetic grid cell closest to the catchment centre (Fig. 3).
algorithm (GA), followed by Powell’s (P) Tables 1 and 2 give details on the gauge stations.
Table 1. Names, coordinates, altitudes and annual precipitation statistics of the gauge stations in the Ubaye study area.
ID Name Lat Long Altitude Mean StDev 5th percentile 25th percentile 75th percentile 95th percentile
(m .a.s.l.) (mm) (mm) (mm) (mm) (mm) (mm)
1 Barcelonnette 44.3915 6.6718 1155 711 140 481 629 795 984
2 Condamine 44.4598 6.7449 1325 663 143 472 569 782 874
3 Saint-Paul 44.5936 6.8424 1903 924 207 569 765 1101 1206
4 Uvernet 44.3186 6.6925 1660 987 214 688 857 1136 1332
HYDROLOGICAL SCIENCES JOURNAL – JOURNAL DES SCIENCES HYDROLOGIQUES 1401
Table 2. Names, coordinates, altitudes and annual precipitation statistics of the gauge stations in the Salzach study area.
ID Name Lat Long Altitude Mean StDev 5th percentile 25th percentile 75th percentile 95th percentile
(m .a.s.l.) (mm) (mm) (mm) (mm) (mm) (mm)
1 Gerlos 47.2269 12.0444 1250 1278 105 1124 1180 1350 1443
2 Paß Thurn 47.2997 12.4222 1200 1206 131 1004 1096 1312 1402
3 Hochfilzen 47.4703 12.6217 960 1781 185 1479 1643 1925 2080
4 Felbertauerntunnel 47.1181 12.5056 1650 1403 133 1184 1287 1488 1624
5 Böckstein 47.0872 13.1158 1140 1342 132 1158 1268 1363 1622
6 Flachau 47.3472 13.3958 910 1154 93 1024 1087 1206 1304
7 Golling-Torren 47.5917 13.1575 473 1474 104 1293 1413 1533 1656
8 Gosau 47.5892 13.5431 765 1742 177 1460 1613 1895 2020
9 Hintersee 47.7611 13.2225 750 2035 246 1592 1850 2240 2388
10 Bischofswiesen-Loipl 47.6525 12.9315 845 1894 175 1575 1786 2044 2168
11 Hüttschlag 47.1772 13.2314 1030 1096 87 951 1038 1160 1239
12 Glanegg 47.7494 13.0175 450 1550 153 1284 1449 1641 1814
13 Enzingerboden 47.1667 12.6333 1768 1438 201 1120 1255 1573 1811
14 Schmittenhöhe 47.1786 12.7381 2102 1518 143 1315 1407 1652 1739
15 Zell am See 47.3267 12.7953 767 1175 98 1023 1114 1242 1351
16 St. Veit 47.3292 13.1550 747 1102 106 913 1052 1159 1301
17 Dienten 47.3908 13.0333 1265 1480 138 1282 1374 1640 1699
18 Hüttau 47.4131 13.3156 720 1138 102 984 1064 1201 1321
19 Salzburg Airport 47.8014 13.0017 424 1278 105 1124 1180 1350 1443
Figure 4. Autocorrelation of catchment precipitation occurrences (occ) and amounts (amt) for Model A, Model A+ and Model B in
both catchments.
HYDROLOGICAL SCIENCES JOURNAL – JOURNAL DES SCIENCES HYDROLOGIQUES 1403
Table 3. Goodness-of-fit tests for the four distribution functions For Model B and the Salzach catchment, the
and for the three model set-ups and two study areas: chi-square Exponential distribution is the worst choice.
test at 5% level, RMSE and Q-Q correlation (average of all fits). Hybrid distributions seem to capture the observa-
Rejections chi- RMSE Q-Q
square test (mm) correlation tions well without any rejection produced by the
Catchment precipitation chi-square tests. Acceptable results are also achieved
Salzach models A/A+ Exponential 11 1.7803 0.9878 with Gamma and Weibull distributions, as the aver-
(12 fits) Gamma 0 0.8113 0.9930
Weibull 3 1.9223 0.9799 age RMSE and Q-Q correlations indicate. For the
Hybrid 0 0.7045 0.9948 Ubaye catchment, the performance of the Weibull
Ubaye models A/A+ Exponential 12 2.8977 0.9792
(12 fits) Gamma 0 0.9308 0.9948
and Hybrid distributions is similar. As in the case
Weibull 1 1.5221 0.9859 of Salzach, the Exponential distribution seems to be
Hybrid 0 1.0112 0.9941 the worst choice. The Gamma distribution performs
Site-specific precipitation better than the Exponential distribution but worse
Salzach Model B (216 Exponential 66 1.6825 0.9867
fits) Gamma 33 1.316 0.9881 than the Weibull and Hybrid distributions. When
Weibull 26 1.2396 0.9887 evaluating the annual maximum (catchment) preci-
Hybrid 0 1.127 0.9911
Ubaye Model B (48 Exponential 24 2.7651 0.9879 pitation (Fig. 5), a large spread in models A/A+
fits) Gamma 15 1.7284 0.9905 becomes obvious. For models A/A+ and both
Weibull 5 1.3826 0.9929
Hybrid 2 1.3845 0.9914 study areas, the Weibull distribution performs
worst. The Hybrid distribution performs consider-
ably better. In general, more bias can be detected
Weibull distributions appear to be less suitable for Salzach. The multi-site Model B is less sensitive
when fitting to lumped precipitation time series. to different parametric curves. In the Salzach area,
In particular the Exponential distribution causes a Hybrid and Weibull distributions outperform the
high number of rejections in the chi-square tests. Exponential and Gamma distributions. In Ubaye,
Figure 5. Return period of annual maximum daily (catchment) precipitation for the three model types, both study areas and the
four different parametric distribution functions implemented.
1404 K. BREINL
Figure 6. Return period of annual (catchment) precipitation totals for the three model types, both study areas and the four different
parametric distribution functions implemented.
all distributions except the Exponential appear to be precipitation. In the Ubaye study area, differences
reasonable. The return periods of annual (catch- between types of models are generally less pro-
ment) precipitation totals are plotted in Fig. 6. nounced (Fig. 8). In models A and A+, mean
monthly precipitation is well reproduced by all dis-
tribution functions. As in the Salzach study area,
4.2 Hydrological output the Weibull distribution tends to overestimate, at
least in the second half of the year. In Model B,
4.2.1 Monthly and daily statistics
the hybrid distribution tends to overestimate preci-
For Salzach and models A/A+, the mean monthly
pitation in the spring season. However, the impact
precipitation amounts are well reproduced by all
of the distribution functions on the resulting mean
distributions except the Weibull distribution,
monthly and daily flows is small. Model B performs
which tends to overestimate, leading to an overesti-
better than models A and A+ in the summer season.
mation of daily and monthly mean flows (Fig. 7).
In autumn, models A+ and B tend to slightly over-
The overestimation is caused by the poor para-
estimate the flows. Model A performs better, but it
metric fits for lumped precipitation time series as
is possible that a poor reproduction of autocorrela-
discussed above. Model B performs well with all
tion compensates for too high precipitation
distribution functions, reproducing the bootstrap
amounts from the parametric distributions. In gen-
precipitation amounts and flows very well. Some
eral, the bias from the HBV model (i.e. bias between
bias of the HBV model can be detected regardless
bootstrap and observations) is less pronounced
of the input precipitation. This bias is represented
compared to the Salzach study area.
by differences between the flows produced from
bootstrap input and the actual observations.
Models A+ and B better reproduce mean summer 4.2.2 Extreme design flows
flows compared to Model A, presumably due to a The performance of the different model types and
better reproduction of the autocorrelation of the four distributions was compared with respect to
HYDROLOGICAL SCIENCES JOURNAL – JOURNAL DES SCIENCES HYDROLOGIQUES 1405
Figure 7. Observed and simulated monthly mean (catchment) precipitation and monthly and daily mean flows for the three model
types (Salzach catchment).
Figure 8. Observed and simulated monthly mean (catchment) precipitation and monthly and daily mean flows for the three model
types (Ubaye catchment).
1406 K. BREINL
Figure 9. Observed and simulated extreme design flows (annual, winter/spring, summer/autumn) for Q10, Q20 and Q50 (Model A,
Salzach). Bars denote the mean, lines the 95th confidence intervals. Results are shown for the observations (Obs), bootstrap
simulations (BS) and the four distribution functions. Black dots denote the distribution that best matches the mean of the
bootstrapping results.
the annual maximum flow, the annual maximum of precipitation compared to Model A. Figure 11
winter/spring flow, and the annual maximum sum- shows the results for Model B. Differences between
mer/autumn flow. As can be seen in Fig. 9, the the types of distributions are considerably less pro-
differences between the distribution functions are nounced, which results in three different distribu-
pronounced. For annual flows, the Weibull distribu- tions (Exponential, Weibull, Hybrid) outperforming
tion tends to overestimate for all return periods. the others. However, the Hybrid distribution per-
The most suitable distribution is the Hybrid distri- forms best in the majority of cases.
bution. Results are similar for winter/spring and In the Ubaye study area, the Weibull distribution
summer/autumn extremes. Some bias between the likewise tends to overestimate extremes when using
observations and bootstrap is obvious, especially for models A and A+ (Figs 12 and 13), particularly for
the winter/spring season. However, this bias appears higher return periods. The Hybrid distribution best
to be small compared to the confidence intervals of reproduces the bootstrap results. Differences between
observations, in particular for Q20 and Q50. Results types of distributions are obvious but less pronounced
are comparable for Model A+ (Fig. 10). The Hybrid compared to models A/A+ in the Salzach study area.
distribution appears to be most suitable. Simulated For Model B (Fig. 14), the differences are even less
values are generally higher, which is very likely pronounced than for the univariate model set-ups. All
related to the better reproduction of autocorrelation distribution functions appear to be reasonable,
HYDROLOGICAL SCIENCES JOURNAL – JOURNAL DES SCIENCES HYDROLOGIQUES 1407
Figure 10. Observed and simulated extreme design flows (annual, winter/spring, summer/autumn) for Q10, Q20 and Q50 (Model A+,
Salzach). Bars denote the mean, lines the 95th confidence intervals. Results are shown for the observations (Obs), bootstrap
simulations (BS) and the four distribution functions. Black dots denote the distribution that best matches the mean of the
bootstrapping results.
although the Hybrid distribution performs best in all changes the statistical behaviour of the site-specific
cases. precipitation, leading to pronounced differences in
the fits of precipitation distributions. This is different
to the findings by Li et al. (2013), where distribution
5 Discussion and conclusion
parameters from areal precipitation and single sites
The study has shown that all three models (A, A+ and “shared the same characteristics”. Furthermore, in this
B) have potential to be used for lumped hydrological study, different weighting methods, including Thiessen
modelling. However, as could also be demonstrated, polygons, IDW, OK as well as KED, did not improve
the classic Richardson type Model A underestimates the fits and led to similar curve parameters. Goodness-
the autocorrelation of precipitation amounts. Model of-fit tests as well as various precipitation related plots
A+ and Model B, as well as other multi-site models, revealed increased bias when fitting parametric distri-
perform better in this respect (e.g. Brandsma and butions to lumped precipitation time series. It is valid
Buishand 1998, Buishand and Brandsma 2001, to assume that this phenomenon is mainly related to
Beersma and Buishand 2003). All model types show the mountainous character of the two catchments:
similar performance in the simulation of wet spells. Time series from raingauges that are located at very
The study has also shown that the averaging/weigh- different altitudes in different micro-climatic condi-
ing process to produce lumped time series obviously tions, and which are exposed to orographic effects in
1408 K. BREINL
Figure 11. Observed and simulated extreme design flows (annual, winter/spring, summer/autumn) for Q10, Q20 and Q50 (Model B,
Salzach). Bars denote the mean, lines the 95th confidence intervals. Results are shown for the observations (Obs), bootstrap
simulations (BS) and the four distribution functions. Black dots denote the distribution that best matches the mean of the
bootstrapping results.
a complex topography, are lumped together. These and Gamma distributions. Differences in the perfor-
site- specific precipitation time series differ signifi- mance of different parametric distributions are gener-
cantly. In the Ubaye catchment, the difference in the ally less pronounced in Model B. This becomes obvious
annual amount of precipitation between sites is up to when simulating monthly and daily flow statistics
49%. The difference in altitude between the lowest and where all distributions produce reasonable results in
highest rain gauge is 748 m. In the Salzach catchment, both study areas. The choice of the parametric distri-
the differences are even more pronounced (86% and bution in models A and A+ has a larger impact on the
1652 m). In this study, the problem of lumped preci- monthly and daily flow modelling results. This parti-
pitation time series particularly applies to the Weibull cularly applies to the Salzach catchment. It also became
distribution, which reproduces precipitation statistics obvious that Model A tends to produce lower flows
poorly in models A/A+ but well in Model B. In general, than Model A+ and Model B (e.g. summer period
for models A/A+, the Exponential and Weibull distri- Salzach), which also affects the simulation of extremes.
butions perform worse than the Gamma and Hybrid As the performance of the three models is similar with
distributions. The latter seems to be the most reason- regard to the generation of wet spells, only the con-
able choice. In Model B, the Weibull and Hybrid dis- siderably lower autocorrelation of the precipitation
tributions tend to perform better than the Exponential amounts in Model A can explain this shortcoming
HYDROLOGICAL SCIENCES JOURNAL – JOURNAL DES SCIENCES HYDROLOGIQUES 1409
Figure 12. Observed and simulated extreme design flows (annual, winter/spring, summer/autumn) for Q10, Q20 and Q50 (Model A,
Ubaye). Bars denote the mean, lines the 95th confidence intervals. Results are shown for the observations (Obs), bootstrap
simulations (BS) and the four distribution functions. Black dots denote the distribution that best matches the mean of the
bootstrapping results.
(see Section 4.1). The key conclusions of this study can ● Model B is less sensitive to the choice of para-
be summarized as follows: metric distribution functions. Potential poor fits
of single sites are averaged out in the subse-
● Fitting parametric distribution functions to quent weighting procedure. All parametric
lumped precipitation in models A/A+ requires curves produce reasonable mean and extreme
additional attention. Poor fits can have strong design flows, with the Hybrid distribution per-
impact on the simulation results. forming best in the majority of cases.
● Poor fits for lumped precipitation time series ● Model A tends to produce lower monthly and
are very likely caused by the mountainous daily flows as well as lower extremes. This
character of the two catchments and thus by becomes obvious in the summer season of the
lumping time series that differ significantly Salzach catchment and can be explained with
(due to different altitudes, different site-speci- the lower autocorrelation of simulated precipi-
fic micro-climatic conditions and orographic tation amounts (compared to models A+ and
effects). B). Floods in the Salzach catchment often
● The Gamma and Hybrid distributions seem to be occur in summer after long wet spells, i.e. the
more suitable for fitting lumped precipitation reproduction of the serial autocorrelation is
than the Exponential and Weibull distributions. important.
1410 K. BREINL
Figure 13. Observed and simulated extreme design flows (annual, winter/spring, summer/autumn) for Q10, Q20 and Q50
(Model A+, Ubaye). Bars denote the mean, lines the 95th confidence intervals. Results are shown for the observations (Obs),
bootstrap simulations (BS) and the four distribution functions. Black dots denote the distribution that best matches the mean
of the bootstrapping results.
● Independent modelling of precipitation and univariate Richardson type weather generator (mod-
temperature may introduce additional bias. els A and A+), curve fitting requires particular atten-
Some bias can be detected in the Salzach study tion. The wrong choice of parametric distribution
area (mean flows, overestimation of extreme function can lead to a large bias in the simulations.
design flows in winter/spring). For this reason, The major shortcoming of the Richardson type
a sensitivity study (not shown in this paper) model, which is the underestimation of the autocor-
with conditioned temperature and the bootstrap relation of precipitation, can be tackled by an
model (using different Fourier series for tem- improved Richardson model type (Model A+). The
perature on wet and dry days, see Semenov multisite model (Model B) produces the least bias
et al. 1998) was conducted for the Salzach and there is considerably less difference between the
study area, but did not noticeably improve the different types of distribution functions. However,
results. there is a trade-off between the increased efforts of
setting up such a multi-site weather generator com-
Altogether, all weather generators have the potential pared to implementing comparatively straightforward
to produce reasonable results, as long as the follow- univariate models, and the potential improvements
ing findings are taken into account: for the of simulations. Additional studies would be required
HYDROLOGICAL SCIENCES JOURNAL – JOURNAL DES SCIENCES HYDROLOGIQUES 1411
Figure 14. Observed and simulated extreme design flows (annual, winter/spring, summer/autumn) for Q10, Q20 and Q50 (Model B,
Ubaye). Bars denote the mean, lines the 95th confidence intervals. Results are shown for the observations (Obs), bootstrap
simulations (BS) and the four distribution functions. Black dots denote the distribution that best matches the mean of the
bootstrapping results.
Bárdossy, A. and Das, T., 2008. Influence of rainfall observa- Coles, S., 2001. An introduction to statistical modeling of
tion network on model calibration and application. extreme values. New York: Springer.
Hydrology and Earth System Sciences, 12 (1), 77–89. Das, T., et al., 2008. Comparison of conceptual model per-
doi:10.5194/hess-12-77-2008 formance using different representations of spatial varia-
Bárdossy, A. and Pegram, G.G.S., 2009. Copula based multi- bility. Journal of Hydrology, 356 (1–2), 106–118.
site model for daily precipitation simulation. Hydrology doi:10.1016/j.jhydrol.2008.04.008
and Earth System Sciences, 13 (12), 2299–2314. Deidda, R., 2010. A multiple threshold method for fitting the
doi:10.5194/hess-13-2299-2009 generalized Pareto distribution to rainfall time series.
Beck, H.E., et al., 2013. The impact of forest regeneration on Hydrology and Earth System Sciences, 14 (12), 2559–2575.
streamflow in 12 mesoscale humid tropical catchments. doi:10.5194/hess-14-2559-2010
Hydrology and Earth System Sciences, 17 (7), 2613–2635. Filliben, J.J., 1975. The probability plot correlation coefficient
doi:10.5194/hess-17-2613-2013 test for normality. Technometrics, 17 (1), 111–117.
Beersma, J.J. and Buishand, T.A., 2003. Multi-site simulation doi:10.1080/00401706.1975.10489279
of daily precipitation and temperature conditional on the Foufoula-Georgiou, E. and Lettenmaier, D.P., 1987. A
atmospheric circulation. Climate Research, 25 (2), 121– Markov renewal model for rainfall occurrences. Water
133. doi:10.3354/cr025121 Resources Research, 23 (5), 875–884. doi:10.1029/
Bergström, S., et al., 2001a. Climate change impacts on run- WR023i005p00875
off in Sweden - assessments by global climate models, Freudiger, D., et al., 2014. Large-scale analysis of changing
dynamical downscaling and hydrological modelling. frequencies of rain-on-snow events with flood-generation
Climate Research, 16 (2), 101–112. doi:10.3354/cr016101 potential. Hydrology and Earth System Sciences, 18 (7),
Bergström, S., et al., 2001b. Climate change impacts on run- 2695–2709. doi:10.5194/hess-18-2695-2014
off in Sweden - assessments by global climate models, Furrer, E.M. and Katz, R.W., 2008. Improving the simulation
dynamical downscaling and hydrological modelling. of extreme precipitation events by stochastic weather gen-
Climate Research, 16 (2), 101–112. doi:10.3354/cr016101 erators. Water Resources Research, 44 (12), W12439.
Beven, K., 1989. Changing ideas in hydrology - the case of doi:10.1029/2008WR007316
physically-based models. Journal of Hydrology, 105 (1–2), Gao, H.K., et al., 2012. Modeling the runoff and glacier mass
157–172. doi:10.1016/0022-1694(89)90101-7 balance in a small watershed on the Central Tibetan
Beven, K., 2001. How far can we go in distributed hydro- Plateau, China, from 1955 to 2008. Hydrological
logical modelling? Hydrology and Earth System Sciences, 5 Processes, 26 (11), 1593–1603. doi:10.1002/hyp.8256
(1), 1–12. doi:10.5194/hess-5-1-2001 Garavaglia, F., et al., 2010. Introducing a rainfall compound
Booij, M.J., 2005. Impact of climate change on river flooding distribution model based on weather patterns sub-sam-
assessed with different spatial model resolutions. Journal pling. Hydrology and Earth System Sciences, 14 (6), 951–
of Hydrology, 303 (1–4), 176–198. doi:10.1016/j. 964. doi:10.5194/hess-14-951-2010
jhydrol.2004.07.013 Gilleland, E. and Katz, R.W., 2011. New software to analyze
Brandsma, T. and Buishand, T.A., 1998. Simulation of how extremes change over time. Eos, Transactions
extreme precipitation in the Rhine basin by nearest-neigh- American Geophysical Union, 92 (2), 13–14.
bour resampling. Hydrology and Earth System Sciences, 2 Grimaldi, S. and Serinaldi, F., 2006. Asymmetric copula in
(2/3), 195–209. doi:10.5194/hess-2-195-1998 multivariate flood frequency analysis. Advances in Water
Breinl, K., Turkington, T., and Stowasser, M., 2013. Resources, 29 (8), 1155–1167. doi:10.1016/j.
Stochastic generation of multi-site daily precipitation for advwatres.2005.09.005
applications in risk management. Journal of Hydrology, Haberlandt, U., 2007. Geostatistical interpolation of hourly
498, 23–35. doi:10.1016/j.jhydrol.2013.06.015 precipitation from rain gauges and radar for a large-scale
Breinl, K., Turkington, T., and Stowasser, M., 2015. extreme rainfall event. Journal of Hydrology, 332 (1–2),
Simulating daily precipitation and temperature: a weather 144–157. doi:10.1016/j.jhydrol.2006.06.028
generation framework for assessing hydrometeorological Haberlandt, U., et al., 2011. Rainfall generators for application in
hazards. Meteorological Applications, 22 (3), 334–347. flood studies. In: A.H. Schumann, et al., eds. Flood risk assess-
doi:10.1002/met.1459. ment and management. Dordrecht: Springer, 117–147.
Brissette, F.P., Khalili, M., and Leconte, R., 2007. Efficient Hagg, W., et al., 2006. Runoff modelling in glacierized
stochastic generation of multi-site synthetic precipitation Central Asian catchments for present-day and future cli-
data. Journal of Hydrology, 345 (3–4), 121–133. mate. Nordic Hydrology, 37 (2), 93–105.
doi:10.1016/j.jhydrol.2007.06.035 Harmel, R.D., Richardson, C.W., and King, K.W., 2000.
Buishand, T.A. and Brandsma, T., 2001. Multisite simulation Hydrologic response of a small watershed model to gen-
of daily precipitation and temperature in the Rhine Basin erated precipitation. Transactions of the ASAE 43 (6),
by nearest-neighbor resampling. Water Resources 1483–1488. doi:10.13031/2013.3047
Research, 37 (11), 2761–2776. doi:10.1029/2001WR000291 Hartigan, J.A., 1975. Clustering algorithms. New York: Wiley.
Burn, D.H., 2003. The use of resampling for estimating Hartigan, J.A. and Wong, M.A., 1979. Algorithm AS 136:
confidence intervals for single site and pooled frequency a K-Means clustering algorithm. Journal of the Royal
analysis. Hydrological Sciences Journal, 48 (1), 25–38. Statistical Society. Series C (Applied Statistics), 28 (1),
doi:10.1623/hysj.48.1.25.43485 100–108.
Clark, M.P., et al., 2004. A resampling procedure for gener- Hashmi, M.Z., Shamseldin, A.Y., and Melville, B.W.,
ating conditioned daily weather sequences. Water 2011. Comparison of SDSM and LARS-WG for simula-
Resources Research, 40 (4). doi:10.1029/2003WR002747 tion and downscaling of extreme precipitation events
HYDROLOGICAL SCIENCES JOURNAL – JOURNAL DES SCIENCES HYDROLOGIQUES 1413
in a watershed. Stochastic Environmental Research and System Sciences, 15 (7), 2259–2274. doi:10.5194/hess-
Risk Assessment, 25 (4), 475–484. doi:10.1007/s00477- 15-2259-2011
010-0416-x Mehrotra, R., Srikanthan, R., and Sharma, A., 2006. A
Haylock, M.R., et al., 2008. A European daily high-resolution comparison of three stochastic multi-site precipitation
gridded data set of surface temperature and precipitation occurrence generators. Journal of Hydrology, 331 (1–2),
for 1950–2006. Journal of Geophysical Research- 280–292. doi:10.1016/j.jhydrol.2006.05.016
Atmospheres, 113 (D20), D20119. doi:10.1029/ Merz, R. and Blöschl, G., 2008. Flood frequency hydrology: 1.
2008JD010201 Temporal, spatial, and causal expansion of information.
Hofstra, N., et al., 2009. Testing E-OBS European high- Water Resources Research, 44 (8), W08432. doi:10.1029/
resolution gridded data set of daily precipitation and surface 2007WR006744
temperature. Journal of Geophysical Research-Atmospheres, Mudelsee, M., et al., 2004. Extreme floods in central Europe
114 (D21), D2110. doi:10.1029/2009JD011799 over the past 500 years: role of cyclone pathway
Holzmann, H., et al., 2010. Auswirkungen möglicher “Zugstrasse Vb’’. Journal of Geophysical Research-
Klimaänderungen auf Hochwasser und Wasserhau- Atmospheres, 109 (D23), D23101. doi:10.1029/
shaltskomponenten ausgewählter Einzugsgebiete in 2004JD005034
Österreich. Österr. Wasser- und Abfallwirtschaft, 56 (1–2/ Nash, J.E. and Sutcliffe, J.V., 1970. River flow forecasting
10), 7–14. doi:10.1007/s00506-009-0154-9 through conceptual models part I — a discussion of prin-
Hughes, J.P., Guttorp, P., and Charles, S.P., 1999. A non- ciples. Journal of Hydrology, 10 (3), 282–290. doi:10.1016/
homogeneous hidden Markov model for precipitation 0022-1694(70)90255-6
occurrence. Journal of the Royal Statistical Society: Series Papalexiou, S.M., Koutsoyiannis, D., and Makropoulos, C.,
C (Applied Statistics), 48 (1), 15–30. doi:10.1111/1467- 2013. How extreme is extreme? An assessment of daily
9876.00136 rainfall distribution tails. Hydrology and Earth System
Hundecha, Y., Pahlow, M., and Schumann, A., 2009. Sciences, 17 (2), 851–862. doi:10.5194/hess-17-851-2013
Modeling of daily precipitation at multiple locations Paquet, E., et al., 2013. The SCHADEX method: a semi-
using a mixture of distributions to characterize the continuous rainfall-runoff simulation for extreme flood
extremes. Water Resources Research, 45 (12), W12412. estimation. Journal of Hydrology, 495, 23–37.
Jenks, G., 1967. The data model concept in statistical map- doi:10.1016/j.jhydrol.2013.04.045
ping. International Yearbook of Cartography, 7, 186–190. Photiadou, C.S., Weerts, A.H., and van den Hurk, B.J.J.M.,
Katz, R.W., 1977. Precipitation as a chain-dependent process. 2011. Evaluation of two precipitation data sets for the
Journal of Applied Meteorology, 16 (7), 671–676. doi:10.1175/ Rhine River using streamflow simulations. Hydrology and
1520-0450(1977)016<0671:PAACDP>2.0.CO;2 Earth System Sciences, 15 (11), 3355–3366. doi:10.5194/
Katz, R.W., 1981. On some criteria for estimating the order hess-15-3355-2011
of a Markov-chain. Technometrics, 23 (3), 243–249. Press, W.H., et al., 2002. Numerical recipes in C++: the art of
Keef, C., et al., 2010. A multivariate model for the broad scale scientific computing. 2nd ed. Cambridge: Cambridge
spatial assessment of flood risk. In: BHS third international University Press.
symposium, managing consequences of a changing global Qian, B., Gameda, S., and Hayhoe, H., 2008. Performance of
environment, July, Newcastle. British Hydrological Society. stochastic weather generators LARS-WG and AAFC-WG
Khalili, M., Brissette, F., and Leconte, R., 2009. Stochastic for reproducing daily extremes of diverse Canadian cli-
multi-site generation of daily weather data. Stochastic mates. Climate Research, 37 (1), 17–33. doi:10.3354/
Environmental Research and Risk Assessment, 23 (6), cr00755
837–849. doi:10.1007/s00477-008-0275-x Racsko, P., Szeidl, L., and Semenov, M., 1991. A serial
King, L.M., McLeod, A.I., and Simonovic, S.P., 2014. approach to local stochastic weather models. Ecological
Simulation of historical temperatures using a multisite, Modelling, 57 (1–2), 27–41. doi:10.1016/0304-3800(91)
multivariate block resampling algorithm with perturba- 90053-4
tion. Hydrological Processes, 28 (3), 905–912. Raff, D.A., Pruitt, T., and Brekke, L.D., 2009. A framework
Leander, R. and Buishand, T.A., 2009. A daily weather gen- for assessing flood frequency based on climate projection
erator based on a two-stage resampling algorithm. Journal information. Hydrology and Earth System Sciences, 13 (11),
of Hydrology, 374 (3–4), 185–195. doi:10.1016/j. 2119–2136. doi:10.5194/hess-13-2119-2009
jhydrol.2009.06.010 Refsgaard, J.C., Storm, B., and Clausen, T., 2010. Systeme
Li, Z., Brissette, F., and Chen, J., 2013. Finding the most Hydrologique Europeen (SHE): review and perspectives
appropriate precipitation probability distribution for sto- after 30 years development in distributed physically-based
chastic weather generation and hydrological modelling in hydrological modelling. Hydrology Research, 41 (5), 355–377.
Nordic watersheds. Hydrological Processes, 27, 3718–3729. Reiss, R.-D. and Thomas, M., 2007. Statistical analysis of
doi:10.1002/hyp.v27.25 extreme values. Basel: Birkhäuser.
Lucio, P.S., Silva, A.M., and Serrano, A.I., 2010. Changes in Richardson, C.W., 1981. Stochastic simulation of daily pre-
occurrences of temperature extremes in continental cipitation, temperature, and solar-radiation. Water
Portugal: a stochastic approach. Meteorological Applications, Resources Research, 17 (1), 182–190. doi:10.1029/
17 (4), 404–418. WR017i001p00182
Ly, S., Charles, C., and Degré, A., 2011. Geostatistical Romanowicz, A.A., et al., 2005. Sensitivity of the SWAT model
interpolation of daily rainfall at catchment scale: the to the soil and land use data parametrisation: a case study in
use of several variogram models in the Ourthe and the Thyle catchment, Belgium. Ecological Modelling, 187
Ambleve catchments, Belgium. Hydrology and Earth (1), 27–39. doi:10.1016/j.ecolmodel.2005.01.025
1414 K. BREINL
Schädler, G., et al., 2012. Flood hazards in a changing climate. Texas under Climate Change, with Implications for Water
Karlsruhe: Center for Disaster Management and Risk Resources Planning and Management”. Texas A & M
Reduction Technology (CEDIM). University, Department of Biological and Agricultural
Schwarz, G., 1978. Estimating the dimension of a model. The Engineering.
Annals of Statistics, 6 (2), 461–464. doi:10.1214/aos/ Steele-Dunne, S., et al., 2008. The impacts of climate change
1176344136 on hydrology in Ireland. Journal of Hydrology, 356 (1–2),
Seibert, J., 1997. Estimation of parameter uncertainty in the 28–45. doi:10.1016/j.jhydrol.2008.03.025
HBV model. Nordic Hydrology, 28 (4–5), 247–262. Turkington, T., et al., 2014. Empirical atmospheric
Seibert, J., 1999. Regionalisation of parameters for a conceptual thresholds for debris flows and flash floods in the
rainfall-runoff model. Agricultural and Forest Meteorology, Southern French Alps. Natural Hazards Earth Systems
98–99, 279–293. doi:10.1016/S0168-1923(99)00105-7 Sciences Discussions, 2 (1), 757–798. doi:10.5194/
Seibert, J., 2000. Multi-criteria calibration of a conceptual nhessd-2-757-2014
runoff model using a genetic algorithm. Hydrology and Uhlenbrook, S., et al., 1999. Prediction uncertainty of con-
Earth System Sciences, 4 (2), 215–224. doi:10.5194/hess-4- ceptual rainfall–runoff models caused by problems in
215-2000 identifying model parameters and structure. Hydrological
Seibert, J. and Vis, M.J.P., 2012. Teaching hydrological mod- Sciences Journal, 44 (5), 779–797. doi:10.1080/
eling with a user-friendly catchment-runoff-model soft- 02626669909492273
ware package. Hydrology and Earth System Sciences, 16 Vlček, O. and Huth, R., 2009. Is daily precipitation Gamma-
(9), 3315–3325. doi:10.5194/hess-16-3315-2012 distributed? Adverse effects of an incorrect use of the
Semenov, M.A. and Barrow, E.M., 1997. Use of a stochastic Kolmogorov-Smirnov test. Atmospheric Research, 93 (4),
weather generator in the development of climate change 759–766. doi:10.1016/j.atmosres.2009.03.005
scenarios. Climatic Change, 35 (4), 397–414. doi:10.1023/ Vrac, M. and Naveau, P., 2007. Stochastic downscaling of
A:1005342632279 precipitation: from dry events to heavy rainfalls. Water
Semenov, M.A., et al., 1998. Comparison of the WGEN and Resources Research, 43, 7. doi:10.1029/2006WR005308
LARS-WG stochastic weather generators for diverse climates. Wilby, R.L., Tomlinson, O.J., and Dawson, C.W., 2003.
Climate Research, 10 (2), 95–107. doi:10.3354/cr010095 Multi-site simulation of precipitation by conditional
Serinaldi, F., 2009a. Copula-based mixed models for bivariate resampling. Climate Research, 23 (3), 183–194.
rainfall data: an empirical study in regression perspective. doi:10.3354/cr023183
Stochastic Environmental Research and Risk Assessment, 23 Wilks, D.S., 1998. Multisite generalization of a daily stochas-
(5), 677–693. doi:10.1007/s00477-008-0249-z tic precipitation generation model. Journal of Hydrology,
Serinaldi, F., 2009b. A multisite daily rainfall generator 210 (1–4), 178–191. doi:10.1016/S0022-1694(98)00186-3
driven by bivariate copula-based mixed distributions. Wilks, D.S., 1999a. Interannual variability and extreme-value
Journal of Geophysical Research-Atmospheres, 114, 10. characteristics of several stochastic daily precipitation
doi:10.1029/2008JD011258 models. Agricultural and Forest Meteorology, 93 (3), 153–
Shrestha, R., Tachikawa, Y., and Takara, K., 2006. Input data 169. doi:10.1016/S0168-1923(98)00125-7
resolution analysis for distributed hydrological modeling. Wilks, D.S., 1999b. Multisite downscaling of daily precipita-
Journal of Hydrology, 319 (1–4), 36–50. doi:10.1016/j. tion with a stochastic weather generator. Climate Research,
jhydrol.2005.04.025 11 (2), 125–136. doi:10.3354/cr011125
Singh, V.P. and Mishra, A.K., 2011. Report as of FY2010 for Zambrano-Bigiarini, M., 2012. “hydroTSM” - R package
2009TX334G: “Hydrological Drought Characterization for version 0.4-2-1.