Professional Documents
Culture Documents
available at www.sciencedirect.com
a
Department of Civil Engineering, University of Castilla-La Mancha, Spain
b
Department of Civil Engineering and Engineering Mechanics, and Center for Sustainability of Semi-Arid Hydrology and
Riparian Areas (SAHRA), University of Arizona, USA
Received 12 June 2007; received in revised form 5 September 2007; accepted 14 September 2007
KEYWORDS Summary Using rain-gauge station records for the statistical characterization and simu-
Regional frequency lation modeling of spatio-temporal precipitation field involves many issues and simplistic
analysis; assumptions. One major issue is related to dealing with uncertainty at-site sample statis-
Orographic tical inference, because of the limited length of records. Regional frequency analysis uses
precipitation; the idea of substituting space for time in order to reduce uncertainty. It assumes equal
Smoothing shapes of the precipitation statistical distributions in a region. However, this assumption
regionalization; limits the area of the analyzed region where this assumption is valid. The extension is
Precipitation dependent on terrain complexity.
geostatistic; This work presents a new approach for the statistical regionalization of a large precipitation
Precipitation simulation field, replacing the shape constancy assumption for the hypothesis of smooth spatial varia-
tion. The approach accounts for every uncertainty on site information, using an L-moment
method for inference analysis. Additionally, the orographic effect is introduced in the region-
alization, which substantially improves the interpolation performance and estimation of areal
precipitation. The approach is used for modeling the monthly precipitation field in the Júcar
River Basin Authority Demarcation (Spain), incorporating its stochastic structure, and spatial
dependency coming from a geostatistical analysis. Issues related to the estimation of regional
precipitation, and mean areal precipitation are discussed in the exposition.
ª 2007 Elsevier B.V. All rights reserved.
Introduction
* Corresponding author. Tel.: +34 926295422; fax: +34 926295391.
E-mail addresses: Javier.Gonzalez@uclm.es (J. González), Regionalization is the inclusion in frequency analysis of
jvaldes@u.arizona.edu (J.B. Valdés). data from sites other than the site at which statistic
0022-1694/$ - see front matter ª 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.jhydrol.2007.09.059
28 J. González, J.B. Valdés
characterization is required. The method assumes that the Sinclair, 1986). Acreman and Wiltshire (1989) noted that
frequency distributions of other sites are similar to the site the explicit construction of a region is not necessary in a
of interest. Most of the statistical regionalization applica- fractional-membership procedure. This lead to the concept
tions are concerned with reducing the uncertainty of ex- of region of influence, proposed by Burn (1990), where the
treme quantile estimation. The index-flood procedure of weights must account for the influence of each site in the
Dalrymple (1960) is an early example. A popular reference estimation of parameters or quantiles for any particular
in applying regionalization methods for flood flow frequency ‘‘site of interest’’. The largest disadvantage of the method
analysis in the United States is the Bulletin 17B of the United is in the weight definition, where no universal definition ex-
States Water Resources Council (US Water Resources Coun- ists. Another related approach is the mapping procedure,
cil, 1976, 1977, 1981). The assumption of ‘‘homogeneous which is effective when a smooth relation to site character-
region’’ was suggested as invalid for US streamflow data istics can be found. Parameters or quantiles arising from re-
(Benson, 1962). Thus there is reason to doubt whether gional frequency analysis are mapped or plotted against
regionalization is worthwhile. However, research has shown typical site characteristics in each region (Schaefer, 1990;
that even though a region may be moderately heteroge- Fill, 1994). A mapping approach can also be used with at-
neous, regional analysis will still yield much more accurate site estimates (McKerchar and Pearson, 1990). The major
quantile estimates than at-site analysis (Lettenmaier and disadvantage when compared with a simpler regionalization
Potter, 1985; Lettenmaier et al., 1987; Hosking and Wallis, procedure, such as the index-flood procedure, is the diffi-
1988; Potter and Lettenmaier, 1990). culty in estimating the accuracy of the final quantile
The effect of serial cross-dependence on frequency anal- estimates.
ysis has been investigated by Landwehr et al. (1979) and The different procedures proposed to overcome region-
MacMahon and Srikanthan (1982). They found that serial alization limitations seek to avoid the idea of regions de-
cross-dependence causes a small amount of bias and a small fined by borders where abrupt changes occur in favor of a
increase in the standard error of quantile estimates. How- smooth continuity between regions. The real world pro-
ever, a small amount of serial cross-dependence in data ser- duces a continuous field, where uniformity may be a suffi-
ies has little effect on the quality of quantile estimates cient approximation in some places. In this work a new
(Hosking and Wallis, 1997). Therefore, regionalization is regionalization approach is presented in which the uniform
considered a powerful tool for improving the quality of sta- assumption is replaced by a more realistic assumption when
tistical characterization and quantile estimation. The major large areas are analyzed: smooth spatial variation in the fre-
hypothesis is the uniformity in the frequency distribution quency distribution parameters. The different L-moments
parameters. The serial cross-dependence may limit the coefficients which define the frequency distribution func-
ability to reduce uncertainty by increasing spatial tion at each site are approximated by surfaces, which are
information. fitted from at-site estimations, taking into account the sam-
The uniformity and independence required in a regional- ple estimation uncertainty and the orography factor. This
ization process limit the applicability of the method. Unifor- provides a regional approximation of a random field distri-
mity restricts the maximum area covered by the database: bution. The approach is used to model monthly precipita-
distant sites are less probable to be similar in their fre- tions in the Júcar River Basin Authority Demarcation
quency distributions. The second assumption, however, (Spain), producing a spatio-temporal model for spatial pre-
may require distant sites to be serially independent and to cipitation simulation.
add information to regional statistical analysis. The space
to meet these two restrictions depends much more on the The precipitation database
nature and type of variable to be analyzed (e.g. streamflow
or precipitation depth, and monthly or annual maximum), The Júcar River Basin Authority Demarcation is located in
and often, on the terrain complexity (e.g. complex orogra- the east of the Iberian Peninsula, along the Spanish Mediter-
phy, geology, and soil uses distribution introduces spatial ranean coast. It comprises a vast area of 42,989 km2, and it
heterogeneity in the hydrological processes). is composed of the aggregation of several river basins with
When large study areas are to be analyzed, the regional- outflow to the Mediterranean Sea: Cenia, Mijares, Palancia,
ization approaches fail to achieve spatial homogeneity. Turia, Júcar, Serpis, and Vinalopó. The principal mountain-
Uniform values of dispersion or shape frequency distribution ous system is the Iberian System, in the northwest of the
coefficients are not to be expected, and applying regionali- demarcation, where the maximum altitude of 2013 m above
zation for the whole area is not valid. Several approaches the mean sea level is achieved. The terrain provides high
may be found in the literature to overcome this issue. orographic complexity in the Iberian System and lower com-
Fiorentino et al. (1987) and Gabriele and Arnell (1991) pro- plexity in the Betic Mountainous System, which comprises
posed a procedure that involved a hierarchy of regions. part of the south basin borders (see Fig. 1).
Relatively large regions are defined over which the shape The Iberian System plays an important role in the atmo-
parameters are assumed to be constant, and these regions spheric dynamics in the region. It is exposed to the Mediter-
are subdivided into smaller regions over which the disper- ranean (east) and Atlantic (west) fronts. The mountain
sion parameter is assumed to be constant. Hierarchical system provides a barrier so that few Atlantic fronts reach
regions method may produce abrupt changes in the param- the Mediterranean coast, excepted in the south (Rodrı́-
eters when passing from one site to a neighboring site. guez-Puebla et al., 1998). Convection processes from the
Fractional-membership procedures considers a site to Mediterranean Sea and topographical configuration charac-
provide information on several regions, rather than belong- terize the rainfall distribution over the coast (Millán
ing to a particular region (Wiltshire, 1986; Acreman and et al., 1995).
A regional monthly precipitation simulation model based on an L-moment smoothed statistical regionalization approach 29
4,500,000 2000
1800
1600
4,400,000
1400
Latitude (UTM)
1200
1000
4,300,000
800
600
4,200,000 400
200
Figure 1 Digital elevation model for the Júcar River Basin Authority Demarcation (thick magenta line), and the rain-gauge stations
distribution (black points). The study area has been divided into hydro-homogeneous regions (thin magenta lines). Elevation is
expressed in meters above the mean sea level.
The precipitation database comprises the historical re- the different regions show high variability, depending on
cords of 1583 stations. It includes monthly precipitation the rainfall front that reaches each region, and produced
depths, and extends from 1856 to 2006. During this period by local effects induced by the topographic configuration
the number of active stations has been changing. Only a and sea coast proximity. The orographic effect in the flat re-
few stations were constructed before 1910, and the pace gions is not so clearly evident in the figure because of the
increased after that date until 1940, when almost one higher precipitation heterogeneity among flat regions.
hundred stations were active. Then increased interest in There the orographic effect is also presented, however
improving the spatial precipitation information produced a due to the lower differences in altitude within each flat re-
progressive increase in the number of stations until 1990, gion, it has lower importance than other local factors.
when almost 750 stations were active. Thus, the Júcar Demarcation database is characterized
No uniformity was found in the spatial distribution of by irregular spatial and temporal sampling of the rainfall
rain-gauge stations. A higher station density is presented distribution. High variability in the mean annual precipita-
near to the coast (Fig. 1). Fig. 2 shows the 100 · 100 m2 tion depth is present, with local precipitation ranging from
DEM empirical cumulative probability distribution function less than 200 mm to greater than 1000 mm. A regional sta-
(c.d.f.), which represents a bimodal distribution composed tistical analysis of the monthly precipitation distributions
by the mountainous regions (i.e. higher than 700 m mean is required to account for the whole data set, in order to ob-
sea level) and the flat regions (i.e. lower than 700 m mean tain a better spatial description. However, because of the
sea level). The empirical c.d.f. of the stations elevation is high variability of at-site precipitation, the sample uncer-
also plotted. Comparing both curves, a higher station spatial tainty must be quantified and accounted for.
density around the coast is noticed. The flat region, which
occupies 20% of the study area, contains more than 50% of
the rain-gauge stations. However, a very low spatial density The L-moment smoothed statistical
exits along the mountainous regions, where only a few regionalization approach
stations characterize the precipitation regime in higher
elevations. These regions are of great interest for water re- Statistical regionalization looks to improve at-site statistical
sources evaluation. characterization by incorporating spatial data. Several
When comparing mean annual precipitation depths methods have been described above, including regional
against elevations, two effects are observed (see Fig. 3). shape estimation and index-flood to hierarchical regions,
On one hand, orographic effects are shown in the precipita- fractional-membership, region of influence, mapping, or
tion distribution, noted specially in the mountainous re- Bulletin 17. However, none of these methods provide a uni-
gions, with an increasing precipitation trend for higher versal approach to produce a smooth continuous spatial
altitudes. Alternatively, annual precipitation depths along field of the statistical distributions. The application of these
30 J. González, J.B. Valdés
0.9
0.8
0.7
Empirical c.d.f. F(x)
0.6
DEM
0.5
Rain—gauge stations
0.4
0.3
0.2
0.1
0
0 200 400 600 800 1000 1200 1400 1600 1800 2000
Elevation (m)
Figure 2 Empirical cumulative probability distribution function for the 100 · 100 m2 DEM and the rain-gauge stations elevation in
the Júcar River Basin Authority Demarcation.
1400
1200
Mean Annual Precipitation Depth (mm)
1000
800
600
400
200
0
0 200 400 600 800 1000 1200 1400 1600 1800
Elevation (m)
Figure 3 Mean Annual Precipitation Depth again the rain-gauge station elevation, taking from a 100 · 100 m2 DEM in the Júcar
River Basin Authority Demarcation.
methods to the monthly precipitation distribution in the Jú- itly obtains the continuous spatial field of the statistical dis-
car Demarcation is inappropriate because of its complexity. tributions of the regional precipitation. The goal is not only
Below, a new approach is described for the statistical to characterize each site’s statistical distribution, but to
regionalization based on L-moments. This approach explic- analyze the spatial statistical distribution of monthly pre-
A regional monthly precipitation simulation model based on an L-moment smoothed statistical regionalization approach 31
Spatial L-moments distribution by smoothing Here, a cubic smoothing spline was used to fit each sta-
surface fitting tistic, (k1 and s), which requires a regular grid of observed
points. Since rain-gauge stations do not follow a regular
After estimating the parameters of the distribution of the grid, the spatial information was translated to a designed
sample statistics (Eq. (7)) for the set of Ns rain-gauge sta- regular grid. A quadrangular 10 km by 10 km grid was se-
tions (Ns = 1583) for every month, corresponding surfaces lected. The value of the statistic in each node in the grid
must be fitted to represent the spatial distributions. The was estimated by Eq. (9), assigning more importance to
process must account for a smooth spatial variation of nearest stations by the inverse square distance procedure
the statistics (i.e. low surface roughness), while simulta- and accounting for the uncertainty. Consequently, the
neously the resulting field must be likely in the set of sta- weight for smoothing spline fitness was evaluated by Eq.
tistics distributions coming from the inference process at (10).The weights are inversely related to the square dis-
each site (i.e. ‘1 2 N(k1, rl), and t 2 N(s, rt)). Thus the fit- tance to the stations, and their standard deviation
ting approach must adapt to the spatial variability at the PNs vðkÞ
k¼1 h i2
same time that it smooths sample errors. Additionally, ðkÞ
d ði;jÞ sv ðkÞ
the orographic effect must be able to be included if vði; jÞ ¼ PNs ð9Þ
1
significant. k¼1 h i2
ðkÞ
Kriging methods are nowadays a preferable option in the d ði;jÞ sv ðkÞ
the distance and precipitation heterogeneity. A suitable res- to the smoothing parameter, p, the slope q‘1 in Eq. (12)
olution in the regionalization may filter these local effects plays a role. Thus, both variables must be adjusted simulta-
and provide a better estimation of the area average precip- neously. Fig. 4 presents the resulting values of q‘1 , and its
itation. The filter effect in the regionalization is controlled, seasonal behavior. The largest orographic effects are pro-
in this case, by the grid resolution for smoothing spline duced in May and June, when the impacts are not negligible.
fitness. The mean annual slope implies an increment in the annual
The objective in this case was to characterize the precip- precipitation depth of 116 mm/km of elevation in a given
itation for water resources management. Therefore, a grid area.
of 10 km by 10 km was considered suitable, given the objec- Fig. 5 shows the map of mean monthly precipitation in
tive and the expected spatial heterogeneity in the region. April after smoothing regionalization for the area of study.
Three variables were used to characterize the statistical That month shows a high orographic effect in precipitation
distribution of monthly precipitation, ‘1 and t for nonzero distribution, with the impact in the map.
precipitation, and P0 for zero precipitation.
L-CV nonzero monthly precipitation fitness
Mean nonzero monthly precipitation fitting Traditional regionalization methods often hypotherize that
Fig. 3 shows the orographic effect in the mean annual pre- the L-CV coefficient is constant. However in this work, this
cipitation. Similar figures were obtained when analyzing hypothesis is replaced by a continuous smooth variation of
for each month the average precipitation against elevation. the statistic in the region of interest. After computing the
However, the orographic effect in the precipitation is sea- standard deviation of the estimation of t at each station
sonally important, a function of the kind of fronts that pro- and month, using Eq. (6), it was checked that the hypothesis
duce rainfall. of uniformity of the L-CV coefficient in each month is unex-
To account for the orographic effect in the mean non- pected. No value exceeds the 99% confidence intervals of
zero monthly precipitation the next model is proposed the distribution of ts for the stations.
Therefore, a smoothed spline surface was fitted each
‘1 ðx; y; zÞ ¼ ‘01 ðx; yÞ þ q‘1 z ð12Þ
month for the region. In the case of the t statistic, no oro-
where ‘01 ðx; yÞ
is the estimated mean precipitation in the graphic effect was found to be significant in any month, so
point (x, y), considering no orographic effect (i.e. z = 0), only the q parameter was necessary to maximize the likeli-
which is taken into account by a linear function of the hood given by Eq. (11) for each month. Fig. 6 shows the
elevation z. resulting distribution of t in April for the Júcar Demarcation.
Eq. (12) separates the orographic effect from the local The smoothness of the statistic is much more important
effect. To maximize the likelihood (Eq. (11)), in addition than in the case of ‘1 (Fig. 5).
20
18
16
14
12
ρ (mm/km)
10
0
Oct Nov Dec Jan Feb Mar Apr May Jun Jul Aug Sep
Month
Figure 4 Seasonal orographic effect in the mean nonzero monthly precipitation, evaluated by the slope q‘1 (bars), and annual
mean slope (line).
34 J. González, J.B. Valdés
80
4,500,000
70
Latitude (UTM)
4,400,000 60
50
4,300,000
40
4,200,000 30
Figure 5 Spatial distribution of the mean (‘1) precipitation depths (mm) in April, in the Júcar Demarcation, by smoothing
regionalization, taking into account the orographic effect.
4,500,000 0.5
0.45
Latitude (UTM)
4,400,000
0.4
4,300,000
0.35
4,200,000
0.3
500,000 600,000 700,000 800,000
Longitude (UTM)
Figure 6 Spatial distribution of the coefficient of L-variation (t) of the precipitation depths in April, in the Júcar Demarcation, by
smoothing regionalization.
Zero monthly precipitation probability fitness those events, and is estimated by dividing the number of
The final statistics to define the monthly precipitation fre- zero precipitation events (m) by the total number of obser-
quency distribution function along the study area is the vations (M) at a station. The distribution of the estimation
probability of zero precipitation P0. The event of producing depends on both values, m and M, but it does not follow a
zero precipitation during a month, with probability P0, fol- normal distribution. Instead, its distribution is related to
lows a binomial distribution. P0 provides the frequency of the Fisher distribution (Johnson et al., 1992). Therefore,
A regional monthly precipitation simulation model based on an L-moment smoothed statistical regionalization approach 35
the approach of smoothing regionalization approach pre- the month-sites that reject the hypothesis was accepted
sented here is not applicable. In this work, the estimation uniform, and it does not concentrate in a group of months.
of P0 computed at each stations was considered directly. Additionally, the spatial distribution of the stations is dis-
The larger spatial variability of P0 is produced during perse. Thus the hypothesis of stationarity in the monthly
summer months (with lower precipitation depths), when autocorrelation coefficient with lag = 1 was considered,
small-extension storms introduce larger spatial variability and the stationary r(St)s were computed.
with respect to humid months. Thus, the impact of not com- Smooth spatial variation in the qs is also expected, much
puting regionalization of P0 is small for the objectives of the like it occurs with L-moment statistics. The hypothesis of
study. uniform q over the region of study was not accepted, com-
paring the 95% confidence intervals of the computed qs at
A spatio-temporal simulation model every site. Therefore, a smooth spline surface was fitted
for a smooth regionalization of r. The fitting was performed
The spatial L-moments distribution statistically character- over the transformed Z, taking into account the correspond-
izes the monthly precipitation in the Júcar Demarcation. ing variance at each site, following the above procedure
This characterization is used here to fit a simulation model, (Section ‘‘Smoothing spline surface fitness’’). An orographic
which allows the generation of synthetic monthly precipita- effect was observed for q and Z, and the next model will
tion traces that are consistent with fitted frequency take this into account
distributions. In addition, the model produces likely Zðx; y; zÞ ¼ Z0 ðx; yÞ þ qZ z ð14Þ
spatio-temporal distributions. The temporal and spatial
0
structure of the data is analyzed and modeled next. where Z (x, y) is the estimated Z-transformed monthly auto-
Before starting the analysis, and in order to compute cor- correlation coefficient in the point (x, y), considering no
relation coefficients to analyze temporal and spatial rela- orographic effect (i.e. z = 0), which is taken into account
tionships, the data sets for every month and site have by a linear function of the elevation z. The slope qZ takes
been transformed to a normal distribution. The normaliza- a value of 0.0073 1/km. Fig. 7 shows the spatial distribution
tion was performed using the statistical distribution in each of r after smooth regionalization.
month and site. Each normalized value corresponds to the The largest monthly autocorrelation coefficient is pro-
N(0, 1) variable which cumulative probability distribution duced in the north, with r 0.2, along the Iberian Moun-
function (c.d.f.) coincides with the c.d.f. of the monthly tainous System. No significant autocorrelation is produced
precipitation. These rainfall c.d.f. distributions are given over the southern mountains, because of their smaller size
in Eq. (1). In order to apply the transformation in the lower and proximity to the coast.
tail of the distribution, the c.d.f of zero precipitation was The monthly autocorrelation coefficients regional distri-
considered P0/2 instead of P0. This reduces the truncation bution may be used to model the stochastic structure of
effect in the normalized distribution. With the normalized monthly precipitation in the area, considering its temporal
data set, the stochastic structure was analyzed first, dependence. An autoregressive stochastic model AR(1)
followed by the spatial dependency. was considered to define the model (Box and Pierce, 1976):
X tþ1 ¼ / X t þ atþ1 ð15Þ
The stochastic structure
where Xt would correspond to the normalized monthly pre-
cipitation series in each station, / is considered equal to r,
The stochastic analysis of the normalized data set sought to
and at is an independent random variable, which follows a
define the temporal dependency of the monthly precipita-
normal distribution, a 2 Nð0; r2a Þ, with r2a ¼ ð1 r2 Þ r2X .
tion. First, the Pearson correlation coefficient was com-
The normal assumption of at is checked for every station,
puted for every site, and between each two consecutive
in addition to its independence. The series at is called the
months: (Month, Month + 1) ! r(Month, St); where St repre-
shock series, because it represents the fluctuation effect
sent the station index. Also, to analyze the seasonal behav-
in the model. Thus, the model defined with Eq. (15) repre-
ior in r, the confidence interval of the estimation was
sents the stochastic structure in the monthly precipitations.
computed. The confidence interval of r is obtained by the
Fisher Z-transformation (Fisher, 1915)
Spatial dependency model
1 1þr
Z ¼ ln ð13Þ In order to produce a simulation model of monthly precipi-
2 1r
tation in the Júcar Demarcation, in addition to the stochas-
For the transformed Z, the approximate variance r2(z) = 1/ tic structure it is necessary to account for the spatial
(n 3) is independent of the correlation, and only depends dependence of the at-site simulations. This spatial depen-
on the sample size n. Furthermore, even the distribution of dence is evaluated with the cross-correlation between two
Z is not strictly normal; it converges rapidly to normal as the stations. The variable used for analyzing this spatial depen-
sample size increases for any values of q. dency is the residual of the stochastic model at each sta-
Knowing not only the estimated rs, but also their tion, or shock time series at. Using this variable the
distributions, it is possible to check the hypothesis of sta- temporal dependence model is disaggregated from the spa-
tionary q (null hypothesis, H0), against seasonal q (alterna- tial dependence model, thereby simplifying the simulator
tive hypothesis, H1). Therefore, with a test significance model. This assumes that the shocks follow a correlated
a = 0.05 it was accepted the null hypothesis in a percentage spatial structure that are cross-correlated in the at-site
of month-sites lower than 10%. The temporal distribution of monthly precipitation series.
36 J. González, J.B. Valdés
0.2
0.18
4,500,000
0.16
0.14
Latitude (UTM)
4,400,000 0.12
0.1
0.08
4,300,000
0.06
0.04
4,200,000 0.02
Figure 7 Spatial distribution of the stationary monthly coefficient of autocorrelation (r) of the precipitation depths in the Júcar
Demarcation, by smoothing regionalization.
The hypothesis of stationary, (no seasonal) cross-correla- using the Fisher Z-transformation. The number of terms in
tion between each pair of stations was evaluated using the the Fourier series, n, was chosen following the F-stopping-
5% confidence intervals computed by the Fisher Z-transfor- criteria, i.e. sequentially increasing the number of terms
mation (Eq. (13)). Assuming stationarity, the spatial depen- until the decrease in the mean square error is not significant
dence was analyzed by geostatistical procedures. The with respect the decrease in the degrees of freedom.
objective was to model this dependency by fitting an Resulting coefficients are presented in Table 1.
expression for the correlogram. The distance was found to The obtained correlogram produces a uniform spatial
be a significant variable in the correlogram definition. As dependence model over the region, which represents the
distance increases, the cross-correlation between shock cross-correlation between stations. This is used in the sim-
series decreases, with exponential and gaussian models ulation model, which is validated below.
being the best options for representing the behavior (Deu-
tsch and Journel, 1997). Additionally, anisotropy effect Model validation
was found in the correlogram, so the orientation between
the two sites is relevant in the cross-correlation coefficient. The monthly precipitation smooth regionalization of the
The best model considered was a composite model, frequency distributions and the lag = 1 autocorrelation coef-
which combined an exponential and a gaussian decay of ficients, in addition to the spatial dependency model, pro-
the correlation coefficient C vide the components to simulate monthly precipitation
h2
time series in the Júcar Demarcation. The steps for N-years
h
AðhÞ
Cðh; hÞ ¼ e e BðhÞ2 ð16Þ simulation at a set of stations are:
where h is the distance, h 2 [0, p) is the azimut direction,
• Compute 12 · N random set of values at each station,
and A(h) and B(h)2 are the decay coefficients functions of
following a normal standard distribution, in accor-
the exponential model and the gaussian model, respec-
dance with the cross-correlation matrix produced by
tively. These decay coefficients functions were expressed
Eq. (16).
in the form of Fourier series
• Transform the time series at each station to Nð0; r2a Þ,
X
n with the r2a corresponding to the station, by ra
AðhÞ ¼ a0 þ ½ai21 sinð2 i hÞ þ ai2 cosð2 i hÞ ð17Þ multiplication.
i¼1 • Incorporate the AR1 stochastic structure (Eq. (15)), and
Xn
compute the normalized monthly precipitation time ser-
BðhÞ ¼ b0 þ ½bi21 sinð2 i hÞ þ bi2 cosð2 i hÞ ð18Þ
i¼1
ies at each station.
• Produce the monthly precipitation time series at each
The coefficient ai and bi, i = 1, . . . , n, were fitted to meet station by inverting the normalizing transformation,
the maximum likelihood criteria, using the statistical accounting for the frequency distribution at each sta-
distribution of the sampled cross-correlation coefficients, tion, given by the smooth regionalization (Eq. (8)).
A regional monthly precipitation simulation model based on an L-moment smoothed statistical regionalization approach 37
5.1E15
4.0E9
tains dependency properties for simultaneous simulations in
a set of stations, and can simulate at sites with no previous
16
observations. In order to validate the model it is appropriate
to check the performance over areal averaging precipita-
5.5E15
1.7E9
tions because the model has been fitted from at-site data.
15
The regionalization used in the fitting procedure is expected
to produce good spatial representation of the precipitation
4.9E15
1.0E8
realizations. However, samples are only derived from at-
14
site precipitation. The estimation of areal averaging precip-
itation from at-site data may be computed by the Thiessen
9.7E15
polygon method (Thiessen, 1911), or an inverse square dis-
5.3E9
tion coming from the sample data, and the simulated data
1.1E14
7.0E9
next steps:
1.0E14
4.8E9
8.6E15
8.2E9
age precipitations.
3.7E8
for the full period, and this would impact in the areal
average, only the simulated data corresponding with
3
ing from the sample size and spatial structure, and the
38 J. González, J.B. Valdés
Dalrymple, T., 1960. Flood frequency analyses. Water Supply Paper, Lettenmaier, D.P., Wallis, J.R., Wood, E.F., 1987. Effect of regional
1543–A, US Geological Survey, Reston, VA. heterogeneity on flood frequency estimation. Water Resour.
de Boor, C., 1978. A Practical Guide to Splines. Springer-Verlag. Res. 23, 313–323.
Deutsch, C., Journel, A., 1997. Geostatistical Software Library and MacKee, T.B., Doesken, N.J., Kleist, J., 1993. Drought monitoring
Userś Guide, second ed. Oxford University Press, New York. with multiple timescales. In: Eighth Conference on Applied
Fill, H.D., 1994. Improving Flood Quantile Estimates Using Regional Climatology, Anaheim, Calif. Am. Meteorol. Soc.
Information, Ph.D. Thesis, Cornell University, Ithaca, NY. MacMahon, T.A., Srikanthan, R., 1982. Log pearson type 3 distribu-
Fiorentino, M., Gabriele, S., Rossi, F., Versace, P., 1987. Hierar- tion effect of dependence, distribution parameters and sample
chical approach for regional flood frequency analysis. In: Singh, size on peak annual flood estimates. J. Hydrol. 52, 815–826.
V.P. (Ed.), Regional Flood Frequency Analysis. D. Reidel, Martı́nez-Cob, A., 1996. Multivariate geostatistical analysis of
Norwell, Mass, pp. 35–49. evapotranspiration and precipitation in mountainous terrain. J.
Fisher, R.A., 1915. Frequency distribution of the values of the Hydrol. 174, 19–35.
correlation coefficient in samples of an indefinitely large McKerchar, A.I., Pearson, C.P., 1990. Maps of flood statistics for
population. Biometrika, 507–521. regional flood frequency analysis in New Zealand. Hydrol. Sci. J.
Gabriele, S., Arnell, N., 1991. A hierarchical approach to regional 35, 609–621.
flood frequency analysis. Water Resour. Res. 27, 1281–1289. Millán, M., Estrela, M.J., Caselles, V., 1995. Torrential precipita-
Goovaerts, P., 2000. Geostatistical approaches for incorporating tions on the Spanish east coast: the role of the mediterranean
elevation into the spatial interpolation of rainfall. J. Hdrol. 228, sea surface temperature. Atmos. Res. 36, 1–16.
113–129. Phillips, D.L., Dolph, J., Marks, D., 1992. A comparison of
Greenwood, J.A., Landwehr, J.M., Matalas, N.C., Wallis, J.R., geostatistical procedures for spatial analysis of precipitations
1979. Probability weighted moments: definition and relation to in mountainous terrain. Agr. Forest. Meteorol. 58, 119–141.
parameters of several distributions expressable in inverse form. Potter, K.W., Lettenmaier, D.P., 1990. A comparison of regional
Water Resour. Res. 15, 1049–1054. flood frequency estimation methods using a resampling method.
Haberlandt, U., Klöcking, B., Krysanova, V., Becker, A., 2001. Water Resour. Res. 26, 415–424.
Regionalisation of the base flow index from dynamically simu- Raspa, G., Tucci, M., Bruno, R., 1997. Reconstruction of rainfall
lated flow components – a case study in the Elbe River basin. J. fields by combining ground raingauges data with radar maps
Hydrol. 248, 35–53. using external drift method. In: Baafi, E.Y., Schofield, N.A.
Hevesi, J.A., Flint, A.L., Istok, J.D., 1992. Precipitation estimation (Eds.), Geostatistics Wollongong’96. Kluwer Academic, Dordr-
in mountainous terrain using multivariate geostatistics. Part i: echt, pp. 941–950.
structural analysis. J. Appl. Meteorol. 31, 661–676. Rodrı́guez-Puebla, C., Encinas, A.H., Nieto, S., Garmendia, J., 1998.
Holawe, F., Dutter, R., 1999. Geostatistical study of precipitation Spatial and temporal patterns of annual precipitation variability
series in Austria: time and space. J. Hydrol. 219, 70–82. over the Iberian Peninsula. Int. J. Climatol. 18, 299–316.
Hosking, J.R.M., Wallis, J.R., 1988. The effect of intersite depen- Schaefer, M.G., 1990. Regional analyses of precipitation annual
dence on regional flood frequency analysis. Water Resour. Res. maxima in Washington State. Water Resour. Res. 26, 119–131.
24, 588–600. Thiessen, A.H., 1911. Precipitation averages for large areas. Mon.
Hosking, J.R.M., Wallis, J.R., 1997. Regional Frequency Analysis: An Weather Rev. 39 (7), 1082–1084.
Approach Based on L-Moments. Cambridge University Press, NY, US Water Resources Council, 1976. Guidelines for determining flood
USA. flow frequency. Bulletin 17, Hydrology Committee, Washington,
Johnson, N.L., Kotz, S., Kemp, A.W., 1992. Univariate Discrete DC.
Distributions, second ed. Wiley. US Water Resources Council, 1977. Guidelines for determining flood
Landwehr, J.M., Malatas, N.C., Wallis, J.R., 1979. Probability- flow frequency. Bulletin 17A, Hydrology Committee, Washing-
weighted moments compared with some traditional techniques ton, DC.
in estimating gumbel parameters and quantiles. Water Resour. US Water Resources Council, 1981. Guidelines for determining flood
Res. 15, 1055–1064. flow frequency. Bulletin 17B, Hydrology Committee, Washing-
Lettenmaier, D.P., Potter, K.W., 1985. Testing flood frequency ton, DC.
estimation methods using a regional flood generation model. Wiltshire, S.E., 1986. Regional flood frequency analysis I: homoge-
Water Resour. Res. 21, 1903–1914. neity statistics. Hydrol. Sci. J. 31, 321–333.