вход по аккаунту



код для вставкиСкачать
Int. J. Climatol. 18: 873–900 (1998)
Atmospheric Science, ETH Zürich, Switzerland
Recei6ed 4 February 1997
Re6ised 19 September 1997
Accepted 19 September 1997
A new precipitation climatology covering the European Alps is presented. The analysis covers the entire mountain
range including adjacent foreland areas and exhibits a resolution of about 25 km. It is based on observations at one
of the densest rain-gauge networks over complex topography world-wide, embracing more than 6600 stations from
the high-resolution networks of the Alpine countries. The climatology is determined from daily analyses of
bias-uncorrected, quality controlled data for the 20 year period 1971 – 1990. The daily precipitation fields were
produced with an advanced distance-weighting scheme commonly adopted for the analysis of precipitation on a
global scale. The paper describes the baseline seasonal means derived from the daily analysis fields. The results depict
the mesoscale distribution of the Alpine precipitation climate, its relations to the topography, and its seasonal cycle.
Gridded analysis results are also provided in digital form.
The most prominent Alpine effects include the enhancement of precipitation along the Alpine foothills, and the
shielding of the inner-Alpine valleys. A detailed analysis along a section across the Alps also demonstrates that a
simple precipitation–height relationship does not exist on the Alpine scale, because much of the topographic signal
is associated with slope and shielding rather than height effects.
Although systematic biases associated with the rain-gauge measurement and the topographic clustering of the
stations are not corrected for, a qualitative validation of the results, using existing national climatologies shows good
agreement on the mesoscale. Furthermore a comparison is made between the present climatology and the Alpine
sections of the global climatology of Legates and Willmott and the Greater European climatology from the Climate
Research Unit (University of East Anglia). Results indicate that the pattern and magnitude of analysed Alpine
precipitation critically depend upon the density of available observations and the analysis procedure adopted. © 1998
Royal Meteorological Society.
region; precipitation averages; rain-gauge networks; meso-climatology; mountain climate
Precipitation constitutes one of the most important meteorological and climatological parameters for our
ecosystems and civilization. This is, in particular, the case in mountainous regions such as the Alps, which
are able to extract ambient moisture from the atmosphere by various orographic precipitation mechanisms. In such regions, water erosion has formed our landscape, and the availability of precipitation and
soil moisture is one of the prime factors that determines the vegetation cover and thus the basis of our
agriculture. Also, runoff from precipitation and snow-melt in mountainous regions usually provides
freshwater resources for large areas, which include the adjacent flatland neighbourhood. On the other
hand, precipitation is also responsible for a range of natural disasters, ranging from flash-flooding,
landslides, avalanches to serious cases of hail damage.
* Correspondence to: Atmospheric
Contract grant sponsor: Swiss Priority Programme; Contract grant number: SPP-U 5001-044602
CCC 0899–8418/98/080873 – 28$17.50
© 1998 Royal Meteorological Society
It is thus not surprising that the monitoring and analysis of precipitation has always been central to
mountain meteorology and climatology. Accurate knowledge of the spatial and seasonal variations of
long-term mean precipitation is required for a variety of planning tasks in civil engineering, agriculture
and forestry. The development of high-resolution numerical weather prediction models, and the threat of
global climate change, has further increased the motivation to better observe the distribution of
precipitation and understand the underlying process. For example, the investigation of the relationship
between regional precipitation and large-scale flow conditions (e.g. von Storch et al., 1993), and the
detection of climate trends (Widmann and Schär, 1997) requires distributions of precipitation as input for
statistical analyses. Again the development of numerical weather forecasting and climate models critically
depends upon accurate analyses of observed precipitation for validation purposes (e.g. Hulme, 1994a;
Lüthi et al., 1995). Yet, although recent efforts have led to global and continental-scale climatologies that
are well-suited for climate dynamics and modelling studies (e.g. Arkin and Xie, 1994; Hulme et al., 1995),
adequate analyses on the regional scale are often not available.
For the region of the European Alpine mountain chain two categories of climatologies exist. On the one
hand there are analyses for certain sectors of the Alpine region, mostly delimited by national or
departmental borders, which exhibit a spatial resolution as fine as a few kilometres, and for which
extensive use is made of dense rain-gauge networks (see Table I). The design of these analyses is adapted
to local planning purposes, and their use for research activities in meteorological and climate research is
restricted by the limited spatial extent, the lack of temporal resolution (beyond the mean seasonal cycle),
and the fact that such information is commonly reproduced in climatic atlases but not accessible in digital
form. On the other hand the Alpine sections from global and continental-scale precipitation analyses (e.g.
Legates and Willmott, 1990; Hulme et al., 1995) embrace the entire region with a homogeneous analysis
procedure, sometimes include information on interannual variability, and also can be accessed digitally.
Yet, suffering from restrictions in the international data exchange (Hulme, 1994b), the observational basis
Table I. Recent officially available national high-resolution charts of annual mean precipitation (some also include
mean seasonal and monthly distributions). The references have been consulted for a qualitative validation of the
Alpine-wide mesoscale precipitation climatology (see section 5.1). (Note that the number of stations refers to the
stations used for the entire country, rather than for the part of the Alpine region)
Stations, period and comments
Steinhauser, 1953
Behr, 1993
1415, 1901 – 1950
ca. 1400, 1972 – 1990, polynomial interpolation, uniform height
Direction de la
Météorologie, 1988
Direction de la
Météorologie, 1989
ca. 1500, 1951 – 1980
Schirmer and
Vent-Schmidt, 1979
ca. 4000, 1931 – 1960
ca. 1200, 1921 – 1950
4121, 1951 – 1980, elevation detrended kriging, variable height
Servizio Hydrographico,
ca. 2400, 1921 – 1950
Landeshydrologie und
Geologie, 1992
ca. 400, 1951 – 1950 bias corrected data, elevation detrended
kriging, uniform height gradient
Slovenia, Croatia,
Dinaric ridge
Rankovic, 1980
Kari-Krei, 1991
1931 – 1960
1951 – 1980
Vosges and
Reklib, 1995
1951 – 1980, kriging, regression with multiple parameters of
regional topography
© 1998 Royal Meteorological Society
Int. J. Climatol. 18: 873 – 900 (1998)
Figure 1. Observing sites of the SYNOP network in the Alpine region. Observations from these stations are subject to an
international data exchange via the global telecommunication system. The poor spatial resolution should be compared with the
density of Alpine rain-gauge networks displayed in Figure 3
is rather limited. Often the underlying station sample is similar to that of the conventional SYNOP
network (cf. Figure 1) and this seriously limits the spatial resolution and expected accuracy of such
analyses. An extensive observational basis appears, in particular, essential over complex terrain such as
the Alps, where the representativity of point measurements is very limited.
Although it has not been the prevalent perspective so far, viewing the mountain range as a whole was
not completely uncommon in the history of Alpine precipitation climatology. Some of the Alpine-wide
analyses have even established new climatological concepts and significantly influenced the research on
mountain meteorology. For example Raulin (1879) in his early study ‘U8 ber die Vertheilung des Regens im
Alpengebiet 6on Wien bis Marseille’ demonstrated the regional variability of the seasonal precipitation
cycle and this was one of the first comprehensive studies to recognise the various climatic ‘spheres’
influencing the region. Raulin’s study comprised multiyear data series at a notable 249 observing sites
(station spacing around 40 km), which he assembled from printed reports. Again, interested in a climatic
regionalization, Knoch and Reichel (1930) have exploited long-term monthly averages at 412 stations
across the Alps. Their study was one of the earliest to take into account the uncertainty of observations
due to the measurement bias of rain-gauges. Later, in his pioneering isohyetal analysis, Ekhart (1948) was
able to characterize the Alpine precipitation climate in terms of two wet zones along the northern and
southern slopes and drier conditions in the Alpine valleys. These findings, were confirmed by the seminal
studies of Kubat (1972) and Fliri (1974), who together assembled an impressive data base of monthly
time-series (1931 – 1960) from more than 1000 stations between ‘Mont Blanc und Hohen Tauern’ all
collected from printed publications. For some of the stations even daily totals were manually digitized and
this led to a synoptic climatology of Alpine rainfall variability (Fliri and Schüepp, 1983; Fliri, 1986). In
fact Fliri’s analysis is still widely used as a reference today, because it is the most recent Alpine-wide
climatology derived directly from high-resolution rain-gauge data. Any further improvements require the
consideration of the full network and hence —for practical reasons—digital access to the data. As an
alternative, Baumgartner et al. (1983) combined existing national climatological maps into a composite
map of mean annual precipitation.
As alluded to above, the currently existing climatologies cannot provide together the full spatial
coverage of the Alpine region and/or the temporal variability down to a daily resolution, as is needed for
studies on regional climate dynamics, mesoscale meteorology, and high-resolution model validation. Thus
there is a pressing need for a new climatology that makes use of as many of the available rain-gauge
observations as possible. In this article we present a new Alpine precipitation climatology. The analysis
© 1998 Royal Meteorological Society
Int. J. Climatol. 18: 873 – 900 (1998)
covers the entire region of the Alpine range including a broad ribbon of its adjacent foreland. With
respect to the spatial resolution (about 25 km) this climatology is intermediate to the two categories
mentioned above, and appropriate to resolve precipitation signals from various topographically controlled
mesoscale atmospheric processes (see Bergeron, 1968; Browning et al., 1974; Smith, 1979; Banta, 1990;
Schär et al., 1997).
Use is made of an extensive observational data base, which consists of time-series with daily rain-gauge
totals at more than 6600 stations during the 20 year period 1971–1990. The spatial analysis is performed
on a daily basis, resulting in a long-term data base of daily mesoscale precipitation fields. In the current
paper we describe the data base, the quality control and analysis procedures, and discuss the baseline
mean seasonal climatology as deduced from the daily precipitation fields. In subsequent studies we will
use the data base for further analysis on the frequency distribution of daily precipitation totals (for some
preliminary results, see Frei, 1995).
With a typical station spacing around 10 km the Alpine rain-gauge system constitutes one of the densest
in situ observing systems over a mountain range of its size in the world. The various regional networks
belong to and are operated by governmental and regional agencies (meteorological and hydrological
services), which are also responsible for archiving the data (e.g. Müller and Joss, 1985; Weingartner,
1992). The majority of the rain-gauges are still operated manually by spare-time observers, yet in recent
years subnets were equipped with automatically recording instruments (e.g. Gutermann, 1986). The high
costs of such extensive measurements and the awareness of their economic value has obstructed the
exchange of data for research purposes, and constitutes one of the main obstacles for the establishment
of Alpine-wide climatologies. The regular exchange of weather data via the Global Telecommunication
System (GTS) is confined to selected stations —the so-called SYNOP network—which has a poor spatial
resolution (cf. Figure 1).
For the present climatology we have made large efforts to obtain access to data from the high-resolution networks of the Alpine countries. In most cases extensive negotiations were necessary to establish the
confidence of the data providers. The establishment of an international Alpine research programme
(MAP; Binder and Schär, 1995), which circumstantially was initiated at the time of these negotiations,
helped to support the awareness about the scientific needs for high-resolution observational data, and to
build the relevant contacts. Access could be granted by most of the data centres contacted through
contracts regulating the non-distribution of the raw data, and the purely scientific purpose of its use.
Some of the centres also charged a fee for the expenses of the extraction and transfer of the data. A final
effort was then made by integrating the numerous data portions into a homogeneous data format. The
resulting data base allows us to produce an Alpine-wide climatology with the densest data coverage
available at present.
The structure of the paper is as follows: section 2 introduces the analysis domain and the characteristics
of the Alpine-wide rain-gauge data set. The analysis procedure is then presented in section 3 together with
a discussion of the error sources. Results for the annual and seasonal mean distributions are discussed in
section 4. Comparisons of these fields with climatologies from other studies (i.e. regional high-resolution
maps and the Alpine sectors from global/continental analyses) are detailed in section 5. Possible
applications and extensions of this work are briefly sketched in the concluding section 6.
2.1. Domain and topography
Figure 2(a) shows a physiographic map of the domain utilized for this climatological analysis. It is
defined as a latitude by longitude window (2°–17°E, 43°–49°N). In the west–east direction it has a
horizontal scale of about 1200 km and extends from central France across Switzerland to eastern Austria.
In the north–south direction the domain measures about 700 km and ranges from the Ligurian- and the
Adriatic coasts to southern Germany. To the north and east of the Adriatic Sea, the domain also
comprises parts of the territories of Slovenia, Croatia and Bosnia–Herzegovina.
© 1998 Royal Meteorological Society
Int. J. Climatol. 18: 873 – 900 (1998)
The arc-shaped mountain range of the European Alps constitutes the major topographic feature within
the analysis domain. It has an average ridge height of about 2500 m, a length of 800 km and a mean width
of about 200 km. The ridge is intersected by a series of major valleys, dividing the range into numerous
mountain massifs. Most of the valleys run northwards or southwards on to the foreland, but a few
prominent valleys, such as the Valais and the Inn valleys, are aligned along the main ridge over many tens
of kilometres. Figure 2(b) depicts the location and local names of some main geographical features, which
will be referred to later in the text.
Figure 2. (a) Physiographic map of the European Alps and the surrounding foreland area. The window displayed corresponds to
the analysis domain of the climatology. (b) Location of major geographic features labelled with terms referred to in the text
© 1998 Royal Meteorological Society
Int. J. Climatol. 18: 873 – 900 (1998)
Table II. Components of the Alpine rain-gauge data set. Listed are the time periods covered by the respective data
sets (time period), the total number of time-series available within the analysis domain (station total), and the average
number of quality proofed observations available per day (station average). Eventual quality testing performed by the
data provider is indicated
Data set: region covered (data source)
Time period
AU: Austria (Hydrogr. Zentralbüro and ZAMG,
CH: Switzerland (Schweiz. Meteorolog. Anstalt,
DL: Southern Germany (Deutscher Wetterdienst,
Offenbach a.M.)
RO: South-eastern France (Météo-France,
IT: Northern Italy (Servizio Idrografico e
Mareografico Nazionale, Roma)
UC: Northern Italy (Ufficio Centrale Ecologia
Agraria, Roma)
SY: Northern Italy and Bosnia–Herzegovina
(SYNOP data, Seewetteramt, Hamburg)
CR: Croatia (Meterol. and Hydrol. Service,
SL: Slovenia (Hydrometeorological Institute,
Alpine region (data set version 4.0, combined from
above components)
Quality control
1971 – 1900
Partly (status 1/95)
1971 – 1990
1971 – 1990
1971 – 1990
Yes, spatial
Yes, after 1972
1971 – 1986
(1987 – 1990)
1971 – 1990
557 (48)
446 (46)
Range test only
1977 – 1990
1971 – 1990
1971 – 1990
1971 – 1990
Range test,
spatial consistency
The analysis domain also comprises a typically 300 km wide belt of flatter terrain adjacent to the main
Alpine body. In general the Alpine foreland is characterized by smoothly undulating or flat terrain,
intersected by several smaller scale hill ranges, which divide the area into interconnected plains and river
basins. Some of the smaller scale topographic features of the Alpine foreland are readily resolved with the
grid-spacing of the climatological analysis (about 25 km). These are the Swiss Middleland, the Jura
Mountains (France and north-western Switzerland), the Schwarzwald (south-western Germany), the
Vosges hill (eastern France), the Dalmatian range (along the eastern Adriatic coast), the Appennino
(Italian peninsula) and the Massif Central (France).
2.2. The precipitation data set
The basis for this climatological analysis is a data set of daily rain-gauge observations with high spatial
resolution over most of the Alpine region. It was established by combining several sectorial data sets that
have been archived by governmental weather services and hydrological agencies. Our aim was to make the
best possible use of all available existing networks. Upon request, most of the data centres that were
contacted have kindly provided an extensive portion of their data for our research.
The nine components of the data set each comprise time series from stations located within the national
boundaries of some Alpine countries. Table II lists the individual data sets and indicates the respective
regions and time-periods, and the number of rain-gauge series. Originally our requests were made for data
between 1966 and 1992. An optimal coverage in the combined data set was achieved for the 20 years
1971–1990, which was then chosen as the reference period. The spatial distribution of rain-gauge stations
from which observations are available during the reference period is shown in Figure 3.
For the national territories of Austria, France (south of 49°N and east of 2°E), Germany (south of
49°N) and Switzerland, data could be obtained from the Austrian Hydrographical Office, Météo-France,
the German Weather Service and the Swiss Meteorological Agency, respectively. (Note that parts of the
© 1998 Royal Meteorological Society
Int. J. Climatol. 18: 873 – 900 (1998)
Figure 3. Location of all observing stations from which rain-gauge time-series are used for the climatological analyses
networks are operated by institutions other than the data provider.) These data sets (referred to as AU,
RO, DL and CH) cover the entire reference period and together comprise observations at several thousands
of stations (see Table II). This results in an exceptional data coverage over the western, northern and eastern
sections of the analysis domain, with a typical station spacing as close as 10 km (see Figure 3). During
the reference period some stations have been closed and others newly opened, such that the average number
of daily reports is lower than the total number of stations (a6erage and total columns in Table II). The
station density in the four data sets, however, exhibits only slight variations in time (see Figure 4).
Figure 4. Evolution of the number of daily observations (yearly average) available in the various component data sets (see Table
© 1998 Royal Meteorological Society
Int. J. Climatol. 18: 873 – 900 (1998)
It was necessary to include data sets from several providers for the northern Italian region. The main
contribution stems from the Italian National Hydrological Service (Servizio Idrografico e Mareografico
Nazionale), which has recently — supported by a working group of Italian climatologists and hydrologists—initiated a central data base for the Italian Alpine region, and released part of their time series for
the present climatological study (IT, see Table II). With this component a dense spatial coverage is
achieved over north-eastern Italy and parts of the Appennino hill range (Figure 3), although some areas
of the western and central Po valley exhibit lower station density. Most of the observational records from
the IT data set currently available to us end in 1986 (except for 48 records of the northern Appennino,
see Figure 4). The spatial and temporal gaps could partly be compensated with the inclusion of two
additional contributions (UC: UCEA, Rome; SY: SYNOP network), both of which, however, exhibit a
significantly lower resolution than IT. The SY data set was retrieved from original SYNOP reports
collected by the Seewetteramt Hamburg, and was available for the subperiod 1978–1990. Together the
three datasets provide a spatial coverage which is, at least regionally, comparable to the northern parts of
the study domain for the first 16 years of the reference period, but significantly poorer for the final 4
The main contributions for the south-eastern part of the Alpine region have been obtained from the
Meteorological and Hydrological Service in Zagreb (CR) and the Slovenian Hydrometeorological
Institute in Ljubljana (SL). Together these networks comprise rain-gauges over the Julian Alps, along the
Adriatic east coast and the northern sector of the Dalmatian hill range (Figure 3). Although less dense
than over the northern and western portions of the Alps, the networks are much better than SYNOP
coverage, especially after 1980 when the number of Croatian stations has significantly increased (see
Figure 4). For the Southern Dinaric Alps (territory of Bosnia–Herzegovina) the analysis is based on 12
time-series from the above mentioned SY data set (see Table II).
The resulting aggregated database covers the entire Alpine mountain ridge, including its foreland
regions. The actual version (referred to as version 4.0) comprises time-series from a total of more than
6600 rain-gauge stations, with typically 5000 daily reports for a particular day of the reference period.
There is still considerable inhomogeneity in the spatial coverage across the domain, particularly for the
years after 1986, and this will be accounted for by using an analysis scheme with a regionally variable
spatial resolution (see section 3). During the interpretation of the climatological results it will be
important to bare in mind the variable density of the observational basis.
Between some of the networks and even between individual stations there are notable differences in the
observing practices. For example, various types of rain-gauges are simultaneously in operation: in Austria,
Germany and Switzerland various types of the Hellmann rain-gauge are widely in use, whereas in France
a cone-shaped model (Pluviomètre SPIEA) is very common. During the last decade additional instruments
suited for automated operation were adopted in most countries. Technical specifications of rain-gauges
used in the Alpine region can be found, for example, in Blumer (1994) and in Champeaux and Tamburini
(1995). Differences also pertain with regard to the daily reading times: readings at 07:00 CET (AU, CR,
DL, RO, SL, SY), at 07:30 (CH) and at 09:00 (UC, IT) are common. Inconsistencies emerging from the
use of different gauge types and daily reading times were ignored in the spatial analysis and the
climatological results described in this paper.
2.3. Quality control
The data sets obtained from the various data centres exhibit variable quality status (see Table II).
Whereas at least part of the series from the high-resolution networks were thoroughly controlled by the
data providers, the SYNOP data (SY) for Northern Italy and the Southern Dinaric Alps constitute
unproved raw data. The poor data quality of SYNOP precipitation reports due to typing, transmission
and coding errors is well known (Schneider et al., 1992). Again our own experience shows that the
declaration of missing observations is not handled properly everywhere. In some of the original data sets,
time series with dry reports over several months have been detected, which apparently should have been
indicated as measurement failure.
© 1998 Royal Meteorological Society
Int. J. Climatol. 18: 873 – 900 (1998)
In order to exclude the most obvious errors from the Alpine data set, a quality control procedure was
applied to all the daily observations prior to their spatial analysis. For this purpose we used a simple
range test and a spatial consistency test. Daily reports queried by the quality testing were then simply
dropped and no attempt was made to estimate appropriate replacements.
The range test filters from the data set reports larger than an upper limit of 300 mm per day. Although
observations exceeding such an amount have been observed in the Alpine region in extreme cases (e.g.
Lauscher, 1971; Jacq, 1994), their numerous occurrences in the SY data set (typically 30 per year!) are
mostly bogus reports. Other data sets, such as the complementary Italian data set, UC (see Table I), have
already been screened with a similar threshold by the data providers. The range test was only applied to
unproved components of the Alpine data set.
For the spatial consistency testing we adopted, with minor modifications, a procedure that is
operational at the German Weather Service (Behrendt, 1992). Each daily report is verified against some
tolerance bounds derived from neighbouring stations. These tolerance bounds depend upon both the
long-term monthly mean and the daily value at the stations under consideration. The consideration of
monthly means allows systematic anomalies to be taken into account, for example, exposure of the
stations. In a modification of the standard procedure, the tolerance bounds used in our study also depend
upon the proximity of surrounding observations, in order to cope with the variable network density. The
testing procedure deals separately with cases of isolated precipitation surrounded by dry reports, and cases
of isolated dryness surrounded by rainy reports.
Whereas for the unproved SY-dataset (SYNOP-reports) about 0.8% of the original data was rejected.
The rejection rate was less than 0.05% for the high-resolution data sets. Thus, there is no significant
reduction in the data amount, and the a6erage station numbers listed in Table II also hold for the quality
proved data sets.
The fields of the Alpine precipitation climatology are derived by spatial analysis of the irregularly
distributed station data on to a regular grid specified by latitude and longitude circles. The grid spacing
is 0.3° and 0.22° in the west – east and north – south directions, respectively, and corresponds to a mesh size
of about 24 km. The basis for the mean monthly and annual precipitation (section 4) are primary daily
analyses, calculated for each day of the reference period (1971–1990) from quality proved station data.
In comparison with the direct griding of station means, this procedure allows the optimal use of
observations from stations that exhibit a high rate of missing reports. The analysis scheme is designed to
calculate local area means from a set of station values in the neighbourhood of the analysis grid-points.
The following subsections describe the analysis scheme and discuss the possible sources of error for the
resulting climatological fields.
3.1. Spatial analysis scheme
The spatial analysis of station data on the regular grid is undertaken with a modified version of the
SYMAP interpolation algorithm of Shepard (1968, 1984). The procedure estimates a local area average
by appropriately weighting observations at stations within a search radius from the grid-point.
Two features of SYMAP are especially advantageous for the present application. First, the weight
assigned to an observation depends not only on the station distance from the grid-point, but it also
accounts for the directional isolation with respect to other stations in the search neighbourhood. Clusters
of observations to one side of a grid-point are appropriately down weighted. This particularly helps to
improve the performance of the analysis along the boundaries between high- and low-resolution networks.
Second, the search radius is adapted to the local observational density, requiring a minimum number of
stations to be included. This feature allows a homogeneous treatment over the entire analysis domain,
© 1998 Royal Meteorological Society
Int. J. Climatol. 18: 873 – 900 (1998)
regardless of the local station density. Although a coarse resolution analysis results for areas with sparse
data coverage, optimal use is made of the high-resolution networks.
Intercomparisons of various spatial analysis schemes have demonstrated the potential of Shepard’s
algorithm. Using surrogate daily precipitation fields from radar, Bussières and Hogg (1989) found that
Shepard’s scheme is superior to the simpler Cressman iteration and commensurate with an ordinary
kriging analysis. Similarly Weber and Englund (1992) found that a distance weighting scheme similar to
Shepard’s method could reach the accuracy of kriging techniques for their intercomparison case. The
favourable performance, the flexibility with station density and the computational efficiency have made
SYMAP a popular scheme for large-scale climatological analyses. A spherical version of the method
(Willmott et al., 1985) has been adopted for a number of recent global precipitation climatologies (Legates
and Willmott, 1990; Huffman et al., 1995).
For the present application some modifications were made to the standard SYMAP procedure. With
the original algorithm area-mean quantities are determined by locally averaging point estimates, which
have been calculated with an inverse square distance law on a primary fine-resolution grid (see e.g.
Legates and Willmott, 1990; Rudolf et al., 1992). For simplicity in this application we use a ‘flat’ distance
weighting function to arrive at a direct estimate of area means on the final analysis grid. The chosen
distance weighting function wd has the form:
wd(r̃i ) =
(1 + cos(p × r̃ i ))/2
for r̃5 1i
for r̃\1i
r̃i =
((xs −xg )/iDx)2 + ((ys − yg )/iDy)2
is the scaled distance of the station (xs, ys ) from the grid-point (xg, yg ). The scaling factor is a multiple (i )
of the mesh size (Dx, Dy), and it defines the size of the search neighbourhood. For each grid-point the
smallest possible from i= (1, 2, 3, 4) is chosen such that at least three observations are contained in the
search area. If the minimum requirement is not met with i= 4, no grid-point values is returned. Note that
our setting for the radial weights is not compatible with the idea of the ‘gradient corrections’ applied in
the original SYMAP procedure (Shepard, 1984). This feature was omitted in our analysis. However, for
the directional component of the weights the original SYMAP formula is applied.
The regional adaptation of the search neighbourhood controlled by coefficient i results in variations of
the effective resolution across the analysis domain. Figure 5 displays the spatial distribution of the search
radius for the two subperiods with different data coverage for Northern Italy (see section 2.2). The full
resolution of the underlying grid (25 km) is attained in the data dense areas where typically five to 25
stations values contribute to each grid-point estimate. However, for most parts of Northern Italy
(especially during the last 4 years) and for the Southern Dinaric Alps the search radius is two to three
times larger than the mesh size and only three to eight stations are available per grid-point estimate. The
coexistence of different search radii has some notable influence on the analysis fields. First, there are
extrapolations from data-dense into data-sparse regions, in some few locations with distances exceeding
75 km. This can lead to an artificial broadening of local patterns at the boundary between high- and
low-resolution coverage (see also Rudolf et al., 1992). Directional weighting can partly compensate for
this effect. Second, the analysis will not be able to resolve smaller scale features in areas with low
observational density, whereas a feature with a similar scale will be resolved in a data dense region. The
side effects of the adaptive search neighbourhood require some attention during the interpretation of the
climatological fields of section 4.
3.2. Sources of error
The accuracy of area-mean precipitation analyses suffers from the measurement error of point observations and the error due to limited spatial sampling of point observations for the area-mean estimation.
Here we briefly comment on the significance of these two types of errors for the present climatology.
© 1998 Royal Meteorological Society
Int. J. Climatol. 18: 873 – 900 (1998)
Figure 5. Spatial distribution of the search radius used with the SYMAP algorithm for the analysis of station data. Values are given
in units of the basic mesh size (about 25 km). The panels represent the situation for the period (a) 1971 – 1986 and (b) 1986 – 1990,
with different availability of high-resolution data over Northern Italy (see also section 2.2)
Contributions to the error in conventional rain-gauge measurement stem from the deflection of
hydrometeors in the wind field above the gauge orifice, the wetting of gauge walls, evaporation from the
container and snow drift into the gauge. Most significant are the wind-induced losses, from which an
undercatch must be expected of up to several 10 percents (e.g. Neff, 1977; Sevruk, 1985a; Groisman and
Legates, 1994). The magnitude of the undercatch depends on the ambient wind speed, the drop-size
distribution (i.e. precipitation intensity) and the type of precipitation (rain or snow). In principle, a
correction of the systematic gauge bias could be applied using an appropriate methodology (Sevruk,
1982). This, however, would require access to wind and temperature data in as high a resolution as
possible, and to include metadata on gauge details (device type, local wind exposure) important for the
correction procedure (Sevruk and Zahlavova, 1994a). Although highly desirable, an Alpine-wind correction of the systematic gauge bias is thus far beyond the scope of the present study, and would require a
multiyear effort.
© 1998 Royal Meteorological Society
Int. J. Climatol. 18: 873 – 900 (1998)
Sevruk (1985b) and Richter (1995) have made estimates of the long-term bias for Hellmann rain-gauges
in Switzerland and southern Germany, respectively, by applying correction procedures for sample stations
with additional data available. For the annual mean their results indicate a systematic undercatch of
about 7% at low-elevation or protected sites and up to 25% at wind-exposed sites at higher elevations (i.e.
with a significant fraction of snowfall). The bias exhibits a distinct annual cycle, with a maximum of 8%
(30%) in winter and a minimum of 5% (10%) in summer at protected (exposed) sites. For southern
Switzerland (i.e. to the south of the main Alpine crest) a seasonally independent bias of about 4% was
estimated. On average, for Switzerland the annual mean undercatch is calculated to 8%.
The high spatial variability of rain- and snowfall over short distances (see e.g. Gutermann 1974; Blumer
and Lang 1994), together with the limited spatial sampling, is associated with random and systematic
errors in the area-mean estimation. For example, it has long been recognized that a systematic
underestimate can result from an unrepresentative clustering of rain-gauges at low elevations (i.e.
populated mountain valleys) in combination with an increase of mean precipitation with altitude (Ekhart,
1948; Peck and Brown, 1962). Even within today’s high-resolution networks in the Alpine region, there
is a significant bias in station distribution relative to the area-height distribution (see Figure 6). Shepard’s
analysis scheme does not take account of this altitudinal clustering.
Methods have been proposed to compensate for this by making use of digital topography data and
linear precipitation – height relationships (Kirchhofer, 1993; Behr, 1993; Müller-Westermeier, 1995). Yet a
proper application of such schemes for a large area like the Alpine region is complicated by the high
spatial variability of this relationship, its non-linearity and the dependence on the spatial scale under
consideration (Uttinger, 1951; Lauscher, 1976; De Montmollin et al., 1980; Lang, 1985; Phillips et al.,
1992). Figure 6 gives an example with a weak, possibly insignificant height dependence. Based on
observations from dense local networks, recent studies (Blumer, 1994; Desurosne et al., 1996) demonstrate
that the local variability of precipitation at several Alpine massifs is often poorly cast with a function of
topographic height and strongly depends on other physiographical factors (slope, exposure, broad-scale
topographic environment), and that linear regression analysis over larger regions overestimates the
precipitation–height gradient. Again it has been stressed that the altitudinal precipitation trends are
sensitive to the correction of gauge undercatch (Sevruk and Zahlavova, 1994b). Thus an adequate account
of the effect of non-representative gauge distribution in the Alpine region would presume an objective
regionalization and a proper quantification of the local precipitation–topography relationship using an
extended description of topography (see e.g. Bénichou, 1994; Daly et al., 1994), as well as the
consideration of the measurement biases. Schemes suitable for application over large areas, such as the
Alpine region, have yet to be evaluated.
Figure 6. Altitudinal distribution of the fraction of rain-gauges (full circles, right ordinate) and the fraction of surface area (light
squares, hypsometric distribution, right ordinate). Results are for 200 m altitude intervals and a 150 × 80 km2 domain of the eastern
Swiss and western Austrian Alps comprising 129 rain-gauge stations in total. Bars represent the average annual mean precipitation
at stations in the corresponding altitude interval (mm day − 1, left ordinate). The hypsometric distribution was determined from a
topographic data set with a resolution of about 650 m (0.0083°)
© 1998 Royal Meteorological Society
Int. J. Climatol. 18: 873 – 900 (1998)
Figure 7. Relative difference (%) between spatial analysis of mean annual precipitation with modified and regular weighting of
high-evaluation sites. The thick line represents the 800 m MSL topographic contour
In order to estimate the analysis error associated with the unrepresentative height-distribution of the
gauges (and to roughly assess the significance of applying precipitation–height gradients in the analysis),
the precipitation analysis was repeated with a modified weighting of high-elevation sites. To this end,
stations above 1800 m MSL are given six times, and stations between 1200 and 1800 m MSL are given
twice, the normal weight in the regular analysis scheme. (Note that these settings would roughly
compensate for the elevation bias of the station sample used in Figure 6.) As alluded to above such a
procedure is too crude to account adequately for topographic clustering, but the magnitude of possible
analysis errors would be reflected in a comparison with the unmodified analysis. The relative difference
between modified and regular weighting (see Figure 7, mean annual precipitation) is mostly positive and
indicates some precipitation increase with altitude in the inner Alpine region. Except for two local patches
the relative difference is below 10%, which is very low when compared with the spatial variations of the
annual mean distribution (see Figure 9). In fact the two distributions from the regular and modified
weighting (not shown) are visually identical. This result illustrates that the characteristic mesoscale
patterns of the present climatology are solid, at least qualitatively, despite the presence of a biased
network and some precipitation – height gradients. This conclusion is also confirmed from a slightly
different viewpoint (see section 4.1 and Figure 10).
Finite spatial sampling is also associated with random errors in the area mean estimates. Here we briefly
illustrate the spatial variations and the order of magnitude of this error component. The relative sampling
error e sampl
for the (climatological area-mean at grid-point j, can be expressed as:
e sampl
=N −
× j.
P( j
where Nj is the sample size (number of available point observations), and P( j and sj denote, respectively,
the true mean and true standard deviation of point measurements in the search neighbourhood (see e.g.
Rice, 1995). (For simplicity we have assumed here that the area mean is calculated as the arithmetic mean
over point observations.) Clearly an accurate estimate of P( j and sj from the available station sample is
© 1998 Royal Meteorological Society
Int. J. Climatol. 18: 873 – 900 (1998)
Figure 8. Relative sampling error e sampl (%) for the 1983 annual precipitation mean according to Equation (2). The thick line
represents the 800 m MSL topographic contour
again complicated by the expected non-representativeness of this sample. Nevertheless for these illustrative purposes we can use the available point observations to estimate e sampl
Figure 8 displays the estimated sampling error e sampl
of the year 1983 (note that
there is high data density for north-eastern Italy during this year (Figure 4)). The spatial variations are
dominated by the contrast between areas of high observational density where the estimated sampling error
is generally below 5%, and areas of poor data coverage where errors in the order of 10% and more are
to be expected. Large errors also can be found in some locations with dense station coverage, indicating
high local variability (e.g. Massif Central, Vosges–Rhine valley). In summary, the accuracy of the present
climatology could be improved by a further extension of the data base over data-sparse areas/periods,
while for regions with dense networks the measurement bias and non-representative gauge distributions
are likely to dominate over the random error contribution.
In this section the annual, seasonal and monthly precipitation climatologies are presented. Except for
precipitation frequency in section 4.3, these were obtained by averaging the respective daily precipitation
fields obtained with the spatial analysis scheme described in section 3.1. This procedure, in comparison
with the spatial analysis of long-term station means, allows a more optimal use of data from interrupted
time series. All results refer to the 20 year reference period 1971–1990.
4.1. Annual mean
The mean annual precipitation is displayed in Figure 9. Despite the implicit spatial smoothing over
areas of typically 500 km2, the climatology still exhibits remarkable variations, with values ranging from
1.3 mm day − 1 to 6.4 mm day − 1. On the scale of the entire mountain range, characteristic features of the
distribution are an elongated wet anomaly extending along the northern rim of the Alps, two major wet
© 1998 Royal Meteorological Society
Int. J. Climatol. 18: 873 – 900 (1998)
Figure 9. Mean annual precipitation for the period 1971– 1990. The thick line represents the 800 m MASL topographic contour
zones to the south of the main crest, intermittent dryness at several inner-Alpine regions, and generally
dry conditions over the adjacent flatland areas.
The northern wet anomaly is typically 50 km wide and spans over more than 600 km. Distinct smaller
scale variations of this band can be identified with prominent mountain massifs (e.g. Glarneralpen in
eastern Switzerland, Allgäuer Alps in western Austria, for geographical terms refer to Figure 2). Across
the St. Gotthard transect, a narrow pass across the mountain ridge, the northern anomaly is connected
to the wet zone centred over the Lago Maggiore south of the main crest. A second major wet zone of the
southern rim is located north of the Adriatic Sea over the Julian and Carnic Alps (north-eastern Italy and
Slovenia), and is connected to an anomaly extending along the ridge of the Dinaric Alps. The two
southern anomalies exhibit the highest mean values in this mesoscale climatology. A band of enhanced
rainfall spans between them following the characteristic bow-shape of the mountain flank in the Dolomiti
area, and slightly moister conditions also prevail along the eastern rim of the western Alps. The wet
anomalies along the Alpine rims contrast with comparatively dry conditions at inner-Alpine sectors,
mostly in regions with major east – west running valleys (e.g. Valais and Aosta Valley, upper Inn valley
and Kärnten). At the driest inner-Alpine spot of the climatology, near the Venosta valley in southern
Tirol, mean annual precipitation is below 2 mm day − 1, as low as typically observed remote from the
Over the Alpine foreland regions the climatological distribution (Figure 9) shows distinct wet anomalies
related to some of the smaller-scale hill ranges (e.g. Jura Mountains, Vosges, Schwarzwald, Appennino
and the Massif Central). However, much like the Alpine ridge, the Massif Central also exhibits rim-type
anomalies oriented along its western and south-eastern slopes. For the other cases the anomalies are
found approximately centred over the topographic features. Especially, the Jura Mountains, the Vosges
and Schwarzwald receive a remarkable amount of precipitation, given the comparatively small height of
these topographic features. The flatlands adjacent to the Alpine region exhibit little spatial variation.
Southern Germany appears to be a little wetter than the flatlands of north-eastern Austria, the Po Valley,
central France and the lower Rhône basin, where the lowest values are registered in the present analysis.
© 1998 Royal Meteorological Society
Int. J. Climatol. 18: 873 – 900 (1998)
Some of the features in Figure 9 should be interpreted with caution because of the variable data density
and associated limitations of the analysis scheme (see section 3). For example, the weak spatial variability
along the western Italian Alps (Piemonte, Ligurian Alps), or the moderate contrasts to the western Po
Valley, must be considered with reservations because of the comparatively coarse station sample. Again
the dry sector between the two major anomalies of the southern Alps could be influenced by the
extrapolation of dry inner-Alpine (Tirol) and flatland conditions (Po Valley), compensating for the lack
of in situ data (see Figure 3). Also, the generally broader scales of features along the Dinaric ridge may
be the result of the moderate observational resolution in combination with the topographic complexity of
the region. These conjectures are mostly supported from existing higher resolution climatologies that
cover a fraction of our domain (see section 5.1).
The characteristic features of Figure 9 imply that the distribution of the Alpine precipitation climate on
the mesoscale is governed by the main topographic slopes at the rim, which are responsible for
comparatively large-scale atmospheric ascent, enhanced precipitation and associated rain-shadowing of
inner-Alpine sectors. In Figure 10 we further illustrate this characteristic pattern in terms of a north–
south section across the eastern ridge from southern Germany to the Po Valley. The diagram depicts the
analysed mean annual precipitation (zonal average between 10.2°E and 12.5°E) together with the
zonally-averaged topography and the statistics (minimum, maximum and interquartile range) of the
long-term averages from rain-gauges in the corresponding zonal belts.
Approaching the ridge from the flatland (either from north or south), both the annual precipitation
mean and the variability between stations increase. This effect starts at about 50–80 km ahead of the base
of the main topographic slope (near 47.75°N, and 45.25°N). The centre of the wet northern anomaly is
found near the base, centred at about 800 m topographic height. The southern anomaly is at an elevation
of roughly 1200 m. In the inner Alpine area, mean annual values decrease despite the steepness of the
large-scale topography, and there is an attendant decrease of the zonal minimum and lower quartile in the
Figure 10. Mean annual precipitation (symbols) and topographic height (solid line) along a north – south section across the eastern
Alps. The data represents the zonal mean of the analysis between 10.2°E and 12.6°E. The symbols (see insert for legend) depict the
interquartile range and the minima and maxima of the distribution of mean annual precipitation at stations in the corresponding
latitude belts. The number of stations considered for each of the statistics is between 45 and 80 north of 47°N and between 10 and
30 south of 47°N
© 1998 Royal Meteorological Society
Int. J. Climatol. 18: 873 – 900 (1998)
station sample, which expectedly reflects the rain-sheltering of stations, for example, in deep valleys. A
similar tendency is also observed for the maximum and upper quartile, i.e. for quantities that represent
the situation at exposed sites of the station sample (mountain tops or slopes). This result demonstrates
that the inner-Alpine dryness evident in the present climatology is not merely an artefact of the
unrepresentative station sample (i.e. biased towards sheltered conditions, see section 3.2). Indeed, this
dryness is observed not only in sheltered valley stations, but also is equally found for the full range of
station exposures. It follows that the wet rim and interior dryness anomalies are solid features of the
mesoscale distribution despite some quantitative contamination from the clustering of stations near valley
4.2. Annual cycle
The features of the mean annual climatology exhibit characteristic seasonal variations. Figure 11
depicts the four seasonal mean fields. In order to better depict the whole range of mean daily precipitation
amounts, the contour interval is selected differently to that used for the annual mean in Figure 9.
Pronounced differences are evident between winter and summer. The winter season is generally dry,
shows an excitation of the anomalies over the northern rim and the forefront hill ranges (Jura Mountains,
Schwarzwald, Massif Central, Appennino). The Alpine southern wet zones seen in the mean annual field
are only weakly discernible, especially that over the Lago Maggiore region. On the other hand, summer
is the main rainy season in the Alps. A broad moist anomaly covers all but the south-western part of the
ridge, exhibits mean values typically more than twice as large as in winter, and has embedded the familiar
rim-type features. The wet conditions in the mountains contrast with the dry summer climate of the
French Mediterranean coast, the lower Rhone Valley, the Italian peninsula and the eastern Adriatic coast.
For spring and autumn the general mean patterns are fairly similar (Figure 11). The major wet regions
known from the annual mean are evident in both seasons, but in contrast to winter and summer, the
southern rim anomalies are particularly prominent. Here spring and autumn means reach values similar
to those in summer, whereas the areas north of the main crest are clearly dryer.
An illustration of the pronounced month-by-month variations in the annual cycle over selected
subdomains is provided in Figure 12. A monomodal cycle prevails north of the Alpine main crest
(southern Germany and northern Alps), with the highest precipitation amount during the main convective
period (May–September). Weaker variations are evident for the Schwarzwald (also representative for the
Jura Mountains) where a secondary smooth precipitation maximum is found between November and
January. The southern and south-western Alpine areas exhibit a bimodal regime, with maxima in
spring/early summer and in autumn. For four of the regions (‘Cevennes’, ‘south-western Alps’, ‘Lago
Maggiore’ and ‘Po valley’) the month of October is a well pronounced maximum and the high
precipitation activity of this month in the Appennino, the Po Valley and the Adriatic east coast is also
reflected in the autumn mean distribution of Figure 11.
4.3. Precipitation frequency
A further interesting aspect of the Alpine precipitation climate is the frequency of days with
precipitation, defined here as the occurrence of days with a total larger than or equal to 1 mm. (Note that
the frequency at lower thresholds may suffer from the observer’s accuracy (Böhm, 1978).) The mean
annual distribution is shown in Figure 13. In contrast to the climatologies of Figures 9 and 11, this
diagram was obtained from the analysis of the long-term mean frequencies at individual stations. Here the
Alpine topography is reflected in the pronounced gradient across the main crest. To the south of the ridge
(Po Valley) the occurrence is below 24%, whereas to the north, rain or snowfall is observed every third
day or more on average. The distribution is quite different from that of the mean precipitation (cf. Figure
9), and some of the characteristic features of the latter, such as the Alpine rim anomalies, are at most
weakly discernible in the precipitation frequency. There is a similarly pronounced demarcation between
frequent and rare occurrences across the Massif Central. It appears that this hill range contributes to the
screening of some precipitation systems from the lower Rhône valley and the south-western Alps, where
© 1998 Royal Meteorological Society
Int. J. Climatol. 18: 873 – 900 (1998)
Figure 11. Mean seasonal precipitation for the period 1971 – 1990. Note that contour intervals are consistent between the four panels but differ from the setting in Figure 9. DJF:
winter; MAM: spring; JJA: summer, SON: autumn
© 1998 Royal Meteorological Society
Int. J. Climatol. 18: 873 – 900 (1998)
Figure 12. Mean annual cycle of monthly precipitation (mm day − 1) for selected subdomains in the Alpine region. Domains are
outlined in the inset
the precipitation frequency shows a gradual north–south transition and reaches even slightly lower values
than over the Po Valley. Finally, note the rare occurrence of precipitation over north-eastern Austria,
which signals the more continental-type climate of this region, possibly amplified by an Alpine-scale
shielding effect.
The remarkable across-ridge variations in the frequency and the concurrent quasi-symmetry of the
mean amounts, point to the high precipitation intensity along the southern rim of the Alpine ridge. In fact
the Lago-Maggiore region and Julian-Carnic Alps are known to be often affected by severe precipitation
events (e.g. Geiger et al., 1991; Bonelli and Pelosini, 1992; Frei, 1995; Cicogna et al., 1996). Similarly over
southern France an enhanced frequency of extreme events has been noted for the south-eastern rime of
the Massif Central (Jacq, 1994). A thorough analysis of the intensity distribution and the climatology of
strong precipitation events from the present Alpine-wide data set will be presented in a forthcoming
© 1998 Royal Meteorological Society
Int. J. Climatol. 18: 873 – 900 (1998)
Figure 13. Annual frequency (%) of days with precipitation ]1 mm. The distribution was obtained directly by spatial analysis of
frequencies determined at individual stations
As outlined in the introduction, the present mesoscale Alpine climatology is intermediate to two groups
of existing climatologies — high-resolution sectorial climatological maps, and coarser resolution continental or global analyses. In this section we compare the results of the present mesoclimatology with fields
from both of these groups.
5.1. High-resolution sectorial climatologies
Here the comparison is based on mean annual precipitation fields mostly published in national climatic
atlases in the form of charts. Table I contains a list of the references consulted. To our knowledge the
selection reflects most recent work officially available from the respective countries. The comparison with
these charts is complicated by the variable reference period adopted and most of all by the incompatible
resolution. With a resolution near the kilometre scale there is a lot of detail provided and some
‘eye-smoothing’ was necessary to estimate the supra-local features for comparison with our climatology.
The present comparison is therefore rather descriptive.
As far as data-dense regions are concerned, the mean annual precipitation displayed in Figure 9 gives
a highly satisfactory qualitative representation of the gross features in the high-resolution climatologies.
A fractured wet zone along the northern mountain chain; large mean values in the Lago Maggiore region
and the Julian-Carnic Alps as well as wet anomalies associated with the Jura Mountains, Schwarzwald,
the Massif Central and Appennino range are features that can be discerned clearly in the national
high-resolution climatologies. This also pertains for the dry conditions in the major along-ridge valleys
(below 2 mm day − 1 for the Inn valley, Valais and Aosta valley).
Discrepancies of a quantitative nature occur for some of the high-altitude mountainous regions. For
example the yearly maximum in the northern wet zone (4.25 mm day − 1, see Figure 10) is lower than a
© 1998 Royal Meteorological Society
Int. J. Climatol. 18: 873 – 900 (1998)
rough estimate from German and Austrian climatological charts (between 4.4 and 5.2 mm day − 1). These
discrepancies may arise from the application of a linear precipitation–topography relationship for the
high-resolution analyses (or the neglect of such dependencies in the present study), as well as from the
incorporation of annual collector data for the national climatologies. Some of the quantitative discrepancies with the climatology for Switzerland may also be attributed to the correction of gauge undercatch
considered for the Swiss climatology, but not the present study.
Regarding the sectors with comparatively poor data coverage, the present climatology offers a
reasonable but partly inaccurate picture of the gross features. For example, the Italian climatology shows
more structure for the moist zone along the Piemonte Alpine flank (2.7–4.1 mm) and generally lower
values for the western Po Valley ( B 2 mm day − 1). Also the moist zone connecting the major southern
anomalies is slightly more pronounced. Again a comparison with sectorial climatologies for the Dinaric
ridge shows that the regional maxima of the present climatology are reasonably well positioned but
slightly too broad in scale, extending too far to the east. In all these cases poor data coverage and/or the
extrapolation from dense networks is likely to be the primary reason for the discrepancies (see also section
One limitation evident in Figure 9 is particularly noteworthy and pertains to the underlying mesoscale
resolution. The sectorial high-resolution climatologies exhibit individual dry conditions (less than 2 mm
day − 1) for the two parallel running Valais and Aosta valleys (50 km apart) and significantly enhanced
mean values for the mountain massif in between (Monte Rosa). The limitations in resolution (of the
station sample and the analysis scheme) impede the depiction of such strong regional variations in the
present climatology, yet an areal average value of 2.3 mm day − 1 is reproduced. A similar situation is
found for the Salzach and Drau valleys and the Hohe Tauern massif in Austria. To accurately capture
such local features at the current resolution would require account to be taken of the local precipitation–
topography relationships (see section 3.2).
Of special interest for the present intercomparison is the Alpine-wide precipitation climatology of Fliri
(1974). His subjective analysis is at a resolution comparable to that of Figure 9, yet had to be based on
a coarser network. The mean distribution (also reproduced in Schär et al., 1997) confirms the general
picture of our analysis, but also highlights the aforementioned inaccuracies in data-poor regions. Fliri
(1974) also computed the annual cycle for a selection of observation sites in the Alpine region. Together
with other studies (e.g. Steinhauser, 1949; Ceschia et al., 1991; Bonelli and Pelosini, 1992; Cacciamani et
al., 1995) these confirm the spatial variations of the annual cycle regimes displayed in Figure 12.
5.2. Coarse-resolution climatologies
The two coarse-resolution precipitation climatologies to be considered are the corresponding sectors of
the global analysis by Legates and Willmott (1990); hereafter LW) and the precipitation climatology for
greater Europe of the Climate Research Unit (Hulme et al., 1995; hereafter CRU). Both these climatologies were established from long-term monthly mean rain-gauge data analysed on to a regular grid with
0.5° resolution (corresponding to about 50 km). In order to restrict the intercomparison with our
climatology to the resolved features, the present Alpine high-density data set was reanalysed with a
comparable resolution, and this low-resolution climatology will be referred to as AL in this subsection.
In comparison with the Alpine analysis the coarse-resolution climatologies had to be founded on a
highly restricted station ensemble. Presumably not more than a few dozen rain-gauges were available in
the Alpine region, and it can thus be expected that this is one of the main reasons for discrepancies with
the present climatology. Differences in the reference period, the data preprocessing and the analysis
procedure may also be important: the CRU climatology strictly relies on station normals for the
1961–1990 reference, whereas LW considered all available means covering some part of the 1920–1980
period. In contrast to the other analyses, LW applied a correction to the station data for systematic gauge
undercatch (see also section 3.2). The spatial analysis method for LW is very similar to that of the present
Alpine climatology (see section 3.1), whereas CRU is based on an analysis method that uses topographic
elevation as an additional predictor variable. Three precipitation climatologies were produced by CRU for
© 1998 Royal Meteorological Society
Int. J. Climatol. 18: 873 – 900 (1998)
Figure 14. Long-term mean annual precipitation in the Alpine region according to the analyses of Legates and Willmott (1990);
(LW), Hulme et al. (1995); (CRU) and the present high-resolution data using a comparable grid resolution (AL). Note that different
grey-scales have been used for the various panels
© 1998 Royal Meteorological Society
Int. J. Climatol. 18: 873 – 900 (1998)
Figure 15. Mean annual cycle of domain-average Alpine precipitation (mm day − 1, domain indicated in the AL panel of Figure 14)
according to Legates and Willmott (1990); (LW), Hulme et al. (1995); (CRU) and the present analysis (AL)
the minimum (CRU-LO), mean (CRU-MN) and maximum (CRU-HI) elevation of the grid pixels as
taken from a digital elevation model. Apart from yielding information on the vertical gradients, Hulme
et al. (1995) anticipate that some adjustment could be made in the CRU-MN fields for the unrepresentative distribution of stations (i.e., the bias towards low elevations, see section 3.2).
Long-term mean annual distributions of the two low-density climatologies are displayed in Figure 14
together with the high-density AL field. With respect to the main pattern there is surprising similarity
between the LW and AL fields. The global analysis resolves the northern Alpine rim zone, the inner
Alpine dryness and there is also enhanced precipitation along the southern rim. Discrepancies with AL are
found in a wet zone over central Switzerland, the lower prominence of the Lago Maggiore southern
anomaly and the connection of the rim anomalies across the eastern Alps. The CRU analysis also
represents an Alpine precipitation anomaly, yet without smaller scale details. The signal takes the form of
a strong anomaly centred over the high Alps of eastern Switzerland and a smaller scale, weaker anomaly
over the south-eastern Alps. A similar general pattern is found in the CRU-LO and the CRU-HI fields
(not shown). In quantitative terms both the CRU and LW climatologies give substantially higher peak
values over the mountain ridge compared with AL, but over the flat foreland areas there is close
agreement. Averaged over the entire Alps (frame in AL-panel of Figure 14) the mean annual values are
3.1 mm day − 1 for AL and 3.5 mm day − 1 ( + 13%) for both the CRU and LW climatologies.
An intercomparison of the monthly mean fields (not shown) shows similar results. The global analysis
(LW) exhibits characteristic seasonal variations comparable to AL (see also section 4.2), although there
are differences with respect to the amplitude of individual features. For CRU the annual variations in the
pattern are less pronounced, the major anomaly being evident throughout the year. Averaged over the
entire Alps, all three climatologies exhibit an annual cycle with a summer maximum and a minimum in
late winter (see Figure 15). Generally CRU and LW precipitation values are larger than those in AL.
Maximum discrepancies are in wintertime, where the relative differences to AL amount to 20%, but there
are also remarkable differences for up to 25% for individual summer and autumn months (for CRU in
August and for LW in October).
The generally smooth appearance of the CRU pattern indicates that its effective resolution is somewhat
coarser than the spacing of the underlying grid. This can be expected, at least partly from limitations in
station density, which probably in this area did not allow for a resolution at the scale of the mesh. An
important factor of the CRU is the large and uniform vertical gradient (1.4 mm day − 1 per 1000 m)
evident in the three-dimensional analysis (not shown). Its magnitude appears to be at the upper limit of
what can be observed locally in the Alpine region (e.g. Lang, 1985; Blumer, 1994). It is likely that this
gradient is responsible for the prominent imprint of the Alpine ridge into the climatology, as well as the
lack of inner-Alpine dryness, and consequently the generally larger domain-mean values compared with
AL. We presume that sparse data coverage has forced the CRU analysis scheme to infer vertical gradients
from variations between stations on the foreland and elevated sites along the mountain ridge.
© 1998 Royal Meteorological Society
Int. J. Climatol. 18: 873 – 900 (1998)
The higher domain-mean values of LW compared with AL may be attributed at least partly to the
corrections of the systematic gauge undercatch (carried out for LW, but not for CRU and AL). This
interpretation is supported by the generally larger differences in winter compared with summer (see Figure
15), consistent with larger corrections for snow undercatch in the cold season. The relative difference in
the annual and domain mean between LW and AL (13%) is slightly larger than the correction factor (8%)
found on average over Switzerland (Sevruk, 1985b).
The intercomparison of the domain-mean annual cycle (over the area indicated in the AL panel of
Figure 14) indicates a critical aspect of the station density for Alpine climatologies. The various features
of the Alpine precipitation pattern exhibit pronounced variations in the annual cycle (see section 4 and
Figure 12). Hence an accurate representation of the annual cycle, even in the domain-average, relies on
a proper representation of the relative magnitude and size of these features. This goal is obviously difficult
to achieve from a station sample that is at the limit of resolving such features.
A new mesoscale climatology has been constructed for the area of the European Alps and the adjacent
foreland. The design of the climatology is adapted to the specific needs of mesoscale climate dynamics and
model validation purposes. The coverage of several countries and the attempted spatial resolution and
accuracy of the analysis have demanded that a special effort be made in the establishment of an adequate
data base. The resulting climatology is the first that is derived from a combination of many national
high-resolution rain-gauge networks since the seminal study of Fliri (1974), and it is the only one of its
kind that is available in digital form (for access over the Internet see Frei, 1997).
In the climatology presented, most of the Alpine domain is covered with an observational density that
is exceptional for mountainous terrain. A spatial analysis was undertaken on a 25 km grid, separately for
each day of the period 1971 – 1990, and long-term monthly and seasonal means were derived from the
resulting set of daily grids. Possible error sources include the systematic rain-gauge bias, the clustering of
rain-gauges at low elevations, and errors in some regional features pertaining to the areas/periods with
coarse station density.
The primary precipitation characteristics in the Alps are the moist zones extending along the primary
flanks, and dry conditions in the interior of the mountain range. This pattern is evident throughout the
year, but with remarkable and non-synchronous seasonal variations in the amplitude of individual
anomalies. The processes that appear most relevant for this range-scale pattern are the orographically
forced upslope precipitation downwind sheltering (Smith, 1979; Banta, 1990) and topographically
triggered summertime convection (e.g. Cacciamani et al., 1995; Schiesser et al., 1995). Altitudinal extrema
of mean precipitation (as evident from the foreland–rim–ridge contrasts) have also been attributed to the
rapid decrease of atmospheric moisture with altitude (e.g. Lang, 1985; Alpert, 1986). The Alpine features
and processes contrast with the precipitation signal as observed over some smaller-scale hills, which
possess an orographic precipitation anomaly centred over the topographic high that is more characteristic
for seeder–feeder-type rainfall enhancement (Bergeron, 1968; Browning et al., 1974). Yet for many
real-case situations in the Alpine region the above prototype mechanisms provide a rather limited picture
of the orographic influence on precipitation. Dynamic interactions between weather systems and the
mountain ridge engender a wealth of mesoscale dynamic phenomena, many of which are associated with
preferred regions and seasons of occurrence as well as characteristic precipitation signals (e.g. Schär et al.,
1997). Furthermore the proximity of both Mediterranean and the Atlantic modulates the moisture flux
during southerly and westerly flow conditions. The presented distribution of long-term mean precipitation
is therefore the result of the combined influence from several large-scale climatic regimes, the specific
geographical nature and shape of the Alps, and a complex multiscale interaction of the mountain range
with a variety of weather and flow systems.
The specific characteristics of the Alpine precipitation anomaly also pose a serious challenge to the
objective analysis of precipitation data. The occurrence of inner-Alpine regions of dryness (as is well
© 1998 Royal Meteorological Society
Int. J. Climatol. 18: 873 – 900 (1998)
illustrated in Figure 10) clearly demonstrates that the use of precipitation–height correlations in
interpolating station data should be used only with great care. Ideally such statistical relations should
distinguish between the exposure of the station (summit and valley stations, upstream and downstream
facing stations), allow for variability within the Alpine region, and ensure that horizontal precipitation
variations across the Alpine ridge are not mistaken as vertical gradients.
An intercomparison of the Alpine sectors from a coarse-resolution global (Legates and Willmott, 1990)
and a continental climatology (Hulme et al., 1995) with the present analysis points to some limitations of
the former, primarily associated with a restricted observational network. Certainly an accurate representation on the 50 km scale was beyond the scope of these climatologies. Rather they were designed for
validating global climate models on the 100 – 200 km scale. Also, the rain-gauge sample available may
suffice over flat terrain. The limitations over complex terrain are nevertheless worth noting because there
is a tendency to use coarse-resolution analyses for validating regional climate models. The differences
between our analysis results and those of the large-scale climatologies also points to possible uncertainties
in large-scale analyses for other mountainous regions.
The use of the present analysis for mesoscale meteorological and climatological studies is not only
supported by the adequate spatial resolution, but also from the availability of daily analysis fields, which
enable investigations of the temporal variability of Alpine precipitation. In subsequent papers we will
exploit the data base for the frequency analysis of heavy precipitation events (for preliminary results see
Frei, 1995). The data could also be used for a range of other purposes such as the analysis of past
precipitation variability and its relation to the synoptic-scale flow (Widmann et al., 1995; Widmann,
1996), the investigation of the sensitivity of the hydrological cycle with respect to putative climate changes
(e.g. Frei et al., 1997), the construction of down-scaling methods to infer regional precipitation patterns
and changes from larger scale information (e.g. Bretherton et al., 1992; Gyalistras et al., 1994; Von
Storch, 1995; Fuentes and Heimann, 1996) the evaluation and improvement of high-resolution numerical
weather prediction models (as planned within the Mesoscale Alpine Programme, see Binder and Schär,
1995), and regional climate models (e.g. Marinucci et al., 1995; Lüthi et al., 1995). A further area of
possible application is the validation and calibration of remote sensing techniques for the monitoring of
precipitation from satellites (e.g. Arkin and Ardanuy, 1989; Rasmusson and Arkin, 1993) or radar (e.g.
Joss and Lee, 1995).
For all the aforementioned applications, the Alpine region could provide an ideal mountainous testing
framework, due to the presence of one of the densest high reliability observational networks covering one
of the Earth’s major mountain ranges.
We gratefully acknowledge fruitful discussions with and valuable information for this climatology from:
Ingeborg Auer (ZAMG, Wien), Huw Davies (Atmospheric Science ETH, Zürich), Franz Fliri (University
of Innsbruck), Thomas Gutermann (SMA, Zürich), Herbert Lang and his co-workers (Geography ETH,
Zürich), Joze Rakovec (University of Ljubljana), Bruno Rudolf (GPCC, Offenbach), Max Schüepp
(SMA, Zürich), Heinz Wanner (University of Bern) and Martin Widmann (University of Washington,
Seattle). We are indebted to the following institutes for providing access to daily precipitation data:
Deutscher Wetterdienst, Offenbach a. M.; Hydrographisches Zentralbüro des Bundesministeriums für
Land- und Forstwirtschaft, Wien; Zentralanstalt für Meteorologie und Geodynamik, Wien; MétéoFrance, Toulouse; Ufficio Centrale di Ecologia Agraria, Rome; Swiss Meteorological Institute, Zürich;
Servizio Idrografico e Mareografico Nazionale, Rome; the Italian MAP working group on climatology;
the Meteorological and Hydrological Service, Zagreb; and the Hydrometeorological Institute of Slovenia,
Ljubljana. The following individuals were particularly helpful in providing the relevant personal contacts
and helping to obtain access to the data: Guillaume Baillon, Mauro Bencivenga, Philippe Bougeault, Peter
Binder, Carlo Cacciamani, Tanja Cegnar, E. Fischer, Thomas Gutermann, Branca Ivancan-Picek,
Andreas Kaestner, Manfred Kurz, Janja Milkovic, Franz Nobilis, Marc Payen, Luigi Perrini, H. Schmidt,
Tomas Vrhovec. The gridded surface climatology CRU95 used in section 5 has been supplied by the
© 1998 Royal Meteorological Society
Int. J. Climatol. 18: 873 – 900 (1998)
Climate Impacts LINK Project (Department of the Environment Contract PECD7/12/96) on behalf of the
Climate Research Unit. The visualization of the analysis fields was done with the graphics package IVE
developed at the University of Washington (Seattle). The research was supported by contributions from
the Swiss Priority Program (contract SPP-U 5001-044602).
Alpert, P. 1986. ‘Mesoscale indexing of the distribution of orographic precipitation over high mountains’, J. Appl. Meteorol., 25,
532 – 545.
Arkin, P.A. and Ardanuy, P.E. 1989. ‘Estimating climatic-scale precipitation from space: a review’, J. Climate, 2, 1229 – 1238.
Arkin, P.A. and Xie, P. 1994. ‘The Global Precipitation Climatology Project: first algorithm intercomparison project’, Bull. Am.
Meteorol. Soc., 75, 401–419.
Banta, R.M. 1990. ‘The role of mountain flows in making clouds’, in Blumen, W. (ed.), Atmospheric Processes o6er Complex Terrain,
Meteorol. Monograph, 23, 229–283.
Baumgartner, A., Reichel, E. and Weber, G. 1983. Der Wasserhaushalt der Alpen, Oldenburg, München.
Behr, O. 1993. ‘Methods for national and regional precipitation distribution assessment’, in Analysis Methods for Precipitation on
a Global Scale, GEWEX Workshop, Koblenz, WMO, WCRP-81 A77.
Behrendt, J. 1992. Dokumentation ROUTKLI: Beschriebung der Prüfkriterien im Programmsystem QUALKO (to be obtained from
Deutscher Wetterdienst, Abteilung Klimatologie, Offenbach a.M., Germany), 12 pp.
Bénichou, P. 1994. ‘Cartography of statistical pluviometric fields with an automatic allowance for regional topography’, in Debois,
M. and Désalmand, (eds), Global Precipitations and Climate Change, NATO ASI Series, Vol. I26, Springer-Verlag, Berlin, pp.
187 – 199.
Bergeron, T. 1968. Studies of the Orogenic Effect on the Areal fine Structure of Rainfall Distribution, Meteorological Institute,
Uppsala University, Report No. 38.
Binder, P. and Schär, C. 1995. MAP Mesoscale Alpine Programme, Design Proposal, 63 pp. (to be obtained from Swiss
Meteorological Agency, CH-8044, Zürich, Switzerland).
Blumer, F.P. 1994. Höhenabhängigkeit des Niederschlags im Alpenraum. PhD thesis, ETH-Zürich, No. 10784, 242 pp.
Blumer, F.P. and Lang, H. 1994. ‘Spatial difference of precipitation in the Eastern Swiss Alps’, International Conference on Alpine
Meteorology, 1994 (Lindau, Germany), Annal. Meteorol., 30, 237.
Bonelli, P. and Pelosini, R. 1992. ‘Intense precipitation on the Italian side of the Alps: analysis and forecasting’, CIMA 1992,
Toulouse, pp. A1 –A9.
Böhm, R. 1978. Erste Erfahrungen mit der Datenkorrektur und EDV-Aufbereitung 6on 75 Ombrometerstationen in Wien und
Umgebung auf Tagessummenbasis. Annual report of ZAMG, Appendix 5, D36 – D56 (available from ZAMG, Hohe Warte 38,
A-1191 Wien).
Bretherton, C.S., Smith, C. and Wallace, J.M. 1992. ‘An intercomparison of methods for finding coupled patterns in climate data’,
J. Climate, 5, 541 –560.
Browning, K.A., Hill, F.F. and Pardoe, C.W. 1974. ‘Structure and mechanism of precipitation and the effect of orography in
wintertime warm sector’, Q. J. R. Meteorol. Soc., 100, 309 – 330.
Bussières, N. and Hogg, W. 1989. ‘The objective analysis of daily rainfall by distance weighting schemes on a mesoscale grid’,
Atmos. Ocean, 27, 521–541.
Cacciamani, C., Battaglia, F., Patruno, P., Pomi, L., Selvini, A. and Tibaldi, S. 1995. ‘A climatological study of thunderstorm
activity in the Po Valley’, Theor. Appl. Climatol., 50, 185 – 203.
Ceschia, M., Micheletti, St. and Carniel, R. 1991. ‘Rainfall over Friuli – Venezia Giulia: high amounts and strong geographical
gradients,’ Theor. Appl. Climatol., 43, 175–180.
Champeaux, J.-L. and Tamburini, A. 1995. Zonage climatique de la France à partir des séries de precipitations (1971 – 1990) de réseau
climatologique d’état. Note No. 33 de la centre nationale de recherche métérologiques, Météo-France, Toulouse, 177 pp.
Cicogna, A., Salvador, M. and Micheletti, S. 1996. ‘Heavy rainfall in Friuli – Venezia Giulia; new data for a climatic extreme?’ MAP
Newsl., 4, 2 – 6.
Daly, C., Neilson, R.P. and Phillips, D.L. 1994. ‘A statistical – topographic model for mapping climatological precipitation over
mountainous terrain,’ J. Appl. Meteorol., 33, 140–158.
De Montmollin, F.A., Olivier, R.J. and Simard, R.G. 1980. ‘Evaluation of a precipitation map using a smoothed elevation – precipitation relationship and optimal estimates (kirging)’, Nordic Hydrol., 11, 113 – 120.
Desurosne, I., Ribot-Bruno, J., Watremez, S. and Oberlin, G. 1996. Interactions pluies/rélief à l’échelle temporelle des intensités, guide
pratique des données plui6ographiques et des résultats d’un réseau préalpin, Le TGP. Report published by Cemagref, Dision
Hydrologie-Hydraulique, CP 220, Lyon.
Direction de la Météorologie 1988. Précipitation en France. Direction de la météorologie national, Service central d’exploitation
météorologique, I.G.N., Paris.
Direction de la Météorologie 1989. Atlas Climatique de la France (réedition 1989). Ministère des Transport, Direction de la
Ekhart, E. 1948. Die Niederschlagverteilung in den Alpen nach dem Anomalienprinzip. Geogr. Ann., 728 – 739.
Fliri, F. 1974, ‘Niederschlag und Lufttemperatur in Alpenraum’, Wissensch. Alpen6ereinshefte, 24, 110 pp.
Fliri, F. 1986. ‘Synoptisch-klimatologische Niederschlagsanalyse zwischen Genfersee und Hohen Tauern’, Wetter Leben, 38,
140 – 149.
Fliri, F. and Schüepp, M. 1983. Synoptische Klimatographie der Alpen zwischen Mont Blanc und Hohen Tauern,’ Wissensch.
Alpen6ereinshefte, 29, 686 pp.
© 1998 Royal Meteorological Society
Int. J. Climatol. 18: 873 – 900 (1998)
Frei, C. 1995. An Alpine precipitation climatology based on high-resolution rain-gauge observations. MAP Newsl., 3, 46 – 47.
(Available from Swiss Meteorological Institute, Krähbühlstr. 58, CH-8044 Zürich.)
Frei, C. 1997. A precipitation climatology of the Alps from high-resolution rain-gauge obser6ations. Digital documentation and access
to the Alpine precipitation climatology: – clim (periodically updated).
Frei, C., Widmann, M., Lüthi, D., Davies, H.C. and Schär, C. 1997. ‘Response of the regional water cycle to an increase of
atmospheric moisture related to Global Warming’, Proceedings 7th Conference on Climate Variations, American Meteorological
Society, 2 – 7 February, Long Beach, California.
Fuentes, U. and Heimann, D. 1996. Verification of Statistical-dynamical Downscaling in the Alpine Region, Report No. 47, Institut
für Physik der Atmosphäre, DLR-Oberpfaffenenhofen, Germany.
Geiger, H., Zeller, J. and Röthlisberger, G. 1991. Starkneiderschäge des schweizerischen Alpen- und Alpenrandgebietes. Eidgenössische Versuchsanstalt für Wald, Schnee und Landschaft, WSL, CH-8903 Birmensdorf, Switzerland.
Groisman, P.Y. and Legates, D.R. 1994. ‘The accuracy of United States precipitation data’, Bull. Am. Meterol. Soc., 75, 215 – 227.
Gutermann, T. 1974. Zur Niederschlagsmessung in der Schweiz: Aktuelle Beobachtungsnetze und grundsätzliche Messprobleme.
Arbeitsber. Schweiz. Meteorol. Zentralanst., 45, 33 pp. (Available from Swiss Meteorological Agency, Zürich, Switzerland).
Gutermann, T. 1986. Das automatische Wetterbeobachtungsnetz der Schweiz (ANETZ), SMA, Bericht, 17 pp. (Available from SMA,
CH-8044 Zürich, Switzerland.)
Gyalistras, D., von Storch, H., Fischlin, A. and Beniston, M. 1994. ‘Linking GCM-simulated climatic changes to ecosystem models:
case studies of statistical downscaling in the Alps’, Climate Res., 4, 167 – 189.
Huffman, G.J., Adler, R.F., Rudolf, B., Schneider, U. and Keehn, P.R. 1995. ‘Global precipitation estimates based on a technique
for combining satellite-based estimates, rain gauge analysis, and NWP model precipitation information’, J. Climate, 8, 128 – 1295.
Hulme, M. 1994a. ‘Validation of large-scale precipitation fields in General Circulation Models’, in Debois, M. and Désalmand, F.
(eds), Global Precipitation and Climate Change, NATO ASI series, Vol. I26, Springer-Verlag, Berlin, pp. 387 – 405.
Hulme, M. 1994b. ‘The cost of climate date—a European experience’, Weather, 49, 168 – 174.
Hulme, M., Conway, D., Jones, P.D., Jinag, T., Barrow, E.M. and Turney, C. 1995. ‘Construction of a 1961 – 1990 European
climatology for climate change modelling and impact applications’, Int. J. Climatol., 15, 1333 – 1363.
Jacq, V. 1994. In6entaire des situations a precipitations dilu6iennes Languedoc-Rousillon/PACA/Corse, Report No. 3 of MétéoFrance, SCEM, 190 pp. (To be obtained from: Météo-France, Direction Interrégionale Sud-Est, 2 Boulevard Chateau-Double,
13098 Aix en Provence, France.)
Joss, J. and Lee, R. 1995. ‘The application of radar-gauge comparisons to operational precipitation profile corrections’, J. Appl.
Meteorol., 34, 2612–2630.
Kari-Krei, 1991. ‘Klima’, in Enciklopedija Slo6enije, Vol. 5. (Available from Zalozba Mladinska knjiga, Ljubljana, Slovenia.)
Kirchhofer, W. 1993. ‘Mapping of corrected mean annual precipitation amount with respect to orography’, in Analysis Methods of
Precipiation on a Global Scale, GEWEX Workshop Koblenz, Germany, WMO, WCRP-81, A87.
Knoch, K. and Reichel, E. 1930. Verteilung und jährlicher Gang der Niederschläge in den Alpen’ Veröff. Preuss. Met. Inst., 9.
Kubat, O. 1972. Die Niederschlagsverteilung in den Alpen mit besonderer Berücksichtigung der jahreszeitlichen Verteilung’,
Veröffentl. Uni6. Innsbruck, 73, (Alpenkundliche Studien), 52 pp.
Landeshydrologie und Geologie 1992. ‘Mittlere jährliche korrigierte Niederschlagshöhen 1951 – 1980’, in Hydrologischer Atlas der
Schweiz, EDMZ, Bern.
Lang, H. 1985. ‘Höhenabhängigkeit der Niederschläge’, in Sevruk, B. (ed.), Der Niederschlag in der Schweiz Beitr. Z. Geol.
Schweiz— Hydroloie, 31, 149–157.
Lauscher, F. 1971. ‘Globale und Alpine Klimatologie der Starkregen’, Hochwasser Raumplanung, 11 (Amt der Kärntner Landesregierung), 37 – 39).
Lauscher, F. 1976. ‘Weltweite Typen der Höhenabhängigkeit des Niederschlags’, Wetter Leben, 28, 80 – 90.
Legates, D.R. and Willmott, C.J. 1990. ‘Mean seasonal and spatial variability in gauge-corrected, global precipitation’, Int. J.
Climatol., 10, 111 –127.
Lüthi, D., Cress, A., Frei, C. and Schär, C. 1995. ‘Interannual variability and regional climate simulations’, Theor. Appl. Climatol.,
53, 185 – 209.
Marinucci, M.R., Giorgi, F., Beniston, M., Wild, M. and Tschuck, P. 1995. ‘High-resolution simulations of January and July
climate over the western Alpine region with a nested regional modelling system’, Theor. Appl. Climatol., 51, 119 – 138.
Müller, G. and Joss, J. 1985. ‘Messnetze und Instrumente’, in Sevruk, B. (ed.), Der Niederschlag in der Schweiz, Beitr. Z. Geol.
Schweiz — Hydrologie, 31, 17–47.
Müller-Westermeier, G. 1995. ‘Numerisches Verfahren zur Erstellung klimatologischer Karten’, Bericht Deutschen Wetterdienstes,
193, 25 pp. (To be obtained from German Weather Service, Offenbach a.M., Germany.)
Neff, E.L. 1977. ‘How much rain does a rain gauge gage?’ J. Hydrol., 35, 213 – 220.
Peck, E.L. and Brown, M.J. 1962. ‘An approach to the development of isohytal maps for mountainous areas’, J. Geophys. Res., 67,
681 – 694.
Phillips, D.L., Dolph, J. and Marks, D. 1992. ‘A comparison of geostatistical procedures for spatial analysis of precipitation in
mountainous terrain’, Agric. For. Meteorol., 58, 119–140.
Rankovic, S. 1980. Karta II-13 of Atlas Klime SFR Jugosla6ije (exact year of publication not known). (Available from
Hydrometeorological Institute of Slovenia, Ljubljana.)
Rasmusson, E.M. and Arkin, P.A. 1993. ‘A global view of large-scale precipitation variability’, J. Climate, 6, 1495 – 1522.
Raulin, V. 1879. ‘U8 ber die Verteilung des Regens im Alpengebiet von Wien bis Marseille’, Z. österreichischen Gesellschaft Meteorol.,
14, 233 – 247.
Reklib, 1995. Klimaatlas Oberrhein Mitte-Süd, vdf Hochschulverlag, Zürich.
Rice, J.A. 1995. Mathematical Statistics and Data Analysis, 2nd edn. Duxbury Press, Belmont, CA.
Richter, D. 1995. ‘Ergebnisse methodischer Untersuchungen zur Korrektur des systematischen Messfehlers des Hellmann-Niederschlagsmessers’, Bericht Deutschen Wetterdienstes, 194, 93 pp. (To be obtained from German Weather Service, Offenbach a.M.,
© 1998 Royal Meteorological Society
Int. J. Climatol. 18: 873 – 900 (1998)
Rudolf, B., Hauschild, H., Reiss, M. and Schneider, U. 1992. ‘Die Berechnung der Gebietsniederschläge im 2.5°-Raster,’ Meteorol.
Z., N.F. 1 (Special Issue: Global Precipitation Climatology Centre).
Schär, C., Davies, T.D., Frei, C., Wanner, H., Widmann, M., Wild, M. and Davies, H.C. 1997. ‘Current Alpine climate’, in Cebon,
P. et al. (eds), A View from the Alps: Regional Perspecti6es on Climate Change. MIT Press, in press.
Schiesser, H.H., Houze, R.A. and Huntrieser, H. 1995. ‘The mesoscale structure of severe precipitation systems in Switzerland’,
Mon. Wea. Re6., 123, 2070–2097.
Schirmer, H. and Vent-Schmidt, 1979. Das Klima der Bundesrepublik Deutschland; Lieferung 1: Mittlere Niederschlagshöhen für
Monate und Jahr, 1931 – 1960, German Weather Service, Offenbach am Main, Germany.
Schneider, U., Henning, D., Hauschild, H., Reiss, M. and Rudolf, B. 1992. ‘Zur Berechnung der monatlichen Niederschlagshöhen
aus synoptischen Meldungen’, Meteorol. Z., N.F. 1 (Special Issue: Global Precipitation Climatology Centre).
Servizio Hydrographico 1959. Precipitationi medie per il trentennio 1921 – 1950, Publication 24, Ministero dei lavori pubblici, servizio
hydrographico, Roma.
Sevruk, B. 1982. Methods of Correction for Systematic Error in Point Precipitation Measurement for Operational Use, WMO-No. 589,
Operational Hydrology Report No. 21, World Meteorological Organization, Geneva.
Sevruk, B. 1985a. ‘Correction of precipitation measurements, summary report;, in Proceedings of the ETH/IAHS/WMO Workshop
on the Correction of Precipitation Data, pp. 13–23.
Sevruk, 1985b. ‘Systematischer Niederschlagmessfehler in der Schweiz,’ in Sevruk, B. (ed.), Der Niederschlag in der Schweiz, Beiträge
Geol. Schweiz— Hydrologie, 31, 65–75.
Sevruk, B. and Zahlavova, L. 1994a. ‘Classification system of precipitation gauge site exposure: evaluation and application’, Int. J.
Climatol., 14, 681 –689.
Sevruk, B. and Zahlavova, L. 1994b. ‘The effect of systematic error of precipitation measurement on vertical gradients of
precipitation in Switzerland’, in International Conference on Alpine Meteorology, Lindau, Ann. Meteorol., 30, 48 – 51.
Shepard, D.S. 1968. ‘A one-dimensional interpolation function for irregularly spaced data’, Proceedings 23rd ACM National
Conference, Brandon/Systems Press, pp. 517–524.
Shepard, D.S. 1984. ‘Computer mapping: the SYMAP interpolation algorithm’ in Gaile, G.L. and Willmott, C.J. (eds), Spatial
Statistics and Models, Dordrecht, pp. 133–145.
Smith, R.B. 1979. ‘The influence of mountains on the atmosphere’, Ad6. Geophys., 21, 87 – 230.
Steinhauser, F. 1949. ‘U8 ber die Struktur des Jahresgangs der Niederschläge am Zentralalpenkamm’, Wetter. Leben, 2, 1 – 4.
Steinhauser, F. 1953. ‘Niederschlagskarte von O8 sterreich für das Normaljahr 1901 – 1950’, Beitr. Hydrogr. O8 sterreich, 27, Wien.
Uttinger, H. 1951. ‘Zur Höhenabhängigkeit der Niederschläge in den Alpen’, Arch. Meteorol. Geophys. Bioklimatol., B2, 360 – 382.
Von Storch, H. 1995. ‘Spatial patterns: EOFs and CCA’, in Von Storch, H. and Navarra, A. (eds), Analysis of Climate Variability,
Applications of Statistical Techniques, Springer-Verlag, Berlin.
VonStorch, H., Zorita, E. and Cubasch, U. 1993. ‘Downscaling of global climate change estimates to regional scales: an application
to Iberian rainfall in wintertime’, J. Climate, 6, 1161–1171.
Weber, D. and Englund, E. 1992. ‘Evaluation and comparison of spatial interpolators’, Mathematical Geol., 24, 381 – 391.
Weingartner, 1992. ‘Niederschlagsmessnetze’, in Hydrologischer Atlast der Schweiz, Landeshydrologie und -geologie, EDMZ, Bern.
Widmann, M. 1996. Mesoscale 6ariability and long-term trends of Alpine precipitation and their relation to the synoptic-scale flow.
PhD thesis No. 11769, ETH, Zürich.
Widmann, M., Bresch, D., Frei, C. and Schär, C. 1995. ‘Daily Alpine precipitation and its relation to the synoptic scale flow — a
principal component and singular value decomposition approach’, Proceedings of the 6th International Meeting on Statistical
Climatology, Galway.
Widmann, M. and Schär, C. 1997. ‘A principal component and long-term trend analysis of daily precipitation in Switzerland’, Int.
J. Climatol., 17, 1333–1356.
Willmott, C.J., Rowe, C.M. and Philpot, W.D. 1985. ‘Small-scale climate maps: a sensitivity analysis of some common assumptions
associated with grid-point interpolation and contouring’, Am. Cartogr., 12, 5 – 16.
© 1998 Royal Meteorological Society
Int. J. Climatol. 18: 873 – 900 (1998)
Без категории
Размер файла
992 Кб
Пожаловаться на содержимое документа