A new long-term gridded precipitation data-set for the alps and its application for Map and Alpclim

Phys. Chem Earth (B), Vol. 26, No. 5-6, pp. 421-424, 2001 © 2001 ElsevierScience Ltd. All rights reserved 1464-1909/01/$ - see front matter

Pergamon

PII: S1464-1909(01)00029-6

A New Long-Term Gridded Precipitation Data-Set for the Alps and its Application for Map and Alpclim I. Auer', R. Biihm ~ and M. Maugeri 2

~Central Institute for Meteorology and Geodynamics, Vienna, Austria 'Instituto di Fisica Generale Applicata, Milan, Italy

Received 25 April 2000; accepted 9 October 2000

Abstract. One of the objectives of EU-project ALPCLIM is

the generation of a gridded data-set of monthly instrumental precipitation data. The area of investigation covers the Alps and wide regions of the surroundings from 4 ° to 18°E and 43 ° to 49 ° N. Grid distance is 1 deg longitude and 1 deg latitude. The project is not finished yet, more than 140 single series have been collected by now and are in the state of homogeneity testing and adjusting. The average linear distance of the stations is 75 km. Furthermore first results can be shown, dealing with the main purpose of the generation of the instrumental data-set within ALPCLIMto use the instrumental data (both temperature and precipitation) to create a longer temperature proxi-data-set based on stable isotope ice core data from high elevation sites in the Monte Rosa and Mont Blanc region. The ice core temperature proxis are supposed to be at least 500 years long. The precipitation series will be used to analyse the problem that ice-cores in principle carry information only for precipitation days, not tor all days. Analyses on daily temperature and precipitation data of 50 years series have shown already a way to construct "precipitation-mass weighted mean temperatures" for the longer series based on monthly values. The degree of correlation of those "precipitation-mass-weighted" temperatures with real temperatures will decide on the possibility to use stable isotope proxis from high level sites as temperature information. In addition to the described use within ALPCLIM, also projects like MAP may profit from the outcome of the ALPCLIM precipitation data-set. It will provide the MAP community with carefully homogenised monthly precipitation series to see the situation of MAPresults in a long-term context. ¢, 2001 Elsevier Science Ltd. All rights reserved

Correspondence to: Dr. Ingeborg Auer, Central Institute for Meteorology and Geodynamics, Hohe Warte 38, 1190 Vienna, Austria

1. Introduction

The Alps are a region of high potential in climatological research. On the one hand they offer a large scale of different climates from Mediterranean and Atlantic influences in the South and West to continental features in the East and from low elevation plains, valleys and basins to high elevation mountain climate in the regions above tree line and snow line. On the other hand they offer a wealth of climate data not easily obtainable elsewhere. The paper discusses the way to a data set of alpine wide coverage based on homogenised monthly precipitation series. It shows also some first applications within the Projects MAP (Mesoscale Alpine Project) and ALPCLIM (Environmental and Climate Records from High Elevation Alpine Glaciers).

2. The long term precipitation data set

The area of investigation covers the Alps and wide regions of the surroundings from 4 ° to 18 ° E and 43 ° to 49°N. Grid distance is 1 deg longitude and 1 deg latitude. Unfortunately, practically all long term instrumental data series are disturbed by inhomogeneities caused by changes in instrumentation, station moves, relocations of instruments and also changes in the local environment, etc. Such inhomogeneities have to be removed, otherwise the results of climate change studies can be erroneous. The procedure of homogenisation is bases on two steps. Step one is the elimination of documented breakpoints with the help of metadata information, step two the application of relative homogeneity tests. For precipitation series breakpoints with documented adjustment values are most commonly related to gauge relocations with parallel measurements or instrumental changes with parallel measurements. In the Austrian precipitation network (Auer et. al, 2000) a general tendency from higher installations (above ground) of rain gauges to lower ones could be found, thus causing a systematic bias of increasing precipitation in the original data.

422

I. Auer et al : Data-Set for the Alps

After step one (the elimination of all quantitatively well known breakpoints) in step two relative homogeneity tests are applied and the adjustment factors are added to the time series. Most of the homogenisation procedure lies in the authority of national Weather Services or Institutes, the used procedures are MASH, HOCLIS, CAUSSINUSMESTRE and SNHT (all tests are well described in WMO, 1999 and Peterson el. al, 1998,). More than 140 single series have been collected by now, currently, nearly one half of the series have been homogenised, the others are in the state of homogeneity testing and adjusting. 95% of the series are longer than 100 years, the earliest starting in the end of the 18 'h century. The mean distance of the stations is about 75 km. The temporal increase of the number of stations since 1778 is shown in Fig.2.1. Number of stations 160 150 140 130 120 110 100 90 80 70 60 50 40 30 20 10 0

distance information transport where no series exists near a grid point ( this is typical for the older parts of series with sparse spatial resolution) the interpolation will be truncated at the starting year of the longest single series within 200 km distance from a grid point. Additional borders of truncation will be defined due to regional different long term trends of precipitation. The Austrian precipitation series which have been already homogenised (Auer et al., 1998) do not show a spatial consistency at all. The warming since 1890, which is the spatially uniform characteristic of the temperature change, has been accompanied in some regions and in some periods by trends towards drier conditions and in others towards humid conditions. For mean air temperature such a gridded data set (grid distance 1 deg lat and 1 deg long) is already existing covering the whole alpine territory (B6hm et al., 2000), but regional and vertical differences turned out to be of minor importance. The correlation within the alpine area is rather high, 90% of all sub-regions are inter-correlated by more than 0.8, only 4% by less than 0.7. The strongest decoupling occurs in winter between high alpine sites and the low level regions. The precipitation data set will be of more interest due to the very high spatial and temporal variability of the precipitation series.

° ~ ~ ° ~ 4. Application for A L P C L I M - Temperature signal in isotopic ice cores

Fig.2.1. Time series of the long term alpine precipttation stations.

3. Application for M A P - Mesoscale Alpine Climate (MAC). Studies on Climate Variability within the alpine region utilising homogenised long term climate time series have been and will be of great importance within the future too. Currently, MAC (comp. Meeting Report of MAP Climate Workshop, 2000) is an initiative to establish an alpine wide climate data bank with daily resolution consisting of elements like precipitation, air temperature, cloudiness/sunshine duration and many others more of at least 40 years lengths. The 1 to 1 deg gridded long term precipitation data set covering a time period of more than 200 years could be an important supplement to the daily data sets which most probably will not reach centennial time scales. For an easier and more systematic mathematically handling the single series will be interpolated to grid points. The interpolation will be performed by a Gauss weighing function with weight 1 at the grid point quickly decreasing to 0.1 at 200 km distance. In principle all single series contribute to each grid point, but with very small weights at larger distances from the grid point. In order to avoid large

Isotopic data from ice cores are often used as temperature proxis. But ice cores do not carry the signal of mean temperature in climatological sense (based on daily measurements) - what is stored in the ice mass can only be a temperature signal at precipitation days. As a few heavy precipitation days use to be responsible for most of the snow accumulated in the snow pack, a weighted temperature due to daily precipitation amounts should have the highest correlation to isotopic data. deg C

mm

n~

10 8 6 4 2 0 -2 -4

20 18

~AA IAV~ \ A II,',V,'X' I W I.~ /3, .IV l' ' . ' ; \ II;'.~\ /d~ ,//_\~/,..~.

ki/_IR~

rC,~ll/'

' ~," ',.\ b: ": ~ , ' ; V ; ,v,

-,

I

\,

16

1,

~2

'

_

6

-6 -8 10

,

4 2

12

J~ Y July

rrsum

#

August ~ t m a x

0

September .....

tmln

.......

trait

Fig.4.1. Case study: Time series of temperature and precipitation at Pian Rosa (3480 m asl) during summer 1981 (July to September).

423

1. Auer et aL: Data-Setfor the Alps The example of Figure 4.1. (time series of temperature and precipitation measured at Pian Rosa (3480 m) during summer 1981) shall underline that. For the three months period the respective values in question were: Mean temperature: -0.5 deg C Mean temperature on precipitation days: -1.1 deg C Mean temperature on dry days: +0.1 deg C Precipitation mass weighted temperature: -1.2 deg C In this single case the mean temperature on dry days turned out to be 1.2 deg warmer than that on precipitation days and 1.3 deg C warmer than the precipitation mass weighted temperature. To achieve a general feature about the described different temperatures in statistic terms the respective calculations were carried out for a sample of 19 different Alpine sites with daily temperature and precipitation data for the period 1961-1990. At high elevation sites mass weighted temperatures of precipitation days turned out to be cooler than temperatures at precipitation-free days all the year round (most pronounced in October and April in north and central alpine sites, least in southern alpine winter months). In general the effect is strongest in the Northern Alps, weaker in the Central Alps and weakest in the Southern Alps. At low level sites precipitation mass weighted temperatures are cooler than temperatures at dry days from March to October but warmer from December to February. Usually daily data sets are not available in homogenised form for long term series. Also the ALPCLIM temperature and precipitation data sets are based on monthly data. So we need a step from daily to monthly data. A concept for this step has been developed within the project and is going to be applied on the ALPCLIM data as soon as the homogenised precipitation data set is ready for use. The general ideas can briefly be shown here: The mean monthly differences between "mass weighted precipitation corrected temperatures" and the mean temperatures at precipitation free days (both based on daily data 1961-1990) be defined as AT(100%) and can be used to correct long term temperature series. The index "100%" stands for the 1961 to 1990 mean of precipitation amounts the single month-values of the homogenised long-term series of the ALPCLIM data set are given in percent of this standard period. To calculate the monthly mean precipitation-corrected temperature of any single month n of the series (Tcorr n) we used as a first guess the linear interpolation model: Tcorr n = T n + corrn (1) with T ° being the monthly mean temperature of all days and corrn = corr~c,, + AT(prec) (2) AT(prec) being linearly interpolated between the means of the standard period of: AT(100%)=0 (for prec = 100%) and AT(0%) at dry days (shown in Figure 4.2 for four regional means).

Examples of results for regional sub-groups are presented in Fig.4.2. There is a N-S-gradient of the corrections from higher values in the north to lower values in the south. In the northern Alps precipitation is more frequently connected with northern wind components and advection of colder air whereas for south Alpine sites there is also a considerable southern component at precipitation days which decreases the temperature difference to dry days. For low level sites winter precipitation frequently is combined with a mixing and therefore warming of shallow cold air masses ("masked cold fronts"). This effect causes higher temperatures at precipitation days than at dry days in the winter lowlands whereas high level sites are colder at precipitation days during the whole year. Fig.4.3 illustrates the described linear interpolation method for relative monthly precipitation totals from 0 to 200% (relative to the 1961 - 90 average). The simple linear model is in good accordance with the real values (based on daily data) in the shown range. For months with more than 200% precipitation it over-estimates the effect and will be replaced by more sophisticated methods with upper limits. Currently we test some models with truncation at climatologically defined upper limits for corr. delta T (K~

°

i

1

[

r

~___...

~

__¢,-

,," \

..~ -2

-3

B Jan

Feb

:

I Jul

Aug

mean north 1666m

" " -

mean centrat 2337m

mean south 1889rn

~ * " mean

Mar

Apt

May

dun

Sep

Oct

low level n o r t h

Nov

Dec

681 m

Fig.4.2. Mean monthly temperature at dry days minus precipitation-mass weighted temperaturefor the period 1961-1990, derived from 19 stations ( = AT(0%)used for formula2)

As final step monthly precipitation corrected temperature time series have been calculated for a subset of 19 sites and tested against the original temperatures in respect to high frequent variability (monthly correlation) and low frequent variability, in order to identify possible existing different trends. The high frequent variability of precipitation corrected and original temperature series turned out to be highly correlated (>0.9 in each case), what supports the assuriaption that a temperature signal in ice cores is repregentative not only for corrected but also for original temperatures.

424

[. Auer et al.: Data-Set for the Alps

The long term differences of precipitation corrected and original temperature series showed that their long term trend is not significantly different to zero. This means that using original temperature time series for comparison with isotopic ice core sites is admissible because original temperature series explain 80 to 90 % of the high frequent variance of precipitation corrected temperatures (which would be a better measure for isotopic ice cores). The differences of corrected and original long term series are not biased, their long term trend is not significantly different to zero. South-Alpine corr (K) 5 relat prec~p. (%)

4

Conclusions At present a long term homogenised gridded precipitation data set covering the whole alpine area and surroundings is in work. Time resolution is one month, the grid distance is 1 deg longitude and 1 deg latitude. Up to now nearly one half of the 140 original series have been tested for homogeneity and adjusted. This data in combination with the already existing gridded long term temperature data set will be a suitable basis for multiple alpine climate change studies. It offers also the possibility to adjust the long term temperature series to "precipitation mass weighted temperature series" which are assumed to represent the temperature signal stored in ice cores better than the normally used mean temperature series.

3 ---e-- 0

2

"',,,-20

1

_? , ~ . : .

0

, .

40 "t,-60

: . . . .

-0-80

-1 -2

100 --=-120 -o-140

-3

- x - 160

-4

- = - 180 --+-200

-5 Jan Feb Mar Apr May Jun

Jul

Aug Sep Oct Nov Dec

North-Alpine corr (K) relat prec=p (%)

4

,

;

,

:

;

,

:

i

¢

,

*

I

~

I

~

,

0

,

Auer, 1., B6hm R. and Schoner, W., 2000: Long Chmatic Series from Austria. Accepted for Chmatic Impacts." The Last 1000 Years. Kluwer Academic Publishers

--c- 80

MAP, 2000: MAP Climate Workshop 31 March 2000, Meeting Report: http://map.ethz.ch

-..*- 40 -.-o-60 ¢

Auer, I., Bohm R., Schoner, W. and Hagen, M., 1998: Endbertcht des Pro]ekts ALOCLIM (Austrian Long Term Climate). Centr Inst. f. Met. a. Geodyn. GZ 308.938/3-1V/B/3/96. Vtenna

--~- 100 ---t,- 120 -o--140

--.,,-20

]

References

B6hm, R., Auer, I., Hagen, M., and Schoner, W., 2000: Sctenttfic Report of ZAMG m: ALPCLIM. Envtronmental and Chmattc Records from High Elevation Alpine Glaciers, 2nd Year Progress Report, EC programme Environment and Climate 1994-1998, EC - contract: ENV4-CT-0693, co-ordinated by D. Wagenbach. (http://www.uphys.uni-heidelberg.de/glacis/ALPCLIM/homepage.html)

-4-0

0

Acknowledgements. Our thanks for data providing goes to the following institutions: Meteo France Toulouse, NOAA-GHCN Ashevdle, DWD Offenbach, Meteo Swiss Zurich, CNR-ISAO Bologna, UNI-Milano, SMS Tormo, HA-APB Bolzano, MHSC Zagreb, SHMI Ljubljana, HMS Budapest, SHMI Bratislava, ZAMG Vtenna

---x.- 160

:2

~180 --*-200 Jan Feb Mar Apr May Jun

Jul Aug Sep Oct Nov Dee

Fig.4.3. Mass wetghted precLpitation corrections (corr) for South-Alpine hlgh-elevaUon and North-Alpine high elevation temperatures for different relative precipitation amounts from 0 to 200% (1961-90 = 100%).

Peterson, T.C., Easterling, D.R., Karl, T.R., Groisman, P., N]choll, N.. Plummet, N., Torok, S., Auer, I., Boehm, R., Gullet, D.,Vincent, L., Heino, R., Tuomenvirta, H., Mestre, O., Szenumrey, T., Salinger, J., Forland, E., Hanssen-Bauer, I., Alexandersson, H., Jones, P., and Parker, D., Homogenett), ad)ustments of tn situ atmospheric chmate data: A revtew, Int. J. ChmatoL 18: 1493-1517, 1998 WMO, 1999: Proceedings of the 2nd Semmar for Homogemzatton of Surface Chmatological Data (Budapest, Hungary, 9-13 November 1998). WCDMP-No. 41. WMO-TD No.962. WMO, Geneva