Influence of changes in the station location and measurement routine on the homogeneity of the temperature, wind speed and precipitation time series/Ilmajaama asukoha ja mootmistingimuste muutmiste moju temperatuuri, tuule kiiruse ning sademete hulga aegridade homogeensusele.

Keevallik, Sirje ; Vint, Kairi

1. INTRODUCTION

The value of meteorological data, recorded all over the world, cannot be underestimated. Therefore very strict prescriptions are elaborated for measurement fields and routines [1]. Unfortunately, these guidelines are not always followed. Therefore even questions arise if the records are reliable and could be used at climate applications [2]. Problems also arise when measurement times are changed as most of the meteorological elements show daily cycles [3,4]. Precipitation records are sensitive to the measurement equipment and corrections for wind, wetting and evaporation [4,5]. Wind speed is affected by local orography and the openness of the measurement field. The recorded wind speed may be different for a traditional wind vane, anemorhumbometer or an automatic device [6].

Even in case when there are no changes in the measurement equipment and routine, the time series of meteorological data may be contaminated. One such widely known problem is urbanization that may contribute to the fake rise of surface air temperature [7]. Trends in cloud cover and wind are affected by growing forest: this is the case for two Estonian meteorological stations, Ristna and Tiirikoja. Cloud amount, recorded at Ristna, showed unrealistic decreasing trend because the growing forest shadowed the part of the sky near the horizon where the ground based observer could have recorded larger values of the cloud amount. That is why this station was left out of the analysis of Estonian cloud cover during 1955-1995 [8]. The Tiirikoja wind data does not serve as good input for models to get realistic wave regime on the Lake Peipsi. Most probably the questionable quality of the data can also be attributed to the growing trees around the measurement site [9].

A poor station location is rather widely analysed in the context of temperature measurements. Several attempts have been made to correct the errors with various adjustment methods [10,11]. On the basis of data from Taiwan, in [12] it is demonstrated how artificial discontinuities in the temperature time series occur due to station relocation. Such problems lead to elaboration of the methods to detect inhomogeneities in the observed data series [13-15]. To prepare homogenization of the time series, attention should be paid to the metadata [16]. An analysis of the changes in the station location, measurement instruments or observation routines is the basic step by the detection of discontinuities in the meteorological time series and their homogenization.

Tarand [17] has made an attempt to reconstruct a homogeneous time series of the air temperature in Tallinn for the period of 1756-2002 on the monthly basis. For the period of 1850-2002 (that may be called a period of modern observations), the following operations were conducted: reducing all different time observations to a 24-hour average, filling in the gaps in the data by means of the recordings at the neighbouring stations, and eliminating meso-scale impact at measurement sites by means of parallel observations. The attempts to homogenize older data were based on parallel measurements and recordings at Paldiski and St. Petersburg.

The aim of the present paper is to present metadata for three Estonian meteorological stations and to check if the changes in the location, instruments and observation schedule have introduced inhomogeneities into the time series of the principal meteorological parameters--daily and monthly mean temperature, wind speed and precipitation sums.

2. MATERIAL AND METHODS

The stations under consideration are situated in different parts of Estonia: Tallinn at the coast of the Gulf of Finland, Tartu in South Estonia and Parnu at the coast of the Gulf of Riga (Fig. 1). For these stations all changes have been registered. The dates of the changes are labelled as possible breakpoints and the homogeneity of the time series is checked by means of a standard test. For this purpose, the time series were divided into different periods that cover the time intervals between the instants of changes. A two-sample Student's Z-test has been applied to determine whether the averages of the meteorological parameters during different periods are equal. To decide whether or not the samples had equal variance, the .F-test was applied. The differences were detected on the significance level of 0.05.

The analysis of the possible breakpoints was only performed for cases when only one non-climatic change occurred at the breakpoint - either a relocation of the station or a change of the equipment or a change in the observation times. The periods between the instants of changes only contained whole years in order to eliminate possible systematic differences due to the annual cycles of the meteorological parameters.

The periods when no change in the location, instrumentation or measurement routine took place, were sometimes very short. In the present paper only time intervals with the length of at least ten years were considered. A comparison of shorter time periods may reveal changes due to the natural variability of meteorological parameters that could be ascribed to the consequences of artificial changes.

It can be said that all changes in the measurement times took place at all meteorological stations simultaneously. Until 1966 all measurements were carried out according to the Local Mean Time (GMT + 2 h). On the 1st of January 1966 the Moscow Time was introduced (GMT + 3 h). In 1992 the Moscow Time was replaced by the GMT (Greenwich Mean Time), but this did not change the measurement routine as the difference between these two time systems is three hours. Such reorganization only changed the division of measurement instants between subsequent calendar days: for the sake of homogeneity, a new date starts at 21 GMT.

[FIGURE 1 OMITTED]

To get daily average values of temperature from the existing inhomogeneous data set, special correction coefficients have been proposed [18]. These coefficients represent average differences between the mean value, calculated from the temperature of three (or four) observations a day, and the average daily temperature, calculated from the thermograph data. In the present paper no corrections have been introduced.

It is not recommended to directly compare the daily average wind speed for the periods with different observation times [4]. Therefore, wind data are analysed for a shorter period of 1966-2010 when no changes in observation times took place.

Precipitation is not measured according to the same scheme as temperature and wind. The rain gauges are checked up to 4 times per day, sometimes with unequal intervals. A change in the number of observations per day may introduce systematic errors into the daily and monthly sums in case no wetting correction is applied [19]. In Estonia, the wetting correction was introduced in 1966. When the strict homogeneity of the time series is necessary, the wetting correction should be added to all earlier measurements [4].

Traditional manual measurements were replaced by automatic weather stations at the beginning of the last decade. These stations record data continuously and special filtering is needed to simulate traditional measurements to keep the time series as homogeneous as possible [20]. Automatic weather stations in Estonia use the following Vaisala equipments: temperature sensor DTS-12A, averaging wind display unit WAD21M, anemometer WAA151, rain and precipitation sensor RG13H and weighing gauge VRG101.

3. RESULTS

3.1. Metadata for the Tallinn meteorological station

Although the instrumental observations date back to the end of the 18th century [17], the present paper describes changes for the period of 1931-2010 (tables 1 to 3). During this period the station has moved from the vicinity of the Tallinn Lower Lighthouse to Kose, later to Ulemiste and finally to Harku. Tallinn Lower Lighthouse (59[degrees]26'N, 24[degrees]48'E) is situated on the cliff above the eastern part of the city. Kose (59[degrees]28'N, 24[degrees]49'E) is located lower, not far from the cliff. Ulemiste meteorological station (59[degrees]24'N, 24[degrees]36'E) was established at the airport, near the southern border of the city. Harku (59[degrees]38'N, 24[degrees]58'E) is situated some kilometres to the west of Tallinn.

3.2. Metadata for the Tartu meteorological station

The time series of temperature starts in 1821, but in the present paper the period of 1881-2010 is considered (tables 4 to 6). During the time span of 129 years, the measurements were carried out at the Tartu Meteorological Observatory that was situated at different places inside the town (58[degrees]23'N, 26[degrees]43'E). In 1950 the station was moved out of the town, 7 km to the south to Ulenurme (58[degrees]18'N, 26[degrees]41'E). Since 1997 the station is situated 25 km to the south-west of Tartu at Toravere (58[degrees]16'N, 26[degrees]28'E).

3.3. Metadata for the Parnu meteorological station

Meteorological observations started at Parnu in 1842, but in the present study the period of 1901-2010 is considered (tables 7 to 9). During this time the station was relocated four times. Mostly this took place in the boundaries of the city, e.g. 58[degrees]23'N, 24[degrees]30'E when the observation site was in a courtyard of Nikolai Str. 21. In 2004 the station was moved out of the town to the airport (58[degrees]25'N, 24[degrees]28'E). Until the 23rd of December 2004 the measurements were carried out using the mercury thermometer, but this was placed differently during different periods. The same can be said about the older versions of rain gauge.

3.4. Breakpoints in the time series of daily data

From Tables 1-9, only six changes were identified that provided a possibility to compare periods of the length of ten or more years where only one change took place. They are shown in Table 10 together with the changes introduced to the meteorological time series of daily data.

In both cases that could be checked (Tallinn and Parnu) the relocation of the station introduced statistically significant inhomogeneities into the daily temperature time series. The change affected both the variance and average value. The average temperature after the breakpoint was higher and the variance was lower than before the breakpoint.

Relocation of Tallinn and Parnu stations was also accompanied by an increase in the average values of the daily precipitation sum. In these cases the variance after the change was higher than before.

The Tallinn station was moved from Ulemiste (flat landscape of the airport) to Harku (on the cliff). It is difficult to find the reason why temperature has risen and precipitation sums increased after the relocation. The Parnu station was moved from the beach to the city centre. In this case the rise of temperature and the increase in the precipitation sums could be expected due to the urban conditions.

The change in the observation times from 01, 07, 13, 19 Local Time to 00, 03, 06, 09, 12, 15, 18, 21 GMT at Tartu-Ulenurme led to an increase in the average temperature and a decrease in the variance.

The only possibility to check the impact of the breakpoints on the wind speed data was at Tartu-Ulenurme where replacing the weather station UATGMS by anemograph M-12 led to an increase in the average wind speed from 3.8 m/s to 4.0 m/s.

3.5. Breakpoints in the time series of monthly data

For the same 10-year periods also monthly values of the meteorological parameters were tested. It turned out that these time series were considerably less affected by changes in the station location, equipment or measurement routine (Table 11).

4. DISCUSSION

Many papers by different authors address the climate change in Estonia and report increasing trends in temperature and precipitation and a decreasing trend in wind speed [2122]. These findings are based on the time series, collected at different meteorological stations that may have undergone changes in the location, instruments and observation times. The discontinuities detected in the present paper may partly overlap with natural trends in the meteorological parameters, especially when the time periods under comparison are long. Long periods are traditionally believed to guarantee stable statistical estimates, a feature that is not necessarily true for shorter periods, especially if the data sets exhibit large variability. In the present analysis, in which we have used at least ten years long time periods, all changes in temperature and precipitation showed an increase. Therefore, natural and artificial changes have the same sign and it is difficult to separate them.

One might argue that differences in the average values of meteorological elements detected in the present work are totally caused by natural trends. It would be so if the behaviour of other statistical parameters was homogeneous. In the present case the changes in the average values were always accompanied by changes in the variance.

The method applied in the present paper assumes that there exist one or several homogeneous populations of temperature, wind speed and precipitation sums. Actually, the time series do not consist of independent random values and the entire climate system should be regarded as the concurrent array of weather patterns [23]. Therefore it is highly possible that the variability of the time series is sample-dependent.

To separate the signal of climate change from artificial influences, special procedures are recommended that are also valid for non-stationary series [24]. From this point of view, climate change should not be regarded as a change at the moments of the relevant probability distributions. One possibility is to analyse deviations from the average annual cycle and to estimate the frequency of the outliers [25]. If climate change is not defined by means of average values and variances, it may turn out that the artificial changes do not affect the signal of the changing structure of the variability of the weather (i.e. climate) even if the data is recorded by different means.

doi: 10.3176/eng.2012.4.02

ACKNOWLEDGEMENTS

The meteorological data were drawn from the archives of the Estonian Meteorological and Hydrological Institute. The research was supported by the targeted financing by the Estonian Ministry of Education and Research (grant SF0140017s08).

REFERENCES

[1.] WMO. Guide to Meteorological Instruments and Methods of Observation (6th ed.). WMO, Geneva, 1996.

[2.] Davey, C. A. and Pielke, R. A. Sen. Microclimateexposures of surface-based weather stations. Bull. Amer. Meteorol. Soc., 2005, 86, 497-504.

[3.] Vose, R. S., Williams, C. N. Jnr., Peterson, T. C., Karl, T. R. and Easterling, D. R. An evaluation of the time of observation bias adjustment in the U.S. Historical Climatology Network. Geophys. Res. Lett., 2003, 30, Art. No. 2046.

[4.] Keevallik, S., Loitjarv, K., Rajasalu, R. and Russak, V. Meteorological regime of Lake Peipsi. In Lake Peipsi, Meteorology, Hydrology, Hydrochemistry (Noges, T., ed.). Sulemees Publishers, Tartu, 2001, 18-37.

[5.] Ungersbock, M., Rubel, F., Fuchs, T. and Rudolf, B. Bias correction of global daily rain gauge measurements. Phys. Chem. Earth (B), 2001, 26, 411-414.

[6.] Keevallik, S., Mannik, A. and Hinnov, J. Comparison of HIRLAM wind data with measurements at Estonian coastal meteorological stations. Estonian J. Earth Sci., 2010, 59, 90-99.

[7.] Ren, Y. Y. and Ren, G. Y. A remote sensing method of selecting reference stations for evaluating urbanization effect on surface air temperature trends. J. Climate, 2011, 24, 3179-3189.

[8.] Keevallik, S. and Russak, V. Changes in the amount of low clouds in Estonia (1955-1995). Int. J. Climatol., 2001, 21, 389-397.

[9.] Rosin, K. and Keevallik, S. Regional variation of hourly and daily totals of global radiation recorded at automatic weather stations in Estonia. Estonian J. Eng., 2012, 18, 76-86.

[10.] Peterson, T. C. Examination of potential biases in air temperature caused by poor station locations. Bull. Amer. Meteorol. Soc., 2006, 87, 1073-1089.

[11.] Pielke, R. Sr., Nielsen-Gammon, J., Davey, C., Angel, J., Bliss, O., Doesken, N., Cai, M., Fall, S., Niyogi, D., Gallo, K. et al. Documentation of uncertainties and biases associated with surface temperature measurement sites for climate change assessment. Bull. Amer. Meteorol. Soc., 2007, 88, 913-928.

[12.] Hung, C. Temperature discontinuity caused by relocation of meteorological stations in Taiwan. Terr. Atmos. Ocean. Sci., 2009, 20, 607-617.

[13.] Domonkos, P. and Stepanek, P. Statistical characteristics of detectable inhomogeneities in observed meteorological time series. Stud. Geophys. Geod., 2009, 53, 239-260.

[14.] Ducre-Robitaille, J.-F., Vincent, L. A. and Boulet, G. Comparison of techniques for detection of discontinuities in temperature series. Int. J. Climatol., 2003, 23, 1087-1101.

[15.] Brown, P. and DeGaetano, A. T. A method to detect inhomogeneities in historical dewpoint temperature series. J. Appl. Meteorol. Climatol., 2009, 48, 2362-2376.

[16.] Aguilar, E., Auer, I., Brunet, M., Peterson, T. C. and Wieringa, J. Guidelines on Climate Metadata and Homogenization. WCDMP-No. 53, WMO-TD No. 1186, WMO, Geneva, 2003.

[17.] Tarand, A. Tallinnas moodetud ohutemperatuuri aegrida. Publ. Geophys. Univ. Tartu., 2003, 93, 25-36.

[18.] Handbook of the Climate of the USSR, 4, Estonian SSR, Part II, Air and Ground Temperature. Gidrometeoizdat, Leningrad, 1965 (in Russian).

[19.] Handbook of the Climate of the USSR, 4, Estonian SSR, Part IV, Air Humidity, Precipitation and Snow Cover. Gidrometeoizdat, Leningrad, 1968 (in Russian).

[20.] Keevallik, S., Soomere, T., Parg, R. and Zukova, V. Outlook for wind measurement at Estonian automatic weather stations. Proc. Estonian Acad. Sci. Eng., 2007, 13, 234-251.

[21.] Jaagus, J. Climatic changes in Estonia during the second half of the 20th century in relationship with changes in large-scale atmospheric circulation. Theor. Appl. Climatol., 2006, 83, 77-88.

[22.] Keevallik, S. and Soomere, T. Trends in wind speed over the Gulf on Finland 1961-2000. In Proc. 4th Study Conference on BALTEX. Gudhjem, Denmark, 2004. International BALTEX Secretariat, Geesthacht, Germany, 2004, 129-130.

[23.] Bryson, R. A. The paradigm of climatology: An essay. Bull. Amer. Meteorol. Soc., 1997, 78, 449-455

[24.] Karner, O. ARIMA representation for daily solar irradiance and surface air temperature time

series. J. Atm. Sol-Terrest. Phys., 2009, 71, 841-847.

[25.] Karner, O. Tartu ohutemperatuuri tolerants ja kliima muutlikkus. Publ. Inst. Geogr. Univ. Tartu., 2011, 109, 58-75.

Sirje Keevallik (a) and Kairi Vint (b)

(a) Marine Systems Institute at Tallinn University of Technology, Akadeemia tee 15a, 12618 Tallinn, Estonia; sirje.keevallik@msi.ttu.ee

(b) Estonian Meteorological and Hydrological Institute, Mustamae tee 33, 10616 Tallinn, Estonia; kairi.vint@emhi.ee

Received 31 May 2012, in revised form 28 August 2012

Table 1. Changes at the Tallinn meteorological station that
may affect the temperature time series

Date of the    Location    Instrument      Observation
change                                        time

01.01.1931    Tallinn      Mercury       07,13,21
              Lower        thermometer   Local Time
              Lighthouse
01.01.1941                               01,07,13,19
                                         Local Time
01.09.1941                               07,13,21
                                         Local Time
01.01.1945                               01,07,13,19
                                         Local Time
01.09.1948    Kose
01.01.1965    Ulemiste
01.01.1966                               00,03,06,09,
                                         12,15,18,21
01.05.1980    Harku                      Moscow Time/GMT
01.09.2003                 Vaisala
                           DTS-12A

Table 2. Changes at the Tallinn meteorological
station that may affect the wind speed time
series

Date of the   Location      Instrument
change
01.01.1966    Ulemiste   Anemorhumbometer
                         M63-M1, height
                         10.7 m
01.01.1973               Anemorhumbometer
                         M63-M1, height
                         10.0 m
01.05.1980    Harku
01.01.1997               Vaisala WAD21M
01.09.2003               Vaisala WAA151

Table 3. Changes at the Tallinn meteorological station that may
affect the precipitation time series

Date of the    Location      Instrument     Observation time
change

01.01.1931    Tallinn      Rain gauge       Once a day
              Lower        with the
              Lighthouse   Nifer shield
01.01.1945                                  07,19 Local Time
01.09.1948    Kose
01.09.1952                 Tretyakov
                           rain gauge
01.01.1965    Ulemiste
01.01.1966                 Tretyakov        00,06,12,18 GMT
                           rain gauge
                           with wetting
01.05.1980    Harku        correction
01.04.1981                                  06,18 GMT
01.01.1984                                  03,06,15,18 GMT
01.09.2003                 Vaisala RG13H
01.11.2003                 + Tretyakov      06,18 GMT
11.02.2005                                  06,12,18 GMT
03.02.2006                 Vaisala VRG101
01.05.2009                 + Tretyakov      06,18 GMT

Table 4. Changes at the Tartu meteorological station that may
affect the temperature time series

Date of the   Location   Instrument    Observation time
change

01.01.1881    Tartu,     Mercury       07,13,21 Local
              Tiigi 1    thermometer   Time
01.01.1893    Tartu,
              Tiigi 15
01.01.1926    Tartu,
              Liivi 3
01.01.1941                             01,07,13,19 Local
                                       Time
01.09.1941                             07,13,21 Local
                                       Time
01.01.1945                             01,07,13,19 Local
                                       Time
01.01.1950    Ulenurme
01.01.1966                             00,03,06,09,12,
                                       15,18,21
01.01.1997    Toravere                 Moscow Time/GMT
01.09.2003               Vaisala
                         DTS-12A

Table 5. Changes at the Tartu meteorological
station that may affect the wind speed time
series

Date of the   Location      Instrument
change
01.01.1966    Ulenurme   Anemorhumbometer
                         M63-M1, height
                         12 m
01.01.1969               UATGMS weather
                         station, height
                         12 m
01.01.1986               Anemograph M-12
01.01.1997    Toravere   Vaisala WAD21M,
                         height 10 m
01.09.2003               Vaisala WAA151

Table 6. Changes at the Tartu meteorological station that
may affect the precipitation time series

Date of the   Location    Instrument     Observation
change                                       time

01.01.1881    Tartu,     Rain gauge      Once a day
              Tiigi 1    in garden,
                         height 1 m
01.01.1893    Tartu,     Rain gauge
              Tiigi 15   in garden,
                         height 2 m
01.01.1900               Rain gauge
                         with the
                         Nifer shield
                         on the roof
01.01.1926    Tartu,     Rain gauge
              Liivi 3    with the
                         Nifer shield,
                         height 2 m
01.01.1945                               07,19 Local
                                         Time
01.01.1950    Ulenurme   Tretyakov
                         rain gauge
01.01.1966               Tretyakov       00,06,12,18
                         rain gauge      GMT
                         with wetting

01.04.1981                               06,18 GMT
01.01.1984                               03,06,15,18
                                         GMT
01.01.1997    Toravere
01.09.2003               Vaisala
                         RG13H +
                         Tretyakov
01.11.2003                               06,18 GMT
11.02.2005                               06,12,18 GMT
01.05.2009                               06,18 GMT
28.10.2010               Vaisala
                         VRG101 +
                         Tretyakov

Table 7. Changes at the Parnu meteorological station that may
affect the temperature time series

Date of the     Location       Instrument    Observation time
change
01.01.1901    At the          Mercury        07,13,21 Local
              pilot tower     thermometer,   Time
                              height 3.4 m
01.08.1914                    Mercury
                              thermometer
                              on the roof
31.05.1921                    Mercury
                              thermometer,
01.01.1941                    height 2 m     01,07,13,19 Local
                                             Time
01.09.1941                                   07,13,21 Local
                                             Time
01.01.1945                                   01,07,13,19 Local
                                             Time
07.11.1947    Open site
              on the beach,
              100-150 m
01.01.1966    from the                       00,03,06,09,12,
              water line                     15,18,21
23.04.1971    250 m from                     Moscow Time/GMT
              the beach
              house, open
              to SW, W and
              NW
07.09.1990    Nikolai Str.
              21
23.12.2004                    Vaisala
                              DTS-12A

Table 8. Changes at the Parnu meteorological station
that may affect the wind speed time series

Date of the     Location         Instrument
change
01.01.1966    Open site       Wind vane,
              on the beach,   height 16 m
              100-150 m
              from the
              water line
23.04.1971    250 m from
              the beach
              house, open
01.01.1975    to SW, W and    Anemorhumbometer
              NW              M-63M-1, height
                              15.5 m
07.09.1990    Nikolai Str.    Anemorhumbometer
              21              M-63M-1 on the
                              roof, height from
                              the roof 4.35 m
                              and from the
                              ground 25 m
01.01.1997                    New type of
                              M-63M-1 on the
                              roof, height 25 m
23.12.2004                    Vaisala WAA151,
                              height 10 m

Table 9. Changes at the Parnu meteorological station that may affect
the precipitation time series

Date of the          Location                  Instrument
change

01.01.1901    At the pilot tower        Rain gauge, height 6.1 m
01.08.1905                              Rain gauge, height 9.1 m
07.07.1907                              Rain gauge, height 6.1 m
01.01.1921                              Rain gauge with the Nifer
01.01.1945                                shield
07.11.1947    Open site on the beach,
01.04.1952      100-150 m from the      Tretyakov rain gauge
01.01.1966      water line              Tretyakov rain gauge with
23.04.1971    250 m from the beach        wetting correction
01.04.1981      house, open to SW,
01.01.1984      W and NW
07.09.1990    Nikolai Str. 21
01.09.2003
23.12.2004    Sauga airport             GEONOR T-200B
11.02.2005                                precipitation gauge
01.05.2009
14.09.2010                              VRG101

Date of the     Observation time
change

01.01.1901    Once a day
01.08.1905
07.07.1907
01.01.1921
01.01.1945    07,19 Local Time
07.11.1947
01.04.1952
01.01.1966    00,06,12,18 Moscow
23.04.1971      Time/GMT
01.04.1981    06,18 Moscow Time/GMT
01.01.1984    03,06,15,18 Moscow
07.09.1990      Time/GMT
01.09.2003    06,18 GMT
23.12.2004
11.02.2005    06,12,18 GMT
01.05.2009    06,18 GMT
14.09.2010

Table 10. Statistically significant differences detected at
the comparison of time periods of the length of at least 10
years from daily averages of temperature and wind speed and
daily precipitation sums

Station    Periods      Change in     Parameter
           compared

Tallinn    1966-1979/   Location      Temperature
           1981-2002                  Precipitation

Tartu      1950-1965/   Observation   Temperature
           1966-1996      times
           1969-1985/   Instrument    Wind speed
           1986-1996

Parnu      1972-1988/   Location      Temperature
           1991-2003                  Precipitation

Station    Periods      Variance               Average
           compared

Tallinn    1966-1979/   83/70 [C.sup.2]        5.0/5.8 C
           1981-2002    13.5/17.7 [mm.sup.2]   1.7/1.9 mm

Tartu      1950-1965/   94/90 [C.sup.2]        4.6/5.1 C
           1966-1996
           1969-1985/   3.4/2.7 [m.sup.2]/     3.8/4.0 m/s
           1986-1996      [s.sup.2]

Parnu      1972-1988/   84/75 [C.sup.2]        5.7/6.7 C
           1991-2003    15.3/18.3 [mm.sup.2]   1.9/2.1 mm

Table 11. Statistically significant differences detected
at the comparison of time periods of the length of at
least 10 years from monthly averages of the wind speed
and monthly precipitation sums

Station     Periods     Change in      Parameter
            compared

Tallinn    1966-1979/   Location     Precipitation
           1981-2002

Tartu      1969-1985/   Instrument   Wind speed
           1986-1996

Station     Periods          Variance          Average
            compared

Tallinn    1966-1979/   974/1242             53/59 mm
           1981-2002    [mm.sup.2]

Tartu      1969-1985/   0.6/0.5 [m.sup.2]/   3.8/4.0 m/s
           1986-1996    [s.sup.2]