Printer Friendly

Regional variations in hourly and daily totals of global radiation recorded at automatic weather stations in Estonia/Eesti automaatsetes ilmajaamades moodetud summaarse kiirguse tunni- ja paevasummade regionaalne jaotus.


It is well known that diurnal variation of the surface solar radiation fluxes causes diurnal variation in the sea surface temperature and upper layer vertical stratification [1]. The related deepening/shoaling of the upper mixed layer on the daily scale will also affect the biological and material circulation processes. Marine Systems Institute performs high resolution measurements of temperature, salinity and chlorophyll a fluorescence in the Gulf of Finland. Two transects are recorded per day between Tallinn and Helsinki by means of the Ferrybox system [2]. An autonomous buoy profiler, deployed off the Tallinn Bay, records vertical profiles every third hour [3]. To explain the observed diurnal cycles in the surface layer temperature, vertical stratification and chlorophyll a fluorescence distribution, estimation of surface solar radiation fluxes (including diurnal cycle) should be available. It has not been verified yet whether the estimates, based on model outputs, could be applied for this purpose, especially in cloudy conditions.

Estimates of the surface solar radiation fluxes over the sea are important for oceanographic applications in both long-term (from a month to a season) and short-term (diurnal) scales. Modelling of the development of vertical stratification in an estuary needs (among other parameters) knowledge on heat fluxes at the sea surface [4]. The long-term changes, e.g., in the sea surface temperature, are usually correlated with the radiation or sun hours data from coastal stations [5]. A similar approach was applied in a study of inter-annual variation of the late-summer cyanobacteria blooms in the Gulf of Finland, where the radiation data recorded at Tartu-Toravere meteorology station were used to link the observed bloom intensities to the changes in photosynthetically active radiation [6]. Toravere is situated approximately 200 km from the Estonian coastline, but up to recent times it was the best site in Estonia, where all components of the radiation budget are recorded. The other site was Tiirikoja that is situated somewhat closer to the Gulf of Finland, but Tartu-Toravere was preferred as it is a BSRN station where the quality of data is guaranteed [7].

Since 2003, Estonia has step by step replaced traditional measurement routine at the meteorological stations by automatic equipment. Automatic weather stations offer new possibilities to estimate solar radiation parameters by means of certain models that use meteorological information (e.g. [8]). On the other hand, many stations have been complemented with actinometric equipment that measure directly solar radiation. In Estonia, pyranometers have been installed at several coastal meteorological stations that should offer a possibility to get better input to oceanographic models that need radiation data.

It is widely known that conditions for radiation measurements are extremely strict. They need open horizon (especially in winter when the sun is low), periodic calibration of pyranometers, regular control of the condition of the receivers, etc.). Unfortunately these requirements are in many cases not met at Estonian meteorological stations (except Toravere).

The goal of the present paper is to investigate relationships between global radiation at the coastal stations and Toravere. First, this draws attention to the problems of radiation measurements at the automatic weather stations and, second, this might lead to the possibilities of reconstruction coastal global radiation from Toravere data in case no measurements are carried out at the site of interest. The stress is put on daily and hourly totals. Due to the short observation period (2005-2010) and gaps in data series, it was not possible to derive direct climatological estimates. On the other hand, it was still possible to find common periods of 2-3 years when data were available for all or several observation sites. This gave us a possibility to give a rough estimate of the spatio-temporal distribution of the solar radiation characteristics.

Approximate information on the mean distribution of annual totals of global radiation over the territory of Estonia is presented in the Handbook of Estonian Solar Radiation Climate [9], where the long-term average distribution is calculated on the basis of mean cloudiness and albedo values at 31 meteorological stations. Later we compare our estimates with those described in [9].


In the network of Estonian Weather Service there are eight meteorological stations, where global radiation is measured (Fig. 1, Table 1). Radiation measurements at Pakri were terminated in 2009. At Haapsalu and Roomassaare the automatic stations were installed somewhat later than others, in 2007 and 2008, respectively. In the present study, radiation data from six stations during 2005-2010 were used, stations with shorter observation period were left out.

To measure the downward and upward fluxes of solar radiation, Kipp & Zonen pyranometers CM11 and CM21 are used. All stations with the exception of Tartu-Toravere provide hourly mean radiation flux densities (W/[m.sup.2]). At Tartu-Toravere one minute mean values are gathered and processed.

The measurements of the openness of the horizon were carried out at all stations in 2001 [9]. Since then the forest around Tiirikoja has grown and shades the instruments even more. The estimates for Parnu meteorological station are not available, as the station was relocated in 2003. It is situated at the airport and according to visual estimates the horizon is not shaded considerably. From Table 1 it can be concluded that winter data from Tiirikoja and Narva-Joesuu may be underestimated.


The radiation equipment is calibrated regularly at Toravere. At Tiirikoja one calibration was carried out, on 28 May 2008. No changes in sensibility were detected. The pyranometers at other stations are not calibrated during the period under consideration.

The pyranometers are ventilated to prevent dew and frost at Toravere, Harku and Parnu. Thermal offset corrections are applied in measurements at Toravere. Special instructions are given to the personnel of meteorological stations to check the condition of the receivers [10]. This should avoid the situations when the pyranometer is covered with snow or ice, etc.

The data of Toravere passes strict quality control before it is transmitted to the archives of EMHI (Estonian Meteorological and Hydrological Institute), BSRN and WRDC (World Radiation Data Centre). Datasets from other stations are not checked so thoroughly--before sending to the archives, only these values are removed that are obviously erroneous.


The whole dataset (except data from Tartu-Toravere) contained a number of missing values. Major causes of missing values at daytime were changes in the sensor configuration and temporary interruptions of the automatic stations work.

Nighttime values contained several types of anomalies--there were missing, negative and small positive values. Negative and small positive values are related to the zero offset of the sensor [11]. This is a widely recognized problem of pyranometers that may lead to discrepancies between modelled and measured solar radiation [12]. Thermal offset corrections are applied at the Toravere BSRN station.

The initial data from Harku station contained the greatest number of missing values. The reason here was elimination of negative values during the data transmission from automatic station to the archive. Mostly data were missing during nighttime when there is no solar radiation, but often missing data were shown also for daytime hours, especially for morning and evening.

At Vilsandi, the data from 30 July 2008 to 30 June 2010 were obviously erroneous. Due to incorrect sensor installation at the station, radiation values never exceeded 640 W/[m.sup.2] during the period.


To estimate differences between radiation regime at Toravere and coastal stations, daily totals were calculated from "cleaned" data sets. "Cleaning" was carried out separately for every coastal station depending on the detected problems. At all stations the days were left out when at least one hourly measurement was missing. At Vilsandi the period from 30 July 2008 to 30 June 2010 (when sensor problems were detected) was left out.

At Parnu, two periods when solar radiation was systematically shown at night were left out (20 February to 2 November 2008 and 30 January to 14 September 2009). Additional analysis showed that nighttime recordings formed only 0.2% of the long-term average daily total, but it could be suspected that the sensor or recording regime was also biased. At Harku the zero and missing data were distinguished by means of sunset and sunrise times and records with missing data were left out.

At the coastal stations there exist several cases when in winter the recorded daily total was 0. Keeping in mind that the absolute minimum of the daily total during the period under consideration at Toravere was 0.13 MJ/[m.sup.2], these cases were checked by means of historical atmospheric phenomena records that showed fog, rain or snowfall. Therefore, these results should be considered realistic and not be attributed to some mistake in the measurement routine.

As seen from the above, different periods were left out at different stations. This means that the comparison of the radiation regime at different stations could be carried out for shorter periods that are common to all (or at least to most of the) stations. For different months these periods were different (Fig. 2). In October, the common period for all stations was less than two months. Therefore only four stations are considered where the common period was longer. In January, Tiirikoja and Narva-Joesuu were left out due to the restricted openness of the horizon that might introduce systematic errors.

The following (approximate) features of the spatio-temporal distribution of solar radiation can be seen.

--In April there is more sunshine in West-Estonia than in East-Estonia.

--In July the sunniest places are seaside resorts Parnu and Narva-Joesuu as well as the westernmost island of Vilsandi. Harku meteorological station is situated at least 5 km from the sea on a cliff. Here complicated orography and the neighbourhood of a large city Tallinn affect the meteorological regime.

--An interesting feature can be noted concerning two sites on the northern coast: there is more sunshine at Harku in April and at Narva-Joesuu in July.

--In October the radiation conditions in North-Estonia and East-Estonia are similar, most probably due to extensive homogeneous cloud cover.


To get an overview on the distribution of solar radiation over the Estonian territory, a common period of 2005-2007 (with missing data from October 2005 to March 2006, and August 2007) could be found for which annual average daily totals were calculated. Figure 3 shows that the amount of solar radiation at Parnu and Vilsandi exceeds distinctly that on the northern coast and inland. This is partly due to astronomical factors: in January the TOA (Top of the Atmosphere) radiation on the northern coast forms approximately 86% of that above Toravere. In October this percentage is around 94. The annual averages at Tiirikoja and Narva-Joesuu might be underestimated, as winter data are included in annual average calculations. On the other hand, in case only the period from March to November is considered, Harku, Tiirikoja and Narva-Joesuu show similar values (not shown in the present paper).



The same data from common time periods were used to calculate daily cycles of global radiation. Figure 4 shows results for April and July when a common three-year period could be found for all stations. Figure 4 also shows that differences between radiation conditions at different stations are larger in spring than in summer.

At Toravere, the maximum of the solar radiation is measured around local noon (10:00 GMT denotes the hour from 9:00 to 10:00 GMT or from 11:00 to 12:00 winter EET) whereas solar radiation maximum at Vilsandi is about an hour later. The solar time difference between these stations is approximately 19 min. Therefore, this time lag must be due to meteorological conditions.

In July the differences between stations are the largest around 12:00 GMT, i.e., during the afternoon hours. Most probably these differences stem from the cloudiness that is more extensive at the inland sites.



To estimate global radiation at the coastal stations from Toravere data, respective regression equations can be calculated for each station. For this purpose, additional data on cloud cover was used and cases were chosen when the cloudiness conditions were similar at both stations. Cloudiness is recorded with 3-h intervals. Therefore, we chose for comparison the afternoon hour of 11:00-12:00 GMT (13:00-14:00 EET). Unfortunately, since the 1st of May 2009, clouds are recorded at Narva-Joesuu and Tiirikoja only at 06:00 and 18:00 GMT.

Table 2 shows that hourly totals at Tiirikoja and Parnu can be restored from the Toravere data rather well. The square of the correlation coefficient (coefficient of determination) is 0.68 for Tiirikoja and 0.65 for Parnu. This means that 65%-68% of the variability of the hourly totals of global radiation at these sites is determined by the variability of the fluxes at Toravere. Correlation is low for Vilsandi and Narva-Joesuu, showing that approximately 40% of the variability can be ascribed to the variability at Toravere. Figure 5 presents two examples of such regression, demonstrating the best and the worst correlation.

In case only clear conditions were chosen (actually coverage up to 1/10), correlation is perfect as expected. Figure 6 shows that even for Vilsandi, where the coefficient of correlation in overcast conditions was the lowest, global radiation can be well derived from Toravere data.




The results of the present paper can be divided into two groups.

First, comparison of the simultaneous radiation measurements at coastal stations and Toravere enables one to estimate the approximate solar radiation regime in different regions. Another approximate description of the radiation climatology is shown in [9], where global radiation is estimated on the basis of mean cloudiness and albedo values using the formula of Averkiev [13]. The authors of [9] confirm that annual totals, got by such indirect method, describe the radiation climate roughly. Our estimates are based on direct measurements. Although there are problems with the openness of the horizon and a lot of data were labelled as not reliable, the general features of the radiation regime are similar on the annual basis: there is more sunshine on the West-Estonian islands and West-Estonian coast and less on the North-Estonian coast. Although Toravere is an inland station, it seems to be a favorable site for solar radiation [14]. Handbook [9] gives also a possibility to compare calculated and measured radiation: at Kuusiku direct measurements were carried out during 1954-1963 and at Tooma 1956-1963. It seems that calculated annual totals of global radiation are underestimated by 1% at Kuusiku and 4% at Tooma.

In the present paper also seasonal differences of global radiation are described. Here an interesting feature may be noticed concerning two sites on the northern coast: in April the daily totals at Harku are larger than at Narva-Joesuu, and in July vice versa. This phenomenon is worth further analysis, as the comparison is carried out on a 3-month basis only.

As a result, in case the measurement conditions at the coastal stations are improved, direct measurements give the possibility to describe the spatiotemporal distribution of solar radiation more precisely.

Second, application of Toravere radiation data by marine investigations is discussed. It can be said that direct transfer of inland data to marine conditions is not recommended, as the radiation regimes differ significantly. On the other hand, in case there are no measurements carried out at the seaside, it should be possible to reconstruct global radiation at coastal sites using linear regression. This has been checked for afternoon for two states of cloudiness: clear and overcast. Regression gives good results everywhere when both sites are cloud-free--coefficient of correlation is practically 1.0. In overcast conditions the correlation is over 0.8 for Tiirikoja and Parnu, over 0.7 for Harku and less for the most distant sites Vilsandi and Narva-Joesuu.

And last, but not least: the quality of global radiation data from automatic weather stations should be carefully checked as there are many factors that might contaminate the measurements. From the above it follows that periodical checking of all sensors is necessary and attention should be drawn to the maintenance of the equipment. If possible, also the quality control of data should be introduced.

doi: 10.3176/eng.2012.1.06


The radiation data were drawn from the archives of the Estonian Meteorological and Hydrological Institute and prepared for calculations by Ms Epp Juust. The authors of the present paper are grateful to Associate Professor Ain Kallis for valuable consultations. The research was supported by the targeted financing by the Estonian Ministry of Education and Science (grant SF0140017s08).


[1.] Kawai, Y. and Wada, A. Diurnal sea surface temperature variation and its impact on the atmosphere and ocean: a review. J. Oceanogr., 2007, 63, 721-744.

[2.] Lips, U., Lips, I., Kikas, V. and Kuvaldina, N. Ferrybox measurements: a tool to study mesoscale processes in the Gulf of Finland (Baltic Sea). In US/EU-Baltic International Symposium, 2008 IEEE/OES: 1-8. DOI: 10.1109/BALTIC.2008.4625536

[3.] Lips, U., Lips, I., Liblik, T., Kikas, V., Altoja, K., Buhhalko, N. and Runk, N. Vertical dynamics of summer phytoplankton in a stratified estuary (Gulf of Finland, Baltic Sea). Ocean Dynamics, 2011, 61, 903-915.

[4.] Burchard, H. and Hofmeister, R. A dynamic equation for the potential energy anomaly for analysing mixing and stratification in estuaries and coastal seas. Est. Coast. Shelf Sci., 2008, 77, 579-687.

[5.] Van Aken, H. M. Meteorological forcing of long-term temperature variations of the Dutch coastal waters. J. Sea Res., 2010, 63, 143-151.

[6.] Lips, I. and Lips, U. Abiotic factors influencing cyanobacterial bloom development in the Gulf of Finland (Baltic Sea). Hydrobiologia, 2008, 614, 133-140.

[7.] Eerme, K., Kallis, A., Veismann, U. and Ansko, I. Long-term variations of available solar radiation on seasonal timescales in 1955-2006 at Tartu-Toravere Meteorological Station, Estonia. Theor. Appl. Climatology, 2010, 101, 371-379.

[8.] Belcher, B. N. and DeGaetano, A. T. A revised empirical model to estimate solar radiation using automated surface weather observations. Solar Energy, 2007, 81, 329-345.

[9.] Russak, V. and Kallis, A. (compilers). Handbook of Estonian Solar Radiation Climate. EMHI, Tallinn, 2003.

[10.] McArthur, B. Baseline Surface Radiation Network. Operations Manual. Version 2.1. WMO, Geneva, 2004.

[11.] Solar Instruments, CMP Series Pyranometers Manual, Kipp & Zonen,, 17.03.2011.

[12.] Philipona, R. Underestimation of solar global and diffuse radiation measured at Earth's surface. J. Geophys. Res., 2002, 107, D22, 4654.

[13.] Averkiev, M. S. An improved method for calculating total radiation. Vestnik Mosk. Univ., Ser. Geogr., 1961, No. 1, 40-47 (in Russian).

[14.] Keevallik, S. and Loitjarv, K. Solar radiation at the surface around the Baltic Proper. Oceanologia, 2010, 52, 1-15.

Kai Rosin and Sirje Keevallik

Marine Systems Institute at Tallinn University of Technology, Akadeemia tee 15a, 12618 Tallinn, Estonia;

Received 27 October 2011, in revised form 25 January 2012
Table 1. Station positions and
openness of the horizon

 Station Approximate elevation
 of objects above the
 horizon, degrees

 N E S W

Vilsandi 0 1 1 6
Harku 3 4 2 3
Narva-Joesuu 0 10 12 10
Tiirikoja 10 4 5 10
Toravere 3 5 2 2

Table 2. Correlation and regression coefficients for
reconstruction of afternoon (11:00-12:00 GMT) hourly
totals at coastal stations from Toravere data for
overcast conditions

 Coefficient Intercept, Slope
 of correlation MJ/[m.sup.2]

Vilsandi 0.61 0.094 0.73
Parnu 0.81 0.037 0.81
Harku 0.74 0.055 0.70
Narva-Joesuu 0.65 0.158 0.65
Tiirikoja 0.82 0.061 0.83
COPYRIGHT 2012 Estonian Academy Publishers
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2012 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Rosin, Kai; Keevallik, Sirje
Publication:Estonian Journal of Engineering
Article Type:Report
Geographic Code:4EXES
Date:Mar 1, 2012
Previous Article:Performance Factory--a new approach of performance assessment for the factory of the future/Tulemuslik tootmisettevote--uus lahenemine tuleviku...
Next Article:Exploring vibrations of cracked beams by the Haar wavelet method/Praoga talade vonkumiste uurimine Haari lainikute meetodil.

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters