In 1904 monitoring of sunshine duration started at Skansen in Stockholm, Sweden, Westman (1917). SMHI still measures this meteorological quantity but with another instrument and at another site within Stockholm. These type of changes of instruments and sites cause that it is hard to directly compare the results from different periods. In the following is presented a method how this can be done.
A curiosity is that the sunniest year on record occurred in 2018.
The modern definition introduced in the 1980-ties: Sunshine duration is the time when the direct solar irradiance is larger than 120 W/m². The unit is usually hours.
In the late 19th century when instruments for observing the sunshine duration became available they did not directly measure the direct irradiance. Instead, they used a black or dark blue paper or a photographic paper where the direct irradiance made a track as the sun moved over the sky. The track was evaluated to get the length of the time when the sun was intense enough to produce a track.
The instruments have changed over the years and so have the site within the city. Unfortunately, much of the earliest data found in the archive is only monthly data. The original raw observations are lost and thus it is hard to reevaluate the registrations for control and eventually changes as questionable data are found.
In the beginning of the Stockholm series the instrument was a Campbell-Stokes heliograph. But, already in 1908 a rebuilt Jordan sun recorder was used; here called the Hamberg instrument. This was used at Vanadislunden by the City of Stockholm, who made several meteorological observations at this site.
In late 1951 recordings of sunshine duration by SMHI started at Fridhemsplan in Stockholm using a Campbell-Stokes heliograph. This site was unchanged until SMHI moved to Norrköping in 1975. During this period an improvement was made starting from January 1971, viz. that the instrument (the glass sphere) became heated to reduce the influence from frost and dew.
After 1975 the monitoring was moved to the nearby airport of Stockholm-Bromma. Here the heliograph went on monitoring until 1983. From 1983 up to 2007 the observations of sunshine duration was done using a pyrheliometer and data were sampled (more or less every second) and collected by an automatic station located at The Royal Institute of Technology, KTH.
The quality was strongly improved as the measurements of sunshine duration thus could be done according to the definition. The definition for sunshine duration, when the observations started in 1983, was based on WMO (1977), which recommend a threshold of 200 Wm-2. During the process to start the new network the definition was changed to the present 120 Wm-2 ,WMO (1982). This was not noted at once. Therefore, measurements were made using the old threshold for a couple of years. But, from 1986 and onwards the sunshine duration was observed by a pyrheliometer using 120 Wm-2.
This went on until 2007 when an upgrade of the solar radiation network of SMHI was done. With the resources available it was not possible to have suntrackers at more than three sites and Stockholm was not one of these. Therefore, the sunshine duration had to be observed using a simpler device, a Kipp and Zonen CSD.
From the previous it can be understood that the different instruments has not been measuring the sunshine duration exactly in the same way. There might be a difference in the sensitivity, in the horizon, in evaluation methods and also gross errors in the handling of the data.
For example the oldest data were recorded by a heliograph without heating and thus data was affected by both rime and dew. This has been corrected for using correction values from Rodhe (1975) who mad a comparison of unheated and heated Campbell-Stokes heliographs for about six years. This correction was applied to all Campbell-Stokes heliograph data before the date the heating was introduced.
The applied correction for the monthly value measured by non-heated heliographs is a statistical average number of hours for the month. This means that some months will be over- and other months under-corrected. It is most likely that there will be an overrepresentation of low values in the past in particular for the winter months.
As already was noted by the early users of heliographs their sensitivity vary. The particular instrument used at Skansen and also the Hamberg instrument used at Vanadislunden were tested in the past and the result can be found in Ångström (1928). Both these instruments had a lower sensitivity than later versions of the Campbell-Stokes heliographs. Thus a correction for this was made.
A similar correction had to be applied for the 1983-1985 monthly values. As mentioned for these years a higher threshold-value of the direct irradiance (200 Wm-2) was used to monitor the sunshine duration.
The individual data has been scrutinised and comparisons between original raw evaluations and published values have been done. Relatively often errors in calculations and digitalizations have been found and corrected. There has also been some revisions of earlier interpolations where data has been missing.
Of course the methods are often rough but on a monthly level it is assumed that the corrected dataset is more useful for climate studies than the original set.
The monthly and yearly extremes are presented for two periods 1905-1982 and for 1983-2018 (the present location at KTH). Interestingly, the highest yearly and the highest monthly value in the whole series occurred in 2018. That high monthly value is May 2018 and not June or July that have longer days.
The year with the lowest amount of sunshine was 1912, which was a year strongly affected by the eruption of the volcano Novarupta in Alaska.
If we consider the yearly values the first years up to about 1930 is characterised by strong variation. Then follows a period up to the mid 1970-ties with a lower variation and without any really low values.
The following period is what became known as the period of global dimming. Many sites around the world recorded a decrease in solar radiation into the 1980-ties. After that there has been a recovery of sunshine duration. We are now back on a relatively high level. The same features can be observed for the global radiation.
The values of the oldest data from 1905 to 1907 are the most uncertain as this period is corrected both for non-heated and less sensitive instrument. And the quality of the horizon is unknown.
The automatization and introduction of a pyrheliometer as recording instrument, starting 1st January1983, proved to produce sunshine duration data of much higher quality than before.
The uncertainty output presented in this section is based on information and methodology described in Cook (2002) and GUM (1993) and it is evaluated for monthly values. The calculations, briefly presented in Josefsson (2019), are done using an Excel-spreadsheet following the worked example of Cook (2002). Many details in the estimation are open for discussion as they are rough estimates. From the spreadsheets (not presented here) one can easily see which uncertainty components that dominates for each instrument type and measuring period and that others are of minor importance for the overall result.
A rough estimate of the uncertainty during different periods (related to the present monitoring position at KTH) shows that the oldest monthly values have a general uncertainty of ±8% (2σ) and the latest period ±4%.
The uncertainty for yearly values is reduced slightly as random errors tend to level out. But, a large part of the uncertainty is probably not random and there is a large autocorrelation between months. Roughly the oldest yearly values probably have an uncertainty of about ±5% and in later decades (KTH-location) it’s reduced to ±2%.
Data for download
For the long term corrected dataset monthly, yearly and estimated uncertainty data is available for download here.
Format: year, twelve monthly values (hours), yearly value (hours), relative uncertainty of monthly value, relative uncertainty of yearly value, absolute uncertainty of yearly value.
If you using these data we will be glad if you refer to: Source: SMHI and a reference to Appendix 3 in Josefsson (2019).
If you are interested in hourly data of sunshine duration for later years you can download these from "Öppna data" (now in Swedish, soon also in English).
Ångström A. (1928), Recording solar radiation. A study of the radiation climate of the surroundings of Stockholm, Meddelanden från SMHA, Band 4, N:o 3. Stockholm 1928.
CIMO (1981), The definition of sunshine occurrence in terms of a threshold of solar irradiance. CIMO-VIII, Doc. 13, 29 VI 1981.
Cook R.R. (2002) Assessment of uncertainties of measurement for calibration & testing laboratories, second edition. ISBN 0-909307-46-6.
ISO (1993). Guide to the Expression of Uncertainty in Measurement (GUM), First edition, ISO,1993.
Josefsson W. (1987). Solstrålningen i Sverige. Tids- och rumsfördelning. Byggforskningsrådet Rapport R112:1987. ISBN 91-540-4824-9.
Josefsson W, (2019). Long-term global radiation in Stockholm, 1922-2018. SMHI, Meteorology, No 165. ISSN: 0283-7730 .
Lindholm F. (1955). Sunshine and cloudiness in Sweden 1901-1930, Meddelanden SMHA. Serie B. Nr. 11.
Rodhe B (1975). In Measurements of solar radiation in Sweden 1971. Årsbok 53, 2.4, SMHI, Stockholm.
Westman (1917), Sonnenscheindauer im Mittelschwedische Ostseegebiet 1911-1916. Kungliga Vet. Akad. Handlingar, Band 57 N:o 1. 34 p.
WMO (1977), Commission for Instruments and Methods of Observation (CIMO) - Abridged final report of the seventh session (1-12 August 1977; Hamburg, Germany). WMO· No. 490, ISBN 978-92-63-10490-8.
WMO (1982), Commission for Instruments and Methods of Observation (CIMO), Abridged final report of the eight session (Mexico City, 19 - 30 October 1981), WMO· No. 590, ISBN 92-63-10590 - 1