During IPY-1 in 1882–83, 12 main and a large number of auxiliary research stations were operating in the Northern Hemisphere (mainly on the Labrador Peninsula); for more details see Barr (2008) and Barr & Lüdecke (2010). This study makes use of data from nine stations representing the so-called High Arctic (Fig. 1), whose boundaries and division into climatic regions were assigned in accordance with the Atlas of the Arctic (Trešnikov 1985). The evenly distributed stations (Table 1, Fig. 1) of Godthåb, Jan Mayen, Kapp Thordsen, Malye Karmakuly, Sagastyr, Point Barrow, Lady Franklin Bay, Kingua Fjord and Kara Sea are representative of nearly all climatic regions of the Arctic.

The Kara Sea data series includes meteorological observations made during the drift of the Dutch vessel the SS Varna, which moved from Vaygach Island in the east–north-east direction, along the Yamal Current. Between 15 January and 31 July 1883, measurements were taken in the vicinity of a building erected on pack ice in case the ship was destroyed by ice, which actually happened on 21 January that year. Most of the observations and meteorological measurements were made in the area situated between 70°00′N and 71°45′N, and 62°29′E and 65°25′E (Table 2). The Dutch steam ship eventually sank in the icy waters of the Kara Sea on 24 July 1883 (for more details see Snellen & Ekama 1910; Barr 2008).

According to the recommendations of the International Polar Commission (St. Petersburg 1882), meteorological measurements were to be performed at 1-h intervals (Table 1). However, in the harsh polar conditions not all of the expeditions managed to comply with this recommendation. At most of the analysed stations, the measurements were recorded according to mean local time (Table 3), the only exception here being two American stations (Lady Franklin Bay and Point Barrow), where the records refer to Washington local time. This is why the references from Lady Franklin Bay and Point Barrow were corrected by +1 h and −5 h, respectively, to reduce them to the common mean local time. About 20-min deviations from the local time at Kapp Thordsen and on the Kara Sea were considered hardly significant, and remained unadjusted.

The quality and correctness of the data gathered during the IPY-1 expeditions were already being verified during the expeditions themselves. Furthermore, corrections were made to the measurements to accommodate the methodology accepted at the third meeting of the International Polar Commission in St. Petersburg, held from 1 to 8 August 1881 (Wild 1882). It should be noted that properly trained members of the individual expeditions took the measurements themselves. The instruments had been carefully calibrated at national meteorological centres, such as the Kew Observatory near London and the Central Physical Observatory of St. Petersburg in Russia. After the expeditions, the instruments were verified at the same facilities.

Atmospheric humidity can be measured in several ways. During IPY-1, air humidity was determined using both the psychrometric (measuring e: pressure of the current water vapour content of the air) and hygrometric methods (measuring RH: relative humidity, or the ratio of the actual pressure of the water vapour [e] in the air to the pressure of the saturated water vapour [E] at the specific temperature). The August psychrometer comprised standard mercury thermometers (dry-bulb and wet-bulb types). All the thermometers were precisely calibrated at relevant national meteorological centres and, for example, the indication error of the thermometers used at the Siberian station of Sagastyr did not exceed 0.25°C at temperatures above −20°C (Wood & Overland 2006). At most of the stations, the thermometers were placed in instrument shelters designed by Wild at a height of 1.5–3 m above ground level (Table 1). In the hygrometric method, for example, a Saussure-type hair hygrometer was used (Fig. 2).

According to the Programme for the International Polar Expeditions and recommendations of H. Wild (for more details see Supplementary File 1), when the air temperature exceeded 0.5°C on a wet-bulb thermometer, both a hair hygrometer and a psychrometer were used; at lower temperatures, the relative humidity (RH) was determined using a hair hygrometer only, and the corresponding value of e was calculated with the help of psychrometric charts in relation to the temperature of air read on a dry-bulb thermometer. According to Wild's experience, hair hygrometers could entirely replace the psychrometer at 0°C, as well as at low temperatures if they were properly managed and occasionally verified by comparison with a psychrometer above 0°C. He had, moreover, often found that the two instruments remained in perfect accordance between −5°C and −15°C at the Russian stations (Proceedings of the Meteorological Conference at Leipzig 1873 ).

At the end of the 19th century, the psychrometric charts of Wild–Jelinek (1876) were in use. The values of the saturation pressure of water vapour were given here with respect to supercooled water (E _water ). The table of saturation pressure with respect to ice (E _ice ) was added in the next edition, in 1911 (Anonymous 1912). Adjustments to the calculations of e and RH for negative temperatures of the air, made by Ekholm (1890) during his experiments carried out at the Kapp Thordsen station during IPY-1, were not applied to psychrometric charts of the Norwegian Meteorological Institute until 1895 (Birkeland 1944). The methodology, and its inherent difficulty providing reliable measurements of atmospheric humidity at times of negative air temperature, is best reflected in the reports of the members of the Swedish expedition at Kapp Thordsen (Ekholm 1890; see Supplementary File 1), and of the Dutch expedition covering the Kara Sea (Snellen & Ekama 1910; see Supplementary File 1).

Condensation hygrometers, such as Alluard or Regnault (measuring the temperature at the dew point [Td], that is, the temperature at which the water vapour contained in the air becomes saturated), were used together with a chemical approach as control measures for the psychrometric and hygrometric methods. It is interesting to note that the benefits of the measurement of a dew point as an efficient method of determination of atmospheric humidity in the Arctic were emphasized as early as at the end of the 19th century: “Yet I believe that, my experiments have demonstrated that Regnault's instrument may be utilised even in the polar lands, and that it is a device which failed to win due appreciation when considering correlations between the indications of a hair hygrometer and those of a thermometer, whilst it may be well applied to correct the indications of the hygrometer regarding changes in the values of temperature” (Ekama quoted in Snellen & Ekama 1910: 75). The method has been widely used in meteorological stations in the Canadian Arctic since 1950 (van Wijngaarden & Vincent 2005; Vincent et al. 2007).

The meteorological data used in this paper were sourced from the reports of the polar expeditions organized during IPY-1, published in various years between 1886 and 1910 (Table 1). A majority of these reports are available as part of the internet resources of the Central Library of the US National Oceanic and Atmospheric Administration (NOAA; Fiolek 2007). Photocopies of reports kept at the library of the Scott Polar Research Institute in Cambridge were also used, and—in the case of a Russian expedition to Novaya Zemlya (Malye Karmakuly)—a photocopy of the report kept at the library of the Russian Geographical Society of St. Petersburg.

The data were digitized manually, because this kind of approach guarantees the lowest error rates when acquiring data from antique books and prints (Brönnimann et al. 2006). Optical character recognition software could not be used because it did not recognize the 19th-century typefaces. Manual digitization enabled preliminary verification of incorrect values. Moreover, the editors of all the reports which would provide data for this study carried out a quality control to reject any unreliable data. Comparing original diurnal means given in the reports with means calculated from the digitized values also eliminated obvious digitization errors. Prior to any calculations, the collected data were also subject to a final substantive check using standard methods applied in climatological studies.

At Point Barrow, water vapour pressure was not measured. Therefore, this element was calculated on the basis of the air temperature and relative humidity data available in the report, according to the following formula: e = RH * Ewater / 100 ,

where E _water was taken from psychrometric charts (Różdżyński 1995) for the appropriate measured air temperature.

Mean diurnal values of individual parameters of air humidity were calculated as an actual arithmetic mean, whereas maximum and minimum diurnal values were selected from 24 measurements. As a result of difficulties with the acquisition of measurements in winter, there is a lack of data for some IPY-1 stations. Therefore, in the common reference period of IPY-1, and for all stations, this weather element was described only for the 3 months of May, June and July 1883 (Table , 6 ).

See Supplementary File 2 and 3 for corrected measured water vapour pressure (e) and relative humidity (RH) data, with hourly resolutions.

It is important to point out here that although the first digitized meteorological data from IPY-1 were made available on NOAA's website (http://www.arctic.noaa.gov/aro/ipy-1) by Wood & Overland (2006), it comprised only daily and monthly resolutions (air temperature, atmospheric pressure, direction and wind velocity). Krause et al. (2010) have recently developed a complete database for IPY-1, which has been added to the resources of the World Data Center for Marine Environmental Sciences (http://127.0.0.1:8800/PangaVista). The database was compared with the aforementioned NOAA and Nicolaus Copernicus University databases and no significant differences were found.

Comparison of the historical humidity conditions with current ones has been made using the RH parameter. Five of the nine IPY-1 stations for which year-round RH measurements are available were chosen. The second data set includes contemporary RH data (1961–1990) obtained from the historical sites. In four cases (Godthåb, Jan Mayen, Malye Karmakuly and Point Barrow) the locations of the observations in the two periods were identical. For the Siberian station (Sagastyr), the data from the nearest modern station (Saggylah Ary, situated approximately 80 km from the historical site) have been used. Data for modern stations were taken from the following sources: the Danish Meteorological Institute (Cappelen 2012, for Godthåb), the Norwegian Meteorological Institute (http://www.eklima.met.no, for Jan Mayen), the Environmental Working Group Arctic meteorology and climate atlas (Fetterer & Radionov 2000; for Malye Karmakuly and Saggylah Ary), and the National Climatic Data Center (http://www.ncdc.noaa.gov, for Point Barrow).

For purposes of comparison with modern RH data, historical RH _water data from low temperatures used in this paper have not been recalculated to RH _ice . According to the World Meteorological Organization, modern measurements of RH at temperatures less than 0°C are evaluated with respect to water, because: (1) most hygrometers indicate relative humidity with respect to water at all temperatures, and (2) the majority of existing records of RH at temperatures below 0°C are expressed on a basis of saturation with respect to water (WMO 2008). However, the reader must be aware that some sources of errors and biases still remain. For example, as a result of the use of different types of instruments and psychrometric charts to determine the RH calculations.

The mean values of RH in the IPY-1 period (September 1882 to July 1883) for five of the nine stations analysed were higher than today's values, by 2.4–5.6%, because the air temperature was colder at that time than today, by 1.0–1.5°C on average (for more details see Przybylak et al. 2010). Values higher than those currently prevailing were noted mainly in the Norwegian Sea and Barents Sea regions (5.6 and 5.0%, respectively), whereas in the continental part of the Arctic (the Siberian region represented by Sagastyr station), the positive differences were smaller and their variability around the modern long-term (1961–1990) monthly means were the most stable (Table 7).

Because the content of the water vapour pressure in the Arctic is generally small, so the changes in RH are mainly the function, as mentioned above, of air temperature changes. The highest positive anomalies of RH, exceeding +2 SDs in some months (Fig. 8) were noted in the Atlantic Arctic. These noticeable differences were the result of atmospheric circulations carrying, in the vicinity of the analysed stations, air masses characterized by different thermal and humidity conditions. For example, in February of 1883 in Godthåb, the greatest noted positive anomaly of RH occurred while the station was under the influence of cold air masses (Table VIII in Przybylak et al. 2010). On the other hand, in February of 1883 in Malye Karmakuly, the large positive RH anomaly (13.4%) there was connected with the inflow of warmer but very wet air masses (Fig. 3) from the south–SSW direction (Wyszyński 2012). Large positive anomalies (by 10–13%) were also noted at Point Barrow in winter, which was characterized by strong negative thermal anomalies in comparison with present-day conditions. Such variation in RH is also observable today, indicated by the high range of ±2 SDs in the cold half of the year (Fig. 8).

Despite the exemplary cases described above, most of the observed changes between historical and modern RH values are not significant. The majority of historical daily RH values lie between a distance of +2 SDs from the present long-term (1961–1990) monthly means (Fig. 8).

Meteorological data gathered for the Arctic during IPY-1 are the best in terms of coverage, quality, resolution and other characteristics out of all the early instrumental data for the area. Research carried out during the IPY-1 scientific expeditions brought a significant contribution to the development of hygrometry in polar regions at the end of the 19th century.

The spatial distribution of atmospheric water vapour in the Arctic during IPY-1, in the common observational period (May–July 1883), was similar to the present distribution. The highest monthly means of e for the common reference period of May–July 1883 were determined for the area of Baffin Bay (6.8 hPa) and the Atlantic Ocean (6.0 hPa), where the oceanic influences were the greatest. In the north part of the Canadian Arctic, where the continental type of climate prevails, lower values of water vapour pressure predominated (4.6 hPa) in comparison with the areas subject to the oceanic climate.

In the annual course, the highest values of e during IPY-1 were noted in July and August, while the lowest were in the cold half of the year. The annual course of e is evidently asymmetrical. In the Norwegian Arctic, more water vapour (by about 1.3 hPa) was in the air in autumn than in spring, while the opposite relation occurred in Siberia and the American Arctic.

Monthly means of RH were high, and most often exceeded 80%. The highest mean of RH of all the IPY-1 stations was observed in Siberia (91.8%), and the lowest on Spitsbergen (79.8%). A high RH was found in the area of the Norwegian Sea (90%), with the most oceanic climate. Over the year, the course of relative humidity in the areas of the Arctic that are subject to cyclonic activity remained fairly constant, that is, in the Norwegian Arctic, the basin of the Barents and Kara seas, as well as in Baffin Bay, with values within the range of 75–90%. This kind of annual course of RH in the Norwegian Arctic is also observed at present (Araźny 2003).

In comparison to present-day conditions, mean RH values in the IPY-1 period (September 1882 to July 1883) were higher by 2.4–5.6%. The greatest differences mainly occurred in the Atlantic Arctic, whereas in the continental part of the Arctic the positive differences were smaller and more stable (Table 7). Most of the observed changes between historical and modern RH values are not significant. The majority of historical daily RH values lie between a distance of +2 SDs from the present long-term (1961–1990) monthly means.