Air temperature (TA) can be presumed to be one of the most sensitive indicators of climatic changes. The time series of daily TA from the Hornsund station covers the period 1979 to 2018. In the case of daily mean TA, there are no gaps in data, while for maximum and minimum, daily TA data for 1 September 1979, 29 February 1980, 15 June 2012, and 19 June 2017 are missing. Figure 2a presents the variability of the annual mean of minimum, mean, and maximum TA in 1979–2018 at the Hornsund station. An upward trend is clearly visible for the three analysed variables. The significance of the trend was estimated by the modified Mann–Kendall test (Mann, 1945; Kendall, 1975; Hamed and Rao, 1998) taking into account autocorrelation of time series. The slope of the trend was estimated using Sen's method (Sen, 1968), where the slope is calculated as a median of the slopes of all pairs of points. The outcomes of the modified Mann–Kendall indicated that the trends are statistically significant; the estimated p value is very small (less than 1×10-7) for three presented variables.

The estimated slope of trend equal to 1.34, 1.14, and 1.00 ∘C per decade for minimum, mean, and maximum TA respectively. These are one of the highest increases of mean TA on the planet, more than 6 times higher than the global average of +0.17 ∘C per decade (NOAA, 2020). The rest of the world is not expected to experience such changes until the end of this century (Hanssen-Bauer et al., 2019). The results of trend analyses for mean monthly TA (min, mean, and max) are presented in Table 2. In almost all months there are statistically significant trends except March. In all analysed cases the estimated slope of trend has positive values and indicated the increase in TA. A comparison of the results between variables shows that the largest changes were found for minimum daily TA (1.34 ∘C per decade), while the lowest changes were found for the maximum daily TA (1.0 ∘C per decade). Taking into account changes between months, the largest changes were estimated for January, February, and December (larger than 2.0 ∘C per decade for minimum and mean daily TA). The smallest statistically significant are trends in July and August with slopes of the trend around 0.3 ∘C per decade.

Figure 2b shows the box plots of monthly averages of minimum, mean, and maximum daily TA from the period 1979–2018. The variability of TA depends on the season, with the highest amplitudes during winter months. Summer TA is rather constant, with monthly means reaching usually slightly below 5.0 ∘C. Average monthly TA during winter and early spring usually drop below -10.0 ∘C. The results are in general accordance with observations made at other Arctic stations and reveal that winter is characterised by the highest variability of TA (Gjelten et al., 2016; Osuch and Wawrzyniak, 2016, 2017a). The amplitude between the extreme high and low in this season may be several times higher than in summer. These fluctuations are determined by the relatively stable anticyclonic subsidence with extreme cold and the turbulent cyclonic disturbances that bring higher temperatures, greater cloudiness, and heavy precipitation. The lowest recorded TA measured at a 2 m height above solid ground at Hornsund station was -35.9 ∘C on 16 January 1981, while the absolute maximum was 15.6 ∘C on 31 July 2015. Mean annual air temperature (MAAT) in long-term 1979–2018 is -3.7 ∘C. The average coldest month is March with a mean TA of -10.2 ∘C, and on average the warmest month is July with a mean TA of 4.6 ∘C. The coldest month on record with a mean of -17.9 ∘C was January 1981, and the warmest July 2016 with a mean of 6.3 ∘C. Additionally, in the dataset we also provided monthly and annual positive (PDD) and negative degree days (NDD), calculated as the sum total of daily mean temperatures above or below the 0 ∘C respectively.

The water vapour drives multiple atmospheric processes and has a significant influence on the global climate. It is the main greenhouse gas, affecting the surface by feedback cycle through changing energy balance through radiative fluxes and cloud formation. According to general concepts, the Arctic warming of recent decades is accompanied by the hydrological cycle intensification (Vihma et al., 2016; Osuch et al., 2019). To understand the variability of water vapour concentration and its causes is highly important, especially for climate studies as well as in water balance calculations. At the Hornsund station, the air humidity is currently measured by sensor HMP155, which replaced the previously used HMP45D sensor. The observations cover the period 1979–2018, but measurements were performed four times a day (00:00, 06:00, 12:00, 18:00 UTC) within the periods 1 July 1978–26 July 1981 and 16 August 1982–31 July 1986, two times a day (06:00 and 18:00 UTC) from 27 July 1981 to 30 June 1982, and eight times a day (00:00, 03:00, 06:00, 09:00, 12:00, 15:00, 18:00, 21:00 UTC) since 1 August 1986. Daily time series of the relative humidity (RH) was calculated as a mean of all available measurements within a particular day. There is a gap in the measurements from 1 July 1982 to 15 August 1982. Therefore the trend analyses were performed for the period 1983–2018.

The variability of the annual mean RH in the period 1983–2019 is presented in Fig. 3a. The average over the period 1983–2018 is 79.7 %. The range of variability is from 75.7 % (2003) to 82.7 % in 1994. The trend analyses indicated a statistically insignificant trend.

The course of the monthly mean RH in the period 1983–2018 is presented in Fig. 3b. Higher values of mean RH are observed in warmer months of the year and lower during winter. Such high values are attributed to continual dominance of marine air masses. The annual course of the RH is strongly connected with the air temperature and shows typical variability. It generally increases with warmer air temperatures. However, most of the trends are not statistically significant at the 0.05 level except March, June, and October.

The analyses at daily timescale indicated that drops of RH below 50 % are recorded rather sporadically, although these can occur throughout the year. Such situations are connected with advection of strongly cooled air masses, foehn effects, or katabatic winds from Hansbreen (Marsz and Styszyńska, 2013). The minimum observed quantity reached 24 % on 15 January 1981. The maximum of the observed RH is equal to 100 %. Such conditions occurred 27 times in the period 1979–2018.

In the case of precipitation, the daily sum at the Hornsund station is calculated from four measurements obtained from an unfenced Hellmann rain gauge at 12:00, 18:00, and 00:00, 06:00 of the following day, with the orifice 200 cm2, placed 1 m above the ground level. The time series of the daily sum of precipitation cover the period 1979–2018 with a gap in July 1982.

The influence of the West Spitsbergen Current creates a relatively moist climate in SW Spitsbergen region, which is clearly reflected in the amount of precipitation. In comparison to the other meteorological stations in Spitsbergen (Osuch and Wawrzyniak, 2017a; Hanssen-Bauer et al., 2019), the annual amount reaching 477 mm is the highest. The variability of the annual sums of precipitation in the period 1983–2018 is shown in Fig. 4a. The amount of precipitation varies from 230 mm in 1987 to 805.5 mm in 2016. The trend analyses indicated large changes, an increase of 61.6 mm per decade for the annual sums of precipitation.

The annual course of monthly sums of precipitation from the period 1983–2018 is presented in Fig. 4b. The driest months are April and May with averages of 23 and 24 mm respectively. The highest precipitation is recorded in September reaching on average 75 mm. Trend analyses presented in Table 2 indicate statistically significant changes in January (3.51 mm per decade), September (19.67 mm/decade), and October (13.53 mm per decade).

The measurements of the atmospheric pressure (PA) at Hornsund started in July 1978. In the beginning, PA was measured with a mercury barometer every 3 h. Since 2001 measurements have been conducted every 60 s with a Vaisala PTB200A sensor, replaced by BARO-1QML_AV in 2018. The lowest recorded PA reduced to sea level at Hornsund station was 982.2 hPa on 30 August 1994, while the absolute maximum was 1028.5 hPa on 7 August 1987. Mean annual PA in long-term 1983–2018 is 1008.7 hPa, and its variability is presented in Fig. 5a. An increasing trend (0.25 hPa per decade) is visible, but it is statistically insignificant (pval>0.05).

Figure 5b shows the variability of the mean monthly PA over the period 1983–2018. Well-pronounced seasonality is visible, with a mean monthly pressure higher than 1010 hPa from April to August. The month with the lowest mean PA is December with a mean of 1002.7 hPa, and the month with the largest PA is May with a mean of 1015.7 hPa. The variability of mean monthly PA within the observation period also is visible with the largest variability in January and February (larger than 30 hPa) and the smallest in July (13.7 hPa). The trend analyses of mean monthly PA resulted in a statistically insignificant trend for all months.

The wind is a result of atmospheric circulation and is highly correlated with the intensity of cyclonic activity (Przybylak, 2016). The wind regime results from the latitudinal shape of the Hornsund fjord, location near the seashore and local topography. The measurements of wind speed (WS) and wind direction (WD) were performed at Hornsund with different sensors: 1978–2000 with the Fuess 90z wind meter, 2001–2017 with Vaisala WAA151 for direction and wind speed, since 2018 with Ultrasonic Wind Sensor WMT702. At Hornsund station the height of the anemometer is 10 m above the ground, around 20 m above sea level. WS is measured with an accuracy of 0.1 m s-1 and WD with 5∘. The wind rose for the Hornsund station is presented in Fig. 6. Winds blowing from the east, along the fjord, are prevailing.

The variability of the mean annual WS at Hornsund in the period 1983–2018 is shown in Fig. 7a. The average over the period 1983–2018 is equal to 5.5 m s-1. The lowest values of WS were observed in 1985 (4.8 m s-1), while the largest ones were observed in 1998 (6.3 m s-1). There is a statistically insignificant trend in mean annual WS.

The variability of mean monthly WS in the period 1983–2018 is presented in Fig. 7b. WS regime is well visible with smaller average values during summer months (minimum 4.0 m s-1 in June) and larger average values during winter (maximum 7.1 m s-1 in February). Such variability is a result of the extreme cyclone events that often occur during Arctic winters (Rinke et al., 2017).

Sunshine duration (SD) is one of the important meteorological variables that provides data on the time period during which direct solar radiation reaches the Earth's surface and partly on the quantity of total solar energy. Daily SD is measured at Hornsund using a Campbell–Stokes sunshine recorder (CS). It uses a direct optical method with the heat energy of the Sun's direct radiation burning the card. Such a traditional sunshine recorder has been in service worldwide since the 19th century; although there are multiple automatic radiometers used simultaneously at the Hornsund station, the longest data are recorded by CS. The time series of sunshine duration cover the period 1983–2018. At the Hornsund station, the polar night lasts 104 d (31 October–11 February), while the polar day lasts 117 d (24 April–18 August).

Figure 8a shows the variability of the annual sums of SD at Hornsund in the period 1979–2018. The mean value is 1030.8 h, which is about 28 % of the potential SD calculated for the station (Wojkowski et al., 2015). The large span in the annual SD is visible. The minimum value (755.4 h) was observed in 1994 and the maximum (1325.6 h) in 1985. The slightly decreasing trend in SD is visible but statistically insignificant at the 0.05 level.

Monthly total SD is presented in Fig. 8b. Its variability results from the different duration of the day at the location (latitude 77∘ N) with zero SD during the polar night.

Arctic clouds have a warming effect on the surface during most of the year because their effect of increasing the downward longwave radiation dominates their effect of reducing the net solar radiation over high-albedo snow and ice surfaces. In summer, however, clouds typically have a cooling effect on surface types with a lower albedo, such as the open sea, melting sea ice, and ground (Intrieri et al., 2002; Shupe and Intrieri, 2004). Observations of cloudiness at the Polish Polar Station in Hornsund are conducted by meteorologists and describe the predominant sky condition based upon octas (eighths) of the sky covered by opaque (not transparent) clouds. There are many factors that may hinder the heterogeneity and evaluation of cloudiness, due to the annual change in the meteorological observers and the fact that observers might be subjective, although they are provided with clear observable criteria.

Annual averages of cloudiness in the period 1983–2018 are presented in Fig. 9a. The mean over this period equals 5.85 octas. The minimum value of annual mean was observed in 1988 (5.16 octas) and the maximum in 1984 (6.39 octas). An increasing tendency of mean annual cloudiness is visible. The estimated trend (slope 0.13 octas per decade) is statistically significant at the 0.05 level.

The variability of the monthly cloudiness in the period 1983–2018 is presented in Fig. 9b. The annual cycle is characterised by lower mean cloudiness during the cold period from October till April (5.5–6.0 octas), and this period is also characterised by large interannual variability. The period from May till September is on average cloudier (6.0–6.7 octas), and interannual variability is lower.

The horizontal visibility is quantified using observations made by meteorologists in the surroundings of the Hornsund station with a marine scale that ranges from 1 to 9. The visual observations are performed using known distances to the surrounding mountains and other objects. Values 1 and 2 correspond to very bad visibility, 0–50 and 50–200 m, respectively. Bad visibility (200 m–1 km) is represented by a value of 3. Weak horizontal visibility represents conditions with 1–2 and 2–4 km that are quantified as 4 and 5 in the applied scale. Moderate horizontal visibility, described as 6 in the scale, represents conditions when an object or light can be clearly discerned from 4 to 10 km. Good horizontal visibility (7 in the scale) is 10–20 km, very good (8) 20–50 km, and extremely good (9) is for horizontal visibility larger than 50 km. Noted visibility might be reduced by multiple factors, including all products of the condensation of water vapour such as fog, precipitation, and darkness during cloudy conditions throughout the polar night, as there are no artificial lights in the area. There are no anthropogenic factors that would reduce visibility in the vicinity of the Hornsund station as it is located in the middle of the strictly protected South Spitsbergen National Park. Due to that, reduced visibility cannot be an indicator of poor air quality on the local scale.

Figure 10a shows the variability of mean annual visibility in the period 1983–2018. On average in this period, there is good horizontal visibility that amounts to 7.40; minimum mean annual visibility was observed in 2016 (7.08), while a maximum was observed in 1987 (7.70). A decreasing tendency is visible (slope of trend -0.02 per decade); however the trend is statistically insignificant at the 0.05 level. Variability of the mean monthly visibility at Hornsund in the period 1983–2018 is presented in Fig. 10b. It is characterised by both low interannual and interseasonal variability and on average reaches values between 7 and 8.