Long-term variability of occurrence of precipitation forms in winter in Kraków, Poland
The paper discusses long-term change in snowfall, rainfall and mixed precipitation viewed in conjunction with air temperature and North Atlantic Oscillation (NAO) in winter (December–February). In the study of contemporary climate change and its effect on the hydrological cycle it is useful to focus on winter precipitation forms. A 146-year secular observation series from Kraków, spanning the period 1863–2008, was used to extract data on the number of days with precipitation and on precipitation amount broken down by form. Statistically significant trends were found in total and mixed precipitation, but not in snowfall and rainfall. The climate warming effect has contributed to a material decrease in the snowfall to total winter precipitation ratio during the second half of the 20th c. The highest impact of air temperature was found in the wintertime variation in number of days with snowfall while the NAO had a significant influence on the frequency and amount of both rainfall and snowfall.
Studies of contemporary climate change show a strong air temperature increase in recent years. According to the latest report of the Intergovernmental Panel on Climate Change (IPCC 2007) the mean global temperature rose by 0.74°C between 1906 and 2005. It is not so clear, however, what change has occurred in precipitation since this involves a high degree of temporal and spatial variability. This makes any attempt to link it to the global warming phenomenon more complicated.
In Poland the increase in air temperature observed during the second half of the 20th century was accompanied by a decline in annual precipitation totals in lowland areas (Kożuchowski and Żmudzka 2003; Kossowska-Cezak 2009). Projections for Poland derived from the HadCM2 GS climate model involve greater precipitation, especially in winter (Kożuchowski 2004). Potential benefits or drawbacks of the estimated increase in precipitation will depend on the form of that precipitation and on its intensity. A similar question was posed by H. Ye (2008) with relation to the predicted change in precipitation at high latitudes.
In winter, snowfall frequency might be expected to diminish at the expense of rainfall as temperature increases. This proposition may be overly simplistic however. Indeed, regardless of its form (liquid and/or solid) and type (such as thunderstorm precipitation, hailstorm or solidifying precipitation, Twardosz et al. 2011), precipitation depends not only on temperature, but also on the water vapour content in the air and on the topography. H. Ye (2008) aptly pointed out that the complexity of the climate system and the feedback loops involved make it quite difficult to understand dependencies between the type of precipitation and its frequency, especially at high latitudes where the air temperatures in wintertime often remain below freezing point. For this reason research on wintertime precipitation types in moderate and high latitudes is highly significant in the study of contemporary climate change. One example is snowfall, which is regarded as a good indicator for the detection and monitoring of climate change at the global and regional scale (Namias 1985; Jaagus 1997; Hantel et al. 2000; Huntington et al. 2004).
Snowfall research also has a practical aspect related to road and air traffic problems caused by this form of precipitation. At temperatures below 0°C solid precipitation leads to a build-up of snow cover, a natural form of water retention. A number of hazardous phenomena are linked to snow cover, such as flash floods or large-scale flooding caused by rapid thaw.
The form and type of precipitation represent a wealth of knowledge about the climate and its changes, but these data are rarely used in climatological research. The reason is that it is notoriously difficult to categorise precipitation forms as this demands the kind of information about weather phenomena that can only be accurately obtained from round the clock visual observations. In Central Europe few meteorological stations offer this kind of long-term record (Hohenpeissenberg in Germany since 1792, Warsaw since 1803, Prague-Klementinum since 1804 or Kremsmünster in Austria since 1820), and Kraków is among these rare places.
In Kraków weather observations started in 1792 and were conducted by astronomers until recently, when they were replaced by professional meteorologists (Twardosz and Cebulska 2010). The record provides an unbroken homogeneous string of quality data on daily precipitation, both numeric and descriptive, since 1863 (Twardosz and Niedźwiedź 2001). Homogeneity of the monthly and annual totals was verified and tested in earlier studies (Niedźwiedź et al. 2009, Twardosz et al. 2010) and was identified at the confidence level of 0.05. Only for freezing precipitation (rain or drizzle) reliable data is available from 1920 onwards (Twardosz et al 2011). Precipitation fluctuations in Kraków have been shown (Niedźwiedź et al. 2009, Twardosz et al. 2010) to be representative for larger region of Central Europe.
This study is concerned with long-term change in wintertime precipitation in terms of the whole season and individual months (December–February). Overall precipitation was stratified into snowfall, rainfall and mixed precipitation, and was linked with air temperature and the North Atlantic Oscillation. The global warming effect, observed during the last few decades, raises the question whether the air temperature increase is indeed followed by changes in the frequency and amount of precipitation of different forms, and what role the NAO plays in this development.
2 Data and methods
Observations of atmospheric precipitation started in Kraków when a weather station was established in 1792. Until the mid-19th century, observations were limited to the visual method and noted just the presence and the form of precipitation. The era of instrumental measurements of precipitation amount in Kraków began in August 1849.
This has continued ever since, leaving a continuous record, which—uniquely in Central Europe—remained unbroken even during wars. Daily precipitation totals are available from 1863, which is the first year when the measurements can be brought in line with the contemporary precipitation day standard adopted by the Polish weather service, i.e. from 6:00 am to 6:00 pm UTC (Twardosz and Cebulska 2010). The homogenisation of the precipitation observation timing was possible due to a large amount of archive metadata available. Until 1921, measurements were taken three times a day, in the morning, in the afternoon and evening, after which the number of measurements recorded was reduced to just the morning round. Until 1891, observations were scheduled in local time, which was subsequently replaced with the Central European Time (a difference of 20 min). Ever since the establishment of the station, the observations were performed by qualified observers, which is a guarantee of high reliability. Astronomers, who were responsible for the observation during most of that period, were replaced by meteorologists in 1950.
In the 19th century, early pluviometers were replaced with the Hellmann system pluviometers that had smaller catchment surfaces (Twardosz and Cebulska 2010). To bring these measurements to a common denominator daily totals were converted into the area of 200 cm2. This means that the change of the measurement apparatus does not affect the uniformity of the measurement record (Twardosz 1997, 2005).
First letter: P for amount, N for number of days;
Second letter: L for liquid precipitation, M for mixed, S for snowfall and T for total precipitation (L + M + S),
Third letter denotes individual months: December, January or February (D, J, F).
For example, PS would mean the amount of snowfall in winter, while NSD would mean the number of days with snowfall in December.
A single letter T denotes mean winter temperature.
We also used the North Atlantic Oscillation (NAO) Index, the most important macro scale index for the European climate, defined by J.W. Hurrell (1995) as the normalized pressure difference between a station on the Azores and one on Iceland. We applied an extended version of the index based on the normalised pressure difference between Gibraltar and Reykjavik (Jones et al. 1997), updated on the Climate Research Unit website (http://www.cru.uea.ac.uk/cru/data/nao.htm).
A linear trend was fitted to the precipitation characteristics. The statistical significances of trend in air temperature and precipitation characteristics in the study period were also determined using the Mann-Kendall test.
Pearson’s elementary correlation coefficient was applied to determine a statistical relationship between the precipitation characteristics on the one hand and the air temperature and NAO on the other (von Storch and Zwiers 1999). Principal Components Analysis (PCA) was the method employed to analyse dependencies between the precipitation characteristics, while temperature and time (i.e. the year) were included as additional variables.
The question of trends over time has also been approached by using a standard procedure of feature, i.e. variable, selection (Goyon and Elisseeff 2003) (using Statistica v.9). The method is used to search for variables (among many available) that are capable of explaining a given dependent variable. Such independent variables are sorted according to their predictive power and the top-ones then are included in further calculations. In trend investigation, however, we have applied an approach that may be considered the reverse of a typical one. Time was adopted as the dependent variable, i.e. the variable to be explained by other variables. While time is by no means dependent on the amount of snowfall etc., numerically the procedure is correct. The feature selection procedure involves seeking variables capable of explaining the time elapsing, i.e. those exhibiting the trend.
3 Results and discussion
3.1 Number of days and precipitation totals in winter
Average number of wet days ≥0.1 mm (N) and average precipitation amounts (P) by precipitation form in Kraków (1863–2008) with standard error of the mean
Average number of wet days (N)
Average precipitation amounts (P, mm)
6.0 ± 0.3
4.2 ± 0.3
3.9 ± 0.3
14.1 ± 0.5
13.7 ± 1.1
9.4 ± 0.8
8.2 ± 0.8
31.3 ± 1.7
3.1 ± 0.2
2.9 ± 0.2
2.7 ± 0.2
8.7 ± 0.3
10.3 ± 0.8
9.7 ± 0.8
8.9 ± 0.7
29.0 ± 1.4
6.2 ± 0.3
7.8 ± 0.4
6.8 ± 0.4
20.9 ± 0.7
11.6 ± 0.9
12.8 ± 0.9
11.7 ± 1.0
37.2 ± 1.8
15.9 ± 0.4
15.3 ± 0.4
13.8 ± 0.4
45.0 ± 0.7
36.3 ± 1.5
32.1 ± 1.4
29.5 ± 1.4
98.1 ± 2.5
Mixed precipitation was the least frequent form with approximately 19% of all days with precipitation. This form reflects a great variability of weather conditions over Poland that manifests itself in a frequent alternation of warm and cool air masses.
Statistical characteristics of the number of wet days (N) and precipitation amounts (P) by precipitation form in Kraków (1863–2008): standard deviation (first value) and skewness (second value, the standard error of skewness is 0.2)
Number of wet days (N)
Precipitation amounts (P, mm)
Snowfall totals are weakly correlated between adjacent months (r = 0.15 Jan–Feb, r = 0.16 Dec–Jan). As might have been expected, there is no correlation between precipitation in February and December (r = 0.04). A similar correlation is found in terms of the number of days with snowfall and it becomes even weaker with the other precipitation forms.
3.2 Changes in precipitation in relation to air temperature
Time trend of precipitation characteristics as described by linear regression on time (year number). In the last row the result of regression of NS on temperature is given
+1.96 ± 0.43 deg/100 year
+6.4 ± 1.6 days/100 year
+15.8 ± 5.7 mm/100 year
+4.49 ± 0.66 days/100 year
+13.9 ± 3.1 mm/100 year
+2.1 ± 1.2 days /100 year
−0.7 ± 0.4 mm/100 year
+0.2 ± 1.6 days /100 year
+1.4 ± 4.2 mm/100 year
−0.06 ± 0.04%/100 year
−2.38 ± 0.23 days/deg
The best time “predictors” i.e. the characteristics which best exhibit the time trend. Variables are sorted according to the value of the statistic F; the resulting p-value is also given. The characteristics with p > 0.05 are omitted
The Mann-Kendall test of trend. The standardized normal variable z is given (values below −2 or above 2 are significant at α = 0.05)
According to that trend, for an average winter temperature that can go as high as 7.5°C the expected number of days with snowfall is nil (the 95% confidence interval for the temperature is from 6.0°C to 9.5°C).
The temperature increase also influenced the duration of the snow cover. M. Hantel et al. (2000) concluded that a temperature increase by 1°C would reduce the duration of snow cover in the Austrian Alps by approximately four weeks. Changes in air temperature clearly have a greater impact on the frequency than the amount of snowfall, as demonstrated by the 42% of frequency variance mentioned above (Table 3).
As shown by R.E. Davis et al. (1999) and T.R. Karl et al. (1993) snowfall amount may follow either a positive or negative correlation with temperature. Both studies concluded that what determined the index of such a correlation was the location of the measurement point, which was also responsible for the range of mean monthly temperatures. Where the mean monthly temperature is relatively high the correlation is negative and vice-versa. This effect is consistent with the Clausius-Clapeyron formula that defines conditions for water vapour condensation. Indeed, higher steam concentration in warmer air results in higher snowfall amount and, as demonstrated by H. Ye (2008), its frequency.
If more variables are taken into account (namely 21 monthly characteristics) the percentage of the variability contained in the first two principal components cannot be too high. In any case the general features visible in Fig. 7 are repeated in Fig. 8 with insignificant rotation. Virtually all variables connected with snow (S) are in opposition to temperature, however, the February characteristics are less dependent on temperature, becoming closer to the direction of time. The best correlates of time are, consequently, the characteristics of mixed precipitation.
3.3 Precipitation and NAO
Many studies have demonstrated that the influence of the North Atlantic Oscillation, which is an index representing the inflow of air masses from the west (for the positive phase), on precipitation in Central Europe was at its strongest in winter (e.g. Niedźwiedź et al. 2009).
Correlation of precipitation and NAO. Values with significant correlations are printed in bold type
The influence of the independent variables is illustrated by the relative error of the respective regression coefficient; however the simplest method here is to observe the Student’s t values. The highest t, indicating the highest influence, is obtained for temperature (t = 8.5), next is time (t = 3.4), and NAO (t = 2.1). The timing of the influence is significant in this multiple regression since it is to be removed from the most important variable, i.e. temperature. The NAO, steering temperature, is the least important as an independent factor.
The paper used the secular precipitation record of Kraków to illustrate long-term variability of snowfall, rainfall and mixed precipitation in relation to air temperature and North Atlantic Oscillation during the winter season. The results suggest that change can be identified in the overall total and frequency of winter precipitation and in mixed precipitation (positive trend). This means that wintertime precipitation trends are linked to changes in air temperature, which displays the strongest increase during this season.
Rainfall and snowfall are also strongly dependent on temperature, but no significant long-term trend is found. Indeed, these forms of precipitation display a strong variability in time, as a consequence of equally as high variability of circulation conditions in the temperate zone that is dominated, particularly in winter, by cyclonic activity. What this means is that only a very large change in the amount of precipitation has a potential of statistical significance. Another reason for the inconsistencies in the direction of trends of precipitation and temperature may be the location of the station in the centre of a large city. The significant temperature trend may also be a result of a growing urban effect, which has no impact on synoptic situations that bring precipitation. The climate warming effect, however, has contributed to a marked decrease of the PS/PT ratio (snow to total precipitation in winter) in the second half of the 20th c.
The study demonstrates that air temperature has an impact on the frequency and amount of winter precipitation, but this impact is complex and depends also on other factors, including circulation, which is responsible for the water vapour content in the air.
Much of the variability in the number of days with snowfall (42%) is accountable on the basis of air temperature.
Snowfall and rainfall are strongly, albeit differently, correlated with atmospheric circulation. The higher the NAO value the higher the frequency and amount of rainfall, but the lower the frequency and amount of snowfall. This NAO effect is at its strongest in January when it explains 27% of the number of days with rainfall and 25% of snowfall.
Should the growing trend in air temperature continue it should be expected that snowfall will start decreasing in southern Poland and so will the amount of water from that snowfall. To reduce the expected number of days with snowfall to nil, however, air temperature would have to increase up to 7.5°C.
We thank Mr. Paweł Pilch and Dr. Martin Cahn for reviewing the English. This work was partially supported by the Polish Ministry of Science and Higher Education under Grant No. N N306 119936.
This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.
- Goyon I, Elisseeff A (2003) An introduction to variable and feature selection. J Mach Learn Res 3:1157–1182Google Scholar
- IPCC (2007) Climate Change, The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. In: Solomon S, Qin D, Manning M, Chen Z, Marquis M, Averyt KB, Tignor M, Miller HL (eds). Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, 2007Google Scholar
- Kossowska-Cezak U (2009) Warunki termiczne i opadowe w świetle serii obserwacyjnej z Okęcia (1947–2008). Zeszyty Naukowe Szkoły Wyższej Przymierza Rodzin w Warszawie, seria Geograficzno-Turystyczna 2:25–32Google Scholar
- Kożuchowski K (ed) (2004) Skala, uwarunkowania i perspektywy współczesnych zmian klimatycznych w Polsce. Zakład Dynamiki Środowiska i Bioklimatologii UŁ, ŁódźGoogle Scholar
- Kożuchowski K, Żmudzka E (2003) 100-year series of areally averaged temperatures and precipitation totals in Poland. Acta Univ Wratislaviensis 2542:116–122Google Scholar
- Twardosz R (1997) Pomiary opadów atmosferycznych na stacji meteorologicznej w Krakowie. In: Trepińska J (ed.) Wahania klimatu w Krakowie (1792–1995), Instytut Geografii UJ, 51–58.Google Scholar
- Twardosz R (2002–2003) Zmienność opadów śniegu na Pogórzu Wielickim w latach 1971–2000. Folia Geographica, Ser. Geographica Physica 33–34:105–115Google Scholar
- Twardosz R (2005) Dobowy przebieg opadów atmosferycznych w ujęciu synoptycznym i probabilistycznym na przykładzie Krakowa (1886–2002). Instytut Geografii i Gospodarki Przestrzennej UJ, p 176.Google Scholar
- Twardosz R (2007) Opady atmosferyczne. In: Matuszko D (ed) Klimat Krakowa w XX wieku. Instytut Geografii i Gospodarki Przestrzennej UJ, 127–138Google Scholar
- Twardosz R, Cebulska M (2010) Observations and measurements of precipitation in the Polish province of Galicia in the nineteenth century. In: Przybylak R et al (eds) The Polish climate in the European context: an historical overview. Springer Science + Business Media B. V, Berlin, pp 457–471CrossRefGoogle Scholar
- Twardosz R, Łupikasza E, Niedźwiedź T (2011) Zmienność i uwarunkowania cyrkulacyjne występowania postaci i typów opadów atmosferycznych na przykładzie Krakowa. (Temporal variability in the form and type of precipitation in Kraków in relation to circulation patterns). Wydawnictwo Uniwersytetu Jagiellońskiego, KrakówGoogle Scholar
- von Storch H, Zwiers FW (1999) Statistical analysis in climate research. Cambridge University Press, CambridgeGoogle Scholar