Hydrological Dry Periods versus Atmospheric Circulations in the Lower Vistula Basin (Poland) in 1954–2018

The study area is located in the northern part of Poland (Central Europe) in the lower Vistula basin (Poland's largest river). The authors performed analyses of seven rivers: the Wierzyca, the Wda, the Brda, the Zgłowiączka, the Osa, the Drwęca and the Skrwa Prawa, whose locations are depicted in Figure 1. The entire research area fits into one physico-geographical region of the Southern Baltic Lakelands. The topographical relief and the surface geological structure of the catchments relate to the glacial sediments of the last glaciation. Such areas are characterised by significant height differences of terrain, moraine uplands built mostly of clay formations, and outwash plains of sand and gravel formations (Molewski, Weckwerth 2017). The vast areas of the uplands are cut by ice-marginal valleys (urstromtals) and valleys currently used by river-lake systems. The diverse structure of the catchment determines the processes of rainwater infiltration supplying watercourses with groundwater.

Fig. 1

The research area is located in the zone of influence of both the Atlantic Ocean and the Eurasian continent. The lower Vistula basin lies in the temperate, warm transitional climate, separating the maritime climate from the continental one. It is often surrounded by air masses with different temperature and humidity properties, resulting in volatile weather and various types of precipitation (Woś 2010, Twardosz et al. 2011). The average annual area air temperature in Poland in 1991–2020 was +8.73°C and exceeded that in 1951–1980 by 1.54°C (Climate of Poland 2020). The multi-year mean air temperature for the study area ranged between 7.5°C and 8.5°C (Kejna, Rudzki 2021) and showed a positive, statistically significant upward trend amounting to +0.30°C · 10 a⁻¹ (Climate of Poland 2020). It was in the summer that the most pronounced and statistically significant temperature increase was recorded, followed by winter and spring. The trends in temperature changes in autumn are insignificant. According to Okoniewska and Szumińska (2020), a statistically significant increase in air temperature, especially since the 1990s, in conjunction with the decreasing trend of relative humidity, contributes to more intensive potential evaporation in the study area. Its particularly high values have been repeatedly recorded since the early 21st century. Precipitation is a more variable element of the climate than air temperature. The 1961–2009 mean area precipitation total for Poland was 623.7 mm and 594.2 mm for the study area (Limanówka et al. 2012) with the respective monthly maximum and minimum in July and February. Precipitation is highly diverse in terms of spatial distribution, with the lowest totals recorded in the southern part of the study area (Zgłowiączka River), where the mean long-term figures do not exceed 500 mm (Bartczak et al. 2013). In general, annual precipitation totals do not show statistically significant trends (Pińskwar et al. 2019); yet, in the long term, there are periods with higher or lower annual totals compared to the mean values. Fluctuations and short-term, statistically significant trends in precipitation are of greater importance and have a larger impact on the environment than their long-term equivalents (Bartczak et al. 2013).

All the analysed rivers share a similar feeding pattern and are characterised by various degrees of nival (snow) regime. The volume of water resources in the catchments changes both over the long term and in the annual cycle, with the highest level in spring and the lowest in summer and late autumn. The latter is often combined with low flows of varying duration and intensity. Non-existing, or insufficient, water supply in the autumn and winter may extend a low flow well into another hydrological year. Naturally, the rivers are dominantly supplied by groundwater throughout the year (Gierszewski 2000, Jokiel 2004, Bartczak 2007, Brykała 2009, Szumińska 2014). The smallest water resources are typically identified in the southern part of the study area, where the average annual specific discharges in dry years drop below 1 dm³ · s⁻¹ · km⁻² (Zgłowiączka River) (Bartczak et al. 2014a, b).

The analysis of dry periods was based on mean monthly discharges of seven main tributaries of the lower Vistula in 1954–2018 (Table 1). The data, obtained courtesy of the Institute of Meteorology and Water Management – National Research Institute [in Polish: Instytut Meteorologii i Gospodarki Wodnej – Państwowy Instytut Badawczy (IMGW-PIB)], mostly came from the water gauges located at the lower end of the studied catchments. However, as the selection of the stations was dictated by the need to maintain the temporal homogeneity of the analysis, some of them do not close the examined catchments (Fig. 1).

The analysis of air circulation over central Poland used Maszewski's catalogue of air circulation types (Przybylak, Maszewski 2009) based on the classification proposed by Niedźwiedź (1981, 2013). It distinguishes 21 kinds depending on the direction or lack of air advection, and the pressure system. The situations with advection were marked with capital letters indicating the direction of the air inflow (N, NE, E, SE, S, SW, W and NW). The type of the pressure system is marked with the index ‘a’ for anticyclonic or ‘c’ for cyclonic situations. Additionally, the authors distinguished two non-advective situations: Ca (centre of high pressure) and Ka (high-pressure wedge or ridge), and two situations with varied advection: Cc (centre of low pressure) and Bc (cyclonic trough over central Poland). Cols and indifferent situations that are difficult to define are given an ‘x’ mark. The types of air circulation were determined using the calendar of daily surface weather maps by Deutscher Wetterdienst. Similar calendars based on 21 types of circulation have already been successfully applied to assess the impact of atmospheric circulation on climate change in Poland (e.g. Twardosz, Niedźwiedź 2001, Niedźwiedź et al. 2009, Twardosz et al. 2011) and the Arctic (e.g. Isaksen et al. 2016, Araźny et al. 2018, Araźny 2019, Łupikasza, Niedźwiedź 2019).

The overall atmospheric circulation over the research area was presented with circulation indices which enable the determination of the degree of dominance of the western zonal (W), the southern meridional (S) or the cyclonic circulation (C) (Niedźwiedź 2000, 2001). The W index determines the intensity of the western zonal (positive values) or eastern (negative values) circulation. To calculate its value, the following scoring was used for particular directions of air advection: +2 for the W direction, +1 for NW and SW, −2 for E and −1 for NE and SE. The remaining types of circulation were assigned 0 points. The S index (southern circulation) is a measure of the intensity of the meridional circulation, with its positive values attesting to the dominance of air inflow from the southern sector, and negative – from the northern one.

The considered circulation types displayed the following distribution of points: +2 for S types, +1 for SW and SE directions, −2 for N and −1 for NW and NE. The cyclonicity index C indicates intensified low- (positive values) or high-pressure (negative values) activity. The points corresponding to particular situations are ascribed as follows: +2 for Cc and Bc, +1 for the other cyclonic types, −2 for Ca and Ka and −1 for the remaining anticyclonic types (Niedźwiedź 2000, 2001, Przybylak, Maszewski 2009). The calculated values of the W, S and C indices were then grouped into 12-month periods using the so-called moving sum windows. Hereafter, they are referred to as W-12, S-12 and C-12.

The study of the direction of changes in mean annual discharges and atmospheric circulation indices (estimation and significance) was carried out using the non-parametric Mann Kendall (MK) test (Hirsch et al. 1982). The null hypothesis of the H0 test assumes that the data come from a population of independent data and are evenly distributed. On the other hand, the HA hypothesis of the test assumes that the data follow a monotonic trend. The significance of the MK test was assessed at the level of 0.05.

Dry periods were determined with the Standardised Streamflow Index (SSI), whose advantage is the possibility to calculate values for different time periods. In this study, we opted for the 12-month SSI-12. The considered series were normalised using the general Box–Cox transformation (Box, Cox 1964, 1982), which allowed the authors to obtain a vector of normalised values. The Box–Cox transformation of the variable, indexed by the parameter λ, is given by the formula: u(λ)={xλ−1λ(λ≠0)lnx(λ=0); for x>0{u^{\left( \lambda \right)}} = \left\{ {\matrix{{{{{x^\lambda } - 1} \over \lambda }} \hfill & {\left( {\lambda \ne 0} \right)} \hfill \cr {\ln x} \hfill & {\left( {\lambda = 0} \right)} \hfill \cr } ;\;{\rm{for}}\;x > 0} \right.

The number of values that λ can take in the formula is virtually infinite, which maximises the effect of the transformation.

As the normalised series had different mean values of u and different values of the standard deviation σ, their standardisation was necessary, with the aim being to obtain series with the mean value (u) = 0 and the standard deviation (σ) = 1. Hence, the series with the normal distribution were subjected to the linear transformation according to the procedure: z=u−u¯σu~N(0,1)z = {{u - \bar u} \over {{\sigma _u}}} \sim N\left( {0,1} \right)

The normalisation and standardisation processes made it possible to compare numerous series with a different scale of input data. This procedure positively influenced the homogeneity of the series, which was verified and presented in Bartczak et al. (2014b).

Finally, dry periods were identified following a suggestion put forward by McKee et al. (1993) and WMO (2012), where a drought starts when the SSI value is equal to or below −1.0 and ends when the value becomes positive. Next, the indices were classified so that the degree of their intensity could be assessed. In the original classification presented in McKee et al. (1993) and later Guttman (1999), the index thresholds indicating shortages are −1.0 (moderate drought), −1.5 (severe drought) and −2.0 (extreme drought). Standard deviation values between −1.0 and +1.0 inform about average conditions.

The relationship between the SSI-12 during dry periods and the corresponding indices of a specific atmospheric circulation was expressed using Spearman's rank correlation coefficient. The statistical significance of the coefficient was verified with the null hypothesis H0: ρ = 0 and the alternative hypothesis HA: ρ ≠ 0. The significance was assessed at the level of 0.05.

Smoothing of the calculated SSIs and circulation indices was presented with a moving trend (Hellwig 1967). This method is commonly applied for the analysis of long series with irregular and frequent variations in the direction of trends. In order to perform the smoothing process, a smoothing constant k (k<n) is arbitrarily set. In the next step, the structural parameters of the trend linear function are estimated on the basis of successive fragments of the series with a fixed length k. Then, the structural parameters of the functions in each segment are estimated with the least squares method. The number of segments in a series equals n−k+1. Finally, arithmetic means of the theoretical values are calculated. The analysis was performed for the constant k = 10 years.

All the analyses were carried out within the space of a hydrological year, i.e. 1st November–31st October.

The analysis of mean annual flows concerned seven main tributaries of the lower Vistula basin. The findings as to the characteristics of the mean annual flows and mean annual specific discharges are presented in Table 1, and Figure 2 depicts the basic parameters of the distribution of mean annual specific discharges. The analysed rivers differ both in terms of their catchment area and the mean annual flow (Table 1). For the entire period of 1954–2018, the trends of the mean annual flow are statistically insignificant for all of the rivers (Table 1).

Fig. 2

The annual SSI for the rivers with the highest, average and lowest values of the mean annual flow in 1954–2018 are shown in Figure 3. Although the trends of the mean annual flows are insignificant for the period as a whole, some characteristic sub-periods can still be identified. The first one commenced in 1954 and continued until the early 1980s. At that time, it gave way to the other one, which lasted until the end of the study period. It may be emphasised that the differences in trends and the intensity of the years classified as dry or wet can be observed in both of these sub-periods.

Fig. 3

One of the research objectives was to investigate the possible dominance of the western zonal (W), southern meridional (S) or cyclonic (C) atmospheric circulation over the study area. In the analysed period, the advection from the west decidedly prevailed over the one from the east (index W = 114.6 per year), with a statistically significant growing trend for the index (p-value of 0.0062 for the MK test). The annual index W took a negative value only twice during the studied time (1969 and 1996) (Fig. 4), and the 21st century has only seen the dominance of the western circulation. Additionally, air inflow from the south was slightly more common than that from the north. Interestingly, however, the mean multi-year value of the index S stands at 0, indicating proportional (approx. 50% frequency each) inflow of air masses from the south and the north in the research period. That said, since the second decade of the 21st century, the intensity of the meridional circulation started to increase, with the maximum in 2014 (Fig. 4). Further, the frequency of anticyclonic synoptic situations was higher than that of cyclonic types (index C = −24.7 per year). However, the C index is also characterised by considerable volatility. Its trend is not significant (Table 2), with two characteristic sub-periods identifiable:

Since the mid-1950s to the early 1990s (increase in the frequency of low-pressure systems), and

Fig. 4

The course of the SSI-12 on all the studied rivers in the analysed period is shown in Figure 5.

Fig. 5

The total number of the identified dry periods ranged from 6 (Zgłowiączka River) to 10 (Wierzyca River) (Table 3). Interestingly, the total duration of dry periods in individual catchments differed substantially. The total share (months) of dry periods ranged from 163 (Drwęca River) to 217 (Osa River).

In all but two analysed rivers (Drwęca and Zgłowiączka), the longest dry period occurred between 1990 and 1994 (Table 4), with drought spells lasting from 49 months to 69 months. The Brda River is a particularly telling example, where a dry period, according to the adopted methodology, lasted 69 months, i.e. from February 1990 to October 1995, which makes it the longest one in the entire study period. At the other end of the spectrum is the Drwęca River with no dry periods observed. Although the SSI-12 for this river did happen to be in the negative area, on no occasion did they reach the value of −1.0. It should be noted, however, that although 1990–1995 was quantified as the longest dry period, its intensity on most rivers was not the highest.

The second characteristic dry period occurred in the 21st century between 2014 and 2017 (except the Zgłowiączka – drought starting in January 2013 and the Skrwa Prawa in January 2015) (Table 4). The duration of dry spells in that period ranged from 33 months to 61 months. The analysis of the number of months classified as moderately, severely or extremely dry (Table 4) indicates that the intensity of this drought was higher than that of 1990–1995 (Table 4). In no other dry spell throughout the entire research period were there so many months in which SSI-12 was lower than two standard deviations, which stands for extreme drought. The number of months marked as extremely dry ranged from 12 (30% of the total duration) on the Wda River to 22 (51% of the total duration) on the Drwęca River. The mean value of the SSI-12 also indicates a much greater overall intensity of those dry periods.

During dry periods, the SSI-12 and the C-12 circulation index expressed with the correlation coefficient are convergent as to their course. At the onset of a dry period, when the SSI-12 decrease, the value of the C-12 index tends to go down, as well. And conversely, at the end of a dry period, when the SSI-12 increase, so does the C-12 index. This relationship is exemplified by the Brda River (Fig. 6). Ergo, the occurrence and duration of dry periods are heavily influenced by the presence of anticyclonic systems (Table 5). That said, however, one cannot claim that the incidence of a dry period is up to any specific values of the C-12 index.

Fig. 6

In turn, the relationship between SSI-12 and S-12 and between SSI-12 and W-12 is not so unambiguous. We note that the values of the correlation coefficient tend to be negative, which suggests the opposite directions of the SSI-12 and the circulation indices S-12 and W-12. However, not all of the results are statistically significant, and they do not unequivocally determine the existence of such a relationship (Table 5). This may mean that the occurrence of dry periods over such a long time is heavily influenced by the masses of air flowing in from the north or south. Similarly, the development of droughts may be determined by weakened advection of air masses from the west and intensified inflow from the east. This means that the influence of air circulation should be considered individually and independently for each drought that occurred in the studied multi-year period.

The ambiguous results of the relationship between the direction of the SSI-12 and S-12 and SSI-12 and W-12 indices suggest that the development of all dry periods may be affected by air masses flowing in from various directions. Located in the transitional climate zone, the studied catchments are alternately influenced by marine and continental air from the Atlantic Ocean and Eurasian continent, respectively. We believe that in such areas characterised by variable inflow of air masses, each dry period should be considered and analysed individually.

During the dry period that started at the beginning of the hydrological year 1990 and lasted until 1994, or even until the end of 1995 (Brda and Wda), the S circulation indices took negative values, indicating the domination of the inflow of air masses from the north. On the other hand, the latest drought – the most intense in the entire study, which for most of the analysed rivers began in the mid-2014 hydrological year (except Zgłowiączka – in January 2013 and Skrwa Prawa – in January 2015) – was characterised by positive and high values of the indices, meaning that the inflow of air masses from the south prevailed. Although it was not the longest drought in the studied period, the SSI-12 indices reached extremely minimal values for virtually all the analysed rivers. Additionally, the number of months quantified as extremely dry was the highest and ranged from 11 (Skrwa Prawa River) to 22 (Drwęca River).