Long-term changes of the ice regime of rivers in Latvia

Climate change and the varying climate conditions are closely related to the changes in hydrological processes. Nowadays, several studies have proved that the increase of the global air temperature can cause essential changes in the hydrological cycle and its processes (Rummukainen et al. 2003; Andreasson et al. 2004; Beldring et al. 2008; Yang et al. 2010; Thorsteinsson & Björnsson 2011), including, inter alia, the Baltic Sea catchment basin (Jaagus et al. 1998; Kriaučiūnienė et al. 2008, 2009; Bethers & Seņņikovs 2009; Apsīte et al. 2011, 2013; Latkovska et al. 2012). Research regarding the changes in the Baltic Sea catchment basin shows that during the last 100 years the annual mean air temperature increase has exceeded global trends (0.10–0.07 °C in the Baltic Sea region and 0.05 °C globally) (Bolle et al. 2008). Moreover, based on several different future climate scenarios, a global temperature increase of 1.0–6.0 °C is also forecast for the end of the 21st century (Meehl et al. 2007), and in the Baltic Sea region it is forecast to be, on average, 3–5 °C (Bolle et al. 2008). In Latvia, the annual mean air temperature has increased by 1.4 °C during 1950–2003 (Lizuma et al. 2007), the mean amount of precipitation is increasing and the duration of sunshine is decreasing (Kļaviņš et al. 2008).

One of the indicators of climate change is ice phenology observations of lakes and rivers (Šarauskiene & Jurgelenaite 2008). The seasonal data of rivers, for example, the freeze-up date, the break-up date, the length of the ice cover and the maximum ice thickness serve as good indicators for assessing climate change, in particular, long-term climate change (Magnuson et al. 2000; Beltaos & Burrell 2003). Moreover, changes in the ice regime of rivers are also related to various changes in the physical, biological and social economic area (Beltaos & Burrell 2003; Vuglinsky 2006). The ice cover on rivers may impact the possible use of hydrotechnical constructions, for example, water locks and dams of hydro power stations (Zīverts et al. 2000; Vuglinsky 2006). The ice cover also influences the biological processes in rivers by defining the oxygen concentration under the ice (Chambers et al. 2000; Lima Neto et al. 2007). The decrease of the period with ice cover or non-formation of ice cover has an essential impact upon the ecosystem of the river (Scrimgeour et al. 1994; Wrona et al. 2006; Prowse et al. 2011).

In Latvia, the latest research regarding changes of rivers’ hydrological regimes has mainly been conducted regarding only the runoff of rivers and the ice regime for bigger rivers (Frisk et al. 2002; Kļaviņš et al. 2002, 2004, 2006, 2007, 2009; Kļaviņš & Rodinov 2008). The previous studies have mainly evaluated the long-term changes in water runoff, ice freeze-up/break-up dates and the duration of ice cover; however, they do not analyse the long-term and seasonal changes in the parameters characterising the split of ice regime of rivers and regional differences in rivers of various sizes. Earlier studies on the ice regime of rivers and lakes and their regional differences were conducted by Glazacheva in the 1960s (Glazacheva 1964, 1965).

Therefore, the objective of this study was to analyse the long-term changes in the ice regime of variously sized rivers in Latvia from 1945 to 2012, i.e., freeze-up date, break-up date, ice cover duration and thickness, and their regional peculiarities.

The date of the beginning of the first lasting (20 days and above) ice cover was assumed to be the date of the formation of the ice cover. If stable ice cover is interrupted one to three times by an ice-free period just lasting for a few days (for a period which is considerably shorter than the period with the ice cover), such interruptions were disregarded. The date when wash-outs, a coastal ice-free area appears, the ice breaks up and flows away was assumed to be the date of the break-up of ice. The length of the ice cover was estimated on the basis of the total number of days between the freeze-up and 1 day after its disappearance. The length of the ice cover includes the days with the movement of ice, ice-free areas and water on top of the ice.

The ice thickness was measured in centimetres in the middle of the river on the 5th, 10th, 15th, 20th, 25th and the last date of the month. The mean and the maximum ice thickness were calculated afterwards.

The dates of freeze-up and break-up of the ice cover were transformed into a numerical format, i.e., the day as from 1st October (1st October = day 1, 15th December = day 76, etc.), thus making the handling of the data easier. In order to analyse long-term trends for years where ice covers had not been recorded, 10th March (day 160 as from 1st October) was assumed to be the freeze-up date and 10th December (day 70 as from 1st October) was assumed to be the break-up date. These values are slightly above/below the actual recorded extreme values at each monitoring station.

In order to interpret the obtained results we used three regions (Figure 1): Western (the River Venta basin and small rivers along the coast of the Baltic Sea), Central (the River Lielupe basin and small rivers in the central part of Latvia and basins of the rivers Salaca and Gauja and also small rivers along the Riga Gulf coast) and Eastern (small and medium sized rivers in the River Daugava basin). In the Western hydrological region, there is a greater impact of meteorological processes occurring over the North Atlantic and the Baltic Sea on the river hydrological regime than for other regions in Latvia, particularly in comparison to the Eastern region (Kļaviņš et al. 2002). In this hydrological region, a comparatively shorter ice cover period can be observed (Glazacheva 1980). Geological and climatic conditions have determined the hydrological regime and dense river network in the southern part of the Central region. The northern part of the Central region is characterised by the highest amount of precipitation per year (800 mm and more) (Lizuma et al. 2010) and the shortest duration of vegetation period. The rivers are characterised by high level snowmelt floods, which account for 40–52% of the total annual runoff, and comparatively less pronounced runoff due to rainfall in autumn (amounting to an average of 22% of total annual runoff) than compared to the Western region. The Central region forms a transition territory where climatic conditions of both western and eastern Latvia are present. The Eastern region is characterised by more continental climate conditions than the others, i.e., warmer summers and colder winters with a thick snow cover.

Pearson correlation analysis was performed to test the relationship between the absolute value of the negative temperature sums and ice parameters (freeze-up date, break-up date, number of days with ice).

The multivariate Mann–Kendall test (Hirsch & Slack 1984; Lettenmaier 1988; Loftis et al. 1991) was used to identify the trend shift in the air temperature and ice data analysis. The test also allows the analysis of data rows if there are missing values in them, as well as data rows with non-typical (very low or very high) values. The test was applied to each variable at each site separately, at a significance level of p < 0.05. The trend was considered statistically significant at the 5% level, if the test statistic was above 1.96 or below −1.96.

If the sum of negative air temperatures is calculated, it can be seen that the sum of the negative air temperatures tends to decrease and this is related to the global climate deterioration during the last decades. The analysis of the relationship between the sum of the negative air temperatures and the parameters of the ice regime resulted in the conclusion that there was statistically significant correlation at most HMS, except one HMS Daugava–Krāslava in correlation with the freeze-up date, in three HMS (Lielā Jugla–Zaķi, Lielupe–Kalnciems, Daugava–Krāslava) in correlation with the mean ice thickness and two HMS (Mazā Jugla–Stariņi, Daugava–Krāslava) in correlation with the maximum ice thickness. The sum of the negative air temperatures shows a positive correlation with the freeze-up date, but the break-up date and the duration of ice cover show a negative correlation with the absolute negative air temperature sum which is statistically significant for all the rivers analysed within the study. Also, the ice thickness has a negative correlation with the sum of the absolute negative air temperatures.

The most favourable period for the formation of the ice cover in rivers is mid-December (the 10th to the 20th of December) in 36.5% of cases, followed by the end of December and the beginning of January in 29.5% of cases and also 4.5% of cases in February.

The calculated regression equation for the date of break-up of the ice cover indicates that over the period 1945–2012, break-up of the ice cover has shifted to earlier dates by 4–8 days per 10 years in the Eastern and the Central regions of Latvia. The biggest changes may be observed in the Western region where break-up of the ice cover shifted to earlier dates by 8–15 days per 10 years.

The calculated regression equation for the duration of the ice cover indicates that over the period of 1945–2012 the duration of ice cover has been decreasing by 6–15 days in the Eastern region of Latvia, by 1–11 days in the Central part and by 6–14 days in the Western region per 10 years.

Analysing the time period from 1945 to 2012, it can be seen that, starting from the 1970s, there are more cases when there was no ice cover. In the Eastern region of Latvia this trend is less significant, i.e., one to three cases per decade, however, the biggest changes are related to the Western region where the ice cover might not form four to six times during a decade.

Assessment of the calculated regression equation for the mean ice thickness shows that the mean ice thickness has been decreasing by only 1–5 cm in the Eastern and Central rivers and by 2–6 cm in Western rivers per 10 years.

The calculated regression equation regarding the date of the maximum ice thickness indicates that in the Eastern and North Eastern regions the date of the maximum ice thickness is recorded at 1–5 days earlier, but in the Central and Western regions by 3–6 days earlier per 10 years.

Based on the analyses of the winter severity index, the following seasons can be referred to as severe winters: 1945/1946–1946/1947, 1954/1955–1959/1960, 1968/1969, 1975/1976, 1983/1984, 1993/1994, 1995/1996, 2002/2003 and also 2009/2010–2010/2011. This also coincides with duration of the ice cover, where the longest durations can be mainly seen during the first decades over the period studied. The above-mentioned years are also the years with the latest break-up dates of the ice cover. The following seasons can be referred to as moderate winters: 1962/1963, 1973/1974, 1978/1979, 1987/1988–1988/1989, and the following as warm winters: 1964/1965–1965/1966, 1974/1975, 1980/1981, 1990/1991–1994/1995 and also 2004/2005–2007/2008, when formation of the ice cover was late and the break-up was early.

Air temperature is an indicator of climate change which is most often used for describing global climate change. The variability of the air temperature indicates that air temperature has been rising over the last decades in particular. According to the results of the study by Lizuma (2008) in Latvia, both the annual mean temperatures and mean minimum and mean maximum air temperatures have increased. In Latvia, the most significant increase in the mean air temperature may be observed in winter and spring (Lizuma et al. 2007; Lizuma 2008). Although the increase in summer and autumn is not as high, it is also statistically significant (Lizuma et al. 2007; Lizuma 2008). The increase in air temperature in the winter season has a direct impact upon the ice regime in Latvian rivers because, as stated by Livingstone (1997), statistically in 60–70% of cases, the changes in the ice regime can be explained by the air temperature. Thus, in order for ice cover to form, negative air temperatures are among the major preconditions; however, if negative temperature in the winter season decreases, it has an essential impact upon the ice regime, and duration of the ice cover in rivers decreases or the ice cover does not form at all and this is also confirmed by the results of our research. Moreover, as stated by Beltaos & Prowse (2009), the dates of formation and break-up of ice cover correlate very well with the air temperature approximately 1 month prior to the formation or breaking-up of the ice cover.

The data regarding formation of the ice cover, breaking-up of the ice cover, duration of the ice cover as well as ice thickness allow evaluation of the long-term and also the seasonal climate variability, in particular, in relation to global climate change. According to the results obtained by research and studies performed worldwide (Magnuson et al. 2000; Prowse et al. 2002; Beltaos & Burrell 2003; Vuglinsky 2006; Lemke et al. 2007; Beltaos & Prowse 2009), the later formation of the ice cover, its faster break-up and decrease of the duration of the ice cover can be seen in most regions of the Northern hemisphere. Regarding most rivers of the European part of Russia and West Siberia, as well as the Danube, there is a statistically significant trend regarding the later formation of the ice cover and its earlier break-up, resulting in decrease of duration of the season with ice cover by approximately 20 days over the period 1893–1991 (Beltaos & Prowse 2009). In Canada, Doyle & Ball (2008) in 22 monitoring stations on 19 rivers have found statistically significant later formation of the ice cover and earlier break-up, and increase of air temperature by ∼1.6 °C in winter during the period 1976–2005. Lacroix et al. (2005) also identified a much faster break-up of the ice cover in the Western part of Canada. White et al. (2007), regarding eight rivers in Alaska for the time period 1912–2001, found faster break-up of the ice cover by ∼6 days, and for the rivers of Maine State (USA) even by 16–37 days. Also in Lithuania, Stonevicius et al. (2008), in the course of analysis of data on the Nemuna River, found statistically significant later formation of the ice cover and its earlier break-up by approximately 10 days in both cases. The results of the aforementioned studies also coincide with the results of our research according to which the duration of the ice cover in Latvia has decreased on average by 6–15 days over the period 1945–2012.

There are few studies regarding trends of change in the ice thickness because such data cannot be obtained as easily as, for example, data of the dates of formation and break-up of the ice cover (Beltaos & Prowse 2009). However, as stated by Batima et al. (2004), ice thickness data would be much more accurate indicators of the climate change than the dates of the formation and the break-up of the ice cover, in particular, in the territories where the climate is more continental. The results of our research indicate that the mean and the maximum ice thickness in Latvian rivers have statistically significantly decreased, although the decrease is not comparatively high, i.e., 1–5 cm in the Central and Eastern regions of Latvia and 2–6 cm in the Western region as regards the mean ice thickness. The decrease in the maximum ice thickness is also similar, i.e., 1–4 cm in the Central and Eastern regions of Latvia and 2–8 cm in the Western region. This also coincides with the results of other studies such as, for example, in a study about Russian rivers, Vuglinsky (2006) reported that, generally, in all of the biggest rivers the ice thickness has decreased by 2–14 cm over the period from 1980 to 2000, and in the rivers that are located in the European part of Russia the decrease has been 2–7 cm. Similar trends were also identified in studies in other countries of the Northern hemisphere. For example, in the USA, Huntington et al. (2003), regarding the period from 1912 to 2001, reported the decrease of maximum ice thickness to be as much as 23 cm, and Batima et al. (2004) found a decrease of 20 cm in Mongolian rivers.

Both ice regime data and the winter severity index indicate climate warming in the territory of Latvia. Moreover, trends in changes of the ice regime are also similar in neighbouring countries; for example, in Lithuania (Dubra & Grecevičius 2007; Stonevicius et al. 2008; Šarauskiene & Jurgelenaite 2008) and in Russia (Vuglinsky 2006), as well as in Scandinavia (Magnuson et al. 2000). Analysis of the results reveals statistically significant changes in the ice regime, and the estimated winter severity index indicates much warmer and more humid winters that determine the later formation, earlier break-up and a shorter duration of the ice cover, and a lower ice thickness in rivers. However, as stated by Kļaviņš et al. (2009), the ice regime does not only depend on the meteorological conditions of the particular year and the distance from the Baltic Sea, it also depends on global climate change.

The research results also indicate the regional differences between the Latvian rivers as regards the ice regime. In rivers located in the Western region, the meteorological conditions have a much higher impact upon the ice regime, which is mainly determined by the transfer of humid air from the Atlantic Ocean over the Baltic Sea, mainly due to the operation of cyclones. The comparatively flat terrain promotes the inflow of air masses inland and, although the highlands of Latvia are not very high, they still have an impact on the split of air temperatures and precipitation at local scale. Hence, in the Western region, the climatic conditions are much milder, which is also confirmed by the winter severity indices obtained. Therefore, also the ice regime in rivers in the Western region differs from that in the Northern and Eastern regions. The research results demonstrated that in rivers in the Western region, the ice cover forms approximately 2 weeks later and break-up takes place approximately 3 weeks earlier resulting in a much shorter duration of ice cover. Moreover, the ice thickness data indicate a lower ice thickness in the Western region compared to the Eastern part, and the biggest ice thickness in the Western region may be observed as early as February, while in the Eastern region this mainly happens in March. Similar trends regarding regional differences were also recorded by Glazacheva (1965). As regards formation of the ice cover, during the period studied from 1926 to 1960 in the Western region of Latvia, it was observed that the formation of the ice cover took place 3–4 days later and in the rest of Latvia 1–2 days later. However, Glazacheva (1965) indicated general trends, in particular, in the Eastern region of Latvia formation of the ice cover takes place as early as the end of November to the beginning of December, and in the Western region (at the coast of the Baltic Sea and Gulf of Riga) this happens from the end of December to the beginning of January. A similar situation may be observed regarding the break-up of the ice cover: first, it takes place in the rivers of the Western region, with the latest break-up in the Eastern and the North Eastern regions. Glazacheva's (1965) study stated that break-up of the ice cover took place from mid-March to the beginning of April within a period slightly longer than 2 weeks. However, the results of our research indicate a much longer period of break-up of the ice cover, i.e., from the end of February to the end of March. As regards the ice thickness, Glazacheva (1965) did not observe considerable changes, only stating that the ice thickness in Latvian rivers depended on the climate conditions of the particular winter to a great extent; i.e., in warm and humid winters there may be no ice cover at all and in severe winters the ice thickness may reach as much as 1 m and above in certain rivers.

The long-term series of ice regime parameters (date of freeze-up and break-up, numbers of days with ice cover, annual average and maximum ice thickness) of Latvian rivers from 1945 to 2012 were examined by using statistical methods.

Analysis of the long-term changes of parameters characterising the ice regime indicates statistically significant later freeze-up, earlier break-up and decrease of duration of ice cover and the thickness of ice cover, which is mainly determined by the reduced sums of negative air temperatures (in the period from November to April) during the last decades.

The results of the research reveal a positive trend regarding the freeze-up date, however, a negative trend also emerges in some cases. A negative trend is mainly related to the small rivers of Vidzeme highland where there is high inflow of underground water and lower long-term mean water temperature. A negative trend can also be seen for the rivers flowing from Rietumkursa and Austrumkursa highlands. The earlier break-up of the ice cover is also confirmed by the results of the Mann–Kendall test. A statistically significant positive trend for the freeze-up date and a statistically significant negative trend for break-up dates indicate the decrease of the duration of ice cover in Latvian rivers. The results of the Mann–Kendall test demonstrated a statistically significant negative trend in 88.4% of cases and a negative trend in 11.6% of cases, which means that there is a statistically significant ongoing decrease in the duration of the river ice cover in all regions of Latvia. The longest durations of ice cover are related to the first decades of the research period until approximately the 1970s when the duration of ice cover is gradually getting shorter. Processes above the Northern part of the Atlantic Ocean have an essential impact upon the climate conditions in the Baltic region, in particular, in the cold season (November–April).

Based on the calculated regression equation for the period 1945–2012, the formation of the ice cover in Latvian rivers takes place, on average, 1–7 days later and break-up takes place approximately 4–15 days earlier; duration of the ice cover has decreased by 1–15 days. As regards the ice thickness, the mean ice thickness has decreased by 1–6 cm and the maximum thickness by 1–8 cm.

The estimated winter severity index during the last decades indicates warmer and more humid winters which is also confirmed by later freeze-up and earlier break-up, as well as a shorter duration of ice cover and a lesser thickness of ice. There are regional differences also regarding the winter severity index which is higher for the rivers located in the Eastern and Central regions of Latvia and lower for the rivers located in the Western region of Latvia and near the Baltic Sea.

Both the ice regime data and the winter severity index indicate climate warming in the territory of Latvia. Moreover, the trends of the change of the ice regime are also similar in neighbouring countries, confirming the hypothesis regarding the impact of climate change upon the ice regime in Latvian rivers.

We greatly acknowledge the support of the European Social Fund within the project ‘Support for Doctoral Studies at University of Latvia’ and the Latvian Council of Science (grants No. 526/2013 and No. 514/2012) for this research.