Relationship between Air Temperature Change and Southern Baltic Coastal Lagoons Ice Conditions

The relationship between air temperature (mainly winter, December-March) in Świnoujście, Gdynia, and Elbląg and ice parameters (dates of the first ice and disappearance of the last ice, the length of the ice season, number of days with ice, maximum ice height) of southern Baltic coastal lagoons (Szczecin, Puck, and Vistula) was investigated. Trends in these parameters were determined, too. The observation material comes from the archives of the Institute of Meteorology and Water Management and spanned the winters from 1950/51 through to 2019/20. Relationships between the selected ice parameters for the study basins and the values of air temperature were examined using correlation and regression methods. The regression equations and trends, as well as their correlation and determination coefficients, were determined. The statistical significance of these relationships was examined using the Fisher-Snedecor test. Strong correlations between ice parameters values and air temperature were obtained, characterized by high values of both correlation coefficients and statistical significance. All trends of ice parameters indicate mitigation of ice conditions. An acceleration in both temperature and ice condition mildening occurred in the late 1980s, and especially in the last years of the study period. These trends, except the first ice date, are statistically significant, some even at α < 0.001. The length of the ice season becomes significantly shorter, the number of days with ice and the maximum thickness is smaller, and the last ice is disappearing early. An increase in the correlation and determination coefficients and a characterized trend of ice parameters values towards the East was found. It shows the increased impact of a warming climate in this direction on the southern Baltic coast. Strong correlations and trends may be of prognostic significance.


Introduction
Southern Baltic coastal lagoons are important basins for the maritime economy. These basins are crossed by shipping lanes leading to large ports: to Police, Stepnica and Szczecin via Zalew Szczeciński; to Puck, Kuźnica and Jastarnia via Puck Lagoon, and to Baltiysk, Kaliningrad, and Elbląg via Vistula Lagoon. In addition to the ports mentioned above, situated along the coasts of the lagoons are numerous smaller harbors, mostly fishing ports and marinas. The lagoons are sheltered basins, separated from the sea by islands (Szczecin Lagoon) or spits (Puck Lagoon and Vistula Lagoon). Only Puck Lagoon is in direct contact with marine waters via the Gulf of Gdańsk. The remaining lagoons are linked to the sea via channels. In the present paper, the inner part of Puck Bay (which is the western part of the Gulf of Gdańsk) is termed Puck Lagoon due to the considerable shallowing of Rewa Mewia (in some papers referred to as Seagull Shallow or Mewia Shallow), protruding toward the SW from Hel Peninsula, and Rewa Cape, protruding from the mainland coast toward the NE opposite the Seagull Shallow. Rewa Mewia is an elongated shoal, only about 1 m deep along its whole length, and in some places (especially close to the shore), it is entirely emerged.
Both these features distinctly separate the NW part of the bay from the remainder, effectively making it a lagoon with respect to water circulation, temperature and ice Ice conditions were described by five ice parameters, including first ice (F), last ice (L), ice season duration (S), number of days with ice (N), and maximum ice thickness (H) in a given winter season. There are standard parameters, commonly accepted as the best descriptors of ice conditions during a given winter in a study region cf. [14][15][16]. The first ice is the date when any ice phenomenon occurred in a given season. Last ice is the last day of ice occurrence in a given winter season. Ice season duration is the number of days from first ice to last ice, inclusively, and the number of days with ice is the sum of days, during which ice occurred on a given basin in a given winter season.
In order to assess the diversity of distribution among these ice parameters, we used a coefficient of variability (i.e., the standard deviation to the arithmetic mean ratio for a given variable). For regression equations concerning calendar dates, the actual dates of F and L were converted to numerical values. 1 November was assigned number 1, 2 November-number 2, and so on, until the latest date (19 April), which was assigned the consecutive number 170. In order to examine the relationship between AT and ice parameters in more detail, we used mean AT values not only for the standard winter period spanning December through March (Dec-Mar) but also for the periods Nov-Dec, Dec-Jan, Feb-Mar, and Mar-Apr. This was prompted by the observation that the first ice tended to occur within the former two intervals, while the last ice tended to disappear from the study basins in the latter two intervals.
Empirical relationships between ice parameters (F, L, S, N, H) and average winter AT were investigated using correlation and regression methods. Ice parameters were the assumed dependent variables (predicted variables, y), and AT was regarded as the independent variable (predictor variable, x). We applied linear regression expressed by the equation y = ax + b, where a is the slope of the regression line and b is the intercept. The In order to assess the diversity of distribution among these ice parameters, we used a coefficient of variability (i.e., the standard deviation to the arithmetic mean ratio for a given variable). For regression equations concerning calendar dates, the actual dates of F and L were converted to numerical values. 1 November was assigned number 1, 2 Novembernumber 2, and so on, until the latest date (19 April), which was assigned the consecutive number 170. In order to examine the relationship between AT and ice parameters in more detail, we used mean AT values not only for the standard winter period spanning December through March (Dec-Mar) but also for the periods Nov-Dec, Dec-Jan, Feb-Mar, and Mar-Apr. This was prompted by the observation that the first ice tended to occur within the former two intervals, while the last ice tended to disappear from the study basins in the latter two intervals.
Empirical relationships between ice parameters (F, L, S, N, H) and average winter AT were investigated using correlation and regression methods. Ice parameters were the assumed dependent variables (predicted variables, y), and AT was regarded as the independent variable (predictor variable, x). We applied linear regression expressed by the equation y = ax + b, where a is the slope of the regression line and b is the intercept. The obtained relationships were evaluated statistically. Correlation (R) and determination (R 2 ) coefficients were computed. Statistical significance of the relationships was examined using the Fisher-Snedecor test [17].
Ice occurrence probability was calculated according to the rules set out in the papers by Jevrejeva et al. [18], Leppäranta [19], and Karetnikov et al. [20]. Ice occurrence probability Atmosphere 2021, 12, 931 5 of 20 on a given basin in consecutive years (p), and standard deviation of the estimator (SD) were computed using the following formulas: where: N represents the total number of seasons (years), n-the season number, I(n)binary variable, I(n) = 0-season n is ice-free, I(n) = 1-ice occurs during season n. Ice condition value variability coefficients (i.e., the standard deviation to the arithmetic mean ratio) were also computed in the present work.

Description of Ice Conditions, and Probability of Ice Occurrence on Coastal Lagoons
Coastal lagoons, in comparison to the open southern Baltic Sea, are much shallower, less saline, and more sheltered, which is reflected in the considerably more frequent (common, in nearly every year) occurrence of ice phenomena on the lagoons cf. [21][22][23]. Ice conditions on the coastal lagoons are also more severe than on the Odra River, and slightly milder than on the coastal lakes cf. [13,24,25]. Fast ice cover is the most frequent ice form on the coastal lagoons (especially on Vistula Lagoon), and fine ice forms such as grease ice, shuga, and pancake ice, as well as ice floes, are dominant on lakes and rivers.
Coastal lagoons are characterized by relatively weak water dynamics. The strongest motion of water masses is due to wind-induced wave action. Water currents are very weak, and no tides occur [22,23]. Mean depths of the study basins range from 2.6 m for Vistula Lagoon to 3.8 m for Szczecin Lagoon. For each study basin, the surface area to average depth ratio is high, as reflected by the very high exposure index values (Table 1). Coastal lagoons freeze considerably earlier than the unsheltered, coastal marine waters of the southern Baltic. Ice occurs on the coastal lagoons nearly every winter. The first ice occurs on the lagoons rather early, even as early as November. On Vistula Lagoon, ice has been observed as early as late October (22 October 1979, Ušakovo). Ice occurs over a relatively long period, in extreme cases even up to 166 days (Krasnoflotskoye), and its maximum thickness reaches up to 70 cm (Tolkmicko) [21]. In the sheltered basins of the southern Baltic, higher ice thicknesses are noted only on the Curonian Lagoon (up to 90 cm) [27,28].
On Szczecin Lagoon, fast ice disintegration and drift occur earlier than on the remaining two lagoons. Fast ice is divided on Szczecin Lagoon by a furrow which is kept ice-free, thus sustaining shipping via a shipping lane that enables seagoing vessels to enter Szczecin Lagoon from Pomeranian Bay (Baltic Sea), en route to the seaport in Szczecin. The furrow accelerates the disintegration of fast ice cover. Further, an extensive current polynya occurs where the waters and ice ouflow to the sea, i.e., at the southern end of Piastowski Channel (an artificial waterway connecting Szczecin Lagoon to the Pomeranian Bay). The inflow of waters from the Odra River also accelerates the disintegration of fast ice cover in the southern part of the lagoon. Fast ice survives the longest in the northeastern part of Szczecin Lagoon. This is facilitated not only by bathymetric conditions (shoals) but also by a clear dominance of winds blowing from the SW in winter.
On Puck Lagoon, the relatively early fast ice cover disintegration is facilitated by an influx of marine waters from the southeast. Winds blowing from the W/NW push ice floes toward the Gulf of Gdańsk. On Puck Lagoon, fast ice survives the longest in the NW part of the basin. This is facilitated by the presence of land sheltering this part of the lagoon, especially Hel Spit.
Ice drift usually occurs on the study basins during the period of fast ice cover disintegration (late winter/early spring). During periods of strong winds, ice floes drift in an eastward direction, inducing rafted ice and piled ice formation. Ice hummocks form along eastern coasts of the lagoons, and on shoals, often reaching a height of several meters, with a maximum height of 10 m. Mathematical models concerning ice rafting and piling and the height of ice piles on southern Baltic coastal lagoons are presented in the paper by Girjatowicz [29]. As shown by Kolerski et al. [30], ice conditions will locally undergo modification due to the ongoing construction of the channel across Vistula Spit, which separates Vistula Lagoon from the sea, and the ongoing construction of artificial islands within Vistula and Szczecin Lagoons.
The earliest ice phenomena observed, mostly during wave action, are coastal grease ice, shuga, and pancake ice, and when conditions are still-coastal ice rind. Such ice conditions occur the earliest in small, shallow bays that are sheltered from wind and waves.
Characteristic values of ice parameters on southern Baltic coastal lagoons are tabulated in Table 2      The last ice phenomena are fine floes and brash ice. The location of the last ice disappearance on the lagoons is determined by wind direction. In spring, the prevailing wind direction is from the SW, which causes the last ice to melt in NE parts of the respective lagoons. The period between F and L (i.e, the ice season) is progressively longer toward the east. On average, S equals 64 days for Szczecin Lagoon to 69 days for Puck Lagoon to 94 days for Vistula Lagoon (Table 2, Figure 2). The longest ice season took place on Vistula Lagoon during the very severe winter of 1962/63 and lasted for 151 days . Similarly, N also increases toward the east, on average from 51 days for Szczecin Lagoon to 80 days for Vistula Lagoon. The maximum observed N values are 123 to 138 days, respectively (Table 2, Figure 3). Additionally, H increases toward the east, on average from 17 cm for Szczecin Lagoon to 28 cm for Vistula Lagoon (Table 2, Figure 4).

Lagoons
An analysis of variability coefficients for ice parameters ( Table 2) indicates that these coefficients are higher for Szczecin and Puck Lagoons, and lower for Vistula Lagoon. Vistula Lagoon is located in the eastern part of the southern Baltic coast. In winter, it is characterized by more severe, and more stable climatic conditions. The variability in ice parameters on Puck Lagoon is strongly influenced by the inflows of marine waters. In this case, contact with open marine waters is considerably larger than in the case of Szczecin and Vistula lagoons. This results in a stronger motion and exchange of waters, and consequently, in relatively high ice parameter variability coefficients (Table 2). In the  Box and whisker plots, presenting descriptive measures for first ice (F) and last ice (L) on Szczecin Lagoon (Sz-F and Sz-L, respectively), Puck Lagoon (P-F and P-L, respectively), and Vistula Lagoon (V-F and V-L, respectively) through the period 1950/51-2019/20. Boxes indicate the positions of the lower and upper quartile, and whiskers represent the maximum and minimum values of ice parameters. Consecutive day numbers (with 1 Nov as day 1) are given instead of dates.    The last ice phenomena are fine floes and brash ice. The location of the last ice disappearance on the lagoons is determined by wind direction. In spring, the prevailing wind direction is from the SW, which causes the last ice to melt in NE parts of the respective lagoons. The period between F and L (i.e, the ice season) is progressively longer toward the east. On average, S equals 64 days for Szczecin Lagoon to 69 days for Puck Lagoon to 94 days for Vistula Lagoon (Table 2, Figure 2). The longest ice season took place on Vistula Lagoon during the very severe winter of 1962/63 and lasted for 151 days (18 November-17 April). Similarly, N also increases toward the east, on average from 51 days for Szczecin Lagoon to 80 days for Vistula Lagoon. The maximum observed N values are 123 to 138 days, respectively (Table 2, Figure 3). Additionally, H increases toward the east, on aver-  2014/15, and 2019/20. It is clear from this overview that ice-free conditions tend to occur in the latter half of the study period. Ice occurrence probability (p) increases toward the east. For Szczecin Lagoon, p equals 0.900, for Puck Lagoon p = 0.943, and for Vistula Lagoon p = 0.986 (Table 3). The standard deviation of ice occurrence probability (SD) is higher in the western part of the study area (Szczecin Lagoon, 0.036) than in the east (Vistula Lagoon, 0.014). The eastward decrease in the number of ice-free days, increase in p (along with a concomitant decrease in SD) are influenced not only by winters being more severe in the east but also by the higher stability of climatic conditions in that direction. This is related to the stronger continental influence in the eastern part of the southern Baltic coast. Notably, relatively strong relationships occur between individual ice parameters on the coastal lagoons, with correlation coefficients usually exceeding 0.80 (Table 4). Only F correlates less strongly with the studied ice parameters (especially with L). The strongest relationship concerns N and S. On Puck Lagoon, the correlation coefficient for both variables equals 0.90. All relationships between the studied ice parameters are statistically significant at α < 0.001 level. Only some relationships with F are statistically significant at a slightly lower level (Table 4). This section may be divided into subheadings. It should provide a succinct, precise description of the experimental results, their interpretation, as well as the experimental conclusions that can be drawn.  The southern Baltic coast is located in a temperate climate zone and is characterized by seasonal changes in temperature and insolation. Two main and two transient seasons occur, determined by the quantity of solar energy influx, and are characterized, especially in the cooler half of the year, by a high intensity of atmospheric circulation. This causes high variability in weather conditions, and diverse, and temporal variable air temperatures. Air temperature amplitudes between summer (Jun-Aug) and winter (Dec-Feb) are very pronounced and exceed 17 • C [31]. Absolute amplitudes, however, equal 61 • C forŚwinoujście (from 37.4 to −23.6 • C), and 66.6 • C for Elbląg (from 36.5 to −30.1 • C) [32]. Mean monthly air temperatures are the lowest in January and range from 0 • C (Świnoujście) to −1.5 • C (Elbląg) [31]. A higher air temperature in the western than the eastern part of the coast influences the diversity of ice conditions on the coastal lagoons. Ice conditions are milder in the western part of the coast, which is manifested in a later ice occurrence and earlier disappearance, shorter ice season duration, a lower number of days with ice, and a lower maximum ice thickness cf. [1,33].

Analysis of Relationships between Coastal Lagoon Ice Parameters and Winter Temperature Conditions
Southern Baltic coastal lagoons are located at a similar latitude and within the same climatic zone. They are also characterized by similar physiographic (surface area, depth) and hydrologic conditions. This influences the high similarity among the strength of relationships between ice parameters and AT, but some diversity is evident. The relationships between ice parameters and winter AT are usually highly statistically significant (α < 0.001, Table 5). They have high correlation coefficients (R), indicating a very strong inverse correlation (except for correlation to F), ranging from −0.81 to −0.93. The strongest relationships, with average correlation coefficients concern N (−0.92). Slightly weaker relationships were observed for L (−0.87), H (−0.86), and S (−0.83). The relationships for F, however, are distinctly weaker. This is because F is influenced not only by AT, but also by water temperature, wind, snowfall, and marine water incursions. The relationships of F and L with AT Nov-Dec and AT Feb-Mar , respectively, were stronger than the relationships obtained when AT was averaged for the whole standard winter period (Dec-Mar). This was because F most frequently occurred in the Nov-Dec period, and L most frequently occurred in the Feb-Mar period. Table 5. Correlation coefficients between the air temperature in individual periods and ice parameters (designations as in Table 2 On the study basins, AT makes the strongest impact on N. Linear regression determination coefficients in these cases range from 0.80 for Vistula Lagoon to 0.86 for Puck Lagoon ( Figure 5). This means that the variability in winter (December-March) AT explains the variability in N on the studied basins as 80-86%. As indicated by the y-intercept of the regression line, a 1 • C increase in AT from December to March will cause N to become reduced by as many as 5 and 19 days on these basins, respectively ( Figure 5). On the study basins, AT makes the strongest impact on N. Linear regression determination coefficients in these cases range from 0.80 for Vistula Lagoon to 0.86 for Puck Lagoon ( Figure 5). This means that the variability in winter (December-March) AT explains the variability in N on the studied basins as 80-86%. As indicated by the y-intercept of the regression line, a 1 °C increase in AT from December to March will cause N to become reduced by as many as 5 and 19 days on these basins, respectively ( Figure 5). The AT influences N more strongly than S because N includes only those days on which ice actually occurred. S is defined by the dates of F and L. Ice-free days may occur in between. Thus, this parameter may not correspond to temperature conditions as closely as N. N, which correlates the strongest with temperature conditions of a given winter, will describe ice conditions on a given basin more accurately (the most accurate among the studied parameters).
Very strong correlations (negative) concern also the relationships of L with ATFeb-Mar. Determination coefficients range from 0.75 for Szczecin Lagoon to 0.77 for Vistula Lagoon ( Figure 6). This means that air temperature variability explains the variability in L on the lagoons as 75-77%. As indicated by the regression equation, a 1 °C increase in AT in the Feb-Mar period will cause L on the lagoon to occur on average 9-11 days earlier ( Figure  6). The AT influences N more strongly than S because N includes only those days on which ice actually occurred. S is defined by the dates of F and L. Ice-free days may occur in between. Thus, this parameter may not correspond to temperature conditions as closely as N. N, which correlates the strongest with temperature conditions of a given winter, will describe ice conditions on a given basin more accurately (the most accurate among the studied parameters).
Very strong correlations (negative) concern also the relationships of L with AT Feb-Mar . Determination coefficients range from 0.75 for Szczecin Lagoon to 0.77 for Vistula Lagoon ( Figure 6). This means that air temperature variability explains the variability in L on the lagoons as 75-77%. As indicated by the regression equation, a 1 • C increase in AT in the Feb-Mar period will cause L on the lagoon to occur on average 9-11 days earlier ( Figure 6).
No spatial diversity was observed in the strength of relationships between ice parameters and winter AT on the studied basins. The values of correlation coefficients of the respective ice parameters among individual lagoons are minor, of the order of several 0.01. AT was observed to make the strongest influence on N (R = −0.93), S (R = −0.84), and F (R = 0.70), in the eastern part of the coast, mainly on Puck Lagoon (Table 5). Variability in N, S, and F is explained by AT variability as 86, 71, and 49%, respectively ( Figure 7). As indicated by the regression equations, a 1 • C air temperature increase on Puck Lagoon will reduce N (and S) by 19 days, and delay F by 12 days (Figure 7). On Vistula Lagoon, L (R = −0.88) and H (R = −0.087; Table 5) are the most strongly dependent on AT changes. Szczecin Lagoon is characterized by slightly weaker relationships. Such differences in relationship strength for individual basins are influenced by the degree of winter severity and the stability of winter temperature conditions. The western part of the coast is under a strong influence of the oceanic climate that is milder and more variable than the climate of the eastern part of the southern Baltic coast.  (Table 5). Variability in N, S, and F is explained by AT variability as 86, 71, and 49%, respectively ( Figure 7). As indicated by the regression equations, a 1 °C air temperature increase on Puck Lagoon will reduce N (and S) by 19 days, and delay F by 12 days (Figure 7). On Vistula Lagoon, L (R = −0.88) and H (R = −0.087; Table 5) are the most strongly dependent on AT changes. Szczecin Lagoon is characterized by slightly weaker relationships. Such differences in relationship strength for individual basins are influenced by the degree of winter severity and the stability of winter temperature conditions. The western part of the coast is under a strong influence of the oceanic climate that is milder and more variable than the climate of the eastern part of the southern Baltic coast. The overall conclusion is that the relationships between ice parameters and AT are very strong. This is indicated by high correlation and determination coefficient values and their high statistical significance. This can be explained by the very shallow depth (high exposure index), and the isolation of the coastal lagoons from marine waters, which together enable rapid reaction of the studied ice parameters to changes in AT.
Of all the ice parameters studied for the coastal lagoons, N displays the strongest relationships with AT. This parameter is the best descriptor of the coastal lagoon ice conditions and is closely linked to the severity of winter. In turn, of all the ice parameters studied for the coastal lagoons, F displays the weakest relationships with AT. This is because the formation of first ice is influenced not only by AT, but also by other factors, such as water temperature, snow cover, wind, or marine water incursions (salinity).
In addition to air temperature, future studies should also focus on the influence of water temperature on the first ice occurrence date, and the influence of solar factors on the last ice disappearance date.

Analysis of Trends in Coastal Lagoon Ice Parameters
Climate warming has influenced the mildening of ice conditions on the southern Baltic coastal lagoons. AT Dec-Mar values measured atŚwinoujście, Gdynia, and Elbląg weather stations display a positive trend, statistically significant at α < 0.01 level. Correlation coefficients for these trends range from 0.35 to 0.39, and determination coefficients, respectively, from 0.12 to 0.16 (Figures 8 and 9). This means that AT increases on the southern Baltic coast are explained as 12-16% over the passage of time. In earlier periods, up to the winter of 1986/87, low AT and high ice parameter values (L, S, N, H) occurred relatively frequently. From the winter of 1987/88, mild winters definitely prevail, which is manifested in a clear mildening of ice conditions. As indicated by the regression equations of linear trends in AT at the studied weather stations, winter air temperature rises along the southern Baltic coast by 0.03-0.04 • C/year, or 3-4 • C per 100 years (Figures 8 and 9). The overall conclusion is that the relationships between ice parameters and AT are very strong. This is indicated by high correlation and determination coefficient values and their high statistical significance. This can be explained by the very shallow depth (high exposure index), and the isolation of the coastal lagoons from marine waters, which together enable rapid reaction of the studied ice parameters to changes in AT.
Of all the ice parameters studied for the coastal lagoons, N displays the strongest relationships with AT. This parameter is the best descriptor of the coastal lagoon ice conditions and is closely linked to the severity of winter. In turn, of all the ice parameters studied for the coastal lagoons, F displays the weakest relationships with AT. This is because the formation of first ice is influenced not only by AT, but also by other factors, such as water temperature, snow cover, wind, or marine water incursions (salinity).
In addition to air temperature, future studies should also focus on the influence of water temperature on the first ice occurrence date, and the influence of solar factors on the last ice disappearance date.

Analysis of Trends in Coastal Lagoon Ice Parameters
Climate warming has influenced the mildening of ice conditions on the southern Baltic coastal lagoons. ATDec-Mar values measured at Świnoujście, Gdynia, and Elbląg weather stations display a positive trend, statistically significant at α < 0.01 level. Correlation coefficients for these trends range from 0.35 to 0.39, and determination coefficients, respectively, from 0.12 to 0.16 (Figures 8 and 9). This means that AT increases on the southern The statistical significance of trends in ice parameters is high for all the studied lagoons, and mostly equals α < 0.001 ( Table 6). The average values of correlation coefficients range from −0.32 (H) to −0.50 (S). The largest changes, i.e., the strongest decreasing trends, concern S. The slope of the regression line (a) values for regression lines range from −0.89 for Vistula Lagoon to −1.08 for Puck Lagoon. These are very high values, indicating a reduction in S by about a day per year (Figure 8). Determination coefficients for the discussed trends equal 0.28 and 0.26, respectively. For Vistula Lagoon, the reduction in S is thus explained as 28% over the passage of time. A similar trend was obtained for N on Vistula Lagoon (Figure 9). The regression equations indicate that N and S are both being reduced on average by 0.9 days/year. The first ice occurs on average 0.3 days/year later, and the last ice disappears on average 0.5 days/year earlier. H is diminishing on average by 0.3 cm/year (Figure 9). frequently. From the winter of 1987/88, mild winters definitely prevail, which is manifested in a clear mildening of ice conditions. As indicated by the regression equations of linear trends in AT at the studied weather stations, winter air temperature rises along the southern Baltic coast by 0.03-0.04 °C/year, or 3-4 °C per 100 years (Figures 8 and 9).  Nearly all trends in ice parameters, except for F, display an increase in intensity toward the east (Table 6). Correlation coefficients for the relations between individual ice parameters and passage of time increase from the west (Szczecin Lagoon) to the east (Vistula Lagoon) by 0.07 for S; 0.09 for L, 0.10 for N, and 0.19 for H, respectively. However, absolute values of average correlation coefficients increase by 0.08, that is, from 0.36 for Szczecin Lagoon to 0.44 on Vistula Lagoon (Table 6). An increase in correlation coefficients for trends in ice phenomena toward the east indicates the impact of climate warming is stronger in the eastern than in the western part of the southern Baltic coast. Negative values of correlation coefficients for trends in ice parameters indicate that L will occur 0.36-0.52 days/year earlier, S will become 0.89-1.08 days/year shorter, N will be diminishing by 0.72-0.91 days/year, and H will become 0.14-0.32 cm/year thinner. It is a clear manifestation of climate warming in Europe, like in the whole of the Northern Hemisphere.   In general, both the determined trends in winter AT and in coastal lagoon ice parameters indicate a pronounced mildening of climate conditions. An acceleration in both temperature and ice condition mildening occurred in the late 1980s, and especially in the two last decades of the study period . Trends in AT and individual ice parameter changes are relatively strong and highly statistically significant. The reason for mildening in winter ice conditions may be explained by progressive climate warming. The strongest trends observed for the coastal lagoons concern S and N. H and F are undergoing changes to a considerably lesser extent.

Discussion
The analyses performed as part of this study have shown a very strong influence of AT changes on ice conditions on the studied basins, although the investigated relationships between AT and ice parameters were spatially diverse. Various parameters were dependent on temperature increase to a variable degree, and not always were the relationships very strong. Although AT, dependent on insolation and air mass circulation, is the main factor influencing ice phenomena formation, there are also other factors playing significant parts. These include water temperature, which obviously depends mostly on AT. Most of all, water temperature influences the occurrence of the first ice phenomena in a given season. Water temperature has no significant influence on the remaining ice parameters, because in the period from the occurrence of the first ice all the way until the disappearance of the last ice, water temperature is nearly constant and close to 0 • C. L is to some degree influenced by the insolation and intensity of solar radiation. L is also influenced by H in a given season [19], and the occurrence of rainfall (liquid precipitation) and strong wind [34]. H is in turn influenced to a high degree by the occurrence of snow cover during a given winter [19].
Ice conditions are influenced also by local geographic conditions such as bathymetry, distance from the open sea, or the degree to which a basin is sheltered from the open sea. Shallow and more exposed basins cool more rapidly, which results in a more intense development of ice phenomena cf. [25,35]. Notably, on unsheltered, marine, deeper basins characterized by more intense water dynamics, the influence of factors other than the temperature on ice phenomena is stronger than on lagoons. Such factors do weaken the relationships with air temperature on lagoons, but the degree of weakening is higher on marine basins. For instance, along the southern Baltic coast (in Hel), the correlation coefficient for the relationship between N and AT was lower than those for the studied lagoons and equaled −0.53 [12]. Such relationships may also be weakened by anthropogenic factors, such as icebreaking services, or discharge of heated and saline waters that disrupt the natural development of ice phenomena. These and other factors (wind, marine water incursions, waves, currents) will obviously weaken the relationship between ice conditions and winter AT. Including these factors in the analysis is impossible, for instance, due to the lack of observational data. For this reason, the current work is restricted to the most important factor, i.e., AT variations.
Differences in salinity do not impact the differences in ice conditions among the studied lagoons. Of all the basins, Puck Lagoon is characterized by the highest salinity, on average 5.97 PSU (Puck station) [26]. This value is only slightly lower than average water salinity in the southern part of the open Baltic Sea, equal to as little as 7.5 PSU [36]. The salinity of Vistula Lagoon in Tolkmicko is 2.34 PSU, and in Szczecin Lagoon in Podgrodzie the salinity is only 0.81 PSU [37]. The latter lagoon may therefore be considered a freshwater basin. Even though Szczecin Lagoon has the lowest salinity, it is characterized by the mildest ice conditions. In this case, the water chemistry is a much less significant factor than climatic conditions, specifically an increase in the severity of winters toward the east.
Scientific papers on the contemporary climatic changes emphasize progressive global warming, especially in central and northern Europe. In higher latitudes, it is anticipated that permafrost will recede, glaciers will melt, ice phenomena will develop less intensely, and progressively larger areas of seas and oceans will remain free of ice cf. [38][39][40]. The influence of the greenhouse effect on the mildening of ice conditions was evident on the Baltic Sea as early as the end of the 18th century [41,42]. Since then, ice conditions on the Baltic Sea have been progressively milder cf. [43][44][45]. It is estimated that during the 21st century, the maximum sea ice extent on the Baltic Sea will become reduced by 50 to 80% [46]. Through the 20th century (specifically, from 1896 to 1993), on the basins located along the southern Baltic Sea coast, S was becoming 1-3 days shorter per decade [47]. The largest changes in the Baltic Sea ice cover concern the western part of the basin, and areas east of Bornholm [48]. These changes were not as pronounced in the whole of the Baltic. For instance, in the Gulf of Finland, through the latter half of the 20th century, no statistically significant trend in S was observed [49]. Progressively more intense mildening of ice conditions on the Baltic Sea is manifested by the decrease in values of the basic ice parameters. N is decreasing [18,49], as is H [18,50]. The ice season is becoming shorter, depending on locality, by 14 to 44 days per century [18,50,51]. In addition, ice disintegrates by 8 to 12 days per century earlier [18]. Ice disappearance is also progressively earlier in individual winters [49,52]. Similar to the studied lagoons, changes in S (0.8 days per year) were observed also on the Curonian Lagoon [11], located several tens of km NE from the northern coast of Vistula Lagoon. Further, from the latter half of the 19th century, there has been a clear decreasing trend in the annual maximum sea ice extent on the Baltic Sea [50,53].
The results of the present work show that already in the late 1980s there was a clear acceleration of mildening in temperature conditions and ice conditions on the southern Baltic coastal lagoons. Additionally, in distant Japan (Saroma-ko Lagoon, Hokkaido Island), changes in ice parameter values displayed similar dynamics. A sharp decrease in the number of days with ice has been observed there since 1988. Since then, winters without complete ice coverage occur there frequently. Before 1988, such winters were rare on Saroma-ko Lagoon [54].

Conclusions
Southern Baltic coastal lagoons are shallow, brackish basins characterized by high exposure indices, which facilitate rapid cooling, and intense development of ice phenomena. Their waters are susceptible to changes in temperature and solar conditions, especially to AT variations. In the autumn-winter period, these basins quickly release the accumulated warmth and become covered in ice, predominantly fast ice (ice rind, ice cover). The lagoons are linked to the sea via straits, which may serve as a conduit for marine water incursions in the autumn-winter period. Such marine waters are slightly warmer and more saline than those in the lagoons. Only Puck Lagoon, from the SE, via Puck Bay, is open to marine waters. The close proximity of the Baltic Sea influences the waters of the lagoons by warming them in winter.

-
On coastal lagoons, winters are more severe and ice phenomena are more intense toward the east. In the eastern part of the southern Baltic coast, ice occurs earlier and disappears later, the ice season is longer, the number of days with ice is higher, and the maximum ice thickness is larger. -Additionally, ice cover stability (N/S ratio), p, correlation, and determination coefficients for relationships between ice parameters and AT all increase in an eastward direction. However, coefficients of ice parameter variability, except for F, decrease toward the east. -Physiographic conditions distinguishing the studied basins from the open sea significantly influence the relations between coastal lagoon ice parameters and AT. The relationships are considerably more significant for the coastal lagoons, in comparison to the marine waters of the southern Baltic. - The relationships between ice parameters and AT are weakened by inflows of warmer and more saline marine waters into the lagoons in autumn and winter. Other factors, like strong wind causing water and ice movement, or human activity (discharge of heated and saline waters, shipping traffic, icebreaking services, especially on Szczecin Lagoon), are also significant in this respect. - The strength of trends in ice parameters increases in an eastward direction, which indicates a stronger climate warming toward the east. This pattern is displayed the most clearly by H and N. -Winter AT is clearly increasing along the southern Baltic coast. At the same time, first ice phenomena occur later on the coastal lagoons, and last ice phenomena disappear earlier, S is distinctly shortening, and both N and H are diminishing. - The correlations between ice parameters and AT, and trends in ice parameters and AT, characterized by high correlation coefficients and high statistical significance, may be utilized for forecasting. Institutional Review Board Statement: Not applicable (study not involving humans or animals).

Informed Consent Statement:
Not applicable (study not involving humans).

Data Availability Statement:
The data used in the work are available on the website: https:// danepubliczne.imgw.pl/data/dane_pomiarowo_obserwacyjne/, accessed on 10 March 2021 and in the catalogs provided in the references list.