Tropospheric thermal field variability over Poland in the context of climate change

Based on the results of radiosounding measurements from all Polish aerological stations, an analysis of temperature changes in the troposphere and tropopause over Poland was carried out. The study was performed for data from 1981 to 2018 at 00 UTC for five main isobaric surfaces (MISs): 850, 700, 500, 300 and 200 hPa and for the tropopause. It was found that the tropospheric temperature variability is highest in the cold months of the year and lowest in summer. Annual temperature variability in the upper troposphere is less than in the lower. Temperature trends for the MISs from 850 to 300 hPa for the year and summer are positive and almost everywhere statistically significant at a significance level of .05. For a year, they average from about 0.2°C to 0.5°C/decade. In summer, they are slightly greater and are 0.4–0.5°C. On the MIS of 200 hPa over Poland, however, there are negative thermal trends, both in the year and in summer. They are approximately −0.1°C/decade. In winter, the trends are much smaller than in the summer or for the year. In the lower troposphere they are weak and positive, while in the upper troposphere they are weak and negative. Negative tendencies in the thermal tropopause over Poland were found. For the year and seasons, this trend ranges from −0.2 to −0.3°C/decade. The obtained results regarding the temperature of the troposphere and tropopause over Poland correspond with the results of earlier studies from Europe and the around world. They confirm the warming of the lower and cooling of the upper troposphere with the tropopause, which has been taking place in the last few decades.


| INTRODUCTION
Changing the vertical profile of air temperature in the atmosphere is an important criterion for diagnosing climate change (Santer et al., 1996;Tett et al., 1996;Hill et al., 2001;Duan, 2007;Thorne et al., 2011;Bodeker et al., 2016;Philipona et al., 2018). The troposphere is the most important layer of the atmosphere, because most of the weather events take place there. The tropopause has a complex dynamic in relation to the heat exchange process between the troposphere and the stratosphere, radiation and chemical changes, including changes in the amount of ozone and greenhouse gases-that is why it is a sensitive indicator of climate change (Sausen and Santer, 2003;Santer et al., 2003a;2003b;Seidel and Randel, 2006). All processes occurring in the tropopause zone can change both the physiochemical conditions in the upper troposphere as well as in the lower stratosphere (Holton et al., 1995;Gettelman et al., 2011).
The differences in tropopause location depend on the temperature changes in the upper troposphere and the lower stratosphere (Peethani et al., 2014). The increase in tropopause height observed in recent decades, documented by data from various sources (Sausen and Santer, 2003;Santer et al., 2003a;Seidel and Randel, 2006;Feng et al., 2012), is closely related to the gradual heating of the troposphere and cooling of the stratosphere (Santer et al., 2004;A nel et al., 2006;Austin and Reichler, 2008;Duan, 2007;Feng et al., 2012;Thompson et al., 2012; Intergovernmental Panel on Climate Change (IPCC), 2013; World Meteorological Organization (WMO), 2014; Berkes et al., 2017;Randel et al., 2017;Philipona et al., 2018). The results obtained by Pyka (1990) indicate, among others, that over Europe the trend of free atmosphere temperature according to annual values during 1961-1985 was positive for the three lower troposphere layers (850, 700 and 500 hPa), and negative in the 300-200 hPa layers. Similar results, with a positive trend of up to 200 hPa and with a negative above, were obtained by Santer et al. (2005).
The variability of the tropospheric thermal field as a good indicator of climate variability based on radiosounding data was also noted by Pyka (1990), Duan (2007), Feng et al. (2012) or Philipona et al. (2018). However, there is little work in the climatological literature based on the use of data from atmospheric soundings over Poland for research on climate variability, which is conducted by researchers in many countries. So far, the publication by Pyka (1990) is the pioneering Polish study. Changes in tropopause location at high latitudes were presented by Kowalewski (2003). The article was based on data from reanalysis from the decade 1991-2000, verified only for selected days from 1997 and 1998, with sounding data from Antarctic and Greenlandic stations. The latest publication, based on daily soundings at 00 UTC from Legionowo from 2001 to 2010, determined changes in the average annual and seasonal temperature and the location of selected isobaric surfaces in relation to atmospheric circulation (Urban and Kowalewski, 2019).
Atmospheric radiosounding data is a reliable source of data, free from influencing factors associated with changes in surroundings, or the location of ground measurement points, changing climatic conditions in the topo-or mesoscale, such as from the influence of the urban heat island. In addition, data from soundings are characterized by greater vertical resolution and better identification of the tropopause or main isobaric surfaces (MISs) than reanalysis or satellite data, and daily soundings dating back to the early 1960s allow analysis of their long-term variability (Durre et al., 2005(Durre et al., , 2006Feng et al., 2012).
The main purpose of the article was to characterize the temperature variability of selected isobaric surfaces of the troposphere and tropopause over Poland based on radiosounding data. An additional goal was to examine whether the regularities in the troposphere over Europe (including Poland) during 1961-1985 found by Pyka (1990) still persist and what is their direction and pace of change, taking into account the location of the first tropopause since the 1980s to the present.
This article is the first, comprehensive scientific study based on the results of daily sounding measurements of the atmosphere from all aerological stations in Poland, covering a uniform and sufficiently long period for climatological analysis. It can be an important source of information on the long-term variability of troposphere air temperature over Central Europe (Poland) in recent decades.

| SOURCE DATA AND METHOD
Source data were the results of daily radiosounding measurements of the atmosphere at 00 UTC over all Polish Aerological Measurement Stations of the Institute of Meteorology and Water Management-National Research Institute (IMGW-PIB), that is, in Wrocław (φ 51 07 0 N, λ 16 53 0 E, 122 m a.s.l.), Legionowo (φ 52 24 0 N, λ 20 57 0 E, 96 m a.s.l.), and Łeba (φ 54 45 0 N, λ 17 32 0 E, 1 m a.s.l.). Since they are located in different parts of the country, the stations represent different conditions as part of Poland's climate diversity. Wrocław, located in south-west Poland, with a moderately warm climate; Legionowo, located in central Poland, with a moderate climate and continental features; and Łeba, located in northern Poland, with a moderate sea climate.
Measurement data from 13,879 nighttime soundings from each station were used, which accounted for 456 months in 38 years during 1981-2018. In the analysed period, the locations of the above stations were not changed. Primarily the average annual (January-December) temperature values of selected isobaric surfaces and the tropopause were analysed. All tropopause analyses concerned the first tropopause (with the highest atmospheric pressure, lowest above the ground/sea level) defined as the boundary layer between the troposphere and the stratosphere, where the vertical gradient equals or is less than 2 C/km (WMO, 1957;WMO, 1996;Ivanova, 2013). To capture seasonal differences, their average values were also examined for the winter quarter (December-February) and for the summer quarter (June-August) and for individual months.
Regardless of the duration of the sounding and the horizontal movement of the balloon, the results refer to the vertical atmosphere above the station and to the time of the sounding. Meteorological start data for the radiosounding atmospheric pressure, air temperature, and relative humidity are taken from the Stevenson screen near the start field (WMO, 1996).
The choice of nighttime soundings was dictated by the desire to avoid the impact of solar radiation on the measurement results. Thus, the possible measurement error due to radiosonde heating was eliminated, which should not occur in the study of tropospheric temperature trends (Sherwood et al., 2005).
Many years of continuous measurements used at work, as in all aerological stations across the world, are not without their gaps. The reason may be a failed soundings, lack of archiving data or an archived value found to be incorrect. It was assumed that deficiencies of up to 5 days in a month would not significantly affect the monthly average. To verify this, an experiment was carried out, consisting of the intentional generation of 5-day gaps in 31-day months with a complete set of data, followed by the analysis of mean differences with gaps and averages from the complete set of data. Five days were consecutively rejected from each month (1-5, 2-6, …, 27-31, 28-31 and 1, 29-31 and 1-2, …). Thus, 31 averages of 26 days were obtained for each analysed month. A 95% confidence interval was calculated for each mean. Sample results for July 2010 from Legionowo are illustrated in Figure 1. The graph on the vertical axis shows the tropopause temperature, and on the horizontal axis the numbers of successive averages calculated from incomplete data are marked, as described above. For each mean, its value (large dot) and confidence limits (small dots) are marked. The solid line indicates the monthly average calculated over the entire month. In each case illustrated, the monthly average falls within the confidence interval of the mean of incomplete data. In the entire analysed data set, this situation occurred in 98% of cases. Other cases are situations when the missing data coincides with the warmest or coldest days of the month.
In the second stage of the experiment, five days with the highest or lowest values were rejected. These were not the consecutive days. The next step was as described above. In such cases, only 75% of cases were within the accepted confidence intervals. It was found that in the case of tropopause temperature, 95% of the modelled cases that lacked data resulted in an error of not greater than 0.8 C. This error is smaller than the standard deviation (Table 2). Based on the experiment, it was found that monthly data with deficiencies of up to 5 timely values can be taken and treated as whole, complete data. Possible errors will not be greater than errors caused by other factors or coincidental noise in the distribution of values. However, in a situation where 5 days are missing with the highest or lowest values, the obtained average tropopause temperature will be falsified to a greater extent than the accepted thresholds.
In turn, months with deficiencies over 5 days as well as several whole, individual months were interpolated using the linear correlation method. For more accurate interpolation, the data set was divided into the warm seasons (IV-IX) and the cool seasons (X-III). A separate regression equation and correlation coefficient was determined for each of the seasons, which was used to fill in the missing months. Missing individual (daily) records F I G U R E 1 Impact of missing data on calculated monthly averages. Explanations in the text were not completed. Due to the fact that, in Legionowo there were practically no data deficiencies over 5 days in a month and the correlation coefficients of this station were slightly higher than with the other two in Wrocław and Łeba, it became the reference station for interpolations. Examples of correlation coefficients between Legionowo and Wrocław or Legionowo and Łeba were at the level of 0.98-0.99 for the lower and middle troposphere (850-500 hPa), both for the warm and cool season. In the upper troposphere (300-200 hPa) and tropopause they were slightly lower and amounted to 0.89-0.98.
Deficiencies in measurement data at the tropopause level were almost identical to those on individual isobaric surfaces and represented a small percentage. There were missing data in the whole (e.g., VI-VII.1988-Łeba) or in almost the whole months (e.g., III.1984 andVIII.1988-Łeba;V.1988-Łeba and Legionowo). At the Wrocław station in the period April-December, 1992 there was a longer data discontinuity due to data loss. In turn, individual deficiencies were determined by, for example, failure of the electricity supply or software at a given station. The amount of complete monthly data (without gaps in the daily data/ from 00 UTC) in Wrocław (WR), Legionowo (LG) and Łeba (LB) respectively was: 260 (57%), 326 (71.5%) and 238 (52.2%) out of 456 (100%) possible. However, the distribution of gaps in the data from 0 to 5 records per month was as follows: WR-88.2%, LG-99.8% and LB-90.8% ( Figure 2). Therefore, the measuring material for analysis, taking into account the adopted assumption, in terms of completeness is good.
The temperature of five MISs in the troposphere were analysed: 850, 700, 500, 300 and 200 hPa and the first tropopause. These are among the 16 mandatory surfaces (1,000,925,850,700,500,400,300,250,200,150,100,70,50,30,20 and 10 hPa) in which the World Meteorological Organization recommends measuring meteorological elements in atmospheric soundings (WMO, 1996). The level of 850 hPa, located in moderate latitudes at an altitude of about 1,300-1,500 m, determines the altitude where the daily temperature changes disappear. At the same time, it is the lowest level in the troposphere at which the size of thermal advection can be determined without taking into account the impact of the ground. The average location of the analysed MISs, that is, 850, 700, 500, 300 and 200 hPa for LB, LG and WR stations in the years 1981-2018 was respectively: 1447, 1,463 and 1,471; 2,989, 3,009 and 3, 021; 5,543, 5,569 and 5,588; 90,101, 9,134 and 9,163; 11,706, 11,737 and 11,771 m a.s.l. On the other hand, the average location of the first tropopause for LB, LG and WR was: 10622, 10,743 and 10,899 m a.s.l.
For tropopause temperature and isobaric surfaces, their tendencies of change were determined using simple F I G U R E 2 Distribution of deficiencies in radiosounding data in the tropopause during 1981-2018 regression equations. The statistical significance of the trends determined at the .05 significance level was checked using the t-Student test. A similar approach was used in the study of trends in air temperature in Poland (Tomczyk and Bednorz, 2020).

| Assessment of data homogeneity
In the analysed years 1981-2018, different types of radiosondes were used at Polish aerological stations (Table 1), similarly as across the world. Radiosondes were attached to a latex balloon filled with hydrogen. In the initial period of research, radiosondes from the former USSR (A-22, RKZ-2, RKZ-5 and MARZ) were used. In the first decade of the 21st century, there was a widespread transition to more accurate sensors. The radiosondes RS-92 and RS-92-KL from the Finnish company Vaisala, introduced in Poland since 2004, differed from previous models (RS-80 and RS-90) in housing and electronic solutions. RS-92 radiosondes had factory calibration coefficients stored in them, which in earlier models required manual entry into the system. In 2005, Polish aerological stations were equipped with DigiCora MW21 systems, enabling tracking of radiosondes through GPS navigation, which are still used presently. For this reason, the obtained data from the RS-80, RS-90, RS-92, RS92-KL, RS92-SGPD, RS92-SGP and RS41-SG radiosondes were considered homogeneous. On the other hand, earlier comparison of measurement results from the A-22 or RKZ-5 radiosondes with the reference radiosonde made by the IMGW Aerology Center in Legionowo indicated their high accuracy (Gołaszewska, 1985). The RKZ-2 radiosonde was omitted due to the fact it was used only on individual days in a short period at one of the stations (Table 1).
For example, the random error of calibrating the air temperature measurement with the RS-92 radiosonde over the entire range (from 180 to 310 K), according to the manufacturer's brochure from 2005, is 0.15 K (Vaisala, 2006). The same is true for subsequent radiosonde models, for example, RS-92-SGP, RS-92-KL or RS-41-SG (Vaisala, 2013(Vaisala, , 2017. In addition, each radiosonde is recalibrated prior to take-off by the ground control station (groundcheck). The ground station should be calibrated every two years and certified. This was and is done for example, in the case of Vaisala RS-41 or RS-92 radiosondes and in earlier models of this manufacturer at Polish aerological stations. The same is done for example, at the nearby German aerological station in Lindenberg, operating in the GRUAN network (Global Reference Upper-Air Network) as part of the WMO (Immler et al., 2010;Bodeker et al., 2016).
It should be remembered that all radiosonde systems and types of soundes are approved by the World Meteorological Organization before being allowed to use (WMO, 1996).
Preliminary data analysis and information on changes in the radiosondes used during the research period led to an assessment of the homogeneity of the data sequence used.
For this purpose, the T-Alexandersson test (Standard Normal Homogeneity Test-SNHT) was used to detect changes in the mean value in the time series for variables  .1990-XII.1992IX.1986-XII.1991, III.1992-VI.1993II.1987-XII.1991RS-80 I.1993-VI.1999I.1992-II.1992VII.1993VII. -V.1999VII. I.1992 , 1997). This test was based on differences for corresponding observations. In order to confirm the heterogeneity of the series, when it was initially detected, additional modifications of the test were applied while taking into account observations from the Lindenberg reference station. Reference data series from Lindenberg were obtained from the Integrated Global Radiosonde Archive-IGRA (NOAA/National Centers for Environmental Information. Integrated Global Radiosonde Archive NOAA. IGRA, 2006). The IGRA is the most comprehensive and largest radiosonde data set. It combines different data sources and applies quality control algorithms to remove gross errors (Durre et al., 2006). Its resources are suitable for use in climatological studies (Seidel and Randel, 2006;Añel et al., 2007). Lindenberg's aerological station is closest to the Polish stations (approx. 300-350 km), located in the same geographical and climatic zone.
The tests performed for data from Wrocław showed heterogeneity of observation sequences for the MISs of 200 and 300 hPa, identifying the year 1991 as the moment of change, thus dividing the data sequence into the following two: 1981-1990 and 1991-2018. For further studies, a series of homogeneous observations was taken from 1991-2018, supplemented by corrected observations from the Lindenberg station for the period 1981-1990. It should be noted here that 1990 was the last year the A-22 radiosonde was used (Table 1).
To estimate the temperature values from 1981 to 1990 for Wrocław, a large linear relationship (Pearson's correlation coefficient r is .72 Ä .96, alpha = .05) was used for data from the Wrocław and Lindenberg stations from the period 1991-2018. The prediction was made with the use of linear equations in the form y = ax + b for all analyzed MISs (also for 500, 700 and 850 hPa) and the tropopause. The re-applied Alexandersson tests did not indicate the existence of a jump in the mean value of the series tested.
The T-Alexandersson test was also applied to the data for the Łeba and Legionowo stations. In the case of calculations for data from the Łeba station, all tests showed no grounds to reject the hypothesis of homogeneity of the sample at the significance level of alpha = .01. On the other hand, at the significance level of alpha = .05 in 13 out of 15 tests it indicated homogeneity of the tested data (no grounds to reject the hypothesis of homogeneity). In two cases (VI-VIII, 850 hPa and I-XII, 700 hPa), analysis of the changes in value of the Alexandersson T-test statistics, as well as the division into two periods for which the squares of mean values are calculated, indicates that there is no moment that clearly divides the sample into two series. Higher values of T statistics result more from changes in the variance and test structure (high sensitivity of the test on the value of variance and length of the series) than from the heterogeneity of the sample.
In the case of data from Legionowo station, no years have been shown for which the division into homogeneous periods would be advisable.
Calculations related to the homogeneity of the series were made on the basis of their own procedures in the "R" programming language.

| Average tropospheric temperature and its variability
The average monthly and yearly tropospheric temperature values decrease as the height of the MISs increases. For MISs from 850 to 300 hPa the lowest values are in February. At 200 hPa and in the tropopause they occur in January. The highest, respectively (except 500 and 300 hPa for LB), in August, or for 200 hPa and the tropopause, in July ( Table 2). The average annual MIS temperature values range from about: 2.7-1.6 C for 850 hPa, through −5.2 C to −6.4 C for 700 hPa, −21.1 to −22.2 C for 500 hPa, −47.0 to −47.3 C for 300 hPa and −56.0 to −56.7 C for 200 hPa. However, the average annual tropopause temperature over aerological stations in Poland is in the range −58.5 C in LB to −59.1 C in WR ( Legionowo and other European stations located in similar latitudes (Pyka, 1990). In addition, the results obtained indicate that on the surface of 200 hPa and in the tropopause, the average annual and summer quarter temperature values are the lowest above the WR station, and the highest above LB. So the temperature in the upper troposphere decreases as it moves away from large bodies of water (Atlantic Ocean, Baltic Sea). During the winter quarter, there is practically no thermal differentiation of the upper troposphere and tropopause between stationsaverage differences do not exceed 0.3 C (Table 2). This may be due to the fact that in the cool seasons of the year in south-western Poland, advection of air masses from southern and western directions dominates. Advection from these directions then bring in an inflow of warm air (Urban and Kowalewski, 2019). Average monthly temperatures in the upper troposphere (200 hPa) and tropopause are highest above LB throughout the entire period between April and November ( Table 2).
Analysis of the average temperature of the MISs of the troposphere over Poland in the years 1981-2018 shows that they form cooling surfaces from south to north, that is, from Wrocław (WR), through Legionowo (LG) to Łeba (LB). This regularity occurs up to and including 300 hPa in both annual and seasonal average values (except in winter in LG and LB for 300 hPa). In monthly averages this is noticeable in the middle (500 hPa) and lower (850, 700 hPa) troposphere. However, only between May and November is a surface of 300 hPa indicated. In the December-March period, the upper troposphere at 300 hPa is slightly cooler above the LG station than above the other stations (Table 2). This is most likely conditional upon decreasing and increasing altitude, ground effect and range of advection temperature changes. Thermal changes occurring at the Earth's surface are particularly dependent on changes in the temperature field in the lower and middle troposphere, especially on the so-called advection level, corresponding to the isobaric surface of 850 hPa (Thorncroft and Hoskins, 1990;Wibig et al., 2009a;2009b;Tomczyk and Bednorz, 2016;.
The 850 hPa level, which is at moderate latitudes at an altitude of about 1,300-1,500 m, determines the altitude where the daily temperature changes disappear. At the same time, it is the lowest level in the troposphere at which the size of thermal advection can be determined without taking into account the impact of the ground. For these reasons, it is also the basic baric level taken into account during the air temperature forecast.
The average annual temperature amplitude of the MISs varies in the troposphere. For annual average values, the variability of troposphere temperature in the analyzed stations ranges from 1.1 to 1.2 C in the lower part to 0.6-0.7 C in the upper parts. Among the MISs,  Table 2). The variability of the average monthly, seasonal and annual temperature values of the MISs of the troposphere, expressed in terms of the standard deviation, occurs both in space and in time. The variability is highest during the cool months of the year, respectively, and decreases accordingly as the period increases, that is, for the season and year. Standard deviation values are similarly the lowest for summer months, summer quarter and year. The maximum values of the standard deviation fall in February and are about twice the minimum that occur in June or August. For example, for the surface of 700 hPa, the monthly average is about 2.4 and 1.2 C, respectively (Table 2). This annual standard deviation is typical for the lower and middle troposphere (850, 700 and 500 hPa). The thermal changes for the upper troposphere (300 and 200 hPa) are slightly different than for the lower or middle tropics, due to the changes in thickness of the troposphere itself. Here, the dispersion of the smallest and largest temperature standard deviation is greater. For example, for the MIS of 300 hPa, the smallest value of the standard deviation may occur in May or even in April and the maximum value in autumn or at the end of winter. Nevertheless, the annual temperature variability in the upper troposphere is less than in the lower ( Table 2) variability of tropospheric thermals in Europe were obtained by Pyka (1990).

| Annual and seasonal tropospheric thermal course and their changes in trend
The course of the annual MIS temperature values shows changes from year to year. In the 1980s and 1990s, there was a decrease in the annual troposphere temperature compared to the average for the analyzed multi-year period. It is noticeable for MISs from 850 to 300 hPa. The direction and magnitude of the temperature trends in the lower and middle troposphere in the years 1981-2018 are determined primarily by the temperature course from the beginning of the 21st century (Table 3, Figures 3a,b, 4a,b  and 5a,b). The second analyzed (incomplete) decade of the 21st century is the warmest in the entire period studied, which is noticeable up to and including 300 hPa exclusively (Table 3, Figures 3a,b, 4a,b and 5a,b). In the lower (850 and 700 hPa) and middle (500 hPa) troposphere, throughout the study period, a gradual increase in annual and decade temperatures over Poland is observed (Figure 3, Table 3). Its maximum values occurred in 2014 for the MIS of 850 hPa and amounted to 4.7 C over WR, 4.0 C over LG and 3.3 C over LB. For isobaric surfaces 750 and 500 hPa, they occurred in 2015 (except LB, when Tmax at 500 hPa occurred in 2018). Secondary maxima in annual values in the lower and middle troposphere occurred in 1989. However, the lowest annual temperatures in the lower and middle troposphere over Poland occurred in the mid-1980s, with a minimum in 1985 for MISs in the range of 850-500 hPa (Figure 3a).
Reliable instrumental measurements of the Earth's surface temperature on a large scale have been carried out for over a century. They document a warming of about 0.86 C since 1880, with the warmest last three decades (IPCC, 2013). In perspective of global average, 2015 was the warmest year in terms of average surface temperature (Tollefson, 2016). Most data indicate that the years 2014, 2015 and 2016 have set new global heat records since the beginning of regular meteorological measurements (Rahmstorf et al., 2017). The current decade appears promising as the warmest in the last century and the years 2014-2015, especially 2015, were until recently considered to be the warmest in the history of instrumental observations in Poland. In 2018, in the majority of Poland, the highest average annual air temperature was recorded in the years 1966-2018 (Tomczyk and Bednorz, 2020). The year 2019, in terms of area average temperature for Poland, was warmer by as much as 0.4 C than the last highest average from 2018. For example, the average annual air temperature in Wrocław in 2019 was 11.4 C and was higher by 0.2 C than the last record-breaking amount measured in 2018 as determined from instrumental measurements conducted since 1781. In addition, after 2010 in Europe there was not a single negative extensive thermal anomaly and the area of Europe most often affected by thermal anomalies is its central part, in which Poland is located (Kossowska-Cezak and Twardosz, 2019). In turn, according to NOAA bases (2019), in the years 1880-2018 on a global scale, 9 out of 10 the warmest years are those after 2000, with a maximum in 2016 (average anomaly +0.95 C) and 2018 has been classified in the fourth position, with average anomalies at 0.79 C.
In contrast, in the upper troposphere (200 hPa) as well as in the tropopause, a decrease in temperature is noted, especially noticeable in summer and for the year (Figures 3b and 4b, Table 3). The observed drop in temperature in the upper troposphere over Poland in the  (Pyka, 1990;Santer et al., 2005;Duan, 2007;Quanliang and Lingxiao, 2012;Seidel et al., 2016).
The lowest annual values of temperatures for 200 hPa in 1981-2018 occurred in 1983 in LG and LB and amounted to approx. −57.5 C. In WR, the lowest annual value of −57.9 C was found on this surface in 2003. In 2003, the lowest tropopause temperature over WR and LB was also measured: −60.3 and −59.9 C, respectively. It occurred in LG in 1983 and was −59.9 C (Figure 3b). In turn, the highest annual temperature values on this surface and in the tropopause were measured in 1988 over LB and LG. They were −54.8 C (LB) and −55.2 C (LG) at 200 hPa and −56.2 C (LB) and −56.7 C (LG), respectively. Above the WR station, the highest annual value of −55.3 C per 200 hPa was in 2010 and in the tropopause in 1981, when it was −56.9 C (Figure 3b).
Public interest in the increased concentration of greenhouse gases in the atmosphere, mainly H 2 O and CO 2 , has also increased interest in changes in the atmosphere temperature and their consequences. Most of the research, based on data from ground stations in Poland or Europe, indicates a positive air temperature trend, particularly pronounced in the 20th century, which is still rising (Trepi nska, 1971;Pyka, 1983;Brázdil et al., 1996;Wibig and Głowicki, 2002;Moberg et al., 2006;IPCC, 2013;Migała et al., 2016;Urban and Tomczy nski, 2017). summer quarter, the value of the trend is even higher and amounts to about 0.6 C, for the winter quarter 0.17 C. Trends for year and summer are statistically significant at both stations (Table 4). In addition, their values are close to the corresponding trend values on the 850 hPa isobaric surface (Table 5). The results of ground measurements were confirmed by MIS temperature trends determined from radiosounding measurements from 1981 to 2018. Namely, for MISs of 850-300 hPa for the year and summer, the trends are positive and almost everywhere statistically significant at a significance level of .05. They average for the year from about 0.2 to 0.5 C/decade. Pearson's linear correlation coefficients for the year and summer in the lower and middle troposphere are moderate and strong and are in the range of 0.4-0.7 (Table 5, Figure 3a,b). Similar results of Philipona et al. (2018) for the northern hemisphere in the 30 -60 latitude zone, where the average spatial temperature in the troposphere (determined from data from 29 sounding stations, for example, For the summer quarter, the average thermal trends in the lower and middle troposphere are slightly higher than for the year and range from 0.4 C to 0.5 C/decade (Table 5, Figure 4a). On the other hand, on the MIS of 200 hPa over the whole of Poland, negative tendencies in thermals are observed, both during the year and in the T A B L E 5 The magnitude of change trends, Pearson's linear correlation coefficient (r) and the statistical significance of the isobaric surfaces temperature and the tropopause during 1981-2018 for the year, summer and winter  Figures 3b and 4b).
In the period 1958-2001, the global temperature in the middle troposphere clearly increased. In addition, this change is greater in summer than in winter (Quanliang and Lingxiao, 2012), which is also confirmed by the results of studies of tropospheric thermals obtained over Poland in the years 1981-2018.
In the winter quarter, trends in the troposphere temperature are much smaller than for summer or the year. In the lower troposphere they are weak and positive, while in the upper troposphere they are weak and negative. They are not statistically significant at the significance level of .5 (except 300 hPa and the tropopause for WR). In addition, the values of Pearson's linear correlation coefficients for winter on almost all MISs of the troposphere indicate absent or weak correlation, only locally (Wrocław for 300 hPa and the tropopause) for its moderate strength (Table 5, Figure 5a,b).
Based on data from 109 sounding stations during 1980-2004, a positive (average 0.2 C/decade, that is, comparable to the one in this study on Poland) temperature trend was found over China in the lower and middle troposphere (from 1,000 to 400 hPa) and negative (average −0.6 C/decade) in the upper troposphere and lower stratosphere (200-20 hPa). Due to the morphology of the terrain and the seasonally variable atmosphere circulation, trends vary regionally (Duan, 2007).
Climate change models, including the WACCM (Whole Atmosphere Community Climate Model) chemical-climate model, as well as the results of satellite measurements, show the increase in the troposphere temperature during the years 1978-2014 in monthly values throughout the year in low and medium latitudes, with a relative maximum in the northern hemisphere ($30 -70 N) from April to October (Randel et al., 2017). The results of sounding measurements obtained in this article, for almost identical years, also indicate the largest increase in the troposphere temperature over Poland during the summer (Table 5, Figure 4a).
Negative tendencies during the years 1981-2018 are found in tropopause thermals in the analyzed years over all Polish aerological stations. The rate of tropopause temperature drop for the year, and summer and winter ranges from −0.2 to −0.3 C/decade. However, only for WR and ŁB is the annual trend, and the winter trend for LG, statistically significant at the significance level of .05. The downwards trend is associated with an increase in tropopause location and is strongest in summer, when it is as low as −0.3 C/decade for LG (Table 5). It should also be borne in mind that the heating of the Earth's surface and the increase in the concentration of greenhouse gases in the atmosphere cause the tropopause to move upwards. This effect is mainly influenced by the warmer troposphere, which increases in volume as a result of temperature increase, which leads to an increase in location and thus a decrease in tropopause temperature (Lin et al., 2017). The obtained results for the tropopause temperature over Poland refer to contemporary research and confirm the cooling of the tropopause region (Zhou et al., 2001;Wang et al., 2012).

| SUMMARY AND CONCLUSIONS
The analysis of the tropospheric thermal field variability over Poland based on radiosounding data from 1981 to 2018 authorizes the following statements: 1. The average annual temperature amplitude of the MISs varies in the troposphere. For annual average values, the troposphere temperature variability over Poland ranges from 1.1 to 1.2 C in the lower part to 0.6 to 0.7 C in the upper parts. Among the MISs, the 300 hPa surface seems to be the most stable in this respect, where the standard deviation for the year is about 1.3 C on average. 2. The variability of the troposphere temperature depends on the length of the time step taken. It is inversely proportional to it. Variability, expressed in terms of the standard deviation value, is highest in the months of the cool part of the year and decreases accordingly with the extension of the period, that is, for the season and year. Similarly, the smallest tropospheric temperature variability is characteristic for the summer months, summer quarter and year. The maximum values of the standard deviation fall in February and are about twice the minimum that occur in June or August. This annual standard deviation is typical for the lower and middle troposphere (850, 700 and 500 hPa). The changes in thermals for the upper troposphere (300 and 200 hPa) are slightly different than in the lower or middle one, at least in view of the changes in the thickness of the troposphere itself and the extent of the tropopause. Here, the occurrence of dispersion of the smallest and largest temperature standard deviation is greater. Annual temperature variability in the upper troposphere is less than in the lower. 3. Temperature trends for MISs from 850 to 300 hPa for the year and summer are positive and almost everywhere statistically significant at the significance level of .05. They are on average for a year from about 0.2 to 0.5 C/decade. In summer, they are slightly larger and are 0.4 C-0.5 C. On the MIS of 200 hPa, negative thermal trends can be found over the whole of Poland, both during the year and in summer. They are around −0.1 C/decade. 4. Trends in the winter quarter are much smaller than for the summer or year. In the lower troposphere they are weak and positive, while in the upper one they are weak and negative. They are not statistically significant at the significance level of .5. 5. There are negative trends in tropopause thermals over all Polish aerological stations. The rate of tropopause temperature drop for the year, and summer and winter is in the range of −0.2 C Ä −0,3 C/decade. The downwards trend is associated with an increase in tropopause location and is strongest in summer, when it is as high as −0.3 C/decade for LG. 6. The obtained results regarding the temperature of the troposphere and tropopause over Poland correspond with the results of previous studies from Europe and the world. They confirm the warming of the lower troposphere and the cooling of the upper troposphere along with the tropopause that have taken place in the last few decades.