Temperature and urban heat island effect in Lublin city in Poland under changing climate

This study investigates the characteristics of the temperature regimes at an urban station (Litewski square) in Lublin city in Poland and a nearby rural station (Radawiec), and the Urban Heat Island (UHI) effect in Lublin city. In winter, spring, summer, and autumn at both urban and rural stations frequency distributions of daily minimum (Tmin), and maximum (Tmax) air temperature in 1998–2020 have shifted towards a warmer climate compared to the frequency distributions in 1974–1997. At both stations in 1974–2020, in all seasons, the annual Tmin and Tmax display increasing trends. At Litewski square and Radawiec, Tmax shows increasing trends of 0.083 and 0.088 ºC/year in summer, respectively. This is the largest increase in all four seasons. Furthermore, it is revealed that the heatwaves at both the urban and rural stations have increased in number over time. However, cold waves at both stations show a declining trend. The UHI effect in Lublin city has not increased significantly during 1974–2020. Population in Lublin city has declined over the period 1995–2020, but the population in the surrounding rural counties has increased. It is speculated that this is one of the causes of no clear increase in the UHI intensity. Apart from that, the city’s large green coverage (about 40%) is probably acting as a heating inhibitor. The annual Tmin and Tmax projected by 15 Coupled Model Intercomparison Project Phase 6 (CMIP6) GCMs indicate that the temperature regimes at both urban and rural stations show significant increasing trends during 2015–2100 under the selected SSPs, with the highest increase under high emission scenario (SSP5-8.5) and the lowest increase under the low emission scenario (SSP1-2.6). During 2015–2100, the UHI effect in Lublin city does not show any significant increasing or decreasing trends for the majority of the GCM–SSP combinations.


Introduction
The world population is exponentially increasing since the industrial revolution (Toth and Szigeti 2016). The population growth and migration of people from rural areas to urban areas have caused a rapid expansion of existing cities and the birth of new cities. In Europe, the percentage of the urban population has increased since the 1950s from about 52 to 75% by 2018, and it is expected to increase up to 85% by 2050 (World Urbanization Prospects 2018). The expansion of cities has led to poor urban ventilation as densely built buildings hinder air circulation (Suder and Szymanowski 2014). Green spaces in cities are being converted into builtup areas to accommodate the rising population (Nor et al. 2017). This has consequently altered the moisture and surface energy balance in urban areas leading to an accumulation of heat thwarting the natural cooling effect of green spaces. Cities are also known for their anthropogenic heat generation (Rizwan et al. 2008) due to processes such as air conditioning and internal combustion engines used in vehicles and industries (Sailor 2011). According to Zheng and Weng (2018), energy consumption associated with space heating is the main contributor to anthropogenic heat flux in cities, such as Los Angeles in the mid-latitudes. Also, urban building materials such as concrete, asphalt, and steel absorb solar radiation and release heat back into the urban environment (sensible heat flux) (Radhi et al., 2014). Urban building materials (e.g., asphalt, concrete) have a large thermal admittance (Christen et al. 2012). Furthermore, in urban centers there are large areas available for heat exchange and large volumes where heat can be stored (Ruth 2006;Mohajerani et al. 2017). Above factors results in more effective heat storage in urban areas. This contributes to the relatively elevated temperatures in urban environments compared to surrounding rural environments. This phenomenon where the temperature in urban areas is relatively higher than that in the nearby rural areas is called the urban heat island (UHI) effect (Oke 1987). The intensity of the urban heat island is usually expressed as the difference between the urban and rural temperatures. The space which is warmer than the nearby rural area and surrounds the urban area is called the urban heat island (Mitchell 1961;Oke 1973;Arnfield 2003). Milojevic et al. (2016) found that there is a small but statistically significant increase in mortality due to the UHI effect. Macintyre et al. (2021) studied the contribution of UHI in protecting people against cold-related deaths in winter in the UK. They found that due to the UHI effect in winter, about 15% cold-related deaths are avoided. Thus, analysis of UHI is an important task.
UHI effect can be classified under two main categories; (1) AUHI (atmospheric urban heat island) and (2) SUHI (surface urban heat island) effect (United States Environmental Protection Agency, 2013). AUHI effect is described as the raised air temperature in urban areas relative to that of nearby rural areas, whereas the SUHI effect is described as the raised land surface (or skin) temperature (LST) in urban areas relative to that of nearby rural areas (Gawuc et al. 2020). In the study of the AUHI effect, the air temperature data obtained from weather stations located in urban and rural areas are used (Soltani and Sharifi, 2017), whereas, in the study of the SUHI effect, the surface (or skin) temperature data obtained from satellite remote sensing techniques are used (Tran et al. 2006). In the current study UHI refers to the atmospheric urban heat island effect. Adverse impacts of the UHI include; a decline in thermal comfort, an increase in summer energy demand (air conditioning costs), air pollution and greenhouse gas emissions (more fossil fuel burnt to meet energy demand), heat-related illnesses and mortalities, and degradation of water quality (increase in water temperature affects some aquatic species) (United States Environmental Protection Agency, 2021).
The UHI effect can be considered as the most significant phenomenon in the urban areas of Poland (Kłysik and Fortuniak 1999;Gawuc et al. 2020). In Poland, the first study on the UHI was conducted and presented by Gorczyński and Kosińska (1916). They stated that the average monthly temperatures in the center of Warsaw (the largest city in terms of population in Poland) in summer and winter are about 1.5 °C and 0.5 °C, respectively, higher than those observed outside the city. Differences in thermal conditions in the city and the outside regions were also studied by Gumiński (1930), but using data from Munich, Germany. The thermal conditions (including UHI) of large cities, such as Warsaw and Lodz became the subject of numerous studies. Błażejczyk (2002) found that both the horizontal and vertical structures of Warsaw city influence its climatic and bioclimatic conditions. According to Błażejczyk et al. (2014), in Warsaw UHI effect is most pronounced in autumn (higher than 2.5 °C) and least pronounced in winter (less than 1.5 °C). In Lodz (the 2nd largest city in Poland in terms of population) high building density, artificial surfaces, and narrow urban canyons cause a strong UHI effect and the UHI effect can be as intense as 12 °C at times in the center of Lodz (Kłysik and Fortuniak, 1999). The UHI in Warsaw shows seasonal and daily fluctuations, and the largest intensities are usually observed in the evening and at night (Stopa-Boryczka et al., 2001). Gawuc et al. (2020) studied the UHI effect in Warsaw using air temperature and land surface temperature obtained from satellite data and found that the AUHI effect is more intense than the SUHI. Fortuniak et al. (2006) studied climate data from Lodz city and its rural area (one urban and one rural station) considering the period 1997-2002. They concluded that urban-rural thermal contrast (UHI intensity) begins to increase in the late afternoon and continues till midnight and then peaks between midnight and sunrise before declining with the sunrise. In Lodz city, in winter that there are nights with exceptionally strong episodes of UHI, as intense as 7-8 oC (Fortuniak et al. 2006).
In a study by Szymanowski (2005) in Wrocław, Poland, it was found that the temperature contrasts between different areas of the city can be as high as 5-6 K. Thus, for the study of the UHI effect, the use of data from a large number of stations is highly recommended. However, Lublin city in Poland lacks long-term records of temperature data corresponding to a large number of stations. This is true for many cities in Poland. The influence of topography on the diversity of the city's thermal field in Krakow led to the development of the concept "relief-modified UHI effect" (Bokwa 2010;Bokwa et al., 2015). Bokwa et al. (2018) studied the impact of fog on UHI intensity in Kraków, Poland, and found that fog can decrease the UHI effect by about 1 K under specific weather conditions. Półrolniczak et al. (2017) analyzed the UHI effect in Poznań, Poland, and concluded that the UHI effect is more pronounced, particularly in the evening and nighttime. Furthermore, they concluded that the UHI effect is more frequent than the urban cool (UCI) island effect.
At the beginning of the twenty-first century, researchers in Poland started using modern Geographic Information System (GIS) techniques, which enabled the modeling of Wrocław city's climate with a high degree of accuracy (e.g., Dubicka and Szymanowski 2000;Szymanowski 2004). Szymanowski and Kryza (2012) employed geographically weighted regression (GWR) and multi-linear regression (MLR) to create the spatial structure of the UHI in the city of Wrocław in Poland. They found that GWR is a better technique compared to MLR for constructing the spatial structure of UHI. For studying the structure of the boundary layer over the city of Wrocław, meteorological measurements were obtained using; balloons (Drzeniecka et al. 2003), meteorological towers (Drzeniecka-Osiadacz et al. 2018), and meteorological drones (Szymanowski et al. in. 2019). The data from the above sources allows the creation of a database of numerical parameterizations of the urban boundary layer (Netzel et al. 2012;Kryza et al. 2015), which is much needed for the study of the UHI effect. Recently, in Poland, the use of remotely sensed surface temperature data for the study of SUHI is seen in the literature. Majkowska et al. (2017) studied the SUHI in Poznań in Poland by creating maps of SUHI using remotely sensed LST and air temperature from 9 stations corresponding to the period June 2008 to May 2013.
After World War II, Gumiński (1950) andPaszyński (1957) presented the first few investigations on the UHI effect in Lublin in Poland. Filipiuk et al. (1998), Bartoszek et al. (2014), and Bartoszek and Węgrzyn (2016) used the data from the stations at Litewski square and Felin in Lublin city, and Radawiec and Czesławice in the nearby rural area for analyzing UHI effect in Lublin city. In those studies, it was found that the minimum and maximum temperature is higher in the center of Lublin city in comparison to the nearby rural area. Kaszewski and Siwek (1998) also reported that the center of Lublin (refer to as Litewski square) is warmer than Hajdów, which is located outside the city. This was further confirmed by the fact that the snow cover lasts longer and it is thicker in the western rural region than in Lublin city (Bilik and Nowosad 2000;Nowosad and Bartoszek 2007). Kaszewski (2019) provides a detailed account of Lublin's climate, while Fortuniak (2019) described the history of urban climate and UHI research in Poland.
The aim of this study is to analyze the characteristics of temperature (including heat and cold waves) and the UHI effect in Lublin city in Poland using daily maximum and minimum air temperature (T min and T max , respectively). This is done using data obtained from the past observation archives and the future projections produced by the latest Coupled Model Intercomparison Project Phase 6 (CMIP6) General Circulation Models (GCMs). So far, no study has presented an investigation of characteristics of temperature and UHI effect in Lublin city in Poland using the latest set of CMIP6 GCMs. The study of temperature and the UHI effect can provide important information on its potential impacts on thermal comfort, energy demand, heat-related illnesses, and mortalities.

Study area and data
Lublin city is the largest city in Eastern Poland. In terms of population, Lublin city is the 9th largest city in the country. The administrative area of Lublin is around 148 km 2 and it lies between the longitudes 22°27′-22°41′ E and latitudes 51°08′-51°18′ N (Rysiak and Czarnecka 2018). It is home to about 338.5 thousand residents (Statistics Poland, 2021). Lublin city is located in the northern region of the Lublin Upland. The city is divided into two regions by the Bystrzyca River. The left bank of the river contains deep valleys and old loess gorges, while the right bank is a part of the Świdnicki plateau (Kondracki 2002). Bystrzyca is the main river in Lublin. It is 70 km in length, and 22 km of its length flows through the city. It shows significant seasonal fluctuations in water level, with an average flow rate of 3 m 3 /s at the "Lublin" water gauge. Lublin's highest point (approximately 236 m AMSL) is located in the vicinity of Węglin Park, and the lowest point (approximately 164 m AMSL) is located in the Bystrzyca Valley near Hajdów. Approximately 40% of the city's area is green, which includes forests, parks, lawns, and housing estate greeneries (Kuśmierz et al. 2018).
In this investigation, for studying temperature regimes and the UHI effect, daily T min and T max data corresponding to the urban station at Litewski square in Lublin city and the rural station at Radawiec (approximately 11 km away from Litewski square) for the period 1974-2020 were used. Data for the Litewski square station were obtained from the climate data archive of the Maria Curie-Sklodowska University, while data for the Radawiec station were obtained from the Polish Institute of Meteorology and Water Management-National Research Institute (IMGW-PIB) (https:// www. imgw. pl). The station in Litewski square is the only meteorological station located in the city center with a long data record over the period 1974-2020. The station at Radawiec is the only nearby rural station with a record of data over the period 1974-2020. Półrolniczak et al. (2017) also studied the UHI effect in the city of Poznań, Poland, using data from two observation stations (one station in the city center and the other at the Ławica airport). In addition to Litewski square, data from 4 more urban stations located within Lublin city were also used to study the spatial variations in temperature within the city. These stations were: Dominikanie, Zarnowiecka, Zywnego, and Ofelii. These stations had data sets only spanning over the 2-year period 01-Jan-2009-31-Dec-2010. The 2-year data sets were available at a half-hourly temporal resolution (30-min data sets). It should be noted that the precision of all temperature data is up to one decimal place.
For the study of future likely temperature and UHI effect, daily T min and T max data corresponding to four Shared Socioeconomic Pathways (SSPs): SSP1-2.6, SSP2-4.5, SSP3-7.0, and SSP5-8.5 from 15 CMIP6 General Circulation Models (GCMs) for the period 2015-2100 were obtained from the World Climate Research Programme (WCRP) (https:// esgf-node. llnl. gov/ proje cts/ cmip6/). GCMs represent physical processes of the atmosphere, ocean, cryosphere, and land surface mathematically. CMIP6 presents the latest set of GCMs, which has simulated the past climate and projected the likely climate into the future. These 15 GCMs were selected considering the availability of their data corresponding to the above-mentioned four SSPs. SSPs represent the likely scenarios of projected socioeconomic changes in the future world. SSPs are used to estimate Greenhouse Gas (GHG) emissions in the future world with different socioeconomic conditions and climate policies. SSP1 to SSP5 characterize green to highly fossil fuel intense societies (alternative futures of the world). In other words, SSP1 to SSP5 refers to relatively low GHG emissions to relatively high GHG emissions under specific socioeconomic conditions. Eyring et al. (2016) andO'Neill et al. (2016a, b) provide details about the CMIP6 GCMs and SSPs, respectively. Figure 1 shows the location of the two primary temperature observation stations (Litewski square and Radawiec) and the other 4 stations (Dominikanie, Zarnowiecka, Zywnego, and Ofelii) with Lublin city boundary on a map. Table 1 presents the details of these observation stations and surroundings. Table 2 shows the details of the 15 CMIP6 GCMs used in this study.

Analysis of frequency distributions of historical T min and T max
In this section, the frequency distributions of daily T min and T max (bin size = 0.5 ºC) are studied corresponding to the periods: 1974-1997 (1st half of the data series which covers 24 years) and 1998-2020 (2nd half of the data series which covers 23 years). This comparison of frequency distributions of T min and T max enables the understanding of changes in the frequency distributions in recent years compared to the past years. towards a warmer climate is not as pronounced as in the case of summer. In any given season, in general, changes in the frequency distributions of urban and rural stations look quite similar in appearance. However, the frequency distributions of the urban station signified a warmer temperature regime (relative to the frequency distributions of the rural station). Sachindra and Nowosad (2021) investigated the characteristics of temperature across Poland using hourly observations over the period 1994-2019. They also found that there is a Poland-wide increase in the temperature, and particularly in spring and summer, the rate of warming was higher. Thus, it can be suspected   1974-1997 and 1998-2020) that global warming is the cause of the warming seen at the urban and rural stations.

Analysis of annual T min , T max , and DTR
In this section, the annual T min , T max , and DTR (annual average calculated using observed daily T min and T max ) are studied corresponding to the period 1974-2020. The diurnal temperature range (DTR), which is the difference between the daily T max and T min , is an indicator of the changes in the diurnal temperature cycle. DTR changes are governed by the changes in T max and T min . The analysis of annual T min , T max , and DTR enables the study of overall changes in the temperature regime over time. In addition to that, Sen's slope (SS) (Sen, 1968) of annual T min , T max , and DTR is also calculated for the period 1974-2020. SS is defined as the median of pairwise slopes thus it is less affected by the outliers in data, unlike the least square estimator (Ullah et al., 2019a, b). Therefore, SS is considered a robust estimator of trends in climate data (Rahman and Dawood 2017;Ullah et al. 2018). The magnitude of SS refers to the magnitude of the trend and the sign of SS refers to the nature of the trend (i.e., increasing or decreasing). The larger the value of SS, the larger the trend in the climate variable. Positive values of SS refer to increasing trends whereas negative values of SS refer to decreasing trends (in this study only increasing trends in observed T min , T max , and DTR were seen). The presence of statistically significant (p < 0.05) monotonic increasing trends in annual T min , T max , and DTR is determined using the Mann-Kendall (MK) test. Figures 4, 5, and 6 show the variations in the annual T min , T max , and DTR calculated for the urban and rural stations for each year in the period 1974-2020. In these figures, SS refers to the Sens's slope which is given in ºC/year. The black line in the figures refers to the 3-year running average. It is seen that in all four seasons the annual T min and T max have shown an increasing trend at both stations. The increasing trend in annual T min and T max is largest in summer at both stations. Also, the increasing trend in annual T max is smallest in winter at both stations (also not statistically significant). It is seen that at both urban and rural stations in any given season, the increasing trend in annual T min and T max is similar. For example, in Fig. 5, in summer, for annual T max , at the urban and rural stations, SS is 0.083 ºC/year and 0.088 ºC/year, respectively. Although the overall trend in annual T min and T max is an increasing one, the temperature shows an oscillating pattern, which is clear in the 3-year running average. This pattern is very similar at both stations in a given season. Sachindra and Nowosad Fig. 4 Annual T min at urban (Litewski square in blue) and rural (Radawiec in red) stations during period 1974-2020 (Black line refers to the 3-year running average, SS refers to Sens's slope, if the trend is statistically significant; "significant = yes" otherwise "significant = no")

Fig. 5
Annual T max at urban (Litewski square in blue) and rural (Radawiec in red) stations during period 1974-2020 (Black line refers to the 3-year running average, SS refers to Sens's slope, if the trend is statistically significant; "significant = yes" otherwise "significant = no")

Fig. 6
Annual DTR at urban (Litewski square in blue) and rural (Radawiec in red) stations during period 1974-2020 (black line refers to the 3-year running average, SS refers to Sens's slope, if the trend is statistically significant; "significant = yes" otherwise "significant = no") (2021) showed that there are statistically significant increasing trends in temperature in the period 1994-2019 in Poland in a study based on hourly temperature data obtained from 38 stations. At both stations, in winter, T min shows a higher rate of increase compared to T max whereas, in summer, T max shows a higher rate of increase compared to T min . According to Fig. 6, DTR shows very small statistically insignificant trends at the urban station in all seasons. However, at the rural station in spring and summer, small but statistically significant increasing trends are seen. In summer and spring, at the rural station, DTR is relatively higher. Many studies around the world have reported that T min is increasing more rapidly compared to T max (Easterling et al. 1997;Vose et al. 2005;Ullah et al. 2019a, b). However, some studies have found that T max is increasing more rapidly in certain regions (Kumar et al. 1994;Kothawale and Kumar 2005). T max occurs during the daytime as there is a surplus of incoming short-wave radiation and T min occurs at night-time because of the cooling caused by longwave radiation (Stjern et al. 2020). Since solar radiation is available only during the daytime, it affects T max more than T min , and at nighttime thermal radiation governs T min (Wild et al. 2007). Figure 7 shows the CORINE land use in Lublin city and the surrounding rural area corresponding to the years 2000, 2006, 2012, and 2018. It is understood that the land use around the stations considered in this study has not changed much over time. Also, around the rural station at Radawiec land use has not shown any significant changes. However, the discontinuous urban fabric and rail and road network, and other land use categories immediately out of the city boundary have expanded over time (the red and burgundy areas). It is worth noting that the railway network has not expanded much in the study period. It can be said that land use changes may have had little to no impact on the temperature regimes of the stations in Lublin city. However, the increase in urbanization in the rural area may have contributed to some increase in temperature at Radawiec (see Fig. 5). This is reflected through the small linear trends in UHI intensity. Also, at both Litewski square and Radawiec, temperature regimes show mostly similar rates of increase. Thus, it is concluded that the temperature rise at the Litewski square and Radawiec is due to global warming rather than changes in land use.

Analysis of heat and cold waves
In this analysis, the number of heat waves and cold waves that lasted longer than a given number of days is determined for both urban (Litewski square) and rural (Radawiec) stations. A heat wave is defined as a period equal to or longer than 3 consecutive days with daily T max above the 90th percentile of observations of daily maximum temperature (Perkins and Alexander 2013). A cold wave is defined as a period equal to or longer than 3 consecutive days with daily T min below the 10th percentile of observations of daily minimum temperature. Numerous criteria have been used to define heatwaves (Xu et al. 2016). For example, daily T max ≥ 35 over 2 or more days (Tong et al. 2010), daily T max ≥ 95th percentile over 3 or more days (Xu et al. 2013), and daily T max ≥ 30 over 3 or more days (Hutter et al. 2007). Similarly, cold waves are defined in different manners such as T mean ≤ 95th percentile over 2 or more days (Du et al. 2022).
In this study, the number of heat and cold waves equal to or longer than 3 days, 4 days, 5 days …, and 24 days was identified for each year in the period 1974-2020. Figure 8 shows the number of heat and cold waves for the urban and rural stations. In Fig. 8a c, an increase in the number of heat waves at both urban and rural stations is seen in the latter half of the period 1974-2020 (see the higher prevalence of dark red cells after the year 2009). Furthermore, according to Fig. 8b d, a decrease in the number of cold waves at both urban and rural stations is seen in the latter half of the period 1974-2020 (see the higher prevalence of dark blue cells after the year 2010). Furthermore, there is no clear evidence that at the urban station heat or cold waves last longer than those at the rural station. Thus, it was understood that the level of urbanization in Lublin city has no impact on the intensity of heat or cold waves. In poorly planned cities, urbanization may cause a loss in green coverage and water bodies, which counteracts the UHI effect (Kikon et al. 2016). However, such development activities are not seen in Lublin city. Thus, Lublin city is able to show no significant changes in the intensity of heat and cold waves compared to Radawiec. Majewski et al. (2014) found that in Warsaw the frequency of warm/cold sensations decreased/increased when moving further away from the city center. Warsaw is the largest city in Poland and it is more densely built-up than Lublin.

Variations in temperature within Lublin city
In order to study the spatial variations of temperature within Lublin city, temperature anomalies between the 30-min temperature data at Litewski Square and those at Dominikanie, Zarnowiecka, Zywnego, and Ofelii are calculated (e.g., T Litewski Square -T Dominikanie ). In Table 3, the percentage of temperature data when Litewski square was warmer/cooler compared to the other stations is shown with a positive/negative sign. There is also a small percentage of data points where the temperature at Litewski square is equal to that at the other stations.
In the majority of the instances, the temperature at Litewski square is higher than that at Zarnowiecka, Zywnego, and Ofelii, which indicates that Litewski square is relatively warmer in general. Particularly, this characteristic is more noticeable when Ofelii is considered as the reference station. This could probably be due to the fact that Ofelii station is located about 4.7 km away from the city center (farthest station from the city center). However, the opposite is observed when the anomalies are quantified relative to the temperature data of Dominikanie station. In other words, when temperature data from Dominikanie station are compared with those of Litewski square, Dominikanie station indicated more frequent warmer conditions. The main reason for this phenomenon is that Zarnowiecka, Zywnego, and Ofelii stations are located at a significant distance away from the city center in less densely built-up areas whereas the Dominikanie station is located very close to the city center in a more densely built-up area. Overall, it can be concluded that in Lublin city, there can be warmer areas, such as Dominikanie compared to Litewski square which is in the city center. Therefore, for a more detailed study of the UHI effect in Lublin city, weather stations should be located in more densely built-up areas where heat gets trapped resulting in a higher increase in air temperature. According to Kuchcik and Milewski (2016), the intensity of UHI in Warsaw is clearly associated with the distance from the center of the city. Furthermore, they commented that UHI intensity depends on the building density and nature of buildings. Yadav and Sharma (2018) studied the variation in the UHI effect within the city of Delhi, India (a mega city in South Asia). They commented that high UHI in certain areas of the city could be due to urban structures, which cool slowly and also the influence of dense traffic is a contributor to the UHI effect.

Analysis of frequency distributions of observed UHI
The frequency distributions of UHI calculated in terms of daily T min and T max are studied corresponding to periods 1974-1997 (1st half of the data series) and 1998-2020 (2nd half of the data series). The UHI calculated using T min and T max is indicated by UHI Tmin and UHI Tmax , respectively. This comparison of frequency distributions of UHI enables the understanding of changes in the frequency distributions in recent years compared to the past. Figure 9 shows the frequency distributions of UHI Tmin and UHI Tmax . As seen in Fig. 9a, the whole distribution of UHI Tmin has shifted to higher intensities in spring, summer, and autumn. In winter, UHI Tmin in the two periods does not show any clear sign of shifting. According to Fig. 9b, the whole distribution of UHI Tmax has shifted to lower intensities in spring and summer and shifted to higher intensities in winter and autumn. In spring, the frequency of T min close to 0 ºC decreased at both stations during 1998-2020 compared to 1974-1997 (See Fig. 2). Also, there is an increase in the frequency of T min above 5 ºC in spring at both stations. Overall, the distributions of UHI have not changed much over time. This is because, at both urban and rural stations, temperature regimes have shown similar rates of increase during the period 1974-2020. As seen in Table S1 in the supplementary material, it is understood that the 10th and 90th percentiles, and the median of UHI Tmin and UHI Tmax have not changed much the period 1998-2020 compared to 1974-1997. According to Kaszewski and Siwek (1998), the intensity of the UHI effect in Lublin is very frequent in the range of 0.1-1.0 ºC and very rarely exceeds 3.1 ºC. In that study, they used data from Litewski square (urban station) and Hajdów (rural station) corresponding to the year 1996. In the current study, it is seen that the peak frequency of UHI Tmin occurs at around 1.0 ºC in winter, spring and autumn, while in summer it occurs between 1.5 and 2.0 ºC. Whereas the peak frequency of UHI Tmax occurs around 0.5 ºC in winter and spring, while in summer and autumn it occurs between 0 and 0.5 ºC. Also, according to Fig. 9, it is clear that the UHI effect is more frequent than the urban cool island effect in Lublin (mostly the distributions are on the positive side of the temperature axis). This characteristic is more dominant in UHI Tmin . Also, in general, UHI Tmin is higher compared to UHI Tmax . This indicated that the UHI effect is more pronounced at night and early morning in Lublin (T min occurs at night and early morning and UHI Tmax occurs during daytime). Krüger et al. (2018) also found that at nighttime, the UHI effect is more pronounced than during daytime in a study over Glasgow, UK. Kaszewski and Siwek (1998) also found that the highest UHI effect occurs between 20:00 UTC and 04:00 UTC in Lublin. According to a study conducted in France by Bernard et al. (2017), in spring and summer, UHI variations are governed by the changes in the vegetation cover and in autumn and winter building density plays a major role in shaping the UHI effect.

Analysis of statistical characteristics of observed UHI
In this section, the UHI calculated using annual T min and T max (annual average calculated using daily T min and T max ) is studied corresponding to the period 1974-2020. The analysis of annual UHI Tmin and UHI Tmax enables the study of changes in the UHI regime over the period 1974-2020. As seen in Fig. 10, it is understood that in all seasons throughout the period 1974-2020 temperature at the urban station is higher than that at the rural station, indicating the presence of the UHI effect in the urban environment.
Also, the UHI effect shows a very small monotonic increasing trend in all seasons. However, according to the MK test, the monotonic increasing trends in the UHI Tmin and UHI Tmax effect are statistically insignificant (p < 0.05), except in spring and autumn. As shown previously in Sect. 3.2, at both urban and rural stations, temperature regimes (i.e., T min and T max ) are increasing at almost the same rate. Thus, the UHI effect in Lublin city has not significantly increased over time. According to Fig. 10, UHI Tmin is more pronounced compared to UHI Tmax . This characteristic is more prominent in summer and spring. In Lublin city, the population has gradually declined from the year 1999 (population of about 359 thousand people) to 2020 (population of about 338 thousand people) (Statistics Poland, 2021). This is a decline of about 5.7% in 22 years. The population is one of the indicators of urbanization. Figure 11 presents the population in the rural counties: Głusk, Jastków, Konopnica, Niedrzwica Duża, Niemce, and Wólka (http:// www. gminy. pl/ powia ty/ 67. html) located around Lublin city. The rural station Radawiec is located in Konopnica county. As seen in Fig. 11, in all rural counties, the population has gradually increased over the period 1995-2020. Thus, it can be argued that the rising population may have had a warming effect on the temperature at the Radawiec station. The decline in population in Lublin city and the increase in population at Radawiec are speculated as the reasons for no clear increase in the UHI intensity. In addition to that, Lublin city has a green coverage of about 40%, which is acting as a heating inhibitor. This also is another reason for the no clear increase in the UHI intensity. In a study by Mohammad and Goswami (2022) where UHI effect in 4 Indian cities were studied, they found that in one of the cities, SUHI effect is more pronounced as that city had a rural fringe with large green coverage (higher urban-rural temperature contrast). Lublin city has a large green coverage, thus the opposite phenomenon is seen.

Analysis of statistical characteristics of future
T min , T max , DTR, and UHI

Daily mean and standard deviation bias correction of GCM projected temperature
In this study, the daily T min and T max data for the period 2015-2100 projected by 15 CMIP6 GCMs corresponding to 4 SSPs are used for deriving future projection of temperature at Litewski square and Radawiec. Despite significant improvements, the CMIP6 GCMs still do not explicitly capture city-scale urbanization and climatic processes due to the coarse spatial resolutions at which they are operated. However, GCMs incorporate the effects of urbanization in terms of greenhouse gas emissions, which influence the large-scale climate. Thus, GCMs capture the temperature response to greenhouse gas emissions that helps in extracting the broad warming signal on a coarser scale. The new SSPs are developed by integrating climate forcing and socioeconomic conditions (Gidden et al. 2019;O'Neill et al. 2016a, b). For example, SSP2-4.5 and SSP3-7.0 are embedded with the effects of high urbanization rate and rapid population growth, respectively, while SSP5-8.5 is embedded with high radiative forcing (Gao and Pesaresi, 2021;Jones and O'Neill, 2016;Riahi et al. 2017a, b). Thus, the influence of urbanization and population coupled with radiative forcing is considered in producing large-scale climate projections of CMIP6 GCMs. In terms of land surface modelling, the CMIP6 modelling groups have adopted two different approaches for simulations of land cover, i.e., the simulated land cover approach and the prescribed land cover approach based on a forcing dataset of land use and land cover changes (Hurtt et al. 2020;Song et al. 2021). Both these approaches are used indirectly to capture land surface processes for different land cover classes in earth system models, which have significant effects on global and regional climates under different scenarios (Chen et al. 2020;Lawrence et al. 2019). Since the resolution of the GCM outputs is too coarse for the analysis of city-scale temperature and UHI effect, a bias correction is performed. Under this step, bias in the average and standard deviation of the temperature projections of each GCM is corrected. During biascorrection, characteristics of local temperature (city-scale) are assimilated into these coarser scale GCM outputs which carry the large-scale warming signals. Thus bias-corrected temperature is expected to capture the rural-urban temperature contrast under changing climate. The average and standard deviation of bias-corrected and raw GCM projected T min and T max for the period 2015-2100 are presented in Table S2 in the supplementary material for Litewski square and Radawiec. The theory and application of this bias correction method used in the current study can be found in the studies by Johnson and Sharma (2012) and Sachindra et al. (2014).  The main assumption of this bias correction is that the bias in the average and the standard deviation of the future GCM projections and the past GCM simulations are equal. In other words, the bias in the average and the standard deviation of the GCM outputs is stationary. The bias correction is applied to T min and T max for each station, each GCM and each SSP separately. In the application of this bias correction, first, the future GCM projections of T min and T max are standardized with the corresponding average and the standard deviation of the historical GCM simulated T min and T max . In Eq. 1, T (GCM,F) , μ (GCM,H) , σ (GCM,H) and T � , refer to raw temperature values projected into future by a GCM, average of the temperature simulated by the same GCM pertaining to the historical period 1991-2020, the standard deviation of the temperature simulated by the same GCM pertaining to the historical period 1991-2020, and the standardized temperature values projected into the future by the GCM, respectively.
Then, the standardized temperature values ( T � ) projected into the future by a GCM is retransformed using the average and the standard deviation of the observations pertaining to the historical period 1991-2020 (last 30 years of the records of observations) as shown in Eq. 2. In Eq. 2, μ (Obs,H) , σ (Obs,H) , and T �� refer to the average of observed temperature, the standard deviation of observed temperature, and T �� refer to bias-corrected temperature projected into the future by the GCM, respectively. on fossil fuel intense development (Riahi et al., 2017a, b). Both these scenarios characterize high GHG emissions, thus a relatively larger rise in temperature is expected. However, under SSP1-2.6 the increase in T min and T max is less pronounced particularly in the latter half of the twenty-first century. SSP1-2.6 depicts a world that is characterized by sustainability where a green approach (e.g., use of renewable energy) is followed by the human civilization, thus, the GHG emissions are reduced. Due to the reduction in GHG emissions, the increase in the greenhouse effect is reduced. As a result, the temperature regime (T min and T max ) is showing a reduction in the rate of increase. For example, MRI-ESM-2.0 high-resolution GCM in Figs. 12m and 13m shows a decreasing trend in the temperature regime towards the end of the twenty-first century corresponding to SSP1-2.6 (despite the fact that the overall trend is increasing). The same is valid for the temperature projections of GFDL-ESM4 and EC-Earth3 (finest resolution GCM used) under SSP1-2.6. This is an indication that with the reduction in GHG emissions, warming begins to reduce and even some cooling can occur in the distant future. Also, according to Figs. 12e and 13e, FGOALS-g3 which is the coarsest resolution GCM used in this study has shown the slowest rate of increase in temperature at both stations. However, the rates of increase of T min and T max are similar at a given station so that DTR does not show any clear increasing/decreasing trends (Figs. 12 and 13). As seen in Fig. 14, UHI effect in Lublin city does not show any significant changes under most of the emission scenarios. Table 4 shows the trends in future annual T min and T max in terms of SS corresponding to 15 CMIP6 GCMs under 4 SSPs for the period 2015-2100, for Litewski square and Radawiec. According to Table 4, it is further understood that the temperature regime shows increasing trends in terms of SS for  for all GCMs at both stations. The MK test also confirmed that there are statistically significant monotonic increasing trends in the temperature regimes at both stations for all GCMs corresponding to . It is further noticed that these increasing trends are highest for SSP5-8.5 and lowest for SSP1-2.6. At both stations, in the majority of the instances, the magnitudes of the increasing trends are similar for a given SSP. This hinted that both stations may warm at almost the same rate under a given scenario in the future. A similar characteristic is seen in the past temperature observations at Litewski square and Radawiec. At both stations, the largest increasing trends in terms of SS for T min and T max are seen in the projections produced by ACCESS-CM2 for SSP5-8.5. At Litewski square the ranges of SS for T min and T max corresponding to SSP5-8.5 (SSP3-7.0) are 0.31 to 0.82 (0.23 to 0.63) and 0.26 to 0.88 (0.31 to 0.67), respectively. At Radawiec the ranges of SS for T min and T max corresponding to SSP5-8.5 (SSP3-7.0) are 0.31 to 0.82 (0.22 to 0.62) and 0.27 to 0.91 (0.24 to 0.69), respectively. As shown in Table 5, the UHI effect in Lublin does not show a statistically significant increase or decrease in the future under most of the GCM -SSP combinations. However, UHI Tmax shows statistically significant decreasing trends on quite a few occasions (e.g., under SSP5-8.5 the majority of the GCMs show statistically significant decreasing trends). This is because T max at Radawiec shows higher increasing trends compared to T max at Litewski square (Table 4). Different GCMs use different parameterization schemes and employ different assumptions and approximations (Hwong et al. 2021). Therefore, the projections that they produce vary from one to another giving rise to uncertainties. Sachindra et al. (2016) found that the UHI effect in the city of Melbourne, Australia showed an increase with the rising GHG emissions. They concluded that global warming may exacerbate the UHI intensity in the city of Melbourne. Doan et al. (2019) studied the impacts of land use and anthropogenic heat emissions on the UHI effect in Hanoi, Vietnam using a numerical model. They found that in the period 1990-2010 UHI was governed by land use changes and in the period 2010-2030 anthropogenic heat emissions will make a significant contribution to the UHI effect. According to Macintyre et al. (2021), who used UK Climate Projections (UKCP18) corresponding to RCP8.5 (Representative Concentration Pathway) in a numerical weather model, in the future there is a higher risk of heatrelated mortalities associated with summer UHI through a reduction in cold-related mortalities linked to UHI in winter

Conclusions
Frequency distributions of daily T min and T max for an urban station (Litewski square in Lublin City) and a rural station (Radawiec) for the periods: 1974-1997 (1st half of data) and 1998-2020 (2nd half of data), are analyzed. According to this investigation, it is clear that in winter, spring, summer and autumn, the frequency distributions have shifted to a warmer climate at both stations. However, this characteristic is more pronounced in summer than in other seasons. Also, at the urban station, in a given season, frequency distributions of T min and T max referred to warmer conditions compared to the rural station. According to the trend analysis, in all four seasons the annual T min and T max show increasing trends at both stations in the period 1974-2020 (statistically significant trends except in winter for T max ). It is worth mentioning that, the increasing trends in T min and T max at both stations are highest is summer. In winter, at both stations the smallest increasing trends are seen in T max (statistically not significant). Also, in a given season, increasing trends in T min or T max are similar at the two stations. DTR (T max -T min ) is higher in summer and spring at the rural station, and shows relatively small trends particularly at the urban station (mostly statistically not significant).
It is seen that there is no clear increase in the UHI intensity in Lublin city in the period 1974-2020. Population in Lublin city has steadily dropped from the year 1999 to 2020 (5.7% decline). However, the population in the surrounding rural counties has increased in the period 1995-2020. It is argued that the above two factors and the large green coverage in the Lublin city (about 40% of the area), which acts as a heat sink have subdued the UHI effect. According to CORINE land use data for the years 2000, 2006, 2012, and 2018, land use in the immediate surroundings of the stations has not changed significantly. However, an expansion in urban land use categories outside the city boundary is observed.
All GCMs show increasing trends in future annual T min and T max throughout the period 2015-2100 for relatively high emission scenarios: SSP3-7.0 and SSP5-8.5. However, under SSP1-2.6, which is a low emission scenario, the increase in T min and T max is less noticeable, particularly in the latter half of the twenty-first century. This is because SSP1-2.6 refers to a world which is sustainable and uses green technologies. Thus, it is expected to reduce the emissions and curb the rise in temperature. Also, for most of the GCM SSP combinations, UHI Tmin and UHI Tmax do not show statistically significant increasing or decreasing trends in Lublin city in the future. GCMs do not characterize the city-scale changes in land use and climatic processes, though they capture large-scale warming signals well. In this study, a bias-correction is implemented (can be seen as a simple downscaling approach) to assimilate the city-scale climate (land use changes indirectly included) into large-scale warming signals projected by GCMs. It is worth mentioning that the above approach can introduce uncertainties to the projections.