The assessment of bioclimatic conditions at the national scale remains a highly relevant task. It might be one of the main parts of the national strategy for the sustainable development of different regions under changing climatic conditions. This study evaluated the thermal comfort conditions and their changes in Russia according to gridded meteorological data from ERA-Interim reanalysis with a spatial resolution of 0.75° × 0.75° using the two most popular bioclimatic indices based on the human energy balance: physiologically equivalent temperature (PET) and universal thermal comfort index (UTCI). We analyzed the summer and winter means of these indices as well as the repeatability of different thermal stress grades for the current climatological standard normal period (1981–2010) and the trends of these parameters over the 1979–2018 period. We revealed the high diversity of the analyzed parameters in Russia as well as significant differences between the contemporary climate conditions and their changes in terms of mean temperature, mean values of bioclimatic indices, and thermal stress repeatability. Within the country, all degrees of thermal stress were possible; however, severe summer heat stress was rare, and in winter nearly the whole country experienced severe cold stress. Multidirectional changes in bioclimatic conditions were observed in Russia against the general background of climate warming. The European part of the country was most susceptible to climate change because it experiences significant changes both in summer and winter thermal stress repeatability. Intense Arctic warming was not reflected in significant changes in thermal stress repeatability.
Global warming will lead to an increase in the number of extreme weather events, such as heat and cold waves, as well as bioclimatic changes (IPCC 2013). This issue is important for the assessment of both weather-related health consequences and spatiotemporal variability of bioclimatic conditions.
One of the most negative health consequences is an increase in additional mortality, primarily due to cardiovascular and respiratory diseases (Kovats and Hajat 2008). Air temperature continues to be an important risk factor after the most significant heat events in Chicago, Illinois, in 1995 (Dematte et al. 1998), Europe in 2003 (de’Donato et al. 2015), Moscow, Russia, in 2010 (Konstantinov et al. 2014; Shaposhnikov et al. 2014), and the most recent heat wave in Europe in June and July 2019. The number of days with adverse weather events may rise in the future, which will lead to an increase in climate-related mortality.
The development of timely preventive public health activities requires accurate assessment of thermal sensations. At present, about hundred bioclimatic indices have been developed worldwide to characterize the impact of the thermal environment on human beings in terms of thermal comfort (Blazejczyk et al. 2012). Thermal comfort can be described as the mental and physiological contentment with thermal environment (ASHRAE 1966; Parsons 2003). It is a condition in which an individual would prefer neither a warmer nor a cooler temperature.
Most of the bioclimatic indices are based on empirical meteorological parameters. They are often called simplistic indexes. Small number of indices include human energy balance models (rational indices) (Katavoutas and Flocas 2018). The indices based on direct measurements of meteorological values, such as air temperature, wind speed, and ambient humidity, are more convenient for calculation and more practical to implement. For example, the United States (Weinberger et al. 2018) and Canadian (Smoyer-Tomic and Rainham 2001) weather forecasting services use the heat index, humidex, and wind chill temperature index. Most of these indices, however, have a major limitation—their thermophysiological relevance (Mayer and Höppe 1987; Potchter et al. 2018).
The indices based on human energy balance models provide a more comprehensive and accurate representation of human thermal perception (Blazejczyk et al. 2012; Shartova and Konstantinov 2019); however, they require more input environmental data, including radiation fluxes, which should be measured or estimated, as well as metabolic heat and clothing insulation. This allows physiologically significant evaluation of thermal conditions (Matzarakis et al. 1999) and considers the physical environment and human physiology, as well as associated psychological responses (Vanos et al. 2010). As early as 1938, Büttner (1938) indicated that we need an energetic approach to interactions between the human body and the environment. However, the development of energy balance models of the human body became possible only with the advent of computer technology in the 1970s and 1980s (Höppe 1997). Currently, a variety of powerful tools is applied worldwide, including energy balance models of the human body. The most useful tools include the Comfort Formula energy budget model (COMFA model; Brown and Gillespie 1986), Rayman model (Matzarakis et al. 2007), and Solar Longwave Environmental Irradiance Geometry model (SOLWEIG model; Lindberg et al. 2008) and the detailed three-dimensional “ENVI-met” micrometeorological model (Fabbri et al. 2017) and the Parallelized Large-Eddy Simulation Model (PALM; Fröhlich and Matzarakis 2020). These tools and models are actively used to investigate the methods of creating optimal thermal comfort in real urban environments (Vanos et al. 2010).
Many studies have been aimed at defining thermal conditions for humans in the outdoor environment and grading thermal sensation (Potchter et al. 2018; Bauche et al. 2013; Vitkina et al. 2019; Kajtar et al. 2017). The pioneering predicted mean vote (PMV) model of thermal comfort, which was based on college-aged students, was created by P. O. Fanger in the late 1960s, but it is still used worldwide in environmental engineering. This index requires improved predictive ability, which might be achieved through better specification of the model’s input parameters and accounting for special groups (van Hoof 2008). The use of the PMV index in urban areas in Sweden shown that the steady-state models are not applicable in case of short-term outdoor thermal comfort assessment due to the complication in analysis of transient exposure (Thorsson et al. 2004). A similar study of subarctic climate (Umea, Sweden) showed that local residents are more adapted to the subarctic summer than nonlocals (Yang et al. 2017). Neutral physiologically equivalent temperature (PET) values for summer and winter were calculated according to regression lines that were based on output of interviews and PET values during a field survey in Rome, Italy, using over 1000 questionnaires (Salata et al. 2016). The studies were performed both for the European region and for Russian far east regions, where the combinations of meteorological factors form different thermal comfort conditions (Vitkina et al. 2019).
Although a large number of indices have been developed, only four of them [PET, PMV, universal thermal comfort index (UTCI), and standard effective temperature (SET)] are widely used for outdoor thermal perception studies (Potchter et al. 2018). The PMV index is more widespread as a tool for outdoor and indoor thermal assessments for building design and redesign (Ricciu et al. 2018). SET is similarly used in urban semioutdoor and outdoor environments, for example, streets, railway stations, bus shelters, ferry terminals, and parks (Hwang and Lin 2007; Zhao et al. 2016). The PET and UTCI indices are widely applied in different regions with various spatial resolutions—from the local to regional level (Blazejczyk et al. 2012). Some of these applications have been specifically focused on thermal comfort investigations of resting places (parks and squares) in urban environments (Égerházi et al. 2013; Thorsson et al. 2004). According to Potchter et al. (2018), PET has been constantly used since 2006, and since 2012, it has become the most frequently used index. The UTCI first appeared in 2012, and since then, it has been used more often.
Despite the rapidly increasing use of urban or intraurban assessments of thermal comfort around the world (Mayer et al. 2008; Matzarakis et al. 2009; Grigorieva and Matzarakis 2011; Matzarakis et al. 2013; Chi et al. 2018), obtaining information about bioclimatic conditions at the national scale remains a highly relevant task (Vinogradova 2009; Jacobs et al. 2013; Emelina et al. 2014; Vinogradova and Zolotokrylin 2014; Giannaros et al. 2018; Wu et al. 2019). It is necessary to develop a national strategy for the sustainable development of different regions under changing climate conditions. The use of reanalysis data (gridded meteorological data with global coverage) might be one of the possible solutions (Di Napoli et al. 2018). This issue is relevant, especially for countries with large territories, where the spatiotemporal variations in thermal comfort based on human thermal balance indices have not been investigated on a national scale.
The task of the national-scale assessment of bioclimatic conditions is very relevant, but it is challenging for the Russian Federation, which is the world’s largest state (17.1 million km2) and has very diverse climatic conditions (Kobysheva 2001). According to the Koppen–Geiger climate classification, the types of the Russian climate vary from relatively mild Cfa in the southwest to extremely severe Dfc, Dfc, and Dwd in Siberia and tundra (ET) and ice cap (EF) in the Arctic (Kottek et al. 2006). Wide areas of Russia, especially in its northern regions, are hot spots of recent climate warming (IPCC 2013; Serreze et al. 2009). However, the bioclimatic conditions in Russia have not yet been evaluated in terms of the convenient thermal stress indices on the national scale. The few previous studies that have attempted to evaluate bioclimatic conditions in Russia are limited only by the analysis of simple indices, which are not connected with physiology (Emelina et al. 2014; Vinogradova and Zolotokrylin 2014), or by the analysis of the monthly mean UTCI values (Vinogradova 2019). These generalizations do not allow for more detailed analysis of the spatiotemporal relationships and tendencies. Additionally, these studies are published only in Russian, which decreases the availability of their results for the international scientific community.
In the current study, we aimed to evaluate the spatial patterns of thermal comfort conditions in Russia and to investigate their long-term trends for a contemporary climate based on state-of-the-art biometeorological indices and widely used gridded meteorological data, namely, ERA-Interim reanalysis.
2. Data and methods
The biometeorological indices could be calculated based on specialized microclimatic observations (van Hove et al. 2015), observations at regular weather stations (Urban and Kyselý 2014), the results of high-resolution numerical modeling (Fröhlich and Matzarakis 2020), and other types of meteorological data. However, the assessment of the bioclimatic conditions on such large spatial scales as the area of Russia and on such long temporal scales as decades remains exacting tasks. The network of regular weather stations is too sparse for such a task, especially in remote northern and eastern Russian regions, while high-resolution regional simulations for such a large area would require unprecedentedly large computational resources. The rapid development of gridded meteorological datasets opens up new cost-effective opportunities for such studies.
The novelty of our study lies in the use of gridded meteorological data, namely, the ERA-Interim atmospheric reanalysis for the assessment of the thermal comfort conditions and trends in their changes on a national scale. The global reanalysis products are produced via data assimilation—a process that relies on both observations and model-based numerical forecasts (Parker 2016). The numerical models of the atmosphere, ocean, and other Earth system components provide continuous and physically consistent fields of atmospheric variables on a regular grid, while the data assimilation systems involve observations from a variety of sources to make the model as realistic as possible (Kalnay et al. 1996; Parker 2016). The main advantage of reanalysis products is continuous spatial and temporal coverage over the entire globe for various variables, many of which are practically inaccessible from observational data. They provide comprehensive snapshots of conditions at regular intervals over decades (Parker 2016).
Global and regional reanalysis products have been rapidly developed and improved since the release of the first global NCEP–NCAR reanalysis (Kalnay et al. 1996). Currently, atmospheric reanalysis products are among the most commonly used datasets in weather and climatic studies (Parker 2016). They have been used to study atmospheric dynamics (e.g., Kidston et al. 2010), to investigate climate variability (Ivanov et al. 2019; Semenov and Latif 2015), to evaluate global climate models (e.g., Gleckler et al. 2008) and to supply high-resolution regional climate simulations in initial and boundary conditions (e.g., Varentsov et al. 2018) as well as for ecological applications (Mislan and Wethey 2011). The first pioneering studies have already shown the high potential of using reanalysis data for bioclimatic assessments (Jacobs et al. 2013; Di Napoli et al. 2018).
a. Research data
The study used the ERA-Interim reanalysis (Dee et al. 2011), which was conducted by the European Centre for Medium-Range Weather Forecasts (ECMFW) and is freely available from the ECMWF data portal (https://apps.ecmwf.int/datasets/data/interim-full-daily/levtype=sfc/). It is still not perfect for high-latitude regions but provides the best verification results for wind speed, near-surface temperature, and radiative fluxes in comparison to six other global reanalysis products, as is shown in a detailed evaluation study (Lindsay et al. 2014). The spatial resolution of the ERA-Interim reanalysis data is 0.75° latitude × 0.75° longitude; the temporal resolution is 3 h. The data processing and analysis were performed for territory of the Russian Federation, which was limited to 20°E–170°W, 40°–80°N. The study used the data collected in the period from 1 January 1981, to 31 December 2010 (30 years), which corresponds to the current climatological standard normal period according to the World Meteorological Organization (WMO) for the evaluation of the contemporary thermal comfort conditions, and the data for a longer period from 1979 to 2018 (40 years) for the trend assessment.
Recently, the ERA-Interim reanalysis was replaced by the new ERA5 reanalysis in the chain of ECMWF products (Hersbach et al. 2019). ERA5 data have higher spatial (0.25°) and temporal (1 h) resolutions and likely lower biases. However, as far as the authors are aware, the quality of ERA5 reanalysis has not yet been evaluated for northern Eurasia and specifically for Russia. Additionally, the processing of ERA5 data with hourly resolution on the decadal time scale for such a large area becomes a challenging and computationally demanding task because the data volumes is 12 times higher than that of the ERA-Interim reanalysis. For these reasons, the present pioneering study was limited by the ERA-Interim data, while the migration to ERA5 is planned for further research.
b. Bioclimatic indices
The study used the two most valuable bioclimatic indices based on the human energy balance—PET and UTCI—to evaluate both heat and cold stress. The PET is defined as the equivalent of the air temperature in a standardized indoor setting and for a standardized person, reproducing the core and skin temperatures that are observed under current meteorological conditions (Matzarakis et al. 2010). The UTCI is based on the advanced multinode dynamic model of human thermal physiology and comfort (Fiala et al. 2012), and it is combined with a clothing model (Havenith et al. 2012). The outcome is expressed as a temperature equivalent to that in the reference conditions, implying the same physiological response as under the conditions to be assessed (Jendritzky et al. 2012). The basic difference between the UTCI and PET lies in the fact that UTCI was validated for all climates and seasons, while PET application generally has been limited to western-central European, Mediterranean, and East Asian climatic conditions (Coccolo et al. 2016). In general, there is good agreement and a relationship between the PET and UTCI for warm conditions (Potchter et al. 2018). The UTCI and PET are equally suitable for hot conditions, whereas UTCI is better for warm and humid environments. A study in Guangzhou showed that the linear relationship between the UTCI and relative humidity was more evident and significant (Fang et al. 2018). For cold conditions, the UTCI provides more details about cold stress because of adjusted clothing insulation (Matzarakis et al. 2014). In further analysis, the PET was used to assess heat stress, whereas the UTCI was used to assess cold stress.
The Munich Energy-Balance Model for Individuals and the Fiala multinode model provided the basis for the PET and UTCI calculations, respectively. Using the unique software Rayman Pro 3.1 (VDI 1998; Matzarakis et al. 2007), it is possible to calculate radiation fluxes and bioclimatic indices at a specific point in time and place for individual anthropometric characteristics (age, gender, weight, etc.).
The reanalysis data were used to assess the following atmospheric variables, which are required for further calculation of thermal comfort indices: the 2 m air temperature and 2 m dewpoint temperature, wind speed at a height of 10 m, total cloud-cover fraction, and surface temperature. The preliminary processing of the reanalysis data included calculation of the relative humidity from the dewpoint using the Magnus equation, unit conversation for the cloud fraction, and moving the wind from a height of 10 m to a height of 1.1 m using the logarithmic wind profile (Oke 1987) and fixed roughness length parameter z0 = 0.01 m, which corresponded to a short-cut meadow and is the default for UTCI calculation (Havenith et al. 2012). The radiative fluxes were parameterized by the built-in equations in the Rayman model using the information on location, elevation, time, and total cloud-cover fraction (Matzarakis et al. 2010).
The thermophysiological parameters were set up as follows: male, 35 years old, 1.75 m tall, with a weight of 75 kg, an internal heat production of 80 W, and a heat transfer resistance of clothing of 0.9 clo.
c. Statistical analysis
The developed technology for the PET and UTCI calculations on gridded reanalysis data includes MATLAB software and a conceptual framework for its consistent application from downloading the source data from the ECMWF website to final visualization of bioclimatic values. The technology provides cartographic visualization directly in MATLAB and translates the results into the GeoTIFF raster format for further processing in GIS software.
The output data included the PET and UTCI values for each grid cell and each time moment within the considered period (with a 3-h resolution) as well as a set of statistical parameters on their contemporary climatology. These parameters included the long-term means as well as the repeatability of the different thermal stress categories according to Table 1 for the different months and seasons.
To investigate the contemporary changes in the thermal stress conditions, we performed a linear trend analysis of the seasonal means of the PET and UTCI indices as well as for the repeatability of the different thermal stress categories for the period from 1979 to 2018. We used the Mann–Kendall nonparametric statistical test to evaluate the significance level of the trends and the Sen’s slope estimator (Helsel and Hirsch 1992) to evaluate the rates of changes. These statistical methods can be applied even if the time series do not conform to a normal distribution (Helsel and Hirsch 1992; Meals et al. 2011); therefore, they have been widely used in recent climate and environmental studies (a detailed review is presented, e.g., in Kocsis et al. 2017). Trend analysis was performed in MATLAB software using external libraries (Fatichi 2020; Tilgenkamp 2020). For each reanalysis grid cell and each analyzed parameter, we obtained the Sen’s slope coefficients k and minimum confidence levels p, at which the trends would be statistically significant according to the Mann–Kendall test.
3. Results and discussion
In this section, the maps of the most important bioclimatic parameters and their changes are presented and discussed for two contrasting seasons—summer (June, July, and August) and winter (December, January, and February).
a. Spatial patterns of air temperature and long-term mean PET and UTCI for modern climates
For clarity, we compared the bioclimatic parameters with air temperature, which was the simplest and most understandable indicator of climatic conditions. Considering the average winter (Fig. 1a) and summer (Fig. 1b) temperature distributions over the territory of Russia, it should be noted that the thermal differences in winter were more pronounced than those in summer. The winter temperature magnitude exceeded 40°C. It varied greatly from the coldest regions in eastern Siberia to positive temperatures on the Black Sea and Caspian Sea coasts (up to 0°–5°C). The western part of Russia (Kaliningrad region, which is located between Poland and Lithuania) was slightly colder than southern Russia because of the Baltic maritime climate. Here, the winter mean temperature was approximately 0°C. In summer, the temperature was more dependent on latitude, with the exception of in mountainous areas. The greatest difference was observed between the subtropics in the south (above 25°C) and the Arctic coast (as low as −2.5°C), especially on the islands of the Severnaya Zemlya Archipelago, which separate the Kara Sea and the Laptev Sea.
The areas of extremes of average temperatures in the warmest and coldest months did not always coincide with the areas with extreme comfort indices (Fig. 1). For example, the winter temperature minimums were concentrated in eastern Yakutia, where the mean winter temperature according to reanalysis data reached −40°C and where the lowest air temperature in the Northern Hemisphere had been recorded. Two sites in this region are known as the northern “Pole of Cold”—Verkhoyansk (−67.6°C in 1892) and Oymyakon (−67.7°C in 1933) (Stepanova 1958; Kobysheva 2001; Darack 2013). However, the minimum values of the mean winter UTCI were found on the northern coast of Siberia, where the local extrema fell below −50°C. Such a mismatch of minima in the fields of mean winter UTCI and temperature could be caused by strong and persistent wind at the Arctic coast and, conversely, the typically calm weather in the center of the Siberian high (Yakutia).
The spatial distributions of the mean summer air temperature (Fig. 1b) and PET (Fig. 1d) were very similar: the maximum (up to 25°C) PET were observed near the Caspian Sea coast, and the minimum (as low as −2.5°C) PET were observed on the Arctic coast.
Thus, the long-term means of PET and UTCI demonstrated significant differences within the Russian territory. In summer, such differences were preconditioned first by latitude, whereas in winter, they were shaped first by the contrasts between maritime and continental climates (Kobysheva 2001).
b. Spatial patterns of cold and heat stress frequency for modern climate
The bioclimatic conditions could be characterized not only by the mean values of thermal indices but also by the frequency of the different categories of heat or cold stress (Table 1). Due to the large area of the country, the Russian climate was represented by all categories of heat stress in summer and cold stress in winter. In summer, days with at least moderate heat stress (PET > 23°C) or even with at least high heat stress (PET > 27°C) were not rare, with a country-mean repeatability of 28.7% and 11.5%, respectively. The mean repeatability of the days with at least very high heat stress (PET > 35°C) and the days with extreme heat stress (PET > 41°C) was low, at only 2.5% and 0.3%, respectively. In winter, permanent cold stress was typical for almost the whole area of Russia: the country-mean repeatability of the days with at least high cold stress (UTCI < −13°C) was almost 100%. Stronger cold stress was also frequent: the country-mean repeatability of the days with at least very high cold stress (UTCI < −27°C) was 86%, and the country-mean repeatability of the days with extreme cold stress (UTCI < −40°C) was 49%. Of course, the country-mean values were not very informative for a country such as Russia. We performed a more detailed spatial analysis for those thermal stress categories, which had significant country-mean repeatability and pronounced variations in repeatability within the country, namely, the high heat stress for summer and the very high cold stress for winter.
The spatial distribution of frequency of the days with heat/cold stress was not similar to the spatial patterns of the mean PET and UTCI values (Fig. 2). The wide area, including most of Siberia and the far east of Russia, was inhomogeneous in terms of the mean winter UTCI, but it was characterized by a high (almost 100%) repeatability of days with very high cold stress, with UTCI < −27°C. Within the whole Far East region, the repeatability of such days was noticeably less than 100% only in the south of the Kamchatka Peninsula and in the southeastern part of the region (close to the border with North Korea). In summary, the repeatability of very high cold stress was more than 95% during the winter in approximately 50% of Russian territory, mainly in the east of the country. Relatively “hot” areas were only observed in the Kamchatka Peninsula and extreme south of the Primorskiy Kray territory (close to the border with North Korea). In the European part of Russia, high cold stress repeatability (more than 95%) was observed only in the Arctic (Novaya Zemlya islands and Barents Sea coast). The major part of the east European plain was characterized by high cold stress repeatability, which ranged from 45% to 65%. In the southern Russia (Caucasus, coasts of Black and Caspian Seas only), the repeatability was lower than 5%.
The spatial distribution of the days with high heat stress frequency (PET > +29) for summer, in general, followed the same pattern as the mean temperature and PET values (Fig. 2b). The highest frequency of heat stress was observed in southern Russia, between the Black and Caspian Seas, reaching a maximum of up to 80% near the coast of the Caspian Sea and the Kazakhstan border. In the rest of Russia, the repeatability of the summer days with high heat stress was lower than 40% and further decreased to zero toward the Arctic coast and northeastern Pacific coast. The frequency of such days was almost zero and did not exceed 1%–2% within wide areas in northern Russia, almost anywhere north of 64°N, and in northeastern Russia, including the Kamchatka Peninsula.
c. Contemporary changes and trends
Recent global climate change is associated with temperature growth. Hence, we compared the change rates of the thermal indices and thermal stress repeatability with the temperature change rates. Linear trend analysis clearly showed the pronounced heterogeneity of the rates of the long-term changes of the mean temperature, PET and UTCI values (Fig. 3), and the thermal stress repeatability (Fig. 4) against the background of the general warming trend. For both summer and winter seasons, the spatial patterns of the linear trend slopes for the temperature and PET/UTCI indices were generally similar, which indicated the leading role of the temperature changes in the changes in the indices. However, the rates and significance of the changes in temperature and thermal stress indices could differ. For example, winter warming in northwestern Russia was more strongly expressed in terms of UTCI than in terms of temperature, while the rates of mean summer PET growth in southwestern Russia were higher than the rates of mean temperature growth. If we considered the anomaly of PET and UTCI trends with respect to the temperature trends (Figs. 3e,f), we could see that mean summer PET is increasing slightly faster than mean temperature almost everywhere except in western Siberia, while mean winter UTCI was increasing faster than mean temperature in the European part of Russia and in western Siberia. Such amplified rates of changes in UTCI and PET indicated the additional contribution from the long-term changes in other meteorological factors, which were further analyzed in the next subsection.
The heterogeneity of the change rates of the thermal stress conditions was especially obvious in winter. The highest warming rates in terms of mean temperature and UTCI, up to 2°C decade−1, were registered in northern Russia near the Arctic coast (Fig. 4). Such a pattern as consistent with an overall intensive warming in the Arctic, which is known as one of the areas with the most intensive regional warming rates in the world (IPCC 2013; Serreze et al. 2009). However, wintertime warming in northern and northwestern Russia is accompanied by near-zero changes in the wide areas of Siberia and in the far east of Russia and even by negative trends in the mean temperature and UTCI in southern Siberia. The latter could be explained by the atmospheric response to the strongly reduced Arctic sea ice cover, which was expressed as an anticyclonic surface pressure anomaly and negative air temperature anomaly in southern Siberia (Semenov and Latif 2015).
The spatial patterns of the mean temperature and UTCI change rates were only weakly correlated with the changes in winter cold stress repeatability (Fig. 4a). The rapid warming and increase in the mean UTCI in the Arctic were not expressed in terms of cold stress repeatability because the area is still too cold and windy. The same situation takes place in southern Siberia, where significant cooling trends are not expressed in terms of the cold stress repeatability. A significant decrease in cold stress repeatability was observed only in western Russia, while in most of Siberia and the whole far east of Russia, hardly any changes were observed. This means that the winter thermal stress conditions became softer in those regions where they were already relatively mild, while the regions with the most severe conditions did not exhibit significant changes in cold stress repeatability.
For the summer season, the spatial patterns of the change rates of the mean temperature and PET were more homogeneous in comparison to those in winter. In almost all territories of Russia, we found steady warming trends in terms of PET and temperature, with the only exception of western Siberia, where near-zero changes were observed. The latter was consistent with previous studies on climate change in Russia (e.g., Ippolitov et al. 2014). The hot spots of summer warming were southern Siberia as well as the western and southwestern regions of Russia, where the warming rates exceeded 1°C decade−1. The rates of changes in the heat stress repeatability have more heterogeneous spatial patterns (Fig. 4b), which were only partially correlated with the changes in the mean temperature and PET. The heat stress danger was rising in western Russia, with a hot spot near the western border of the country, and in the southern parts of Siberia and the Far East. The rest of the country, including its northwestern regions, western Siberia, the northern parts of eastern Siberia and the Far East, exhibited near-zero changes in heat stress repeatability. The latter may be due to the absence of warming trends in some regions (e.g., in western Siberia) as well due to the cold climate of other regions (e.g., Arctic coast), where the days with a high thermal stress were not registered in principle, despite the overall warming trend.
d. Trends of the driving meteorological parameters
To investigate the driving factors responsible for faster or slower change rates of the thermal indices in comparison with air temperature change rates, we analyzed the trends for other meteorological variables used to calculate the thermal indices—namely, the wind speed, total cloud cover, and relative humidity (Fig. 5). In winter, the increased rates of the UTCI were registered in the European part of Russia and western Siberia; they were caused by the decreasing wind speed in these regions, which was shown by reanalysis data as well as by regional climate change studies, which were reviewed in Wu et al. (2018). The mean winter cloudiness and humidity were not affected by significant trends. In summer, the areas with amplified rates of PET growth—the European part of Russia and southern Siberia (see Fig. 3f)—were characterized by significant negative trends in total cloud cover, which meant an increase in solar radiation and explained an additional increase in the PET index. A downward trend in the summer cloudiness in the European part of Russia was consistent with the results of (Chernokulsky et al. 2011) based on the cloudiness observations at weather stations, which also showed a decrease in overcast frequency. The high-quality observations at the meteorological observatory of Lomonosov Moscow State University also showed a downward trend in cloudiness and overcast frequency and an upward trend in sunshine frequency (Gorbarenko 2019). The areas with downward trends in cloudiness were also characterized by negative trends in relative humidity, which should partially compensate for the effect of decreasing cloudiness according to the PET index. However, the contribution from decreasing humidity seemed to be smaller than the contribution from decreasing cloudiness, which is consistent with the small sensitivity of the PET index to humidity (Fang et al. 2018).
e. Current trends in the global context
In the present study, we evaluated Russian thermal comfort conditions for the first time using state-of-the-art thermal indices based on convenient gridded meteorological data with a 3-h temporal resolution. The novelty of this approach lies in the possibility of assessing not only the average values of thermal indices but also the repeatability of various gradations of thermal stress. Such studies have not ever been performed for any part of Russia in principle, which complicates the comparison of the results with those of other studies. We could only postulate that our results confirmed the boundaries of territories with maximum cold stress in winter according to a previous study (Bauche et al. 2013; Vinogradova 2019) and compared our results for Russia with results for other regions.
The marked decreasing trend of cold stress in the European part of Russia clearly corresponds to the same cold extremes tendency in the Euro-Mediterranean region (Giannaros et al. 2018) and in northern Serbia (Basarin et al. 2018), as well as the upward trend in heat stress repeatability. An assessment of bioclimatic conditions in western Poland (the Lubuskie Voivodeshi) conducted from 1971 to 2006 (Mąkosza 2013) showed a negative trend in days with categories related to UTCI cold stress. In Crete, the annual number of days with mean daily extreme values of PET/UTCI tended to neither increase nor to decrease during the 30-yr period from 1975 to 2004 (Bleta et al. 2014). An increase in heat stress during summer in the main part of eastern and southern Siberia could be compared to a similar warming trend in China (Wu et al. 2017). The tendency toward increasing heat stress in the territory near the Caspian Sea corresponded to the results of an Iranian bioclimatic condition assessment (Roshan et al. 2018).
We identified the tendency that the mean summer PET was increasing faster than the mean temperature almost everywhere in Russia. This corresponded to the tendency for heat stress that was revealed across Australia using high-resolution ERA-Interim reanalysis data over the period from 1979 to 2010 (Jacobs et al. 2013). Generally, the apparent temperature rose faster than the air temperature, amplifying the expected exposure to discomfort due to global warming in the subtropical region.
In this study, we evaluated the thermal comfort conditions in Russia based on PET and UTCI biometeorological indices and gridded meteorological data from the ERA-Interim reanalysis. The climatological means of the thermal indices and repeatability of the different grades of cold and heat stress were evaluated for the WMO’s current climatological standard normal period (1981–2010); the long-term changes in these parameters were analyzed for the 1979–2018 period.
By analyzing the spatial patterns of the mean PET, UTCI, and temperature fields, we found the mismatch of the winter extreme in terms of temperature and UTCI. While the lowest winter temperatures were found in Yakutia, the lowest winter UTCI values were found on the coast of the Arctic Ocean, where low temperatures were combined with high wind speeds. In summer, the fields of the mean temperature and PET were almost similar.
For the first time, we evaluated the repeatability of the different grades of cold and heat stress in Russia. We found that in winter, nearly the whole area of Russia was permanently subject to at least high cold stress according to the UTCI index, and more than 50% of the territory was permanently under at least very high cold stress, and even extreme cold stress was quite frequent, with a country-mean repeatability of 46%. In summer, days with at least high heat stress according to the PET index were quite frequent south of 60°N, with the probability of such days reaching 80% in the south of Russia. However, the Arctic and northern regions of Russia were almost invulnerable to high heat stress. We registered very few days with stronger heat stress, with a country-mean repeatability not higher than 3%.
Analysis of the summer and winter trends of the two different thermal comfort indices and air temperature showed spatial inhomogeneity of their changes. In summer, warming trends were observed all over the country except for western Siberia, with hot spots in its southwestern parts and in southern Siberia. In winter, warming was observed in the northwestern part of the country in the Arctic. The warming trends in terms of thermal indices generally followed the temperature trends; however, they could have different slopes and significance due to the contributions of other meteorological factors. In winter, decreasing wind speed led to an increase in the mean UTCI in the European part of Russia and in western Siberia, while the summer decrease in cloud cover led to an increase in PET in the European part of Russia and southern Siberia.
The changes in the mean thermal indices did not always correspond to the changes in heat or cold stress repeatability. For example, the intense wintertime warming in the Arctic was not reflected in changes in the cold stress repeatability. The most pronounced changes in the very high cold stress repeatability were observed in the European part of Russia, especially in the northwest of the country, while the most pronounced changes in the summertime high heat stress repeatability were observed in the European part of Russia and in the southern parts of Siberia and Far East. In summary, we can conclude that the European part of Russia is most susceptible to changes in bioclimatic conditions since it experiences both changes in the winter frequency of cold stress and changes in the summer frequency of heat stress.
From a methodological point of view, an important conclusion of the study was the difference between the spatial patterns of contemporary climate conditions and their changes in terms of mean temperature, mean values of bioclimatic indices and the repeatability of heat or cold stresses. Such results clearly illustrate the importance of selecting appropriate indicators for applied tasks related to thermal stress assessment and analysis.
This research was funded by the Russian Science Foundation (Grant 17-77-20070 “Assessment and Forecast of the Bioclimatic Comfort of Russian Cities under Climate Change in the 21st Century”).