Temperature extremes and circulation types in the Czech Republic, 1961–2020

2.1 Temperature data

The basic datasets used in this study consisted of the daily maximum (TMAX) and minimum (TMIN) temperatures measured at 133 climatological stations (Figure 1) run by the Czech Hydrometeorological Institute (CHMI), covering the territory of the CR during the 1961–2020 period. All of stations selected had kept at least 40 years of observations; these were extended and complemented by homogenization procedures to 60 years (Zahradníček et al., 2021).

2.2 Classification of circulation types

The objective classification of circulation types used to analyse synoptic patterns contributing to temperature extremes was based on flow strengths, flow direction, and vorticity (Jenkinson and Collison, 1977; Plavcová and Kyselý, 2011). Circulation types were calculated at daily resolution, using sea-level pressure (SLP) at 16 points (Figure 2) with respect to the geographical centre of the CR, coordinates 49.74°N and 15.33°E. The SLP data from the NCEP/NCAR reanalysis (Kalnay et al., 1996) were employed for calculation of circulation indices for the 1961–2020 period. A SLP value for a given point was obtained through an inverse distance-weighted mean of the nine closest NCEP/NCAR grid points. The classification defines 27 circulation types, of which 9 are anticyclonic (A, AN, ANE, AE, ASE, AS, ASW, AW, ANW), 9 cyclonic (C, CN, CNE, CE, CSE, CS, CSW, CW, CNW), and 8 directional (N, NE, E, SE, S, SW, W, NW). Days remaining unclassified due to low flow strength (<3) and vorticity (<3) were designated type U (for more details of this classification and visualization of the individual types, see Řehoř et al., 2021).

4.1 Circulation patterns in 1961–2020

The occurrence of individual circulation types and their groups appear in Table 1 in terms of the mean relative annual frequencies for the CR in the 1961–2020 period. Anticyclonic types were the most frequent, at 43.4% (A 10.0%, AW 8.0%, and ANW 5.8% were the three most frequent) followed by directional types, at 38.1% (W 7.8%, SW 6.4%, and NW 4.8%). The percentage of cyclonic types reached only 16.9% (C 3.1%, CSE 2.3%, and CS 2.2%). A total of 1.6% of days remained unclassified (type U). Linear trends for individual circulation types and their groups (statistically significant at the 0.05 level) indicated rising trends for anticyclonic types (6.6 days/10 years) and four individual types (A, AN, ASW, and AW). Decreasing trends were significant for cyclonic types (−4.8 days/10 years) and four individual types (CNE, CE, CSE, and CW). Trends in directional types were insignificant, as were for individual types.

Because a number of temperature extremes are almost entirely limited to the SHY and WHY periods, Figure 3 shows the relative frequencies of individual circulation types in these two half-years separately. The relative frequencies of types AS, ASW, AW, SE, S, SW, and W in the WHY were significantly higher, as were types A, AN, ANE, C, CN, CNE, CE, CSE, N, NE, and U in the SHY. Turning to significant linear trends for circulation types in 1961–2020, differences between the two half-years were substantial (not shown). In the WHY, significant trends increased for only AW and ASW (for all anticyclonic types 2.5 days/10 years) and decreased for CSE. In the SHY, significant increases appeared for types A and ASW (for all anticyclonic types 4.1 days/10 years) and decreases for CNE, CE, CSE, CW, and CNW (for all cyclonic types −3.5 days/10 years).

4.2 Maximum and minimum temperatures

Series of mean annual ATMAX indicate a significant increasing tendency in the CR during the 1961–2020 period, with a linear trend of 0.5°C/10 years (Figure 4a). The lowest mean ATMAX figures emerged for 1977–1978, after which a sharp increase took place, to the highest value in 1983 (the difference between the lowest and highest ATMAX reached 7°C). The extreme ATMAX for 1983 was slightly surpassed in 2013 and 2015. In similar fashion, mean annual ATMIN also shows a significantly increasing linear trend of 0.7°C/10 years (Figure 4b). The difference between the two most extreme ATMINs, in 1974 (maximum) and 1985 (minimum), reached 18.1°C. The values for the annual range between ATMAX and ATMIN exhibited an insignificant decreasing trend (Figure 4c).

In the relationship between circulation types and ATMAX/ATMIN, anticyclonic circulation types predominated markedly for ATMIN, at 66.7%, while the percentages for directional and cyclonic types (28.9 and 4.3%, respectively) were deeply below their 60-year means (cf. Table 1). Further to type A (central anticyclone), at 25.1%, the most important types featured cold airflow from eastern Europe (AE 8.9%, E 7.9%, and SE 7.6). For ATMAX, the share of anticyclonic types was below the 60-year mean (37.2%), while cyclonic types (16.8%) and directional types (37.2%) were close to their means. Most frequent were types with warm airflow from the southwest and south (SW 10.6%, ASW 9.8%, and S 9.3%); this also held for unclassified type U, at 8.8%.

When only the highest ATMAX for each of the 133 climatological stations during the 1961–2020 period were selected, it became evident that they were distributed into 18 days in July–August: at 48 stations (36.1%) occurring on July 27, 1983 (type A), at 25 stations (18.8%) on August 8, 2013 (type CW), and at 11 stations (8.3%) on August 8, 2015 (type CSE), always as a parts of HWs. The lowest ATMINs at each of 133 stations occurred on 19 different days during December–February. Related extremes over three consecutive days predominated: at 67 stations (50.4%) on 7–9 January 1985 (types AE, A, A; for 31 stations, that is, 23.3%, on 7 January) and at 33 stations (24.9%) on 12–14 January 1987 (types CNE, SE, SE).

As follows from Figure 5, which shows fluctuations in relative annual frequencies of the occurrence of extreme daily ATMAX and ATMIN, absolute positive extremes concentrated around the second part of 1961–2020, while absolute negative extremes did the opposite. In terms of decadal frequencies, daily ATMAX occurrences were at their highest in 2011–2020 (28.9%; 6.1% in 2015) and lowest in 1971–1980 (9.5%). Conversely, daily ATMIN occurrences achieved a maximum in 1971–1980 (23.7%) and a minimum in 2011–2020 (7.5%). Comparing these annual frequencies with the two “normal” periods of 1961–1990 and 1991–2020, the corresponding values were 31.0 and 69.0% for daily ATMAX, and 68.6 and 31.4% for daily ATMIN.

4.3 Characteristic days

Records of annual means revealed 39.8 summer days and 7.6 tropical days in the CR during the 1961–2020 period; these values decreased with increasing altitude, as documented for the four altitudinal Groups I–IV (Table 2). Series of the mean annual numbers of such days (Figure 6) showed statistically significant increases, with linear trends of 4.4 summer days/10 years and 1.8 tropical days/10 years. These were at their highest for stations located at altitudes of less than 300 m asl (Group I) (5.2/10 years and 2.6 days/10 years, respectively), decreasing with rising altitude (from Groups I–IV); all were, however, statistically significant. Summer days, from May to September, exhibited earlier beginnings and later endings during 1961–2020, as was evident from their negative and positive trends, respectively. These changes were at their most marked for stations in group III (−3.1 days/10 years and 2.6 days/10 years, respectively). The same tendencies also appeared for tropical days occurring, on average, during June–August. Linear trends for these changes were highest for Group I (−4.2 days/10 years and 3.9 days/10 years, respectively). Thus, for both summer and tropical days, a prolongation of the mean period in which they occurred took place (by 18 and 34 days respectively) during the 60-year period.

In comparison with summer and tropical days, the occurrence of days with tropical nights was very low—an annual mean of only 0.25 days in the entire CR during the 1961–2020 period (from 0.5 days in Group I to 0.1 days in Group III), with a maximum in 2015 (2.2 days). However, the related increasing trends were all statistically significant (not shown). Despite the low frequency of these days in the country mean, their occurrence became a serious problem in the larger cities such as Prague and Brno (for the location of places in the CR, see Figure 1), where they were exacerbated by the urban heat island effect, largely in association with HWs (Skalák et al., 2015).

The mean annual numbers of frost and ice days contrasted sharply with the tendencies shown for summer and tropical days, achieving 115.3 and 35.0 days respectively in the CR during the 1961–2020 period. Their mean numbers rose with increasing altitude (Table 2). The significant linear trends for the entire CR were −4.7 frost days/10 years and −2.9 ice days/10 years (Figure 7). In terms of individual altitudinal groups, decreases were at their highest in Group II for frost days (−5.1 days/10 years) and in Group III for ice days (−3.9 days/10 years); Group I figures were lowest for both characteristics. None of the mean beginnings of frost and ice days exhibited significant trends. However, the last occurrence of such days appeared earlier in average terms, constituting a shortening of the mean period in which frost and ice days may occur (by 14 and 25 days, respectively, during the 60-year period).

Examination of circulation types showed that the occurrence of summer and tropical days corresponded to above-mean frequencies of anticyclonic types (46.8 and 47.7%, respectively) and below-mean frequencies of cyclonic (16.2 and 14.8%) and directional (32.8 and 31.7%) types during the SHY. Of the individual types, A and AW made up nearly a fifth of all such days. Also above mean were southerly and southwesterly directions. Days with tropical nights had above-mean frequencies, with cyclonic types (20.6%) and unclassified type U (6.4%) predominant, while anticyclonic (39.5%) and directional (33.4%) types were below their 60-year occurrences in the SHY. AW and SW types appeared most frequently.

Although anticyclonic circulation types remained the most frequent for frost and ice days (44.0% for both), the percentages of all groups of circulation types were close to the 60-year mean of the WHY. However, those with airflow from the northeastern quadrant exhibited above-mean percentages, while those with airflow from the southwestern quadrant lay below mean. Although type A was the most frequent, its frequency was only slightly above the long-term mean.

Notable differences between the two basic groups of days appeared for unclassified type U: for summer and tropical days these were 4.1 and 5.7%, respectively, but for frost and ice days only 0.6 and 0.3%. This is primarily related to substantially higher abundance of the U type in summer due to a tendency towards less distinct air pressure gradients over Europen-Atlantic domain in this season, that is, an unclassified pressure field.

4.4 Heat waves and cold waves

Figure 8 shows fluctuations in the characteristics of HWs over the territory of the CR in the 1961–2020 period. The mean annual number of HW days (Figure 8a) is at its highest in 2015 and in 1994, a fact also reflected in local maxima of smoothed values in the early 1990s and mid-2010s; a significant increasing trend in HW days achieved 1.1 days/10 years. Similar features also appear in the mean maximum duration of HWs (Figure 8b), as well as in mean TMAX, which were at their highest in 2015. Mean duration of HWs extended significantly over the 60 years, by 3.4 days (trend 0.56 days/10 years), while mean TMAX during HWs increased by 1.2°C. With respect to decreases in TMAX with rising altitude, the characteristics shown in Figure 8a,b exhibit decreases from Group I to Group III. As follows from Figure 8c, HWs did not occur at the stations at altitudes of over 900 m asl (Group IV). The year 1994 was exceptional, when all stations in Groups I–III recorded a heat wave. A similar situation also appeared in Groups I and II in 1992, 2007, 2013, and 2015. For the entire CR, altitudinal Groups I–III experienced statistically significant increasing trends for all three HW characteristics mentioned above. Addressing HWs at the 133 climatological stations in the 60-year period reveals that those lasting 3 days constituted nearly half of them (47.9%), followed by 4-day HWs (21.7%) and 5-day HWs (9.9%), i.e., 79.5% altogether. The longest HW, at a length of 19 days, was recorded at the Strážnice station (176 m asl) between 23 July and August 10, 2018, with a mean TMAX of 32.9°C. Anticyclonic types (52.6%) prevailed in this longest HWs, followed by directional (42.1%); of the 11 types that occurred, type E was the most frequent (21.1%), followed by A, AN, ANE, AE, and NE (10.5% each).

Figure 9 shows fluctuations in the characteristics of CoWs over the territory of the CR in the 1961–2020 period. The mean annual number of CoW days (Figure 9a) is at its highest in 1963, followed by three comparable values in 1969, 1996 and 2010. A significant decreasing trend in CoW days stood at −2.6 days/10 years (at its highest in group III, at −3.7 days/10 years). Similar trends also occurred in mean maximum duration of CoWs (Figure 9b), but 1984 is added to 1963 and 1996. This duration became shorter by 6 days for the 60-year period (mean 12 days). Mean TMIN during CoWs exhibited a significant increasing trend (0.35°C/10 years). In accord with decreases in temperatures with rising altitude, the characteristics of CoWs in Figure 9a,b show increases in their values from Group I to Group IV. As follows from Figure 9c, CoWs usually occurred at all stations in all four altitudinal Groups I–IV. The years 1974 and 2020 were exceptions, during which a relatively lower number of stations in Groups I–III recorded a CoW. Some stations located at altitudes of under 600 m asl (Groups I–II) also recorded no CoW in 1988–1989 and 2015. As was the case for the entire CR, altitudinal Groups I–IV experienced significant decreasing trends for all three CoW characteristics mentioned above. Consideration of the frequency of CoWs at the 133 climatological stations over the 60-year period reveals that durations of CoWs were most frequently 9 days (9.8%), 7 days (9.1%), and 8 days (7.8%), that is, 26.7% altogether. The longest CoW, lasting 78 days, was recorded at the Labská bouda station (1,315 m asl) between December 17, 1962 and March 4, 1963, with a mean TMIN of −14.0°C. Anticyclonic circulation types (43.6%) prevailed in this longest of CoWs, followed by directional (37.2%) and cyclonic (18.0%). The most frequent among the 22 individual circulation types that occurred were AE (14.1%), E (11.5%), A (10.3%), and SE (9.0%), that is, 44.9% altogether.

The increased occurrence of anticyclonic circulation types (a difference of 8.7% compared with the 60-year mean for SHY) and unclassified type U (4.2%) is typical of HW days, while cyclonic and directional types were less frequent (differences −6.0 and −6.9%, respectively). In terms of circulation types, with the exception of A (3.5% increase), types with airflow from the eastern sector exhibited the most notable increases (ASE 2.9%, together with ANE, AE, and AS 6.6%; E, SE, and S 4.8% altogether). Notable decreases appeared in the occurrence of types with airflow input from the northwest quadrant (NW −3.7%, N −3.5%, W −2.6%, ANW −2.6%, i.e., −12.4% altogether). Dependence on anomalous circulation types is much less marked for CoW days, a trend that is apparent in smaller differences compared with the 60-year mean for WHY (0.6% for directional, 0.9% for cyclonic, and −1.4% for anticyclonic types). Differences of ≥1% compared with the corresponding means appeared only for types with cold airflow in the winter months: E (1.5%), NE (1.5%), N (1.3%), and ANE (1.1%).

4.5 Warm and cold anomalies

The mean annual number of warm anomaly days, defined with respect to the threshold of the 95th percentile for TMAX in 1961–1990, reached 11.1 days in the CR during the 1961–2020 period (similar means appear in Groups I–III, with 9.0 days in Group IV). In annual distribution, warm anomalies were at their most frequent in summer and at their least in autumn. Annual fluctuations of warm anomaly days showed a clear and significant increasing trend of 2.9 days/10 years (Figure 10a), with a highest annual value of 34.2 days in 2015. Significant trends for the entire CR and the four altitudinal groups occurred in summer and in annual values, and, with the exception of group IV, in winter and summer as well. No significant trend was recorded in spring.

The mean annual number of cold anomaly days, defined with respect to the threshold of the 5th percentile for TMIN, in 1961–2020 in the CR, shows a decreasing linear trend (Figure 10b). This is statistically significant for annual and summer values in the entire CR and all altitudinal groups. A significant decreasing trend also appeared for autumn in Groups I and II. The 60-year mean annual number of cold anomaly days was 4.6 days in the CR, with similar values in Groups I–III; the figure for Group IV was 6.3 days. In annual distribution, the maximum number occurred in winter and the minimum in summer.

Increases in the occurrence of circulation types with warm airflow from the southwest quadrant proved the most strongly associated with warm anomaly days. These consisted of directional types SW (5.5% compared with the 60-year mean), S (2.9%), and W (2.2%), together with anticyclonic types ASW (3.7%) and AW (3.7%), representing an increase of 18.0% overall. Less frequent were circulation types with cold airflow from northern directions (directional types NE −3.1%, N −2.7%, NW −2.3% and anticyclonic types AN −2.5% and ANE −2.0%). For cold anomaly days, there was a strongly marked increase in the occurrence of anticyclonic types: 17.7% compared with the 60-year mean. Particularly involved were types A (difference 8.2%), AE (3.2%), ANE (2.4%), ASE (2.2%), and AS (2.1%), and decreases in cyclonic types by −9.9% (C −2.1%) and directional types by −7.3% (W −4.9% and SW −3.3%).

4.6 Day-to-day temperature change

Day-to-day temperature change served to investigate sudden falls or rises in daily TMAX and TMIN. Mean annual day-to-day change in TMAX for the entire CR in 1961–2020 stood at 2.5°C, with similar values for the four altitudinal groups. The maximum in annual variation appeared in April–May and the minimum in December–January, finding a reflection in related extremes in spring (and summer) and in winter respectively. The majority of the related linear trends were statistically insignificant, although an increasing trend in March for the entire CR and Groups I–III was noted. This situation changed just a little when only the maximum values of day-to-day change were considered. Mean maximum annual change was 11.4°C, with similar values in the four altitudinal groups. Except for August in Group IV, none of the linear trends proved significant. The very highest day-to-day fall in TMAX, of −23.4°C, occurred between December 31, 1978 (type C) and January 1, 1979 (type CNE) at the Prague-Kbely station (282 m asl), while the very highest rise, of 18.4°C, was recorded between January 14, 1968 and January 15, 1968 (type W on both days) for Velké Meziříčí (452 m asl).

The values of mean annual day-to-day change in TMIN remained the same for the entire CR and Groups I and II, at 2.3°C. The maximum in annual variation was achieved in January and the minimum in July, finding reflection in the same extremes in winter and summer respectively. Significant increasing linear trends appeared particularly in May–June, summer, spring, and in annual values. Turning to only maximum values of TMIN day-to-day change, the situation was similar to that for mean values. Mean maximum annual rise was 11.1°C for the entire CR. All related linear trends were insignificant, with the exception of a decrease for September in Group II. The very highest day-to-day falls and rises in TMIN occurred on the same days as for TMAX, values of −25.5°C between December 31, 1978 and January 1, 1979 at the Lysá hora station (1,322 m asl) and 25.0°C between January 14, 1968 and January 15, 1968 in Františkovy Lázně (435 m asl).

Sets of values defined by the thresholds of the 1st and 99th percentiles were created for analysis of extreme day-to-day changes in TMAX and TMIN. On the first days of extreme falls in day-to-day TMAX, there was a notably low frequency of anticyclonic types (16.5% below the 60-year mean), with an above-mean occurrence of cyclonic types (11.8%) and directional types (2.8%). On the second days, a distinctly higher frequency of anticyclonic types (35.4%) emerged, together with a lower frequency of cyclonic types (22.6%) than on the first days. Among individual circulation types, those with airflow from the western quadrant were the most common, often with a clockwise transition (e.g., SW to W, W to NW). On the first days of extreme rises in day-to-day TMAX, anticyclonic types prevailed, at 53.6%, while the frequencies of cyclonic and directional types were below mean (11.0 and 34.7%, respectively). On the second days of these changes, the frequencies of circulation types moved closer to their means, with no notable shifts.

In extreme changes in day-to-day TMIN, the differences between the first and second days were substantial. Extreme decreases started with a very high frequency of directional types (40.0%), followed by anticyclonic types, with 37.9% (5.5% below 60-year mean), and cyclonic types with 21.1% (4.2% above mean). However, on the second days, anticyclonic types took up an overwhelming majority, of 58.7% (15.3% above mean), while the frequency of cyclonic (10.9%) and directional (29.7%) types was below mean. Types with northern airflow were the most common on the first days and type A with 17.5% was the most frequent on the second day. Conversely, extreme rises in day-to-day TMIN began with a very high frequency of anticyclonic types (54.8%); on the second days this decreased to 33.6%, while the frequency of cyclonic and directional types increased. Among individual types, those with westerly and southerly airflow were the most common.

4.7 Comparison of normal periods

In order to compare two normal periods 1961–1990 and 1991–2020, the temperature characteristics may be roughly divided into three main groups, expressing “warm” events, “cold” events, and “extreme” temperatures (Figure 11a–c). Significant differences between 30-year means appeared in all characteristics, except ATMIN, CoW days, and TMAX day-to-day changes. Variance increased, largely in 1991–2020, except in ATMIN, ice days, CoW days, and cold anomaly days. However, these increases were significant only for numbers of tropical days, days with tropical nights, HW days, and warm and cold anomaly days. The same results were confirmed with the non-parametric Mann–Whitney U test with the only exception concerning of numbers of ice days (not significant difference according to Mann–Whitney test). The two normal subperiods show mostly different linear trends. However, their slopes are statistically significant only in case of the number of summer days.

In the light of recent rising temperatures, the box-plots naturally characterize quite opposite changes in the distribution of “warm” (Figure 11a) and “cold” (Figure 11b) events. Closer examination reveals that the shift towards higher (more frequent) values is more evident for “warm” events. For instance, the 1991–2020 boxes (representing the interquartile range, i.e., 50% of all values) stand clearly above the 1961–1990 boxes for most of “warm” events. The same trend appears for ATMAX. A similar shift of “cold” event distribution to lower (or less frequent) values is less marked, especially in median values (Figure 11b for ice days and CoWs). The lesser change in ATMIN distribution accords with this finding. Finally, the 1991–2020 normal exhibited clearly higher positive asymmetry in most of its “warm” event characteristics. This skewness was not so marked for the remaining temperature characteristics.

Based on data from the 133 climatological stations, differences between the two 30-year periods may be described from a spatial point of view (Figure 12). In general, they reflected a dependence of extreme temperatures on altitude and orography. Lower positions tended to return higher positive differences (i.e., in 1991–2020 compared to the previous 30-year period) for numbers of summer days and tropical days (Figure 12a,b) and lower negative differences for numbers of frost and ice days (Figure 12c,d). Higher positions, consisting largely of the mountains around the borders of the CR, showed an opposite tendency. In the spatial expression of differences in TMAX, the greatest increase appeared in the eastern part of the CR and the least in its southwestern part (Figure 12e). In particular, the least increases in TMIN during 1991–2020 appeared north and northeast of a line connecting Brno and Prague; the greatest increases occurred in a broad belt extending from southern to western Bohemia, with a core region around Plzeň (Figure 12f).

In terms of circulation, the relative frequencies of anticyclonic types increased by 5.0% in 1991–2020 compared with the previous 30 years. This was balanced out by a decrease in the occurrence of cyclonic types, by 3.4%, and directional types, by 1.6% (Table 3). All these changes were statistically significant (significance level 0.05). Among individual circulation types, the highest increases in relative frequencies were recorded for anticyclonic types A (1.6%), AW (1.3%), and ASW (1.1%), while the highest decreases occurred in types CSE (−1.1%) and NW (−0.8%). Differences in the relative occurrence of these five circulation types were significant, as were the positive shifts for types AN and ASE, and negative for CNE, CE, CS, and CW. This shows that changes in the occurrence of anticyclonic and cyclonic circulations types are important to changes in trends of temperature characteristics analysed. Since types A and ASW are associated with ATMAX, summer days and tropical days, the increased frequency of these circulation types in the SHY is apparently an important driver of the more frequent occurrence of warm and hot days.