The heat wave duration index (HWDI) averaged at the country level and the fraction of the country affected by a heat wave (AREA) are shown in Fig. 2. This figure reveals that there is strong interannual and decadal variability throughout all summer months (Fig. 2 – left column). For June, there is a statistically significant increase in both HWDI (Fig. 2a and Table S1) and AREA (Fig. 2b), with a much higher frequency of both after the beginning of the 1990s. The longest heat wave was recorded in June 2019 and lasted 10 d, and more than 90 % of the country was affected. Until the beginning of the 1990s there were relatively few HWs, most of them observed between 1960 and 1970, but their duration was much shorter compared to the events recorded from 2000 onwards. Also, in terms of the affected area, after the 1990s most of the heat waves had a larger spatial extent, with an area covered by a HW of more than 80 % in 1996, 2002, 2003, 2010, 2012, and 2019, respectively.

As in the case of June, for July we also observe a statistically significant increase in both the HWDI (Fig. 2c and Table S1) and the AREA (Fig. 2d). At the beginning of the analyzed period (i.e., 1950–1960), there were heat wave events lasting on average 4–5 d (when averaged at the country level) and covering an area up to 80 %. Between 1970 and 1985 no HWs were recorded throughout the country. After 1985 there is a steep increase in the duration of the HWs, with the longest HWs in July 2007 and 2012, when the whole country was affected (i.e., AREA = 100 %). The years 1987, 2002, 2007, 2012, and 2015 have been characterized by HWs with a spatial coverage of more than 80 % (Fig. 2d). For August, the temporal evolution of HWDI (Fig. 2e) and AREA (Fig. 2f) follows the same path as June and July, meaning a significantly increasing trend in both the duration and AREA (Table S1). Over the period 1964–1988 no HWs were recorded in August, while most of the longest and extended HWs were recorded in the last 2 decades of the analyzed period. The longest HW recorded in August was in 2015, followed by 2012, 1992, and 1952. In 1992, 2012, and 2015 the area covered by HWs was higher than 90 % (Fig. 2f). For all analyzed months, the HWs recorded in the last 2 decades were both longer and had a higher spatial extent. If we analyzed all the summer months taken together (JJA), we have a very clear picture (Fig. 2g and h): the period 1950–1970 was characterized by the occurrence of HWs with a duration varying between 3 and 10 d, averaged at the country level, and a spatial extent between 20 % and 80 %, followed by a relatively HW-free period between 1971 and 1985. After this period there was a significant increase in the duration of the HWs, and most of them reached a spatial extent of more than 50 %, especially over the last 2 decades (i.e., 2001–2020).

Since the number of HWs per year is small, especially in the first half of the analyzed period (i.e., 1951–1985), we have aggregated the number of heat waves in decades, to be able to analyze the spatiotemporal changes in their occurrence. We have performed the decadal analysis for each summer month separately (Figs. S1, S2, and S3) and for the whole summer season (JJA) as a whole (Fig. 3). We focused our analysis in this way, to have an equal number of months per decade and also to provide decadal evolution of HW hotspots at the country level. The first analyzed decade is 1951–1960, followed by 1961–1970, and so on until 2011–2020. Figure 3 shows the geographical distribution of the number of HWs per decade summed over the summer months. Overall, there is an increased variability among different regions of the country depending on the analyzed decades. Over the decade 1951–1960 up to 24 HWs per decade were recorded in the southeastern part of the country (i.e., the Dobruja region), while in the northwest part of the country up to 10 HWs per decade were recorded (Fig. 3a). Over the decade 1961–1970, up to eight HWs per decade were recorded mainly in the intra-Carpathian region (i.e., the northwest part of the country) (Fig. 3b). The decade 1971–1980 was almost HW free (Fig. 3c), while for the decade 1981–1990 there were fewer than two HWs per decade at the country level (Fig. 3d). Starting with the decade 1991–2000, the number of summer HWs started to increase all over the country (Fig. 3e). During the 2001–2010 decade, the HW hotspots developed in the western part of the country and the Dobruja region (i.e., southeastern part of the country) (Fig. 3f). Over the decade 2011–2020, there were up to 24 HWs per decade, the most affected areas being the northwest, inside the Carpathian Mountains, and the southeastern part of the country (Fig. 3g). Overall, there were up to 6 times more HWs in the last decade compared to the HWs at the beginning of the analyzed period.

When looking at the decadal distribution of HWs hotspots for each summer month separately, there are some clear differences, especially at the beginning of the analyzed period (Figs. S1, S2, and S3). Over the decade 1951–1960, there were up to five HWs per decade in July (Fig. S2a) and August (Fig. S3a), focused in the northwestern part of the country and the most southeastern corner of the country. In June, a limited number of HWs have been recorded in this decade (about two HWs per decade) over the eastern part of the country (Fig. S1a). The decade 1961–1970 was characterized by up to four HWs per decade in June, over the north and northwestern part of the country (Fig. S1b), while in July (Fig. S2b) and August (Fig. S3b) one HW per decade was recorded in the western part of the country. The decade 1971–1980 was HW free in July (Fig. S2c) and August (Fig. S3c), while in June there was about one HW per decade over a small part of the country (Fig. S1c). The decade 1981–1990 was characterized by up to two HWs per decade, at the country level, in July (Fig. S2d) and August (Fig. S3d). Starting with the 1991–2000 decade, the number of HWs per decade starts to increase at the country level, the most affected months being June (Fig. S1e–g) and August (Fig. S3e–g). Over the decade 2001–2010 there were up to seven HWs per decade recorded in the south and southeastern part of the country in June (Fig. S1f), up to six HWs per decade in the western part of the country and the Dobruja region in July (Fig. S2f), and up to 10 HWs per decade in August, with a focus on the Dobruja region (Fig. S3f). For the last decade (i.e., 2011–2020) the number of HWs per decade increased in all months, but their spatial distribution differs. In June (Fig. S1g), the highest number of HWs per decade was recorded over the northwestern part of the country (up to 10 HWs per decade) and over the Dobruja region. In July, the HW hotspots are over the northern and eastern parts of the country (Fig. S2g), while in August there is a homogenous distribution of up to 10 HWs per decade, throughout the country (Fig. S3g).

From the decadal analysis of the number of HWs, we can clearly state that the decade 2011–2020 was characterized by a significant increase in the number of HWs compared to the previous decade, this increase being the strongest in August. There are preferred hotspots for the HW occurrence, depending on the analyzed decade and month, with] these hotspots being strongly influenced by the geographical distribution of the Carpathian Mountains. The most affected regions by the HW occurrence are the north and northwestern part of the country and the Dobruja region. Dobruja is a region which has been subjected to a significant increase in the mean air temperature and a significant decrease in the summer precipitation (Chelcea et al., 2015; Prăvălie et al., 2017). Overall, there is a significant increase, of ∼ 0.2 HWs per decade in June, over most parts of the country, except some small regions in the northeastern part (Fig. 4a). In July a significant increase of ∼ 0.1 HWs per decade can be observed in the northern part of the country, while for the rest of the country no significant changes have been recorded (Fig. 4b). In August, there is a significant increase in the number of HWs over Romania, especially over the eastern part of the country (∼ 0.2 HWs per decade) (Fig. 4c). When we consider all summer months together, the increase in the number of HWs is significant at the country level, with an increase of up to 0.4 HWs per decade in the eastern part of the country (Fig. 4d).

To analyze the variability and trends of drought conditions, at the country level, we performed a similar analysis like in the previous section: we averaged the SPEI at the country level (Fig. 5), we performed the decadal analysis (Fig. 6), and we performed the trend analysis (Fig. 7). The temporal evolution of June SPEI1 (Fig. 5a), July SPEI1 (Fig. 5c), August SPEI1 (Fig. 5e), and August SPEI3 (Fig. 5g) indicates a strong interannual variability of the drought conditions, at the country level and a statistically significant drying trend for August SPEI1. For June, the driest years, in terms of both amplitude (Fig. 5a) and spatial coverage (AREA, Fig. 5b), were 1950, 1968, 2003, 2006, and 2012. For July, the driest years were 1952, 2007, 2012, and 2015 (Fig. 5c). For these years, the drought conditions extended to more than 60 % of the country (Fig. 5d). In August, the driest years, at the country level were recorded in 1952, 1992, 2000, 2003, and 2018 (Fig. 5e), with the drought conditions covering more than 60 % of the country (Fig. 5f). Moreover, in the case of August SPEI1, only negative values were recorded for 13 consecutive years, from 2008 until 2020. August SPEI3, which is an indicator of drought conditions over the whole summer, indicates that the driest summers, over Romania, were 1950, 1952, 2000, 2003, 2012, 2015, and 2018 (Fig. 5g). For all these summers, drought affected more than 70 % of the country (Fig. 5h).

The drought hotspots, at a decadal scale (Fig. 6), indicate strong spatiotemporal variability between the different analyzed decades and between different regions of the country. Over the 1951–1960 decade (Fig. 6a), the drought hotspots (defined as the number of months per decade when August SPEI3 ≤-1 for each grid point) was focused in the northeastern part of the country. For this period, there were up to three summers per decade characterized by drought conditions over these regions. For the decade 1961–1970 (Fig. 6b) there was a relatively limited number of dry summers (about two dry summers per decade) throughout the country, mostly focused on the northwestern part and south part of the country, while the decade 1971–1980 (Fig. 6c) was drought free. For the decade 1981–1990 (Fig. 6d), there is a rather homogenous pattern at the country level, with up to two (one) dry summers per decade affecting the western (eastern) part of the country. The decade 1991–2000 (Fig. 6e) also indicates a rather homogenous pattern of the drought conditions at the country level, with up to four dry summers per decade in the western part of the country and three dry summers per decade over the rest of the country. The decade 2001–2010 is characterized by an west–east gradient in the drought conditions, with the highest number of dry years (∼ 4) in the southeastern part of the country (Fig. 6f). Over the 2011–2020 decade, the drought hotspots are located mainly over the western part and the southeastern part of the country (Fig. 6g). The highest number of dry summers per decade were recorded throughout this decade (i.e., 2011–2020), with up to six dry summers per decade over the whole western part of the country and over the southeastern part. Overall, the decadal spatiotemporal evolution of the drought conditions (Fig. 6) indicates that drought events are not homogenous throughout the country and that the decades with the highest number of dry summers were 1991–2000, 2001–2010, and 2011–2020. A similar pattern is observed when looking at the SPEI trends, on both monthly (Fig. 7a–c) and seasonal timescales (Fig. 7d). Overall, in June there is a non-significant drying trend over the northwestern and southeastern parts of the country (Fig. 7a and Table S2). In July there is an overall non-significant drying trend over the southwestern and southeastern parts of the country (Fig. 7b), while in August, the spatial trend pattern is rather distinct compared to June and July, being characterized by a general drying trend at the country level, but significant only over the eastern part of the country (Fig. 7c). August SPEI3 trend, which takes into account all summer months, is a combination of the features identified for each month analyzed separately: a drying trend over the whole country, but statistically significant only over the southeastern part of the country (i.e., Dobruja region) (Fig. 7d).

In this sub-section, we analyze the co-variability between hot and dry summers in terms of lagged and in-phase spatial correlation maps between the monthly/seasonal HWDI and monthly SPEI (Fig. 8) and conditional probability maps (Fig. 9). The lagged correlation maps (SPEI leading) have been computed and analyzed in order to test for possible influence of the pre-conditions of dry spring and/or early summer on the occurrence of summer heat waves. To find the best combination, in terms of compatible months for both hot and dry events, we have computed the spatial correlation maps between the monthly SPEI and the monthly HWDI with different time lags (e.g., SPEI leading). For example, June HWDI was correlated with April SPEI1, SPEI3, and SPEI6 (not shown); May SPEI1, SPEI3, and SPEI6 (not shown); and June SPEI1 (Fig. 8a), SPEI3 (not shown), and SPEI6 (not shown). The highest correlations, in terms of both amplitude and spatial extent, have been obtained for the combination of June HWDI and June SPEI1 (Fig. 8a). This finding is also confirmed by looking at the correlation coefficient, both with lag and in phase, between the HWDI index averaged at the country level (i.e., the time series in Fig. 2a) and different combinations of the monthly SPEI averaged at the country level (Fig. S4). Also, for the country-averaged time series, the highest correlation was obtained for June HWDI and June SPEI1. In terms of spatial extent, the highest correlations (e.g., ∼ 0.5) have been obtained for the western and southern parts of the country (Fig. 8a). Nevertheless, a significant correlation between HWDI and SPEI over these regions does not necessarily imply that each drought will always co-occur with a heat wave over that region or vice versa. The conditional probability map for June, which implies the probability of co-occurrence of both a dry (June SPEI ≤-1) and a hot (June HW > 1) month, indicates the most prone regions of a combined hot and dry June are the areas located in the eastern part of the country (Fig. 9a). For July, the highest correlation has been found also for the in-phase relationship (i.e., July HW and July SPEI1) (Fig. 8b). Compared to June, in July the spatial correlations between July HW and July SPEI1 are significant all over the country, with the highest amplitudes over the extra-Carpathian regions and over small areas in the northwest. This dipole-like structure clearly emphasizes the influence of the Carpathian Mountains on the climate variability of Romania, which is in agreement with previous studies over this region which have shown that the Carpathian Mountains play a significant role in the hydroclimatic variability of the country (Busuioc et al., 2015; Ionita, 2015). In terms of co-variability, the most prone regions for a combined hot and dry July are, as in the case of June, the areas located in the eastern part of the country (Fig. 9b). In August, the highest correlations, based on both the amplitude and the spatial extent, have been found between August HWs and August SPEI3 (Figs. 8c and S4). The correlations between August HWs and August SPEI3 reached values up to -0.7 for almost the whole country, with small exceptions in the southeastern corner of the country. The occurrence of HWs in August seems to be influenced not only by the hydroclimatic conditions in August, but also by the previous months. The conditional probability map (Fig. 9c) indicates that hot and dry events have a higher probability of occurrence compared to June and July, and the risk of hot and dry events is distributed all over the country, with small exceptions in the mountain areas (i.e., Apuseni Mountains and Retezat Mountains). When we look at all the summer months together, the highest correlations are obtained between JJA HWs and August SPEI3. The spatial correlation map between JJA HWs and August SPEI3 reaches values up to ∼ -0.8 for almost the entire country (Fig. 8d) and a value of -0.71 when we averaged at the country level (Fig. S4). In terms of co-variability, the most prone regions for combined hot and dry summers are the areas located in the eastern part of the country (Fig. 9d). When considering the 1-month lag, there are still some regions showing significant negative correlations between SPEI and HW, mainly for July HW and June SPEI1 over the southern part of the country and August HW and July SPEII1 over the western part of the country (not shown). Although the lagged correlations are smaller in amplitude compared to the in-phase correlations, the lagged relationship indicates that there might be some influence of the drought conditions on the heat waves in the upcoming months, especially in the case of July and August HWs.

The analysis of the temporal variability of the HWDI and AREA (Fig. 2) has emphasized some extreme HWs for each analyzed month, in terms of both duration and spatial coverage. Thus, in this sub-section, we make a detailed analysis of the longest HW for each month, in terms of extremeness (e.g., rank maps) and large-scale driving factors. The analysis is focused on three distinct cases: July 2012, August 2015, and June 2019.

July 2012 was marked by persistent heat waves, which have determined extremely high temperatures at the beginning of the month in the western part of the country, afterwards extending to all regions, but especially in the plain and plateau areas (Fig. 10a). In some regions of the country (e.g., eastern and central part) the duration of the HWs was up to 24 d. In terms of drought, most of the country was affected by moderate to extreme drought in July 2012 (Fig. 10b), with small exceptions in the northwestern part of the country. July 2012 was the hottest month on record (e.g., over the period 1950–2020) over most of the country (Fig. 10c). In July 2012, 114 meteorological stations through the country recorded temperatures above 35 ∘C (Dima et al., 2016). Over the central part of the country, from the south to the north, July 2012 was both hot and dry (Fig. 10d). The peak of the heat waves was recorded in the first week of the month (Fig. S5). Starting with 2 July, the atmospheric circulation was characterized by a northeasterly flow, which led to an advection of warm air masses, generated over Russia (Fig. S6). At the country level, this large-scale atmospheric pattern resulted in the establishment of an excessive thermal regime and an increase in the number of hot days (i.e., daily temperatures > 35 ∘C), especially in the southern and eastern regions (Fig. 10e and S5). Between 7 and 9 July 2012, the daily maximum temperature up to 10 ∘C was higher, compared to climatology, especially in the eastern part of the country (Fig. S5f–h). These excessive temperatures were driven by the persistence of a high-pressure system over the eastern part of Romania and the presence of an atmospheric blocking center over Fennoscandia and the western part of Russia (contour lines in Fig. S6).

The heat wave and drought event observed throughout the summer of 2015 affected a large portion of continental Europe and was one of the most severe dry and hot summers over the observational period (Ionita et al., 2017; Laaha et al., 2017; Van Lanen et al., 2016). Record high temperatures were observed throughout the whole summer over different parts of Europe. Extremely high temperatures already started to be recorded in June 2015 over the Iberian Peninsula, central and eastern France, the western Alps, and Ukraine. The heat wave and drought conditions extended towards the central part of Europe in July 2015 (Ionita et al., 2017). By August 2015, the heat wave moved and continued to develop in central and eastern Europe, including Romania. For most of the month of August 2015, Romania was under the influence of extremely high temperatures. The first heat waves occurred between 3 and 16 June (not shown). Between 17 and 23 August, a short relief was observed, with temperatures below the climatological mean (not shown). The second and most intense heat wave (e.g., in terms of the temperature anomalies) started to develop on 24 August, reaching its peak at the end of the month (Fig. S7). The longest heat wave was recorded over the northern and eastern parts of the country (Fig. 11a). In some regions in the eastern part of the country, there were up to 24 d which fulfilled the HW definition. Overall, the drought conditions in August 2015 were not as intense as in July 2012. Only the northern part of the country experienced both heat wave and drought at the same time (Fig. 11b and d). August 2015 was also the hottest month on record (e.g., over the period 1950–2020) in the northern and northeastern parts of the country (Fig. 11c). The extremely high temperatures recorded, especially in the last week of August 2015, were mainly driven by the prevailing large-scale circulation. The two long-lasting heat waves in August 2015 were determined by the extension of the North African ridge over most of the European continent (Figs. 11e and S8). During the peak of the second heat wave (i.e., 28–31 August 2015) the eastern part of Europe was affected by a persistent atmospheric blocking system (contour lines in Fig. S8), which was centered over Romania, and by positive values of the water vapor flux divergence (not shown). The anomalous Z500 center over the eastern part of Europe (Figs. 11e and S8) and the divergent water vapor flux over Romania suggest a dominant subsidence and adiabatic warming, reduced cloudiness, and increased incoming solar radiation, thus leading to excessive temperatures over the affected regions.

For the month of June, the longest and largest (in terms of spatial extent) HW event was recorded in June 2019 (Fig. 2e–f). According to Copernicus (https://climate.copernicus.eu/surface-air-temperature-june-2019, last access: 17 December 2021) June 2019 was the hottest June on record both globally and for Europe, with central and eastern Europe being particularly warm throughout the whole month. In June 2019, the northwestern and southeastern parts of Romania were regions the most affected by extreme temperatures (Fig. 12a and c). Record-breaking temperatures were recorded in the most northern part of the country as well as in the Dobruja region (Fig. 12c). These record-breaking temperatures were corroborated with drought conditions (Fig. 12b and d). The eastern, central, and southwestern parts of the country were less affected by extreme temperatures (Fig. 12a and c), and these regions were characterized by wet conditions throughout the month (Fig. 12b). The particular spatial pattern was mainly influenced by the spatial pattern of the large-scale atmospheric circulation (Fig. 12e). The atmospheric circulation at the peak of the heat wave event (Figs. S9 and S10) was characterized by a persistent wave-like pattern extending from the North Atlantic Ocean towards Eurasia (Figs. 12e and S10). Positive (negative) geopotential anomalies were observed over eastern Europe (central North Atlantic and central Siberia), corresponding to the local positive (negative) temperature anomalies underneath (Figs. 12e and S9). The spatial structure of the Z500 field resembles the classic omega blocking circulation (Fig. S10 – contour lines). This pattern favors the advection of warm air from the Sahel towards the eastern part of Europe and enhances the incoming solar radiation, leading to extremely high temperature anomalies underneath the high-pressure system.

All analyzed extreme HWs in this section were mainly driven by the presence of a high-pressure system over the analyzed region, during the peak of the HW event. In order to identify if the presence of a persistent high-pressure system is a necessary ingredient for all HWs identified throughout the period 1950–2020, we have computed the composite maps for the years when the HWDI index (Fig. 2 – left column) was > 5 d and the corresponding Z500 anomalies and the corresponding wind vectors. We performed the analysis for each month separately (Fig. 13). Due to the fact that the relationship between the large-scale atmospheric circulation and the European hydroclimate was found to be limited due to non-stationarity issues (Ionita et al., 2020; Rimbu et al., 2004; Vicente-Serrano and López-Moreno, 2008), we have also computed the stability maps between the HWDI and the monthly Z500. The aim of the composite map analysis is to analyze the relationship between the HWDI and the large-scale atmospheric patterns, but this methodology does not consider whether the relationship between the two variables is stationary in time or not. In order to overcome the problem of non-stationarity and to test if the identified relationship between the HWDI and Z500 is stable over time, we employed a methodology, namely the stability maps, used for the monthly to seasonal prediction of the mean runoff of the Elbe River and in dendroclimatological studies (Ionita et al., 2015; Nagavciuc et al., 2019).

The June composite map of Z500 anomalies and the corresponding wind vectors for years with HWs lasting more than 5 d is characterized by positive Z500 anomalies over the central and eastern parts of Europe and negative Z500 anomalies over the central North Atlantic Ocean (Fig. 13a). Moreover, HWs in June, in Romania, are also associated with an increase in the number of atmospheric blocking days, centered over the southeastern part of Europe (Fig. 14a). The spatial structure of the Z500 anomalies, centered over the eastern part of Europe, leads to the advection of hot and dry air from the northeastern part of Europe. The large-scale atmospheric circulation associated with HWs over Romania, in July, is similar to the spatial structure identified in June, in both the Z500 field (Fig. S13b) and in the case of 2D atmospheric blocking (Fig. 14b). In August, the spatial structure of the Z500 field, associated with the occurrence of HWs over Romania, is characterized by a wave-train like pattern of alternating Z500 anomalies, which extends from the eastern part of the US until Eurasia (Fig. S13c). Extreme HWs, in August, are associated with a low-pressure system over the eastern part of the US, followed by positive Z500 anomalies over the western part of the central North Atlantic Ocean, negative Z500 anomalies centered over the British Isles, and positive Z500 anomalies over the central and eastern parts of Europe. This wave-like pattern suggests a stationary Rossby wave pattern, which is usually associated with heat waves and droughts over the Eurasian continent (Bakke et al., 2020; Barriopedro et al., 2011; Ionita et al., 2012; Schubert et al., 2014). As in the case of June and July, HWs in August are also associated with an increased frequency of atmospheric blocking over the eastern part of Europe (Fig. 14c). The significant relationship between the HWDI and Z500 obtained via the composite map analysis is also confirmed by the stability maps. June HWDI is stably and positively correlated with June Z500 over the eastern part of Europe, centered over Romania (Fig. 15a). The same pattern can also be observed when we compute the stability map between July HWDI and July Z500 (Fig. 15c). In the case of August, the HWDI is stable and positively correlated with Z500 over a region extended from the North Atlantic basin towards the central and eastern parts of Europe and negatively correlated with Z500 centered over the British Isles and North Ocean (Fig. 15e). This dipole structure is reminiscent of the east Atlantic teleconnection pattern, which was found to have a significant influence on the variability of temperature and precipitation over Europe, throughout the whole year (Gao et al., 2017). Based on the monthly stability maps identified in Fig. 15, we defined a Z500 index averaged over the stable regions (black squares in Fig. 15a, c, and e) to analyze the interannual variability of the Z500 over these regions in a long-term context. This analysis was motivated by the fact that it has been suggested that the Z500 over the central and western parts of Europe has increased recently, leading to an increase in the frequency of HWs over these regions (Porebska and Zdunek, 2013; Tomczyk and Bednorz, 2016). The June Z500 index exhibits strong interannual variability over the last 70 years, with the highest amplitudes since the beginning of the 1990s (Fig. 15b). Notably, the highest value of this index was recorded in 2019, which is also the month with the longest June heat wave (Fig. 2a). Over the period 1990–2020 there is a significant increasing trend in the June Z500 averaged over the eastern part of Europe, a trend which closely resembles the one observed for the June HWDI (Fig. 2a). The results of the trend analysis for each month and each analyzed period are given in Table S3. As in the case of June, the July Z500 index also exhibits strong interannual variability over the last 70 years and a significant increasing trend from the 1990s onward (Fig. 15d). The highest values of this index were recorded in 1954, 1987, 1988, 2007, 2012, and 2015. July 2012 is also the month with the longest July heat wave over the analyzed period (Fig. 2c). The time series of the August Z500 index also exhibits strong interannual variability over the last 70 years and a significant increasing trend over the period 1990–2020 (Fig. 15f). The highest value of this index was recorded in 1952, 1962, 1992, 2010, 2015, 2017, and 2019. August 2015 is also the month with the longest August heat wave (Fig. 2c). Overall, the time series of the monthly Z500 presents a strong interannual variability and a significantly increasing trend starting with the beginning of the 1990s, which mirrors the trends observed in the monthly HWDI (Fig. 2). For July and August, the trend of the Z500 indices is significant for both analyzed periods (i.e., 1950–2020 and 1990–2020), while for June the trend is significant only when we consider the 1990–2020 period (Table S3).