Warm‐Season Drying Across Europe and Its Links to Atmospheric Circulation

We study drying trends across the central latitude strip of Europe (47.5–52.5°N and 2.5–27.5°E) during 1980–2019 and their links to atmospheric circulation. Daily differences between potential evapotranspiration and precipitation (PET–P) calculated from the E–OBS data are used to characterize dryness, and atmospheric circulation is represented by circulation types classified using daily sea level pressure patterns from the NCEP/NCAR reanalysis. Circulation types favoring dry conditions in vegetation season (April–September) are identified based on daily PET–P, and their temporal changes, seasonal variations, and links to trends in dryness in individual European regions are analyzed. In the early vegetation season (AMJ), drying trends are observed mainly in Western and Central Europe while in the late vegetation season (JAS), they are located predominantly in Eastern Europe. The dry circulation types include all anticyclonic types in all regions, as well as northeast to south (southwest in Eastern Europe) directional types. Trends of the dry circulation types correspond to those of dryness: the largest increase is found during AMJ in Western and Central Europe but during JAS in Eastern Europe. The results show that trends in dryness in the central latitude strip of Europe in the warm half‐year were associated with changes in atmospheric circulation, as the largest increases in frequency of dry circulation types occurred in the regions and months affected by pronounced drying. The increasing frequency of anticyclonic types in AMJ and reduced inflow of moist air masses from the Atlantic are the key factors supporting intensification of dry conditions in European mid‐latitudes.


Introduction
Drought is among the most hazardous climatic phenomena threatening the human population in recent decades (Van Loon, 2015), especially in connection with heat waves (Barriopedro et al., 2023;Christian et al., 2023).Therefore, many studies have examined drought from different points of view and on various temporal and spatial scales.For example, Dai (2011) assessed drought characteristics from a global perspective, Bastos et al. (2021) and Rebetez et al. (2009) on the European scale with focus on severe summer droughts, Vicente-Serrano et al. (2021) for Western Europe and Lhotka et al. (2020) on a local scale for the region of the Czech Republic.Severe droughts substantially affect the long-term hydroclimatic system (Keyantash & Dracup, 2004;Wilhite, 2000).Physical mechanisms playing roles in long-term changes of the hydroclimatic system and drought development in Europe include atmospheric circulation (Lhotka et al., 2020), often represented by atmospheric blocking (Ionita et al., 2015;Rimkus et al., 2014), modes of variability (Ionita et al., 2022), and circulation types (Plavcová et al., 2014;Stahl & Demuth, 1999;Trnka et al., 2009).
Several studies have focused on the relationship between drought and atmospheric circulation for Central and Western Europe (e.g., Bartczak et al., 2022;Beck et al., 2015;Lhotka et al., 2020;Trnka et al., 2009)  of this type have been relatively rare for most other European regions.The reasons include pronounced drying reported for Central and Western Europe, as well as the fact that this is a highly vulnerable area, where increased frequency of heat waves and droughts during the vegetation season are associated with many adverse effects on ecosystem services such as plant productivity (with related strong economic losses in agricultural crop production) despite relatively high average annual precipitation rates (Hänsel et al., 2019).Increasing trends in drought severity and frequency have been found in several local studies within Central Europe, such as in Germany (Stahl & Demuth, 1999;Träger-Chatterjee et al., 2013), the Czech Republic and Slovakia (Lhotka et al., 2020;Řehoř et al., 2021;Trnka et al., 2009), and Poland (Bartczak et al., 2022;Bogawski & Bednorz, 2016;Urban et al., 2022).Drought development in these areas has been confirmed also by studies analyzing droughts in a broader European spatial context (Bakke et al., 2023;Buras et al., 2020;Garcia-Herrera et al., 2019;Ionita & Nagavciuc, 2021;Ionita et al., 2017Ionita et al., , 2020b;;Nagavciuc et al., 2022;Schumacher et al., 2023;Spinoni et al., 2017;Van der Wiel et al., 2023).These have reported substantial drying especially in the Iberian Peninsula (Garcia-Herrera et al., 2007;Ionita et al., 2017Ionita et al., , 2020a) ) and France (Ionita et al., 2017(Ionita et al., , 2020a)).In the Mediterranean area, dry hotspots have been reported mainly from its southeastern part (the Balkans; Ionita et al., 2015;Mathbout et al., 2021) and Italy (Garcia-Herrera et al., 2019;Ionita, Nagavciuc, & Guan, 2020).In Eastern Europe, the most pronounced drying has been observed in Ukraine (Ionita et al., 2017), the Baltic region (Buras et al., 2020), and Hungary (Ionita et al., 2022).Northern Europe is categorized as a humid area where moist conditions prevail (Ionita et al., 2022), but under specific circulation patterns like those in summer 2018, significant droughts occur also in this region (southern Norway, Sweden and Finland; Buras et al., 2020;Rousi et al., 2023).Worsening drought conditions in Europe have been reported not only in lowlands but also in mountainous areas, such as in the Alps (Haslinger et al., 2019;Scherrer et al., 2022) and the Polish Carpathians (Bokwa et al., 2021;Urban et al., 2022).
From a seasonal point of view, most studies on drought deal with the vegetation season from April to September (Arazny et al., 2021(Arazny et al., , 2023;;Bartoszek et al., 2021;Lhotka et al., 2020;Trnka et al., 2009), when water availability is critical for agriculture, or the summer months of June-August (Bogawski & Bednorz, 2016;Jung et al., 2006;Linderholm et al., 2009;Urban et al., 2022;Wilcox et al., 2018), when temperatures are highest and amplify evapotranspiration.Only exceptionally is a specific month studied (Haslinger & Mayer, 2023).Most studies use data after 1950 to the present (Beck et al., 2015;Bokwa et al., 2021;Ionita et al., 2017;Urban et al., 2022), because longer time series are often unavailable or do not provide the necessary variables.In our analysis, we focused on the period since 1980 characterized by a pronounced change in European climate, with increasing temperature and associated drying.
We extend the study of Lhotka et al. (2020), which focused on the area of the Czech Republic and examined drought trends and their links to circulation for the warm (April-September) and cold (October-March) seasons.The circulation types were divided into anticyclonic, cyclonic, and directional, then subsequently classified into dry and wet.A significant increase of dry and a decrease of wet circulation types was found since the 1950s, coinciding with the observed trends toward drier conditions in both seasons.The record-breaking drought in the 2015-2018 period was also associated with atmospheric circulation favorable for drought (Lhotka et al., 2020).It remained an open question, however, to what extent these relationships between trends in drought and atmospheric circulation hold true for other European regions.
This study aims to extend knowledge of the relationships between dryness and atmospheric circulation.In contrast to Lhotka et al. (2020), the analysis is performed on a daily time scale, using differences between potential evapotranspiration and precipitation and circulation types classifications centered for individual regions using a "moving window" approach.This concept allows linking rapidly changing atmospheric circulation to dry or wet tendencies in each region separately.We also expand the analysis to a broader spatial scale (from Western to Eastern Europe across the central latitude strip), in order to reveal the extent to which the trends in dryness over the past several decades may have been driven by changing atmospheric circulation.We deal with the vegetation season (April-September), when dryness has the most severe impacts, including negative effects on agriculture, with an emphasis on differences between its early (April-June) and late (July-September) parts.

Data and Methods
Links between dryness and atmospheric circulation are evaluated using differences between potential evapotranspiration (PET) and precipitation (P), and circulation types (CTs) derived from sea level pressure data.

Characterizing Dryness
Dryness is characterized by the Climatic Water Balance Index (Wypych & Ustrnul, 2011), defined as difference between PET and P (calculated from gridded E-OBS 24.0e data; Cornes et al., 2018).The same index has recently been used to investigate aridity on a daily scale, for example, by Buras et al. (2020) or Lhotka et al. (2023).PET is defined as hypothetical evapotranspiration with unlimited water access and evaporation needs independent of crop type, crop developmental stage, and management practices.In such a concept, soil factors do not affect PET (Allen et al., 1998;Xiang et al., 2020).To calculate PET, we used the Oudin formula: where PET (mm day 1 ) is potential evapotranspiration, Re (M J m 2 day 1 ) is the top-of-atmosphere solar radiation calculated on the basis of the time of year and geographical location, and TG (°C) is mean daily air temperature (Oudin et al., 2005).The main advantage of the Oudin formula is that it needs only temperature and radiation data while still showing good results compared to more data-demanding methods (Lang et al., 2017;Oudin et al., 2005).

Vegetation Season, Study Period and Regions
The analysis is carried out for the vegetation season (April-September) divided into two parts: early vegetation season (April-June, AMJ) and late vegetation season (July-September, JAS).The reasons for splitting the vegetation season into two parts are that (a) drought during AMJ has particularly large impacts on vegetation in critical stages of its development (Trnka et al., 2009), and (b) trends in dryness and links to circulation may differ between the two parts of the vegetation season.The study period is 1980-2019, for which the trends toward drying are most pronounced (Ionita et al., 2022).Changes in the earlier decades of the 20th century were dominated rather by decadal-scale variability, which makes interpretation of trends less straightforward.Trends in dryness characteristics and frequencies of CTs were estimated by linear regression and the t-test was used to evaluate significance of the slope (at p = 0.05).
While the whole of continental Europe is included in maps of the trends in dryness, the study focuses on regions across the central latitude strip of Europe (along the 50°latitude, Figure 1)., 50°N].Abbreviations for the regions include a numerical value that represents the center of the region in longitude terms.For example, the region centered at 10°E is hereafter labeled E10.

Classification of Circulation Types (CTs)
Atmospheric circulation for the individual regions (Section 2.2) is characterized by circulation types (CTs) derived from circulation indices.The three indices-flow strength, direction, and vorticity (Jenkinson & Collison, 1977)-were calculated in a daily time step using the sea level pressure from the NCEP/NCAR reanalysis (Kalnay et al., 1996).Based on the Lamb (1972) catalog, 27 CTs adjusted for each region were derived from the indices (Blenkinsop et al., 2009).The detailed way of classifying daily patterns into CTs is described in Lhotka et al. (2020) for Central Europe (the E15 region in the present analysis), and the procedure was analogous for the other regions under study.
When the absolute value of vorticity was at least five times the strength, strongly anticyclonic (A, if vorticity <0) or strongly cyclonic (C, if vorticity >0) types were assigned.If the flow strength was greater than the absolute value of vorticity, that day was classified as one of eight directional types (N, NE, E, SE, S, SW, W, and NW).The remaining days were classified into hybrid types based on their direction and anticyclonic or cyclonic vorticity (Lhotka et al., 2020).Those CTs were assigned to one of three groups-anticyclonic (A, AN, ANE, AE, ASE, AS, ASW, AW, ANW), cyclonic (C, CN, CNE, CE, CSE, CS, CSW, CW, CNW), and directional (N, NE, E, SE, S, SW, W, NW).If the sea level pressure patterns had both flow strength and vorticity below 4, that day remained unclassified (U).

Spatial Patterns and Within-Season Development of Trends in Dryness
The temporal development of PET, P, and PET P in the individual regions for 1980-2019 is shown in Figure 2. A pronounced increase in PET was found in both parts of the vegetation season (driven by increasing temperature) while precipitation trends were relatively weak, with slight decreases (increases) prevailing in AMJ (JAS) (Figure 2).A notable decrease of precipitation in the JAS season was observed in the Eastern European region E25 since 2010, thus enhancing the rise of PET P.
Spatial patterns of the PET P trends for 1980-2019 are plotted for the whole of continental Europe in Figure 3. Areas with shades of red show a positive (drying) trend, while blue shades represent areas with a negative (wetting) trend.Black outline denotes areas of statistically significant change at p = 0.05.In the early vegetation season (Figure 3 left), a tendency to drying was found especially in the western and central part of Europe.The area of significant positive trends covers almost the whole of regions E5 and E10, and it extends partially to E15.Significant positive trends are found also in Eastern Europe (E25).Outside the studied regions, positive trends are located in the southeastern coastal area of the Baltic Sea and in northern Italy while negative trends occur only scarcely (e.g., in southwest Norway and southeastern Europe).The area of significant positive trends covers 43.0% (420,617 km 2 ) of the central latitude strip of Europe, while significant negative trends do not occur in the studied area.
In the late vegetation season (Figure 3 right), the area of significant positive trends covers, in contrast to the AMJ season, only 9.3% (91,350 km 2 ) of the central latitude strip.Almost all significant drying trends (within the study regions) are located in E25.Outside the study regions, the area of most pronounced drying occurs mainly in Eastern Europe (Ukraine and Moldova) and to a smaller extent also in other regions (Northern Europe, southwestern Europe, and the northern part of the UK).Similarly to as seen for AMJ, significant trends to wetter conditions occur only in small isolated areas outside the study regions.The development of trends during the vegetation season can be characterized in more detail when analyzing individual months separately (Figure 4).Within the central latitude strip, a significant drying trend was found in April and June; in May and JAS, by contrast, it is limited to only small isolated areas.In April, the drying trend occurred within large areas in Western and Central Europe (E10, E15) and partly also in Eastern Europe (E25) and was significant in 37.3% (364,821 km 2 ) of the central latitude strip.In June, significant drying trends from Western through Central to Eastern Europe covered 36.5% (357,613 km 2 ) of the central latitude strip.
Due to different patterns in PET P trends for April, May, and June, we analyzed these in more detail using 31-day moving windows (Figure 5).Western (E5, E10) and partly also Central Europe (E15) have pronounced local maxima of the drying trends in April and June.The maximum drying trend appears in June also in the other regions (E20, E25), and a secondary maximum is found at the turn of August/September in E25.Wetting trends are less distinctive and were found mostly at the turn of July/August in Western Europe (E5 and E10), and in May and September for Central and Eastern Europe (E15-E25).

Characteristics of CTs and Definition of Dry and Wet CTs
This section summarizes basic characteristics of the CTs, including their frequencies, mean PET P values, and the definition of dry and wet CTs in individual regions under study.As the frequency of unclassified days (U) was small in all regions and both AMJ and JAS seasons (between 0.8% and 2.9%), they were excluded from further analysis.
All anticyclonic types (Section 2.3) are favorable to drying, meaning that they have positive mean PET P values.This applies to all regions (Table 1).The opposite (negative mean PET P values) holds true for cyclonic types except for those associated with easterly or southerly flow (CE, CSE, CS, and CSW), which, due to a relatively warm advection, have mean PET P values close to zero or slightly positive in some regions (  In all regions, the frequency of anticyclonic types (on average 44%) was higher compared to that of cyclonic types (18%).The highest frequencies of anticyclonic types are found between the E10 and E20 regions.The frequencies of directional types ranged between 35% and 40% in all regions, with their lowest values occurring in E10 and E15 (Table 1).
CTs supporting dry conditions (referred to hereinafter as "dry"), with positive mean PET P differences, were identified for each region.Analogously, "wet" CTs were defined as those with negative mean PET P values.Because the differences in CTs favorable to dry conditions were minor between AMJ and JAS (Tables S1 and S2 in Supporting Information S1), the definition was based on mean values for the whole vegetation season (April-September, Table 1).After preliminary analyses, we decided not to include CTs with mean PET P values close to zero (in the interval between 0.5 and 0.5 mm/day) into either dry or wet.This concerned only 1 or 2 cyclonic CTs, and 1 to 4 directional CTs depending on region (Table 1).The reason was that these CTs were associated with little-pronounced PET P characteristics (compared to the other CTs; note the much larger deviations from 0 for all anticyclonic and most cyclonic CTs) and their inclusion into dry or wet CTs might distort the results in case of directional CTs with higher frequency (e.g., N in E25, 6.6%).
Dry CTs include all anticyclonic types for all regions and directional CTs from the northeast through east to the south (NE, E, SE, S), and in Eastern Europe (E20 and E25) also southwest (SW).Only exceptionally is a cyclonic type included within dry CTs (CSE in E10, with mean PET P value 1.0 mm/day).Wet CTs, on the other hand, include most cyclonic types together with directional types from the southwest (except for Eastern Europe) to the north.

Trends of CTs
Trends in the frequencies of anticyclonic (A), cyclonic (C), and directional (DIR) CTs in the individual regions are shown in Table 2.During AMJ (upper part of Table 2), a marked increase of anticyclonic types is found in all regions (statistically significant in E5-E20), accompanied by a decrease of cyclonic types (except in E5) that is significant in E25.Trends of directional types are mostly insignificant, except for a pronounced negative trend in E5, which is primarily due to decrease of CTs with the northern component (significant for the NW type, Table S3 in Supporting Information S1).For the late vegetation season, cyclonic types have slightly positive trends in all regions but they are insignificant (lower part of Table 2).These results indicate different changes in CTs between the two parts of the vegetation season: a significant shift toward more anticyclonic circulation in AMJ across the whole central latitude strip but little-pronounced trends, with insignificant but widespread increases in cyclonic types, in JAS (Table 2).
Increased frequencies of anticyclonic CTs at the expense of cyclonic and directional types in AMJ are illustrated in Figure 6 (top).In Central Europe (E15), for example, the mean frequency of anticyclonic CTs was 39.7% in the 2010s compared to 30.8% in the 1980s.Meanwhile, the frequency of cyclonic types declined from 25.6% to 22.6%.In 2015, a season with particularly pronounced drought and heat waves across the central latitude strip of Europe (Ionita et al., 2017;Urban et al., 2017), the frequency of anticyclonic CTs exceeded 57% in most regions (except for E25) while cyclonic CTs occurred in less than 10% of days during AMJ.In all regions, differences in frequencies between anticyclonic and cyclonic CTs steadily increased from the 1980s to the 2010s.During JAS (Figure 6, bottom), the trends were relatively weak and insignificant, mainly due to reaching a minimum in the frequency of anticyclonic types (and corresponding maximum in cyclonic and directional types) around 2000, followed by increase (decrease).

Characteristics of Dry and Wet CTs and Their Trends
Within-season changes in frequencies of dry and wet CTs from April to September are plotted in Figure 7. Dry CTs are found in lower frequencies around May for most regions.At the same time, in this month the maximum frequency of wet CTs occurs, which corresponds also with the development of wetting in this month (Figures 4  and 5).The dry CTs generally occur more frequently in the late rather than early vegetation season, and mainly in Central and Eastern Europe (E15-E25).In the early vegetation season (AMJ), trends in dry CTs are positive in all regions (Table 3 left) and significant at p = 0.05 between E10 and E20.The magnitude of dry CTs' trends in these regions exceeds +2.5%/decade, corresponding to more than +10% increase over the 40 years under study.In the late vegetation season (JAS, Table 3 right), by contrast, a significant drying trend is found only in Eastern Europe (E25) while trends in dry CTs are relatively weak in the other regions.
The increase of dry CTs in the AMJ season in regions E5 to E20 is in line with the increase of anticyclonic types reported in Table 2 and is statistically significant (except for region E5, where the trend in dry CTs is insignificant).These changes in atmospheric circulation supported drying in these (Figure 3).Some discrepancy can be found only in E20, where the drying trends were less pronounced, mostly insignificant, and in contrast to a pronounced change toward more anticyclonic (and dry) CTs.
In the JAS season, dry CTs increase significantly only in the E25 region (Table 3).This is associated not with changes in anticyclonic CTs (trend close to zero, Table 2) but with directional CTs favorable for drying (namely E and SE; Table S4 in Supporting Information S1).Unlike in the other regions under study, the trend in Eastern Europe toward a higher frequency of dry CTs agrees with the large drying trends observed in the E25 region (Figure 3).Subsequently, within-season variations in the trends of frequencies of dry CTs from April to September were analyzed (Figure 9).Increasing trends dominate in Central and Eastern Europe (E15-E25) throughout the vegetation season.These are largest in June, and a secondary maximum occurs at the turn of August/September.In the western part of the studied area (E5, E10), the trends in dry CTs are positive in AMJ only, with maxima in  Earth and Space Science 10.1029/2023EA003434 April and June.These results are in good agreement with significant April and June trends toward drier conditions in Western and Central Europe as well as maximum drying occurring in June and at the turn of August/September in Eastern Europe (Figure 5).This suggests close relationships between changes of dry CTs and dryness in the studied regions.

Discussion
The main goal of the present study was to better understand trends in dryness and their links to atmospheric circulation across the central latitude strip of Europe, that is, the area affected by pronounced recent drying, based on daily circulation-to-dryness links.We adopted-in a relatively innovative application to our knowledge, we are not aware of another study making use and benefit of a similar approach in drought research-the circulation type classification centered on individual "windows" across the study area.A different view of drynessrepresented by the PET-P index on the daily time scale in our study-is another aspect which adds to existing knowledge.Studies focusing on drought and circulation often use monthly data (e.g., Buras et al., 2020;Faranda et al., 2023;Ionita et al., 2015Ionita et al., , 2022;;Kingston et al., 2015;Vicente-Serrano et al., 2016), or daily data aggregated to a monthly scale (Bakke et al., 2020(Bakke et al., , 2023;;García-Herrera et al., 2019).The daily time scale is crucial in relating drought tendencies to atmospheric circulation patterns, whereas commonly used drought indices such as SPI, SPEI, and PDSI are usually calculated for monthly steps.If high temperatures occur, drought development can accelerate quickly together with heat waves (flash drought), so the daily time step is essential when studying drought dynamics.This is an important aspect in which our analysis differs from most recent drought studies, including those that link dry conditions to atmospheric circulation.1.

Shift in the Development of Drought During the Vegetation Season Across Europe and Changes in Circulation
We found statistically significant drying trends for 1980-2019 that were spatially varying across the central latitude strip of Europe.In the early vegetation season (AMJ), the trends were significant particularly over western parts of this region.In the late vegetation season (JAS), meanwhile, the drying was recorded   Table 1.
predominantly in the east.Similar spatial variability of drying was noted also by Bakke et al. (2023), who focused on the whole of Europe and studied the role of circulation based on high-pressure anomalies for meteorological drought on the monthly time scale.
AMJ drought is an important factor in initiating long-lasting droughts: it poses a major risk for agriculture and, at the same time, affects the vegetation state and its development in the rest of the vegetation season (which is mainly controlled by precipitation in preceding months; Cook et al., 2016).In most studies on drought development, attention is given especially to the summer season (Bakke et al., 2020;Bogawski & Bednorz, 2016;Wilcox et al., 2018) or to the entire vegetation period (Arazny et al., 2021;Rimkus et al., 2014), but that does not provide intra-seasonal insight into trend variability.Significant drought trends in the early vegetation season in Central Europe were found by Trnka et al. (2009), who dealt with station data for the period 1881-2005.In our study's more detailed and updated analysis, we show relatively large variability in drying trends on the monthly scale.While April and June are characterized by significant drying trends over large parts of the domain, almost no regions with drying trends were recorded in May.A similar temporal pattern of drying tendencies was recently reported by Bakke et al. (2023) using SPEI.In Eastern Europe, significant changes toward drier conditions were observed in June and August by Bartoszek et al. (2021), who examined spatiotemporal changes of meteorological drought in agricultural areas.In addition, drying tendencies were also found further east, outside the studied regions, in Ukraine and Moldova (Ionita, Nagavciuc, Kumar, & Rakovec, 2020;Ionita et al., 2017;Potopová et al., 2019).The drying in Eastern Europe during the late vegetation season is related to pronounced warming in late summer within this region and associated warming in the Black Sea region (Řehoř et al., 2024;Sakalli & Basusta, 2018).
Significant increase (decrease) in the frequency of circulation types favorable (unfavorable) to drying was found for AMJ.The increase was widespread across the central latitude strip of Europe and may reflect larger-scale changes in atmospheric circulation.Increased frequency of anticyclonic types may suggest a northward shift of the general circulation, with subtropical highs and ridges becoming increasingly important for Central European climate in late spring and early summer.Some studies have reported decrease of zonal flow in Western, Central, and Eastern Europe (e.g., Garcia-Herrera et al., 2019;Rimkus et al., 2014), but these did not analyze this phenomenon on monthly and/or seasonal scales.We performed an additional analysis of westerly and easterly CTs (Table 4 and Table S5 in Supporting Information S1) and found significant decrease of the westerly CTs in Central and Eastern Europe and significant increase of easterly CTs in Eastern Europe (Table 4).These trends were more pronounced in JAS than AMJ (Table S5 in Supporting Information S1) and suggest a tendency toward decreasing zonal flow across the central latitude strip of Europe.The reduced inflow of moist (and relatively cooler in the warm half-year) air masses from the Atlantic penetrating into the European interior may be one of the important factors supporting drought intensification, and requires further attention.
The increase of dry circulation types and related drying tendencies in Europe may be associated with several other large-scale factors.One of them is atmospheric blocking (Ionita et al., 2015).Atmospheric blocking suppresses zonal flow and strengthens the meridional flow, which is linked to greater presence of tropical air masses in Central Europe (Tomczyk & Bednorz, 2019).The increased frequency of atmospheric blocking over Central Europe is possibly amplified by the weakened Atlantic meridional overturning circulation (AMOC; Bakke et al., 2023;Ionita et al., 2022).Toward the end of summer and in autumn, the development of drought may be related to the North Atlantic Oscillation (NAO; Ionita et al., 2015).Droughts' development obviously depends also on the strength and position of the polar jet stream.The stronger and northerly displaced polar jet stream is related to more frequent anticyclonic conditions in European mid-latitudes (Trouet et al., 2018), leading to prevailing dry and hot weather in the warm half-year.The drought then goes on and is successively exacerbated.By contrast, weaker and undulated jet stream runs further south and is associated with increased cyclonic activity and weather extremes such as heavy precipitation (Garcia-Herrera et al., 2019;Haslinger et al., 2019;Trouet et al., 2018).The development of drought and heat waves on the European continent may eventually be influenced by additional climate dynamic drivers such as El Niño-Southern Oscillation (ENSO; Barriopedro et al., 2023), Quasi-Biennial Oscillation (QBO; Karagiannidis et al., 2008), Atlantic Multidecadal Oscillation (AMO; Faranda et al., 2023), and others (Barriopedro et al., 2023;Domeisen et al., 2023).Combining global climate dynamics with the CTs and their regional effects remains a task for further studies.
Our study is, as a whole, in good agreement with numerous investigations that have found rising temperatures and radiation together with decreasing air humidity to be the main drivers of the accentuated increase of droughts (especially compound dry-hot events) over large parts of Europe.Underscoring these basic relationships, Arazny et al. (2023) reported a statistically significant upward trend in evapotranspiration as a consequence of significantly increasing sunshine duration and air temperature and a decrease in relative humidity during the vegetation season in Poland over 1966-2020.Conditions with little cloud cover, high temperatures, and low relative humidity are typically linked to anticyclonic CTs in the warm half-year.By contrast, a notable increase of water balance deficits (PET P) in recent years within the Eastern European area E25 is driven to a great extent by declining precipitation.
The drought situation in the years 2020-2023 corroborated the main findings of our study, as it was marked by regular early summer droughts in Northwestern Europe and drought hotspots in Eastern Europe in late summer (Faranda et al., 2023;Lhotka & Kysely, 2022;Schumacher et al., 2023;Tripathy & Mishra, 2023;Van der Wiel et al., 2023).A comprehensive analysis of the extraordinary, regionally record-breaking drought, heatwaves, and strong rains in the years since 2020 remains a challenge for the future and may lead to deeper understanding of the presented results.

Links Between Drought Trends and Circulation Types
Good agreement was found between trends in dryness and dry/wet CTs.The substantial drying in the AMJ season across the analyzed domain was associated with increase in the occurrence of dry CTs (statistically significant in E10-E20).In JAS, by contrast, we found a significant increase of dry CTs only in the easternmost (E25) region, which is in accordance with the spatial patterns of drought trends.
While dry (wet) CTs generally consist of anticyclonic (cyclonic) types (Plavcová et al., 2014;Stahl & Demuth, 1999), directional types were found to have different drying tendencies depending on the region.In Western and Central Europe, drier conditions are mainly associated with easterly and southerly directions, which is in line with studies that have analyzed droughts over Central Europe (e.g., Bartczak et al., 2022;Bogawski & Bednorz, 2016;Lhotka et al., 2020;Trnka et al., 2009).In Eastern Europe, by contrast, all directional types (except N) have a dry tendency.This reflects the higher continentality of climate in Eastern Europe, while Western Europe (and partially also Central Europe) is under more pronounced influence of an oceanic climate.
During JAS, the links between trends in dryness and circulation types may be weaker due to a large share of convective precipitation compared to other seasons (Rulfová & Kyselý, 2013).This is amplified in areas with rugged topography, where local convective processes support the occurrence of precipitation.These heavy precipitation events are, however, not evenly distributed in time and space.Vegetation may then suffer long-term drought with occasional torrential rains (Haslinger et al., 2019) and greater losses due to direct runoff.This may contribute to the slight discrepancy between an increase in frequencies of dry types not associated with drought trends in regions E15 and E20 in the late vegetation season.

Choice of the PET−P Index
In our study, dryness was analyzed through the difference between potential evapotranspiration and precipitation, that is, the Climatic Water Balance Index (Wypych & Ustrnul, 2011).This index has intuitive interpretation and provides information on drying on the daily time scale, which is essential for its linking to atmospheric circulation Earth and Space Science 10.1029/2023EA003434 BEŠT' ÁKOVÁ ET AL.
patterns.We used PET P because the variables required for its calculation are available in E-OBS (although approximation is needed in the case of PET), and the same approach can be applied to many other data sets with daily resolution (while such commonly used drought indices as SPI, SPEI, and PDSI are usually calculated for monthly steps).If extremely high temperature occurs, the development of drought may accelerate along with that of heat waves in a short period (flash droughts; Christian et al., 2024;Li et al., 2021;Tripathy & Mishra, 2023;Wang et al., 2023), so the daily time step is essential when studying drought dynamics.Wu et al. (2007) questioned the applicability of SPI in monitoring and analyzing droughts because, in such cases, periods without precipitation are the norm and cannot be considered droughts.Furthermore, SPI is often criticized because it does not take into account temperature and related evapotranspiration (i.e., variables relevant to drought; Vicente-Serrano et al., 2010), causing SPI to underestimate drought especially in areas with a strong warming signal (Ionita & Nagavciuc, 2021;Vogel et al., 2021).The SPEI index includes both precipitation and potential evapotranspiration components but may lead to excessive drying compared to other indicators (Cook et al., 2016) and should be used with caution (Russo et al., 2019).PDSI may be a better choice to characterize atmospheric water demand (Wang et al., 2023), but its main disadvantage is that specific parameters are obtained for areas of the United States, which hinders comparisons among diverse climatic regions (Feng et al., 2021).In PDSI, the data-demanding Penman-Monteith formula is often used to estimate PET (Donohue et al., 2010), making the index highly sensitive to imprecision of the data (Seiller & Anctil, 2016;Weiland et al., 2012).A general disadvantage of these indices is their built-in imitation of soil moisture memory, which is not desirable when linking day-to-day changes between dryness and circulation types.
We derived potential evapotranspiration using Oudin's formula (Oudin et al., 2005) based on temperature and radiation.This method provides better results than does, for example, the Thornthwaite formula based solely on temperature (Thornthwaite, 1948), which tends to underestimate PET (Lakatos et al., 2020;Pereira & Pruitt, 2004;Sentelhas et al., 2010;Xiang et al., 2020).In some cases, better model performance has been reported for more straightforward formulas than for combination formulas like the Penman-Monteith (Weiland et al., 2012), the disadvantages of which were mentioned above.

Relevance of Study Period Lengths and Climatic Seasons
After a preliminary analysis, we decided to focus on the period since 1980, because pronounced multidecadal variability of drought conditions weakens the temporal stability of estimated long-term trends (Hänsel et al., 2019).Figure S1 in Supporting Information S1 shows for illustration trends in PET-P for the entire available period of the E-OBS data since 1950.The trends are much weaker compared to those in Figure 3 (the scale is kept the same in both figures) and the recent drying is not well represented (the trends are strongly influenced by variability in earlier decades).That is why we preferred to confine the study to a shorter period with a clear climate change signal.
The focus on the 40-year period 1980-2019, which is characterized by pronounced warming in Europe (Ionita et al., 2022), allows for better understanding of the links between recent changes in drought and atmospheric circulation in European mid-latitudes, that is, the region with the most pronounced warm-season drying trends in Europe according to the Climatic Water Balance Index (Figure 3).At the same, the period of the last four decades is sufficiently long for estimating trends.The links between dryness and CTs are similar for the whole period of 1950-2019 (Table S6 in Supporting Information S1) as for 1980-2019 (Table 1), and the set of CTs favorable for drying remains almost the same (Table S6 in Supporting Information S1).Using a shorter (e.g., 2000-2019) period for the analysis would result in positive PET P trends in JAS (Figure 2), but such time window would not be sufficient for obtaining reliable trend estimates.
The analysis for individual months provided valuable information on intra-seasonal variations in trends.Trends in both PET P and dry CTs are relatively sensitive to the selected period of the year.This concerns, for example, June versus August.As visible in Figure 5, the 2 months are quite different and yet they are usually considered as part of the same season (Bakke et al., 2020;Bogawski & Bednorz, 2016;Jung et al., 2006;Wilcox et al., 2018).It is questionable to what extent climatic seasons (such as JJA) are suitable for analyzing trends.These findings bring a similar view as seen in a study by Pokorná et al. (2018), who examined the annual cycle of trends in temperature.Generally, we must consider the limitations of trend analysis, which captures only the relationships within the selected temporal and spatial framework and cannot be extended to finer scales without further analysis.
Earth and Space Science 10.1029/2023EA003434

Conclusions
Our study examined trends in dryness and their links to atmospheric circulation across the central latitude strip of Europe (centered upon 50°N) during the vegetation season for 1980-2019.In contrast to previous studies, we used the Climatic Water Balance Index (difference between potential evapotranspiration and precipitation) calculated on a daily time scale, in order to link dry or wet tendencies to circulation types (CTs) centered for each region using a "moving window" concept.The main results can be summarized as follows: 1.In the early vegetation season (AMJ), significant drying trends are observed mainly in Western and Central Europe, while in the late vegetation season (JAS) they are predominantly found in Eastern Europe.The trends are strongest in April and June, and in Eastern Europe also at the turn of August/September.2. The anticyclonic CTs have drying tendencies regardless of region.In case of directional CTs, northeast to south (southwest in Eastern Europe) CTs contribute to drying.3. The trends of the dry CTs correspond to those of dryness: the largest increases in frequencies of dry CTs occur in those regions as well as seasons and months affected by pronounced drying.4. The increasing frequency of anticyclonic types (during AMJ) together with reduced inflow of moist air masses from the Atlantic penetrating deeper to the continental regions of Central and Eastern Europe (mainly in JAS) are the key factors supporting drought intensification in European mid-latitudes.
The results show an eastward temporal shift of the regions most affected by drying during the warm half-year and a close link between changes in atmospheric circulation and trends in dryness over the past four decades.
Increased frequency of anticyclonic types can be interpreted as a signal of a northward shift of the general circulation, with subtropical highs and ridges becoming increasingly important for the Central European climate above all in late spring and early summer.Although atmospheric circulation is only one factor contributing to the overall drying trends (which are primarily driven by increasing temperatures due to climate change), it affects the spatial and temporal patterns of the trends across Europe in the warm half-year.Therefore, it is important to better understand the circulation-to-drought links also in climate model simulations for present and future climates.

•
Significant drying trends in European mid-latitudes since 1980s have been primarily linked to growing potential evapotranspiration • Dry conditions during early vegetation season have become more frequent and severe particularly in Western and Central Europe • Increased frequency of circulation types favoring dry conditions corresponds both spatially and temporally to the drying trends Supporting Information: Supporting Information may be found in the online version of this article.

Figure 1 .
Figure 1.Regions analyzed in the study, centered around 50°N.Numbers indicate longitude of the center of each box (e.g., E5 represents 5°E).

Figure 2 .
Figure 2. Mean seasonal values (AMJ-top, JAS-bottom) of potential evapotranspiration (PET), precipitation (P), and their differences (PET P) in each region for 1980-2019.Solid lines show the 11-year moving averages of the mean seasonal values.

Figure 3 .
Figure 3. Linear trends in PET P for the two seasons (AMJ and JAS) for 1980-2019.The black outline delimits areas where the trends are significant at p = 0.05.The study regions along 50°latitude are marked by red boxes.

Figure 4 .
Figure 4. Linear trends in PET P in individual months for 1980-2019.The black outline delimits areas where the trends are significant at p = 0.05.The study regions along 50°latitude are marked by red boxes.

Figure 5 .
Figure 5. Linear trends in PET P in 31-day moving windows from April to September for 1980-2019.Red points mark trends significant at p = 0.05.

A
statistically significant increase of dry CTs and corresponding decrease of wet CTs during AMJ are analyzed in more detail in Figure8(top).The temporal development of dry CTs shows a steady increase with decadal-scale variations in all regions and reaching a maximum in the early 2000s.Analogous (opposite) changes are found for wet CTs.The differences between frequencies of dry and wet CTs during AMJ increased from 16.5% in the 1980s to 25.5% in the 2010s (on average across the regions).The temporal changes in the JAS season (Figure8bottom) were characterized by decadal-scale variations rather than long-term trends (except in Eastern Europe, where a significant increase of dry CTs was found).

Figure 6 .
Figure 6.Seasonal frequencies of anticyclonic (A), cyclonic (C), and directional (DIR) CTs and their trends in individual regions for 1980-2019.Solid lines show the 11-year moving averages of the seasonal frequencies.

Figure 7 .
Figure 7. Within-season frequencies of dry (brown) and wet (cyan) CTs during the vegetation season.Points mark mean frequencies on individual days (for 1980-2019), while solid lines show 31-day moving averages.For the definition of dry and wet CTs, see Section 3.2 andTable 1.

Figure 8 .
Figure 8. Seasonal (AMJ-top, JAS-bottom) frequencies of dry (brown) and wet (cyan) CTs and their linear trends in individual regions for 1980-2019.Solid lines show 11-year moving averages of the seasonal frequencies.

Figure 9 .
Figure 9. Within-season variations of trends in frequencies of dry CTs for 1980-2019.Points mark trends estimated for 31-day moving windows centered on a given day, and solid lines show 31-day moving windows of those trends.For the definition of dry CTs, seeSection 3.2 and Table 1.

Table 1
Mean Frequencies of I-2019dual CTs and Their MeanPET P Values for 1980-2019 Note.Values refer to the whole vegetation season (April-September).Brown (cyan) shading highlights dry (wet) tendency of a given CT.No shading stands for CTs not listed as dry or wet because their mean PET P values are small (between 0.5 and 0.5 mm/day).BEŠT' ÁKOVÁ ET AL.

Table 2
Linear Trends in Frequencies of Anticyclonic (A), Cyclonic (C), and Directional (DIR) CTs in AMJ (Top) and JAS (Bottom)Seasons for 1980Seasons for  -2019

Table 4
Linear Trends in Frequencies of Westerly and Easterly CTs With Their p-Values During the Vegetation Season (April-Note.Westerly CTs include SW, W, NW, ASW, AW, ANW, CSW, CW, and CNW; easterly CTs include NE, E, SE, ASE, AE, ANE, CSE, CE, and CNE.Bold values show trends significant at p = 0.05.