Since systematic hail observations are not available before 2002 in Switzerland, we use the recently developed hail day reconstruction by . This dataset provides a binary time series indicating daily hail occurrence for two study areas, north and south of the Alps within the Swiss radar domain, which also includes parts of southern France, southern Germany, and northern Italy, during the hail-prone months (April–September) from 1959 to 2022 (Fig. ). For simplicity, these two study areas are hereafter referred to as northern and southern Switzerland, respectively. To reconstruct past hail events, applied an ensemble statistical model that combines a radar-based hail proxy with ERA5-derived environmental predictors. For further details, see .

We use ERA5 reanalysis data to study atmospheric, oceanic, and land-surface variables, using hourly data from 1959 to 2022 at a 0.5° × 0.5° grid resolution, including pressure-level, single-level, and land-surface variables. A list of variables and a short description can be found in Table . Daily anomalies are calculated for all ERA5 variables with respect to a climatological mean and standard deviation estimated from moving windows, following and . To reduce the influence of long-term thermodynamic trends (e.g., warming-driven increases in instability), daily anomalies are calculated with respect to a mean and standard deviation estimated from an 8-year moving window, following and . Additionally, a 30 d moving window centered on each calendar day is used to remove the seasonal cycle. This approach retains low-frequency climate variability, thereby allowing the analysis to focus on circulation-related effects.

Blocking frequencies were computed from 6-hourly ERA5 potential vorticity (PV) fields using the contour-tracking tool ConTrack, developed by building on the approach of . Potential vorticity (PV) cutoffs and PV streamers are identified following the definitions of and , and tracked with the algorithm described by using the same 6-hourly ERA5 PV fields. The detection procedure delineates cutoff cyclones and streamers by tracing the 2-PVU contour on five isentropic levels between 330 and 350 K. Daily frequencies are then computed, whereby a grid point is classified as having a cutoff or streamer if a detection occurs on any of the five levels. To investigate the recurrence and strength of synoptic patterns we compute the R metric as introduced in for the 500 and 300 hPa meridional wind. R is calculated for the latitudes 35–65 and 40–70° N for north and south respectively. High R values indicate a high recurrence of the 500 and 300 hPa large-scale flow.

To assess potential large-scale atmospheric influences on hail activity, we analyze relevant northern hemispheric modes of variability, including monthly time series of the Arctic Oscillation (AO), East Atlantic (EA) pattern, East Atlantic/Western Russia (EAWR) pattern, East Pacific/North Pacific (EPNP) pattern, El Niño–Southern Oscillation (ENSO), North Atlantic Oscillation (NAO), Polar/Eurasia (POLEUR) pattern, Scandinavian (SCAND) pattern, and West Pacific (WP) pattern. These indices are obtained from the Climate Prediction Center of the U.S. National Oceanic and Atmospheric Administration (NOAA) and are computed from NCEP-NCAR1 reanalysis data following and . To quantify relationships between these large-scale atmospheric patterns and regional hail activity, we calculate correlations between the monthly and annually aggregated hail day time series and each mode of variability (see Figs.  and ). Correlations are computed over various temporal periods, including seasonal (i.e., DJF, MAM, JJA, SON), annual, monthly, and preceding periods (i.e., the cold season before the hail season and the previous year's SON).

The reconstructed hail day time series from forms the basis for identifying the years with a high and low hail activity in each study region. To isolate interannual variability and avoid bias toward recent decades, we first remove the long-term positive trend by linearly detrending the annual hail day time series (see Fig. 9 in ). We then select the ten most and least active hail seasons in both the northern and southern regions which correspond to the 15 % least and most active seasons of the investigation period (Table ). The two regions share five of the ten most active seasons (1963, 1970, 1994, 2003, 2017) and four of the ten least active seasons (1961, 1980, 1984, 2021). Notably, although 2021 is widely recognized as an exceptional hail year in Switzerland, it is classified among the least active hail seasons because the season was dominated by only a few exceptionally severe hail outbreaks rather than frequent hail days. The distinction between hailstorm severity and seasonal hail-day frequency of 2021 is consistent with the findings of . In both regions, hail days during the most active seasons are not confined to a single month but occur repeatedly throughout the peak convective period (June–August), with few additional events at the beginning and end of the hail season (April, May, September; see Appendix Fig. ).

To investigate the large-scale atmospheric, oceanic, and land-surface conditions associated with active hail seasons, we compute daily anomalies for selected ERA5 variables (see Table ) across the Northern Hemisphere. Anomalies are calculated relative to a mean and standard deviation estimated from a 30 day, 8-year moving window, following and . We then calculate composite mean fields for two time windows: (i) the full hail season (April–September) and (ii) the preceding cold season (October–March), to identify potential precursors. Hail-season composites are computed over all days in April–September, rather than only hail days, to capture the background conditions that favor frequent hailstorm formation, rather than the instantaneous environment during individual events. Cold-season composites are likewise computed over all days in October–March. All composites are calculated separately for the ten most active hail seasons in the northern and southern regions. Hail-season composites are shown in Figs. , , , and , while cold-season composites are presented in Figs. , , and . Additional time windows were tested (e.g., monthly composites, the peak hail season (JJA), the preceding DJF, and adjacent years), but these did not provide further insight, indicating that the relevant signals are captured within the selected periods. Statistical significance is assessed using a two-sided bootstrap test (1000 samples), comparing composites of the most active and inactive seasons against the 5th and 95th percentile thresholds of randomly sampled means.

Note that the full analysis was also conducted for the ten least active hail seasons. These exhibit largely opposing significant patterns compared to active seasons, indicating that the associated large-scale conditions differ between active and inactive hail seasons. While not discussed in detail to maintain brevity and focus, these contrasting signals are climatologically plausible and support the robustness of the identified relationships. Corresponding additional figures are provided in the Supplement.

To assess differences between these seasonal background conditions and event-scale environments, we additionally compute anomalies separately for hail days and non-hail days. Specifically, Figs.  and  present composites for five subsets: (1) all non-hail days (1959–2022), (2) non-hail days during the ten most active hail seasons, (3) all hail days (1959–2022), (4) hail days during the ten most active seasons, and (5) the seasonal mean over all days in April–September during the ten most active seasons. This comparison allows us to disentangle the environmental conditions on hail days from the broader seasonal background preconditioning during active hail seasons.

Figure  shows that for northern Switzerland, the hail day composites (rows 3 and 4) exhibit a coherent ridge–trough configuration over Europe accompanied by warm near-surface temperatures, enhanced low-level moisture, elevated CAPE, and moderate CIN, which are features that closely resemble those found in the seasonal-mean composites (row 5). As expected, the magnitudes of the seasonal-mean anomalies are substantially weaker than those of the hail day-only composites, reflecting the inclusion of non-hail days that either display weaker anomalies or anomalies of opposite sign (row 1 and  2). This contrast is particularly evident when comparing the two non-hail day composites. The composite of all non-hail days (row 1) exhibits anomaly patterns that are largely opposite to those observed on hail days (rows 3 and 4) and in the seasonal means (row 5), including negative Z500 anomalies over Central and Eastern Europe, weakly negative Mediterranean SST anomalies, and negative anomalies in 2 m temperature, 850 hPa specific humidity, CAPE, and CIN over most of Central Europe; conditions that are generally unfavorable for convective storm development. In contrast, when restricting the composites of non-hail days to the most active hail seasons (row 2), the anomaly patterns more closely resemble those of the hail day and seasonal-mean composites (rows 3, 4, 5), particularly in Z500, SST and 2 m temperature. Z500 anomalies are positive over Europe, Mediterranean SSTs and surface temperature anomalies are predominantly positive across much of the continent, consistent with a background state conducive to hailstorm development (Fig. g,h,i). By contrast CAPE and CIN anomalies remain negative (Fig. k,l). Notably, the two non-hail day composites (rows 1 and 2) differ in sign primarily for Z500, SST, and surface temperature, while they remain similar for CAPE, CIN, and 850 hPa specific humidity. This indicates that large-scale circulation, SST, and surface temperature are already anomalous during non-hail days of active seasons, suggesting that they characterize the seasonal background state rather than conditions unique to hail days. In contrast, CAPE, CIN, and 850 hPa specific humidity remain similar between the two non-hail day composites but become substantially more anomalous on hail days, indicating that they primarily characterize event-scale hail-day environments rather than persistent seasonal preconditioning.

The two hail day composites (rows 3 and 4) are themselves highly similar, demonstrating that hail days in average seasons and hail days in exceptionally active seasons arise from broadly the same type of environment. However, most anomalies are systematically stronger in the hail day composite of the most active seasons (row 4) than in the all-hail day composite (row 3; note the logarithmic color scaling used in the figure). This suggests that the physical processes associated with hail formation are qualitatively similar across seasons, but they occur with greater magnitude during particularly hail-active seasons.

The relatively small differences between Mediterranean SST anomalies on non-hail days in the most active seasons (row 2), on hail days (rows 3 and 4), and in the seasonal means (row 5), together with the weakly negative SST anomalies on all non-hail days (row 1), indicate that warm SST anomalies persist throughout much of active seasons rather than being confined to individual hail events. This persistence is consistent with the large heat capacity of the ocean and the consequently slow evolution of SSTs. These anomalies may therefore either exert a sustained influence on the lower-tropospheric environment, e.g., by enhancing moisture availability, or simply reflect the prevailing large-scale circulation throughout the season. Surface temperature anomalies likewise remain positive in the seasonal-mean and hail day composites (rows 2–5) compared to all non-hail days (row 1).

South of the Alps, the large-scale flow patterns evident in the seasonal means are again recognizable in the hail day composites, although the contrast between hail day and seasonal-mean anomalies is more pronounced than in northern Switzerland (Fig. ). On hail days, Z500 anomalies (rows 3 and 4) reveal a markedly stronger and more meridionally oriented trough-ridge configuration than in the seasonal mean (row 5), with the downstream ridge extending slightly farther into northeastern Europe. As in northern Switzerland, Mediterranean SST anomalies during hail days of the most active seasons are approximately twice as large as in the all-hail days composite and are also positive during non-haildays in the most active seasons (Fig. t, n, h). This indicates that warmer than average Mediterranean SSTs are a robust and persistent feature of hail-active summers in both regions, rather than being confined to individual hail events.

Near-surface temperature, low-level specific humidity, and CAPE exhibit substantially stronger anomalies over the study area and much of Central Europe on hail days in the most active seasons (row 4) than in the all-hail days composite (row 3). This suggests not only that the frequency of hail events may be enhanced, but the higher CAPE could also promote stronger updrafts during the most active hail seasons. CIN anomalies in the all-hail days composite are predominantly negative over the study area, reflecting weaker convective inhibition during average hail seasons compared to the most active hail seasons. In contrast to northern Switzerland, the non-hail day composite within the most active hail seasons (row 2) south of the Alps shows little indication of favorable preconditioning in mean Z500, near-surface temperature, or low-level moisture. This implies that these variables alone do not distinguish active from inactive days during hail-rich seasons. However, Mediterranean SST anomalies remain positive, and CIN anomalies are stronger relative to the all non-hail days composite (Fig. l, f). This combination points to the generally more frequent presence of a capping inversion during active seasons, consistent with repeated advection of warm, relatively dry air aloft and the potential repeated establishment of an elevated mixed layer.

Comparing both non-hail day composites (rows 1 and 2) with the hail day composites (rows 3 and 4) further highlights that differences in anomaly sign over Central Europe, particularly in Z500, CAPE, 850 hPa specific humidity, and 2 m temperature, indicate that the availability of instability, moisture, warm near-surface conditions, and a favorable large-scale circulation must be considered jointly to distinguish hail from non-hail days. Overall, seasonal-mean anomalies south of the Alps are weaker than those found north of the Alps (not shown), reflecting that southern Switzerland is climatologically closer to the convective threshold. Consequently, hail formation south of the Alps appears to be less dependent on strong seasonal-scale thermodynamic preconditioning and more sensitive to synoptic and mesoscale triggers acting within a climatologically already favorable environment.

Our results position annual hail variability within the broader context of large-scale teleconnection patterns that modulate atmospheric circulation over the North Atlantic-European sector. Previous studies have identified a complex and regionally dependent relationship between the NAO and convective storm activity. While and reported reduced observed thunderstorm activity during positive NAO phases over western and central Europe, they also found that positive NAO phases are associated with more thunderstorm-favorable thermodynamic environments. In contrast, reported increased thunderstorm activity during positive NAO phases over parts of western Europe and attributed the apparent discrepancy with primarily to seasonal effects.

Our Z500 anomaly patterns associated with active hail seasons in northern Switzerland project more strongly onto a positive NAO-like structure. However, additional correlation analyses reveal only weak relationships with the NAO, with opposing signs depending on study region and time scale (Appendix Figs.  and ). This weak NAO signal is consistent with previous work showing that the NAO exerts competing dynamical and thermodynamical influences on convection. Negative NAO phases are associated with large-scale ascent through more frequent shortwave troughs but are often associated with reduced low-level moisture availability, whereas positive NAO phases suppress large-scale lifting while enhancing moisture transport into Europe . These competing effects can offset one another, leading to weak or regionally varying relationships between the NAO and convective activity.

In contrast, EA and SCAND patterns consistently exhibit the strongest correlations with hail frequency across all temporal aggregations (monthly, seasonal, annual, and various lead periods), with generally higher correlations south of the Alps. North of the Alps, the strongest correlations are positive for both EA and SCAND, in agreement with previous findings that positive phases of these modes enhance convection-favorable conditions across Europe by modifying meridional temperature gradients and steering large-scale flow anomalies . South of the Alps, the dominant circulation pattern, characterized by a deep trough over the British Isles and a ridge over the Mediterranean, resembles the positive EA phase or the negative SCAND phase, for which the highest correlations are found. The latter contrasts with parts of the existing literature and highlights that the same mode of variability can exert different, and even opposite, influences depending on region. Weaker correlations with the EAWR and WP patterns further suggest that hail variability is influenced by a combination of teleconnections rather than a single dominant mode such as the NAO.

Notably, winters preceding active hail seasons in both regions are characterized by Pacific SST anomalies resembling a positive phase of the PDO. This may either point to a simple co-ocurrence or to a potential role of low-frequency Pacific variability in modulating the North Atlantic–European circulation through large-scale teleconnection pathways, thereby preconditioning the atmosphere for enhanced hail activity in the subsequent summer. While a direct causal link between Swiss summer hail activity and preceding winter anomalies in Pacific SSTs (or related land-surface signals such as Eurasian snow cover or continental surface temperatures) cannot be firmly established, these precursor signals nonetheless suggest potential for seasonal predictability. Because SST anomalies evolve slowly, their state several months in advance can be predicted with greater skill than synoptic-scale atmospheric conditions or local moisture fields. Moreover, the fact that most regions display opposite anomaly patterns during inactive hail seasons compared to active ones (not shown) further supports potential predictive value of these precursor signals.

Surface boundary conditions modulate the thermodynamic environment in which hail-producing convection develops and can therefore influence both the frequency and intensity of convective storms. Mediterranean SST anomalies emerge as a coherent and persistent feature of hail-active seasons in both northern and southern Switzerland. The Mediterranean has been identified as an important moisture source for Switzerland e.g.,, and positive SST anomalies can enhance moisture availability through increased surface latent heat fluxes and by raising the moist static energy of continental air masses, even in the absence of direct advection .

The importance of the Mediterranean as moisture source differs between the two regions. South of the Alps, positive SST anomalies coincide with enhanced latent heat fluxes and southerly low-level flow, suggesting a more direct moisture contribution to hail-favorable environments. North of the Alps, by contrast, lower-tropospheric moisture anomalies over the Mediterranean are weaker, and enhanced moisture availability appears to be driven more by continental processes, such as increased evapotranspiration from wetter-than-average soils in areas adjacent to northern Switzerland (e.g., along the Atlantic coast), or by moisture transported from the Atlantic associated with the incoming trough.

Land–atmosphere coupling may further modulate these environments. Soil moisture anomalies exhibit memory on subseasonal to seasonal timescales and can influence near-surface temperatures through changes in evapotranspiration and sensible heating . In addition, spatial soil-moisture gradients can support convective initiation by inducing mesoscale circulations that enhance low-level convergence and lift , potentially acting in combination with synoptic-scale forcing and orographic effects. In the midlatitudes, these gradients rarely trigger convection on their own but can enhance existing convective potential, whereas in the subtropics, climatological soil-moisture gradients can actively initiate convection .

Despite the presence of statistically significant SST and soil-moisture anomalies during hail-active seasons, it remains difficult to unambiguously determine whether these surface boundary conditions act primarily as drivers of enhanced hail activity or whether they arise as a response to the prevailing large-scale circulation. Resolving this causality would require targeted numerical sensitivity experiments that explicitly isolate the role of surface boundary conditions.

South of the Alps, composites during hail-active periods are consistent with the recurrent presence of an elevated dry layer. The co-occurrence of warm, relatively dry air aloft and moist near-surface conditions, together with positive CAPE and enhanced CIN anomalies, indicates a capped environment in which convective initiation is temporarily suppressed (Fig. ). Such an elevated mixed layer (EML) can delay convection, allowing instability to accumulate and favoring more intense convective development once the cap is eroded by synoptic-scale ascent or orographic lifting. This configuration is dynamically consistent with the prevailing large-scale flow during active hail seasons south of the Alps, characterized by a trough over the British Isles, a ridge over southern Europe, and southwesterly flow advecting warm, relatively dry air toward the Alpine region. Similar environments have been documented for severe and giant-hail events in the Po Valley .

Although the upper-level trough is centered north of the Alps, its downstream influence can affect convective development even south of the Alpine crest. Ahead of the trough, quasi-geostrophic ascent associated with positive differential vorticity and temperature advection promotes large-scale lifting and supports frontogenesis. In Switzerland, up to 45 % of hail events south of the Alps occur in pre-frontal environments, most commonly associated with cold fronts approaching from the northwest . The interaction between synoptic-scale forcing and Alpine orography can enhance low-level moisture convergence, orographic lifting, vertical frontal circulations, and deep-layer wind shear . Thus, even when the trough axis remains north of the Alps, its associated frontal structures and dynamical forcing can play a key role in triggering and organizing severe hail-producing convection in southern Switzerland.

The strong co-occurrence of significant atmospheric, oceanic, and land-surface anomalies across the same spatial and temporal scales highlights the multifaceted nature of hail season variability. No single factor, such as warm Mediterranean SSTs or local soil-moisture deficits, alone determines whether a season becomes hail-active; rather, it is the combined configuration of these elements that creates a conducive environment. This coupling helps explain why five of the ten most active hail seasons north of the Alps coincide with those south of the Alps: certain large-scale features, including anomalously warm Mediterranean SSTs, Central European ridging, and enhanced southerly moisture transport, simultaneously promote convective activity on both sides of the Alpine divide. At the same time, this overlap should be considered when interpreting the regional differences. Since half of the most active seasons are common to both regions, the contrasts identified between the northern and southern composites primarily arise from the remaining five region-specific seasons, which exhibit the strongest regional differences in the atmospheric anomaly patterns. They therefore highlight the atmospheric conditions that distinguish uniquely active seasons in each region, rather than representing entirely distinct circulation regimes.

Despite this overlap, the coherent differences identified in the circulation, thermodynamic, and surface-boundary anomalies indicate that Swiss hail variability is shaped by the interplay of region-specific, recurrent synoptic forcing, thermodynamic preconditioning, and slowly evolving boundary conditions. Multiple teleconnection patterns, particularly the EA and SCAND phases, show strong correlations with hail activity in both regions, while Pacific SST anomalies may provide valuable seasonal precursor signals. Taken together, these findings underscore the potential to improve (sub-)seasonal hail prediction by combining large-scale climate modes with regional circulation and land-atmosphere indicators operating across multiple spatial and temporal scales.