Thunderstorm and hailstorm environments in Mexico

Thunderstorm and hailstorm environments from 1981 to 2018 are analysed using the proximity soundings approach based on the European Centre for Medium‐Range Weather Forecasts Reanalysis v5 (ERA5) reanalysis dataset. Given the nature of the severe weather dataset, four proximity soundings 0000, 0600, 1200, and 1800 local time were retrieved per day with hail activity or thunderstorms. After quality control, a total of 30,760 vertical profiles were used and grouped seasonally and spatially to consider the local variation of the convective environments. Diverse instability, kinematic, and composite parameters were calculated to characterize the environments for these two convective hazards. The results indicate differences mainly modulated by orographic features, continentality, regional climate, and general synoptic characteristics. Higher instability environments of thunderstorms and hailstorms are observed in the subregions near moisture sources such as oceans, principally during summer and autumn when tropical activity occurs. High wind shear environments are concentrated during winter and spring, mainly in central and northern regions of the country, associated with the passage of frontal systems. The covariates are reasonably accurate at predicting severe weather, but their prediction skill decreases during the warm season when severe weather is more common in Mexico. Increased knowledge of severe weather and associated convective hazards will improve weather forecasting and may be helpful in disaster risk reduction due to meteorological hazards in Mexico.

action to mitigate disasters. This is especially important in developing countries, such as Mexico, due to the lack of reliable observation networks and short-range weather forecasting systems. Improvements in severe weather events forecast depend on increasing knowledge about the environmental conditions under which these phenomena occur (Kahraman et al., 2017). Several investigations have employed real atmospheric sounding measurements and proximity reanalysis soundings to examine the characteristics of convective hazards like thunderstorms, hail, and tornadoes (Rodríguez & Bech, 2021;Taszarek et al., 2020Taszarek et al., , 2021Tuovinen et al., 2015;Westermayer et al., 2016). These approaches have provided helpful information about convective environments in several parts of the world and valuable thresholds for different severe weather parameters (Rodríguez & Bech, 2021). Moreover, the results from such studies have shown significant differences in instability, kinematic, and composite parameters in different geographical regions (Brooks, 2009;Púčik et al., 2015;Taszarek et al., 2017).
It is important to note that a significant number of characterizations of severe convective environments studies use reliable severe weather databases such as the European Severe Weather Database (ESWD) or the Severe Weather Data Inventory (SWDI) from the United States. However, other researchers have opted to produce their datasets based on weather radars, fieldwork, or satellite products (Medina et al., 2022;Valdés-Manzanilla, 2022;Zhou et al., 2021). In this sense, no standardization exists to create a severe weather database, thus, representing a limitation in investigating severe convective environments. Regardless of the source of severe weather data, the computation of diverse parameters near the time and location where any convective hazard occurs (e.g., through proximity soundings) provides essential information that might not be obtained when computing thresholds by statistical parameters (Blamey et al., 2017;Le on-Cruz et al., 2022;Li et al., 2020).
Characterizing severe convective environments in countries with large territorial extensions represents a major challenge for different reasons. In the first place, the geographic location determines, to a large extent, the presence of synoptic systems generators of instability environments, such as tropical cyclones, easterly waves, or cold fronts. Likewise, local conditions that can act as trigger mechanisms for storms, such as the orography, can vary significantly in relatively small areas. Such heterogeneity of conditions provides diverse scenarios where severe weather can develop. In this sense, the commonly used approach where convective environments are characterized nationally, continentally, or globally can mask local processes generating severe storms and associated convective hazards. For this reason, several studies on the characterization of severe environments at the local and regional levels have recently been conducted in various parts of the world (e.g., Brown & Jackson, 2009;Dupilka & Reuter, 2011;Medina et al., 2022).
In this context, this research aims to characterize, regionally, the convective environments of thunderstorms and hail in Mexico during the 1981 to 2018 period. Previous climatologies (Vidal Zepeda, 2007) indicate that thunderstorms in the country are related to orographic forcings, such as the Sierra Madre Occidental (SMOc), Sierra Madre del Sur (SMS), the central and western portions of the Trans-Mexican Volcanic Belt (FV), and the Sierra de Chiapas (SCh). Furthermore, their seasonal distribution indicates that thunderstorms usually occur from the end of spring to the beginning of autumn (from May to October) during the afternoon and night hours. Hailstorms are commonly reported over the highest portions of the FV and the SMOc, principally from May to June (Prieto Gonz alez et al., 2020;Vidal Zepeda & Matías Ramírez, 2007). It is important to note that none of the mentioned research examined meteorological parameters related to these convective hazards.
In this regard, some studies have examined deep convective clouds related to severe weather. For example, mesoscale convective systems (MCS) over northwestern Mexico have been widely analysed during the North American monsoon (NAM) (Farf an et al., 2021;Ramos-Pérez et al., 2022;Valdés-Manzanilla, 2015). Likewise, the spatiotemporal distribution of precipitation and deep convective clouds over FV has been analysed, illustrating that both events are related to the orographic features (Brito-Castillo et al., 2022;Le on-Cruz et al., 2021). In addition, other research has examined the spatiotemporal distribution and characteristics of severe local storms, including hail, and some convective modes, as supercell storms (Concepci on et al., 2007;Edwards, 2006;Weiss & Zeitler, 2008).
Although the available scientific research on thunderstorms and hail in Mexico, studies have yet to analyse and compare the convective environments associated with these meteorological hazards at a regional level. The complex orography in Mexico and its geographic location, provide a wide variety of scenarios where severe weather can develop. The examination of the characteristics of such environments in the different portions of the country will provide greater insight into local severe weather-generating processes. In this context, this research uses datasets from the Mexican National Weather Service (SMN) and the European Centre for Medium-Range Weather Forecasts (ECMWF) Reanalysis v5 (ERA5) to generate proximity soundings of thunderstorms and hailstorms during the 1981 to 2018 period. Section 2 describes the study area and subregions defined for the analyses; the datasets and methodologies are also shown in this section. In Section 3, the results are presented. Here, the instability, kinematic, and composite parameters are also discussed. The summary and conclusions are given in Section 4.

| Study area
Mexico is located in North America between the extreme coordinates 32.71 N, −122.17 W, and 11.96 N, −84.64 W (Figure 1). According to the last population and housing census (Instituto Nacional de Estadística y Geografía, 2022), $126 million people live in the country. The mean annual precipitation in Mexico is 745.9 mm, concentrated in the summer months (June-September). However, given the territory extension and its heterogeneity, many different spatiotemporal precipitation patterns and extreme events are present in Mexico (Colorado-Ruiz & Cavazos, 2021).

| Severe weather database and regionalization
The thunderstorm and hail databases were provided by the SMN of the National Water Commission (CONAGUA) of Mexico. These databases contain information from 83 meteorological observatories distributed around the country with different operating periods. The provided databases contain daily reports about the occurrence of thunderstorms and hail based on information provided by weather observers. It is important to note that the records follow the criteria of the World Meteorological Organization (WMO) guidelines. In this sense, a thunderstorm is characterized by a rumbling or thunder accompanied by lightning (luminous glow). Hail (also named hailstorms hereafter) is the precipitation of globules or ice chunks whose diameter can vary from 5 to more than 50 mm (Servicio Meteorol ogico Nacional and Comisi on Nacional del Agua, 2022). It is worth mentioning that the databases do not provide information on the severity of the events.
Data from the 83 SMN observatories were filtered, considering the 1981-2018 period. In addition, the observatories should have 300 months of data, that is, at least 25 years of reliable data. After this simple filter, 50 observatories were  (Cavazos, 1999), the Central Highlands (CH) with 10 observatories (Peralta-Hern andez et al., 2008), and the Southwest (SW) with five observatories (Andrade-Vel azquez & Medrano Pérez, 2021). This previous analysis was necessary because no official Mexican database exists on severe weather.

| ERA5 reanalysis
The vertical profiles of geopotential, specific humidity, temperature, and wind from each thunderstorm and hailstorm day were retrieved from the ERA5 reanalysis (Hersbach et al., 2023a(Hersbach et al., , 2020. This dataset has global coverage with a horizontal resolution of 0.25 × 0.25 and a 1-h time-step. ERA5 covers the atmosphere in 37 pressure levels, from 1000 to 1 hPa, from 1959 up to date. In addition, terrain elevation, surface pressure, and 10 m wind were retrieved from the same database (Hersbach et al., 2023b) and were used in vertical interpolations and other necessary computations. It is important to note that given the complex Mexican orography (altitude ranges from zero to more than 5000 masl), the lowest level of the vertical profiles was compared with the elevation of the sampling point to eliminate values below the sampling point. The reanalysis data provided a better temporal resolution than sounding observations and allowed the generation of vertical profiles in more than 15 sounding stations prevalent in Mexico. Likewise, reanalysis offers advantages to understanding convective parameters with substantial variations in short periods (Potvin et al., 2010;Taszarek et al., 2021). The reliability of ERA5 reanalysis has been proven recently by comparing them with real soundings worldwide; including stations in Mexico and surrounding areas (Li et al., 2020;Taszarek et al., 2021).

| Proximity soundings
The SMN-CONAGUA data do not report the time of the thunderstorm or hail documented. Thus, four representative proximity soundings were computed from ERA5 reanalysis data for each active day with thunderstorms or hail (AD): 0000, 0600, 1200, and 1800 local time (LT). A total of 136,124 proximity soundings were performed using the Sounding/Hodograph Analysis and Research Program in Python (SHARPpy) (Blumberg et al., 2017). This total represents 34,031 AD: 1919 proximity soundings with hailstorms and 32,112 with thunderstorms. Then, a subset of soundings was retrieved by selecting the most unstable sounding from each AD using the most-unstable convective available potential energy (MUCAPE) as a discriminator (Potvin et al., 2010;Zhou et al., 2021). Data quality control was applied to avoid unrealistic data and select those representing an acceptable range of parameters capable of reflecting various convective hazards. Soundings with MUCAPE <150 JÁkg −1 , wind shear values >50 and 100 mÁs −1 (0-1 and 0-6 km, respectively), and lapse rate 0-3 km > 11 C were removed. Furthermore, it was also assumed that the largest and smallest values represent potential outliers that can be related to contaminated soundings (Allen & Karoly, 2011;Potvin et al., 2010;Zhou et al., 2021). After this quality data control, a final set of 30,760 proximity soundings representing 29,268 thunderstorms (91.14% of the original set) and 1492 hailstorms (77.75% of the original set) were considered (Table 1).

| Parameters analysed
The instability, kinematic, and composite parameters analysed were based on previous studies on severe convective environments (e.g., Rodríguez & Bech, 2021;Taszarek et al., 2020Taszarek et al., , 2021Zhou et al., 2021). The convective available potential energy (CAPE) parameter is commonly used to discriminate the severity of weather events (Blamey et al., 2017) and to characterize severe convective storms (Brown & Jackson, 2009;Dupilka & Reuter, 2011). The complement of CAPE is convective inhibition (CIN), which helps identify cases where a mechanism is necessary to force lifting and start the convection (Rodríguez & Bech, 2021). In this research, CAPE and CIN calculations were carried out using the virtual temperature correction (Doswell III & Rasmussen, 1994) in the mostunstable (MU) parcel. The parcel selection allows capturing pre-storm environments, which is helpful because the exact hour of the storm occurrence is unknown (Rodríguez & Bech, 2021;Sherburn & Parker, 2014;Zhou et al., 2021). Furthermore, the mid-level lapse rate (700-500 hPa) calculations were also included. This parameter allows for identifying environments that favour vertical motion and where deep convection is probable.
Wind shear is usually used to identify the storm organization potential. Previous investigations have shown this kinematic parameter is crucial, especially in developing supercell storms and hail growth (Bunkers, 2002;Dennis & Kumjian, 2017). Diverse studies focused on potentially severe convective environment identification employed wind shear to discriminate between severe and non-severe environments (Blamey et al., 2017;Brooks et al., 2003;Dupilka & Reuter, 2011). The storm-relative helicity (SRH) helps to identify the potential for cyclonic updraft rotation in right-moving supercells (Davies-Jones, 1990). SRH is a mostly operational parameter but is also used in the analysis of storm environments (Kahraman et al., 2017;Rodríguez & Bech, 2021;Taszarek et al., 2020). Most studies that reveal relevant information about their role in storm formation were conducted in mid-latitude regions. However, these parameters have been poorly explored in tropical regions, including Mexico.
Finally, the relationship between CAPE and wind shear was investigated using the WMAXSHEAR parameter and two covariates related to severe convective environments. WMAXSHEAR was introduced recently by Taszarek et al. (2017) and is defined in Equation (1). The two covariates selected (Equation (2) and (3)) were also based on previous studies (Blamey et al., 2017;Brooks et al., 2003;Gensini & Ashley, 2011) and were recently applied in Mexico (Le on-Cruz et al., 2022). The computation of both considered only the cases when MUCIN ≥ −75 JÁkg −1 . The 0-6 km wind shear (DLS) and MUCAPE were selected for all the calculations in this section.  Figure 2 shows the seasonal distribution of the relative frequency of thunderstorms and hailstorms days (1981-2018) based on SMN-CONAGUA reports. In general, hailstorms in Mexico are concentrated mainly at the end of spring and summer. The regionalization shows that the NW subregion does not show substantial hail activity. Similar seasonal patterns are observed in CH and FV subregions, with hailstorms concentrated in spring and summer. In these regions, the cold front activity, mainly the winter-spring transition (Le on-Cruz et al., 2021; Pita-Díaz & Ortega-Gaucin, 2020), seems to provide favourable environmental conditions for decreasing the freezing level and increasing the probability of hail formation. The GM subregion shows mild hail activity in spring, summer, and autumn, while the YP subregion exhibits hail activity principally during winter, which can be explained by frontal systems passing over the region (Cahuich-L opez et al., 2020), which tends to decrease the freezing level, favouring the hail formation. The SW subregion shows significant hail activity from spring to autumn, with hailstorms concentrated in the summer months and previously classified as hailstorms in tropical regions with significant orography (Zhou et al., 2021). A similar temporal pattern is observed in the FV and CH subregions but with characteristics of midlatitude elevated terrain hailstorms (Zhou et al., 2021). The seasonal distribution from the database used is consistent with Vidal Zepeda and Matías Ramírez (2007).

| Seasonal distribution
On the other hand, thunderstorms in Mexico ( Figure 2) are concentrated in summer and autumn. This seasonal thunderstorm pattern is supported by other studies on extreme precipitation events (Colorado-Ruiz & Cavazos, 2021;Vidal Zepeda, 2007) and is associated with easterly waves and tropical cyclones (Agustín Breña-Naranjo et al., 2015;Dominguez et al., 2020). The regionalization shows a very similar temporal pattern in almost all subregions, that is, two peaks of high activity during summer and autumn, a transition phase in spring, and low activity during winter. Remarkably, a bimodal distribution is observed only in the NW subregion (active vs. inactive season), which has been related to the influence of the North American Monsoon (Boos & Pascale, 2021;Ramos-Pérez et al., 2022). The highest thunderstorm activity in FV has been associated with the role of complex terrain features in generating deep convection (Brito-Castillo et al., 2022;Le on-Cruz et al., 2021). The GM and YP seasonal patterns are associated with the previously mentioned tropical activity (i.e., easterly waves and tropical cyclones) over the Caribbean Sea. In the SW subregion, not only tropical systems have an influence, but also the northward displacement of the Intertropical Convergence Zone (ITCZ) from spring to autumn (Andrade-Vel azquez & Medrano Pérez, 2021).

| Diurnal distribution
The SMN-CONAGUA dataset does not contain information on the hourly distribution of hail and thunderstorms. However, the methodology used, based on selecting the most unstable sounding, allowed an approximation of the diurnal distribution analysis of these convective hazards (Figure 3). The diurnal patterns of hailstorms and thunderstorm soundings exhibit very similar characteristics with a higher frequency between 1200 and 1800 local time (1800 and 0000 UTC), that is, related to afternoon convection events. In second place are the early morning storms, followed by a contrasting drop in the convective activity during the morning. The diurnal distribution by subregions does not show substantial differences. The dominance of afternoon events presented here is consistent with other investigations on thunderstorms and hailstorms in Mexico (Prieto Gonz alez et al., 2020;Vidal Zepeda, 2007) and with lightning events (used as an indicator of storms) across the United States and Europe . Afternoon convection is a wellknown pattern mainly observed in continental landmasses worldwide related to daytime heating.

| Instability parameters
CAPE presents differences between subregions and over the seasons (Figure 4). Mainly, the highest CAPE medians for thunderstorms are present in GM during Pacific Ocean) could be related to this behaviour (Riemann-Campe et al., 2009). The role of sea breezes in the increase of CAPE values along coastal regions in Mexico could also play an important role and needs to be investigated in depth. The CAPE medians are lower than those observed in the U.S. but more similar and even higher than in several parts of Europe (Kahraman et al., 2017;Taszarek et al., 2017;Taszarek et al., 2020).
Particularly in the NW subregion, the influence of the NAM in summer and autumn is observed with high environmental instability, consistent with previous reports of MCS (Farf an et al., 2021). Over the GM subregion, the highest CAPE environments during spring (medians around 2000 JÁkg −1 for thunderstorms and hail) have been associated with a change in wind patterns from the Gulf of Mexico (Le on-Cruz et al., 2022) coincident with the beginning of a severe weather season in northeastern Mexico and the southern United States. For the SW and YP subregions, the increased instability environments (medians range 1800-2200 JÁkg −1 during summer and autumn) can be associated with tropical activity such as the northward displacement of the ITCZ (Andrade-Vel azquez & Medrano Pérez, 2021; Riemann-Campe et al., 2009) and the role of easterly waves and tropical cyclones in the continental portion of Mexico (Dominguez et al., 2020). Finally, lower CAPE environments in CH and FV (medians lower than 1500 JÁkg −1 ) are related to the locations of these sites with terrain elevations over 2000 masl.
The CIN computations for the different subregions over the seasons do not show substantial differences ( Figure 5). The CIN medians in all seasons are higher than −100 JÁkg −1 for thunderstorms and hailstorms, which is consistent with other studies (Gensini & Ashley, 2011;Taszarek et al., 2020). The threshold of −75 JÁkg −1 proposed for significant severe storms (Bunkers et al., 2010) is not exceeded in any case (except for NW and GM subregions during winter), considering only the medians. The median CIN values higher than the proposed threshold may indicate that an additional mechanism is not necessary to reach the free convection level in the analysed storms. In the cases represented by the values between the 25th and 5th percentiles, which may exceed the −300 JÁkg −1 , a mechanism to force lifting should be important (Rodríguez & Bech, 2021). The complex terrain features in some parts of GM, CH, and FV subregions produce orographic lifting (Le on-Cruz et al., 2021). It is important to note that the lowest CIN values during winter in NW may not be representative of the true distribution due to the small sample size.
The median 700-500 hPa mid-level lapse rates range between 6 C and 7.5 C for both types of activity ( Figure 6). Similar values have been reported previously in the United States and Europe (Kahraman et al., 2017;Taszarek et al., 2020). Seasonal behaviour shows higher lapse rate medians during the winter and spring months, decreasing in summer and autumn. Higher lapse rates in Mexico during the cold and transition seasons can be associated with synoptic patterns , as frontal systems (Luna-Niño & Cavazos, 2018;Pineda-Martinez & Carbajal, 2009). These systems occurred in winter and early spring, principally over CH, FV, GM, and NW subregions, consistent with the computed high medians.

| Kinematic parameters
The 0-6 km deep-layer shear (DLS) from thunderstorms and hailstorms proximity soundings show a similar behaviour among subregions and seasons, except for the SW subregion (Figure 7). The highest wind shear values occur in the winter and spring, when frontal systems are the most significant meteorological phenomena over Mexico (Lagerquist et al., 2020). Hailstorms during winter in CH (24 mÁs −1 ) and GM (27 mÁs −1 ) subregions show higher wind shear medians, which may be related to the passage of cold fronts (Luna-Niño & Cavazos, 2018;Pineda-Martinez & Carbajal, 2009). In the SW subregion, wind shear remains stable (around 5-7 mÁs −1 ) for thunderstorms and hailstorms during the seasons due to its geographic location, away from the influence of frontal systems.
The threshold of 15 mÁs −1 , commonly used in determining the potential for severe storms (Blamey et al., 2017;Brooks et al., 2003), is reached only in a few cases based on the medians (Figure 7). Low wind shear environments characterize hailstorms and thunderstorm environments in Mexico for summer and autumn (medians around 5 mÁs −1 ) with intermediate-high wind shear environments (medians around 13 mÁs −1 ) in the spring. The wind shear medians obtained here are lower than those for the United States, but similar to those in Europe for diverse convective hazards (Coffer et al., 2020;Taszarek et al., 2020). For hailstorms, the results are consistent in the CH, FV, and GM subregions, which are identified as midlatitude elevated terrain hail (Zhou et al., 2021). The relatively low wind shear values obtained in this analysis are consistent with supercell storms being unusual in Mexico, except for the northeastern portion (Le on-Cruz et al., 2022).
The 0-3 km SRH follows the pattern shown in wind shear but with values significantly lower than expected for severe storms (Figure 8). The 0-3 km SRH medians do not exceed the 100 m 2 s −2 thresholds, except for: NW during spring and CH and GM in winter. During summer and autumn, medians are around 20 m 2 s −2 , which can be considered negligible. In cases of considerable hail activity, the SRH medians show values of 50 m 2 s −2 in CH and 40-60 m 2 s −2 in YP during spring. For thunderstorm activity, the values are around 50 m 2 s −2 in GM during spring. The SRH values obtained are significantly lower than those computed in North America and Europe (Dupilka & Reuter, 2011;Púčik et al., 2015;Kahraman et al., 2017;Taszarek et al., 2020Taszarek et al., , 2021. The low SRH values for both thunderstorm and hail environments can be related to the proximity soundings used since this parameter is highly dependent on time and space scales (Bunkers, 2002), and soundings here are approximated to the actual hour of the genesis of the storm. Likewise, the results suggest that the main number of storms analysed are non-mesocyclonic. The SRH values obtained indicate that this parameter is not useful for characterizing thunderstorms and hailstorms in most of the Mexican territory.

| Composite parameters and covariates
The Supercell Composite Parameter (SCP), Significant Tornado Parameter (STP), and Significant Hail Parameter (SHiP) (Thompson et al., 2004) were calculated for each proximity sounding of thunderstorm and hail but did not provide any relevant information. The values are around zero in all the analysed seasons and subregions (not shown). Therefore, such composite parameters were not integrated into the formal analysis. However, the relationship between CAPE and wind shear was investigated using the WMAXSHEAR parameter and two covariates related to severe convective environments. In all the cases, we consider only the soundings where the CIN >−75, as previous research, suggests (Gensini & Ashley, 2011). A total of 29,147 proximity soundings were used for these analyses: 1390 correspond to hailstorms and 27,757 to thunderstorms.
The percentage of proximity soundings for thunderstorms and hailstorms exceeding the proposed threshold value is shown in Figure 9. In the case of the C-composite Index (Gensini & Ashley, 2011), the percentage of soundings that reach the proposed threshold value is under 25% for CH, FV, and SW subregions in all seasons. Interestingly, 100% of hailstorm soundings for NW reach the threshold value selected during winter, and thunderstorm cases are around 37% from spring to autumn. In YP, the highest percentages occurred in spring, reaching 50% when GM showed the highest values ($78%) for hail and thunderstorms ($60%). The low percentages obtained from this index can be related to the more conservative approach in depicting significant-severe environments, which is inherent to the index that weights equally wind shear and CAPE (Gensini & Ashley, 2011). Additionally, such environments in several portions of the country do not follow the typical pattern of midlatitudes.
Generally, the B-composite Index (Figure 9) performs better than the C-composite Index. From this index, the percentages increase to $50% for thunderstorms and hail events during winter and spring over the CH and FV subregions, and it decreases to around 14% for the rest of the seasons. A consistently good performance is observed for the NW subregion, except for hail events during summer. SW subregion shows the lowest performance, with percentages around 25% for all seasons. YP has a similar pattern to the CH and FV subregions but higher percentages, from 25% to 75%, during winter and spring, respectively. Over this subregion, B-composite and Ccomposite indexes reduce their values by 25% in summer and autumn. Finally, the highest percentages in GM are observed during winter and spring (in B-composite Index), ranging from 75% to 100%, and the percentages decrease dramatically under 20% during summer and autumn.
It is interesting to mention how the skill of both indexes to detect severe environments decreases during the summer and autumn seasons, when tropical phenomena prevail. However, the use and the weighting of wind shear could be associated with a dramatic decrease in detecting severe environments during the warm season. In this sense, it is known that tropical phenomena are more common than mid-latitude systems during the summer and beginning of the autumn (Agustín Breña-Naranjo et al., 2015;Dominguez et al., 2020). This seasonal behaviour is also depicted by a decrease in wind shear. The null presence of cold air masses pushing cold fronts over the territory reduces the chances of having high-shear environments.
Recently, the WMAXHSEAR parameter has been pointed out as one of the best discriminators between severe and non-severe thunderstorms (Taszarek et al., 2017). Results from the United States and Europe suggest that this parameter is helpful in the detection of increasing probability of severe thunderstorms and other convective hazards (Rodríguez & Bech, 2021;Taszarek et al., 2020). For the hailstorm proximity soundings analysed (Figure 10), the best performance was obtained for the YP, NW, and GM subregions with 45%, 45%, and 37%, respectively, of the cases exceeding a WMAXHEAR of 500 m 2 s −2 . As the covariates used, the worst performance was detected for the SW (20%), FV (22%), and CH (23%) subregions. For thunderstorm proximity soundings (Figure 11), the observed performance is significantly lower. To this convective hazard, the highest percentage was obtained for the NW subregion, with 45% soundings exceeding a threshold value of 500 m 2 s −2 . Subsequently, the GM, SW, and YP subregions can be found with 24%, 28%, and 23%, respectively. The worst performance is observed in subregions located in high altitudes: CH and FV, with 16% and 21% of proximity soundings with WMAXSHEAR greater than 500 m 2 Ás −2 .
Given the results from the covariates and composite index analysed, the proximity with oceans (which tends to increase CAPE values) and the latitude (which increases the probability of cold fronts and wind shear environments) are relevant. Generally, in the subregions meeting these characteristics, NW, GM, and partially YP, using these parameters is moderately helpful (especially in spring). In the case of the regions located over high altitudes (FV and CH), none of the analysed parameters is a good discriminator for severe storms and related convective hazards. For this portion of the country, intermediate-low shear and low-CAPE environments are typical for the formation of thunderstorms and hailstorms. Previous investigations have suggested the critical role of orography as a primordial mechanism for lifting and trigged convection (Brito-Castillo et al., 2022;Le on-Cruz et al., 2021). Further investigations are required to create a proper severe weather index for regions with such characteristics. Finally, in the case of the SW subregion, severe weather activity is likely related to tropical features, such as easterly waves, tropical cyclones, and the ITCZ.

| SUMMARY AND CONCLUSIONS
The present work aims to characterize the severe convective environments in Mexico. It also seeks to establish a regional analysis that allows identifying local differences for convective hazards, such as thunderstorms and hailstorms, considering different local environments modulated by the orography, continentality, prevalent synoptic systems, and climate. For this purpose, the SMN-CONAGUA database with raw daily data on hailstorms and thunderstorms that occurred during the 1981-2018 period was used. Only 50 sites with at least 25 years of reliable data were considered and assigned to six different climatic regions. Subsequently, vertical profiles were retrieved from the ERA5 reanalysis every 6 hours for computing proximity soundings. They were used to analyse severe weather environments using instability, kinematic, and composite indexes at seasonal and diurnal scales for each subregion.
In general terms, thunderstorms and hailstorms in Mexico are common during summer and the beginning of autumn. In the NW subregion, hailstorms are rare, and thunderstorms show a bimodal distribution, which is explained by the NAM activity. The YP subregion concentrates the hail activity during the cold season, in winter and early spring, possibly modulated by frontal systems activity. The FV, CH, and GM subregions show the high activity of the two convective hazards during spring; the synergy of late cold fronts and moisture advection from the Pacific Ocean and, mainly, the Gulf of Mexico could explain such behaviour. The intense activity during summer in almost all subregions is associated with tropical activity, such as easterly waves and tropical cyclones. The diurnal pattern found, mainly related to afternoon storms, is consistent with other studies and the daytime heating convection pattern typical in continental landmasses.
Thunderstorm and hailstorm instability environments in Mexico show significant differences seasonally and spatially. The subregions with significant influence of moisture advection from oceans (e.g., NW, GM, SW, and YP) show the highest instability from CAPE and lapse rates. These unstable environments are observed particularly during summer and autumn in SW, YP, and NW derived from the tropical activity and NAM influence, An extra mechanism to force convection is not necessary in most cases.
The wind shear indicates a bimodal behaviour in all subregions except for SW, where this parameter seems unimportant. During summer and autumn, the shear environments decline dramatically during the warm season, associated with tropical activity. On the other hand, wind shear environments increase during winter and spring derived from the frontal system activity. Interestingly, high instability and wind shear environments converge during spring, especially in GM and YP, which seems to increase the probability of organized convection and various convective hazards. The SRH parameter does not provide relevant information on thunderstorms and hail activity in Mexico. Significant variations in short periods of this parameter are a possible explanation for the findings.
The composite indexes show good performance, mainly in GM, NW, and YP subregions during winter and spring. The remarkable drop in the performance of these indices in summer and autumn could be associated with the warm season characterized by tropical activity (i.e., easterly waves and tropical cyclones) and low-shear environments. The WMAXSHEAR threshold value (500 m 2 s −2 ) shows better results for hailstorms in the same subregions and similar behaviour but with lower efficiency for thunderstorms. The proximity to oceans, latitude, and terrain height is the most relevant characteristic associated with the performance of these parameters.
On average, the results from instability, kinematic, and composite parameters obtained are lower than those observed in the United States but quite similar to several parts of Europe. It is noteworthy that the thresholds in these parameters should be adapted by subregion and season if they are intended to be used in forecasting severe storms and related convective hazards. Likewise, new convective parameters must be tested to improve severe storm forecasts in subtropical regions such as Mexico. The orographic complexity, the geographic location, and the complex processes observed at local levels make threshold-based convective hazard forecasting a major challenge for weather forecasters in the country. The results from this research are one of the first steps in the national context to improve the knowledge of severe weather and associated convective hazards, which may be helpful in disaster risk reduction due to extreme meteorological events.