Evaluation of the Radar Echo Tops in Catalonia: Relationship with Severe Weather

Abstract

Strong updrafts occur in severe thunderstorms, causing the overshooting tops, an increase in the total lightning activity, and generating a frozen drops nucleus that will produce severe weather when it collapses. The Echo Top is a measurement of the vertical development of the cloud, considering a certain reflectivity threshold: the higher the threshold value, the lower the altitude reached. The present research shows the distribution of the Echo Tops of three reflectivity thresholds (12, 35, and 45 dBZ) in Catalonia for the period 2013–2021, comparing the distribution with the maps of hail, lightning jumps, and the topography of the region. The analysis shows how the maxima occurrence of Echo Tops varies depending on the threshold, indicating that thunderstorms have an initial development at 12 dBZ in the mountainous area, while the 35 and 45 dBZ maxima are shifted to the south, in regions with lower mountains. This last maximum is nearly coincident with the region more hit by large hail.

1. Introduction

Severe weather phenomena (large hail, tornadoes, and strong winds) occur in thunderstorms with a high vertical development [1,2]. In these deep convective clouds the updrafts can exceed 20 m/s [3], which is enough for carrying out the embryos to high atmospheric levels (with temperatures close to −40 °C) and favouring their size enlargement. Once the weight of the freezing particles is large enough, the cloud collapses and the mass reduces producing severe phenomena on the ground.

The Echo Top (TOP) is the remote sensing estimation of the height of the cloud, considering a certain reflectivity (a variable which estimates the energy returned from echo with respect to that emitted by the antenna and measured in dBZ, or decibels relative to the radar reflectivity factor, Z) threshold. In the case of the satellite, the Tropical Rainfall Measuring Mission (TRMM) and the continuing Global Precipitation Measurement (GPM) missions have allowed characterizing the TOP in the Tropic region for 20 years [4,5,6], providing very interesting results for understanding the water cycle and for climatic purposes. The largest TOP estimation differences between InfraRed and radar occurred over Land, implying that the updrafts are stronger in this case than over the oceans. In addition, these analyses showed that Echo Tops of 20 dBZ (or TOP20) over the ocean in convective thunderstorms ranged between 4.5 and 6 km. It was also observed that electrical activity is more probable over land than in oceans [7]. In real-time and operational terms, satellite imagery has a high capacity for identifying the Overshooting-TOP (OT) signature in thunderstorms [8], which correlates well with severe phenomena.

In any case, weather radar has been revealed as the best tool for estimating the Echo Top of thunderstorms, because of the time and spatial resolutions and the capability of scanning a large volume of the atmosphere [9,10]. In fact, Echo Tops (and the derived products) have wide use in meteorology from aviation forecasting to severe weather diagnosis. The previous studies found that the method of estimating the Echo Top affects the final product. The estimation is limited by the radar beam size and the number of elevation scans, producing over- and sub-estimations of the magnitudes. Another important element that affects the calculation is the error in the height estimation, caused by a bad antenna alignment [11]. Because of these errors, which limit the hail diagnosis, especially at long ranges, the previous works suggest the use of the composition of different radars. The capability of detecting the cloud development with Echo Top has revealed the development of a new way of estimating the Z/R relationship, similarly to that considering the reflectivity intensity [12]. Results for different episodes show more reliable rainfall maps because the new method captures the nature of the rainfall structure better.

There are several methods for hail diagnosis using direct or indirect Echo Tops. In the first case, Held [13] found that Echo Tops of 23, 40 and 50 dBZ were notably higher in cases of hailstorms in South Africa in front of rainstorms, with mean values of 9.1 ± 2.5, 6.3 ± 2.6, and 5.0 ± 2.3 km, respectively. Other research in Estonia [14] considered Echo Tops from zero to 45 dBZ with steps of five dBZ, to find the best hail estimator. The skill scores showed that Echo Tops of 15 dBZ (TOP15) presented the best performance. Another result is that Echo Tops of thunderstorms reached heights of 12 km in that region. Other works combined TOP with other derived products. Two of the most common methodologies for detecting hail using weather radar are VIL (Vertical Integrated Liquid) and the Waldvogel criterion [15]. VIL is a conversion of the reflectivity data into water content liquid value based on drop-size distribution and a reflectivity factor along the vertical column. The Waldvogel criterion defines the probability of hail detection based on the difference between the TOP45 and the freezing level. Then, in both cases, the vertical development of the reflectivity in the thunderstorm plays the main role in the increase in the hail probability. In the different analyzed cases the TOP45 ranged between 2 and 9 km, with a high probability of hail in those thunderstorms with pixels exceeding the 7 km height and more than 45 dBZ. Moving to Romania, Cică et al. [16] uses the same parameters among the VIL density (VILD, or the relationship between VIL and Echo Top), and the Hail Kinetic Energy (HKE, estimated in a similar way to VIL, but with different parametrization). The VILD presented the best behavior, with a stepped relationship compared with ground registers. Maximum Echo Tops moved between 12 and 13 km in the study. Similarly, Stefan & Barbu [17] analyzed some radar parameters also for Romania: Echo Top, VIL, and VIL density. Echo Top ranged between 4 and 16 km. Echo Tops over 13 km showed a probability of hail of around 90%. In Belgium, Lukach et al. [18] used radar-estimated parameters for diagnosing hail in thunderstorms. In this case they used the Waldvogel method and the Severe Hail Index (SHI), like the Hail Kinetic Energy. In both cases, the values were highly dependent on the Echo Top estimation. Finally, Nisi et al. [19] evaluated the performance of two hail diagnosis products in Switzerland: The Probability of Hail (POH), based on the Waldvogel method, and the Maximum Expected Severe Hail Size (MESHS), which is also a based-derived product from reflectivity and Echo Top, like VIL or HKE.

The comparison between Cloud-to-Ground (CG) flashes and Echo Top revealed that the firsts usually occur when thunderstorms reached 15.2 km [20], with a clear reduction in the lightning activity as the height decreases. The relationship between VIL and CG is less clear than for Echo Tops. One of the most useful applications of lightning data for nowcasting severe weather in thunderstorms is the detection of the Lightning Jump (LJ), which is a sudden increase in the total lightning activity in thunderstorms [1,2,21]. LJ does not present markedly different electrical and radar behaviors depending on the severe weather phenomena (hail or strong wind). In most of the analyzed LJ in individual thunderstorms, the variation in VIL was notably higher than the Echo Top. Additionally, in most of the cases the LJ occurred just before the maximum Echo Top, while the behavior of the VIL was less clear.

The identification of high values of TOP in severe thunderstorms in Catalonia started with [22], with the analysis of events that occurred in 1993 and 1994. The authors found maximum Echo Tops of 13 and 11 km in both cases. This and other works led to the implementation of the Waldvogel method in Catalonia in 2004 [23], using the Servei Meteorològic de Catalunya (Meteorological Service of Catalonia) radar TOP, the numerical weather prediction isozero height (or freezing level), and the hail ground registers provided by the Hail Pad network of the Catalan Western Depression. As a step further for improving the real-time diagnosis and nowcasting of hail, Farnell et al. [2] presented the work to put in operation the Lightning Jump, with 80% success in the severe weather warnings since 2018. Finally, Rigo & Farnell [24] analyzed thunderstorms with multiple LJ. These storms have high values of radar parameters (including Echo Top) for a longer time than usual in single-LJ thunderstorms.

The present research has as its main goal the spatial characterization of the TOP for different reflectivity thresholds in Catalonia, to better understand the life cycle of the convection in the region, having in mind that each analyzed reflectivity threshold (12, 35, and 45 dBZ) helps to understand different conditions of the convective behavior (Overshooting, convective precipitation development, and severe weather occurrence), according with [25]. For this purpose, it has analyzed a large set of composite files for the period 2013–2021. As secondary questions, we have asked about the relationship between the TOP fields, hail registers, and LJ warnings distributions, and between TOP and the topography of the region. These comparisons intend to improve: (1) the relationship between the vertical development of thunderstorms and hail size; (2) the link between the TOP and the severe weather in general; and (3) the influence of the topography in the height of the thunderstorms.

2. Materials and Methods

2.1. Area of Study

The area of study is Catalonia (see the black rectangle of the left panel of Figure 1). It is in the north-east part of the Iberian Peninsula. The Ebro Valley acts as the connector between Catalonia and the rest of the Iberian Peninsula. The boundaries of Catalonia are in the north with France, with the natural border of the Pyrenees. The Catalan Coast runs from the south-west to north-east bordering the Mediterranean Sea.

The geography of the region is very accidental (see right panel of Figure 1) with many ranges. First, the Pyrenees (heights reaching the 3000 m); second, the Pre-Pyrenees (that do not exceed the 2500 m); third, the Pre-Littoral (with maximum height around the 1800 m); and finally, the Littoral (with heights between 300 and 800 m, but very steeped and close to the sea). Between these ranges there are many river valleys and large, flat depression areas. These flat areas and the coastal regions are where there are found the highest densities of population, and the main economic activities (tourism, agriculture, industry…).

2.2. Data Used

2.2.1. Radar Data

The right panel of Figure 1 also shows the location of the four radars of the Servei Meteorològic de Catalunya (two red capital letters). PBE (Puig Bernat, PB in Figure 1) is the radar located close to the central coast, the most populated area. PDA (Puig d’Arques, PD) covers the north-east of Catalonia. CDV (Creu de Vent, CD) has good coverage of the central part of Catalonia. Finally, LMI (La Miranda, LM) is located in the southern part of the region. They are C-Band Single Polarization systems, generating volumes with a range of 130 km and a spatial and temporal resolution of 500 m and 6 min, respectively. The volumes consist of 15 elevations Plan Position Indicator (PPI) fields. The radar distribution pretends to cover volumetrically all the Catalan territory. All the composite products used in this study have a spatial resolution of 1 km × 1 km.

The composite products used in this study are raster (Geotiff format) files of:

• The volumetric CAPPI (Constant Altitude Plan Position Indicator): three-dimensional reflectivity fields with pixels of 1 km × 1 km × 0.5 km resolution. This product allows an understanding of the precipitating structures from a volumetric point of view;
• The VIL (Vertical Integrated Liquid): It is a product useful for discriminating the occurrence of hail, considered in the region for identifying hailfalls affecting agriculture exploitations [26];
• The Echo Tops: They help to determine the maximum vertical development of the clouds. The operational products of the Servei Meteorològic de Catalunya have three reflectivity thresholds. The first one 12 dBZ (or TOP12), equivalent to 0.1 mm/h rain rate, delimitates the cloud top. The 35-dBZ threshold (or TOP35) is useful for identifying regions with moderate-intensity precipitation. Finally, the 45-dBZ product (or TOP45) allows for discriminating the regions with very intense or hail precipitation.

Figure 2 shows an example of a set of thunderstorms that occurred in Catalonia at 1636 UTC on 6 August 2022. In the top panel the 2 km CAPPI indicates a linear structure with maxima reflectivity between 55 and 60 dBZ (red-colored pixels). The black line between the points labelled “A” and “B” marks the cross section presented in the mid panel of the figure, in which it is possible to see the vertical development of the main thunderstorm and, at the left, a less important structure (with reflectivity not exceeding the 20 dBZ). The bottom panel shows the estimation of the Echo Top for the three thresholds (solid lines, in blue for 12 dBZ, in yellow for 35 dBZ, and red for 45 dBZ). The less intense structure presents only values of TOP12, but no signal of TOP35 and TOP45. On the contrary, the TOP35 and TOP45 observations coincide with the region of the highest values of the TOP12 (known as Overshooting Top, or OT, which is the signature of the maximum height reached by the updraft). Another interesting point is the presence of the anvil area, a horizontal region of low-to-moderate reflectivity at high levels (more than 10 km) and not reaching the ground level, surrounding the OT. The fact that the precipitation does not reach the ground is not reflected in the TOP12 profile, which is only useful for the highest development but not for the lowest regions where reflectivity occurs. The last point to consider in Figure 2 is in the bottom panel. The positive VIL values (purple dashed line) are coincident with the areas with large values of TOP35 and TOP45.

2.2.2. Other Data

To investigate the relationship between the TOP distribution and severe weather, we have considered two different data sources, for the same period 2013–2021:

• Hail registers: observations included in the database of the Servei Meteorològic de Catalunya (see, for instance, [26] for more details), which must contain the fields “date”, “time”, “source” (Spotter, automatic weather station, hail-pad, social network, others), “coordinates”, and “magnitude”. The location of the sources is variable: spotters and social network registers are more usual in the highly populated regions, while the hail-pads network covers the area of the Western Depression marked in green in Figure 1;
• Lightning Jump (LJ) warnings: The Servei Meteorològic de Catalunya runs in real-time and operationally a tool for triggering lightning jumps in the region. In this study we have used the warnings for the period 2013–2021 for comparison with the TOP fields. Each LJ must include the “date”, “time” and the “coordinates” fields. From experimental campaigns, the detection efficiency over the Catalan territory of the lightning location system is between 80 and 90% for cloud-to-ground (CG) flashes and between 65 and 80% for intra-cloud (IC) flashes. The location accuracy is between 0 and 1 km for CG. These values allow for making very precise warnings of severe weather in the region with a lead time between 15 and 90 min, from the last validated years.

The last data source used in this research is the topography presented in Figure 1. It is a raster file (Geotiff format) that has been cropped to the region of study, changing to the spatial resolution of the TOP files (1 km × 1 km) through geo-statistical routines included in the raster package (version 3.6.3) of the R software (version 4.2.2) [27,28].

2.3. Methodology

The research consisted of the generation and analysis of the maps of maxima Echo Top and the Number of counts over a certain height threshold. The height threshold has been different, depending on the Echo Top product: the lower the reflectivity threshold, the higher the height reflectivity (see the mid panel of Figure 2). This means that when the reflectivity is increased, the height observed will decrease because large particles cannot reach the same altitudes as smaller ones. The three thresholds (7.5 km for the TOP12, 5.5 km for TOP35, and 4 km for TOP45) resulted from the analysis of a set of hail-bearing thunderstorms in the area during different seasons of the year. How the maps have been generated is the following:

• A new binary raster, R_BIN, is estimated for each six minutes TOPXX (where XX is 12, 35 and 45) with 0/1 values for the pixels under/over the height threshold. For instance, if the pixel in the row “i” and column “j” TOP^ij = 12.5 km and the threshold is 10 km, then the R_BIN^ij = 1;
• A cumulative raster (R_SUM) increases in one pixel with the non-null value in the binary raster. Then, for the example considered in the previous step: R_SUM^ij = R_SUM^ij + 1;
• Another raster (R_MAX) looks if the pixel of a certain location exceeds the previous total maximum value: if not the procedure continues, and if yes, the new value replaces the old one in this raster position. If in the previous case, the maximum was R_MAX^ij = 12.3 km, then the new value of the R_MAX^ij will be 12.5 km;
• The two previous estimations, R_SUM and R_MAX, are calculated for each raster of the 2013–2021 period (788,880 files), and the three TOP products.

3. Results

This section focuses mainly on the analysis of the maps generated according to the methodology presented in the previous section. Other points considered are the comparison of the same maps with other fields that represent the hail occurrence and the lightning jump activity both for the same period and with the topography presented in Figure 1.

3.1. Echo Top Maps of the Maximum Height and the Cumulative Occurrence

The left panel of Figure 3 shows that the maximum height of thunderstorms in Catalonia reaches 18.5 km, in two different areas: in the eastern part of the Pyrenees and the central-western part of the Pre-Pyrenees. The two regions are surrounded by the black line indicating the percentile of 90% (16.5 km). Other smaller areas are in the southern extreme of the region, in some coastal parts, and residual zones of the Western Depression. The right panel of Figure 3 (the distribution of TOP12 cases exceeding 7.5 km for the whole period) is more adequate for understanding the regions where large vertical developments are usual. It is important to indicate that it can be appreciated as some effects caused by the topography (e.g., beam blockages) or the same composition (with some discontinuities in parts of the fields). In this case, the map shows two maxima regions, marked by the percentile of 75% (black wide lines, 4750 cases in total), one in the eastern part of the Pyrenees (but displaced to the west respecting the maximum of the left panel), but surrounded by practically all the Pyrenees and Pre-Pyrenees, and the second, less important, in the south. This indicates that both regions are where it is more common to find the largest thunderstorms in Catalonia, coinciding with topographic regions (Figure 1, right panel).

Figure 4 presents the same fields as in Figure 3, but for TOP35 and considering the height threshold of 5.5 km. On the one hand, the spatial distribution of the maximum value (left panel) follows a similar shape with the same regions presenting the most relevant observations (again, the black line is the 90% percentile, coinciding with the 13.5 km height). The values, as it was expected, are lower than in the case of TOP12 and do not exceed 15.5 km (15.4 km is the maximum height). On the opposite, the right panel has notably changed respecting TOP12: the south maximum has disappeared while the region with the largest probability of TOP35 exceeding the 5.5 km has been displaced to the southwest. There is a major hotspot coincident with part of the TOP12 (>7.5 km), that is, in the eastern Pyrenees, but also includes parts of the central Pre-Pyrenees. Additionally, a small region of the western Pre-Pyrenees also exceeds the 75% percentile, which corresponds to 300 cases during the period.

The last parameter is TOP45. Figure 5 shows the distributions (total maximum in the left panel and number of cases exceeding 4 km in the right panel). In this case, the maximum value does not exceed 15 km (14.8 km), being the percentile 90% 12 km, like the TOP35 but in this case including more reduced areas of the central Pre-Pyrenees. The maximum TOP45 field has a similar pattern to the two previous, but the maxima in the coastal area have disappeared, focusing all the hotspots on the inland regions. In the case of times exceeding 4 km, the hotspot has concentrated just in the region between the eastern Pyrenees and Pre-Pyrenees, in very tiny areas where the 75% percentile is exceeded (75 cases in total). This is the region where severe thunderstorms are expected to occur more usually, according to the field distribution of TOP45 over 4 km.

Finally,

Table 1. Quantiles (10, 25, 50, 75, 90, and 100) for the maximum TOP field of each product (TOP12, TOP35, and TOP45).

Product	Q10	Q25	Q50	Q75	Q90	Q100
TOP12	14.1	14.6	15.2	15.9	16.6	18.5
TOP35	11.3	11.7	12.3	13.0	13.6	15.4
TOP45	8.8	9.5	10.4	11.3	12.5	14.8

summarizes the distribution of the maximum values for the different products, indicating that the TOP12 has exceeded 14 km height in practically all the Catalan territory at least once during the period of analysis (the 10th Quantile, Q10). In the case of TOP35 and TOP45, the values decrease to 11.3 and 8.8 km, respectively. The median values (Q50) are 15.2, 12.3, and 10.4 km, for TOP12, TOP35, and TOP45, while the 90th Quantile are 16.6, 13.6 h, and 12.5 km.

Table 1 and Figure 4 and Figure 5 show how the development of the convection (represented by TOP12) is very common in a major part of Catalonia, being frequent altitudes of 15 km. Compared with the topography of the right panel of Figure 1, it can be appreciated how the areas of the maxima Echo Tops (and with major frequencies exceeding 7.5 km height) are highly coincident with some different ranges. This is an indication of the role of the topography as the lifting element of the convection, not only in the largest ranges (Pyrenees and Pre-Pyrenees), but also in the Littoral Ranges. TOP35 and TOP45 maps confirm the previous observation but reduce notably the areas with the largest vertical developments. This is because not all the updrafts are in the optimal conditions for the development of the convection (TOP35) and severe weather (TOP45), which is stricter. Again, topography seems to play a role in the development of extreme atmospheric conditions, but it could be secondary, in the sense that the regions are more reduced and the topography is not as high as in the case of TOP12. Finally, it is worth noting the fact that even the absolute values of the percentiles of the heights are notably lower in the case of the TOP35 and TOP45, respecting the TOP12 ones; the significance of having large ice particles (associated with TOP45) at 10 or more km is higher than observing tiny particles (those occurring with TOP12) at 15 km.

3.2. Comparison of Echo Top Maps with Other Fields

The previous maps (Figure 3, Figure 4 and Figure 5) have allowed determining the maximum vertical development of the thunderstorms (left panels) and, on the other hand, the hotspots where thunderstorms usually exceed a certain height value (right panels). The height value used as a threshold considers the altitude for which the clouds are prone to produce severe weather, but even this height is variable depending on the season or the meteorological situation. This section focuses on the comparison of the right panels of Figure 3, Figure 4 and Figure 5 (the spatial distribution of areas with TOP exceeding the threshold height) with other fields, to determine if there exists a good correlation between TOP and the occurrence of severe weather phenomena. In particular, the research has centered on hail cases. Finally, we have compared the TOP over the height figures with the topography shown in Figure 1, to determine the role that this element plays in the triggering of severe thunderstorms.

To do the first comparison, the first step is to convert the punctual registers (of hail or LJ warnings) in raster files like the TOP fields. The technique consisted of transforming the points in 2D fields using the Two-Dimensional Kernel Density Estimation (KDE2D) included in the MASS package of R software [29]. The new maps have the same spatial resolution (1 km × 1 km) and the same bound limits, to be comparable with the TOP distributions. KDE2D consists of applying the kernel smoothing over a sample of points, and producing a density map of the probability of occurrence which ranges between 0 (minimum) and 1 (maximum). This allows observing the repetition of cases over a concrete area, as is the present case, with some records reported by the same spotter or the same hail pad, for example. Figure 6 shows the result for the hail registers. It has considered three categories: the map for all the registers (top panel), the map with only registers over two cm (which is the threshold for severe hail, in the left bottom panel), and finally the observations that have a size exceeding the 4 cm (very large hail, in the right bottom panel). In the first case it can be appreciated three maxima areas: one in the Pre-Pyrenees, the second in the central coast, and the last one in the Western Depression. The last two are the consequence of different factors that usually affect the databases based on direct observations [30]. First, the influence of the highly densely populated regions is appreciated in this case because of the Metropolitan Area of Barcelona, in the central coast and its surroundings. The second one is the use of specific observational networks (the hail-pad network of the Western Depression [23]). However, these effects minimize their impact when there are imposed larger size thresholds. In the case of the severe hail registers (bottom left panel), the maxima associated with the central coast (where there are many small hail registers) has disappeared. If the size threshold increases to four cm (bottom right panel), only one hotspot remains, with two sub-maxima: one in the eastern Pyrenees and the second in the central Pre-Pyrenees. It is important to consider that percentages are not absolute, if not referred to the total dataset of each category. This means that the percentage in an area of severe hail can be lower than of the larger hail because the number of registers is more representative in the second case concerning the total of each dataset.

Figure 7 shows the same field as Figure 6, but for the case of LJ warnings which have a high relationship with larger hailstones events (bottom panels of Figure 7). In this case, the distribution has a clear maximum over practically all the Pre-Pyrenees, with the absolute maximum over the central part of this range. It is worth noting the secondary maximum over the Western Depression, coincident with the one observed in the left panel of Figure 6 (hail over two cm).

Before the comparison of the different matrices, the Shapiro Test has been applied to all the fields to check if they follow a normal distribution. In all cases, the value of the statistical parameter has moved between 0.825 (for big hail registers) and 0.947 (for all hail observations). This indicates the normality of all the datasets. Afterwards, the correlation test has been calculated for all the matrix combinations presented in

Table 2. Correlation values (and the 95% confidence intervals) of the different cumulative Echo Top fields with the different variables: all hail registers (All Obs), severe (≥2 cm) hail (Sev Obs), large (≥4 cm) hail (Big Obs), Lightning jump occurrence (LJ), and the topography (TOPO).

Variable	TOP12		TOP35		TOP45
All Obs	0.074	0.072, 0.076	0.258	0.236, 0.280	0.306	0.286, 0.326
Sev Obs	0.090	0.088, 0.092	0.179	0.170, 0.188	0.297	0.277, 0.317
Big Obs	0.484	0.480, 0.488	0.448	0.410, 0.485	0.588	0.582, 0.596
LJ	0.475	0.466, 0.484	0.789	0.782, 0.796	0.760	0.755, 0.765
TOPO	0.489	0.455, 0.523	0.490	0.485, 0.495	0.204	0.197, 0.211

. In all cases, the p-value is lower than 0.05, which means that all the correlations are statistically significant. Table 2 shows the result of the comparison between the right panels of TOP (this is, the spatial distribution of TOP exceeding a certain height threshold)—Figure 3, Figure 4 and Figure 5 and the distributions of hail (Figure 6), LJ warnings (Figure 7) and topography (Figure 1 right). The correlation (and the 95% confidence interval) has been estimated only for those pixels included in the studied area (the Catalan territory) because it is the area where the comparison is effective. The ground database only includes registers inside this region, while the accuracy of the TOP observations and LJ locations decreases as they are far from the Catalan boundaries. Each variable (all hail registers “All Obs”, severe hail “Sev Obs”, large hail “Big Obs”, Lightning jump warnings “LJ”, and the topography “TOPO”) has been compared with the three distribution maps of Echo Tops: TOP12 > 7.5 km, TOP35 > 5.5 km, and TOP45 > 4 km.

TOP12 presents the worst correlation values for all the variables, except in the case of big hail and TOPO, which is the second-best correlated field. The values of the correlation move between 0.074 (“All Obs”) and 0.489 (“TOPO”), this is, under 0.5 in all cases. TOP35 have correlation values moving between 0.179 (“Sev Obs”) and 0.789 (“LJ”). In all cases except for “Big Obs” the values are larger than for TOP12 and have the maximum in the cases of “LJ” and “TOPO”. In the case of the “Big Obs”, the correlation is the minimum of all three. The correlation between TOP35 and LJ is the absolute maximum for all fifteen values. Finally, TOP45 presents the maximum of all three TOP fields for all the hail registers (“All Obs”, “Sev Obs” and “Big Obs”) but the value is the minimum in the case of TOPO. The correlation for this field ranges between 0.204 and 0.760, constituting the higher range. The mean correlation values are 0.322 for TOP12, 0.433 for TOP35, and 0.431 for TOP45.

Finally, it is important to understand the physical implications of each parameter. Although the three TOP indicate the intensity of the vertical updraft, the consequences of each one is different. TOP12 presents the capability of the updraft for carrying up small particles, while TOP35 and TOP45 are the same but for moderate and large precipitating particles, respectively. This means that the updraft necessary for reaching a TOP45 of 8 km is notably higher than the existing one for the same height of TOP12. It is very common to observe TOP45 connected with TOP12 values 4 or 5 km higher, because even for small hail the TOP needs to be over some kilometres and the results evidence that the correlation will increase as the size of hail and the reflectivity threshold are larger. A similar relationship exists with the LJ warnings, which have a high interaction with severe and big registers (as [1,2] showed). Finally, analyses as [25] inferred that topography plays a major role in cases with low to moderate convection, explained mainly by the TOP12 and TOP35. Then, these variables show a better correlation with the mountain ranges location, while TOP45 does not require the mountains for playing a role as a lifting element.

4. Discussion

Depending on the height of the threshold used for estimating the TOP product (12, 35 and 45 dBZ), the area with the largest values changes in position, size, and frequency. In the case of TOP12, there is a good correlation between the product and the topography: the areas with maximum values are coincident with the most elevated in each part of the region. This would be an indicator that thunderstorms tend to develop in topographic areas in Catalonia where the ranges act as lifting mechanisms of the convective air masses. Additionally, the extension of the areas where convection begins is quite large: 29.1% of the total territory (~30,000 km²) has 4000 or more cases, exceeding the 7.5 km of TOP12 (green to purple colors area in Figure 3). It is important to notice that 4000 cases are equivalent to 0.5% of the total imagery. The values for TOP35 and TOP45 are more modest: 23.6% (19.1%) of the pixels have 175 (44) or more cases exceeding 5.5 km (4 km) of TOP35 (TOP45). Furthermore, there exists a shifting of the maximum area, which in the case of TOP45 is focused on the central Pre-Pyrenees, coincident with the highest severe weather activity in the region [2,3], also according to the ground registers and LJ observations. These results indicate that the most common stormy pattern should be: first, an initial development with the maximum TOP12 in the mountainous areas, then, the thunderstorms increase the TOP35 and TOP45 (implying the growth of the particles inside the cloud) while they move to lower heights. This is coincident with the LJ triggering, or the moment of the largest velocities of the updraft [1,21]. It is clear from this spatial analysis that it is not possible to determine the time relationship between these elements accurately. However, the cited references coincide in confirming the good correlation between them. Finally, some minutes later, the occurrence of severe weather in the vicinities of the LJ location (according to [2], the thunderstorm can travel around 5 and 40 km between both phenomena, depending on the velocity of translation). Again, it is not the objective of this analysis to determine the time differences between the occurrence of both phenomena, but the proximity in space (and the typical speed of the thunderstorms in the region, see for instance [25]) should confirm this point. Figure 8 summarizes the maxima TOP activity (left panel). The profile of the evolution of the different variables (topography, LJ occurrence, large hail registers, and TOP distributions) along the A–B transect (central panel) has allowed the development of the scheme of the thunderstorm’s behavior for the region of analysis. From the central and right panels can be deduced the classic scheme of the convection developed over topographic regions in Catalonia, which is the most usual in this area, while cases as described in [22] are less common. This scheme can be summarized as:

(1)

Warm and moist air mass advected from the sea, moving from point B to A in all the panels of Figure 8;

(2)

A lifting process of the air mass when it collides with those mountains range that is enough for the triggering (depending on the thermodynamic conditions of the environment);

(3)

In some cases, thunderstorms form over the valleys just surrounding the mountains, and later move to the sea;

(4)

Other ranges help to reach the major vertical development, coinciding with the highest TOP35 and TOP45, and with the LJ occurrence;

(5)

Finally, the downdrafts produce severe weather phenomena at ground level, mainly in the Plains and the coastal areas.

Finally, we have compared the TOP values for Catalonia with the estimated ones for other regions. Although the sensor types are different and the thresholds could slightly change, it is possible to obtain a reasonable comparison with other areas of analysis. The values in Catalonia (TOP12 of 15.2 km, TOP35 of 12.3 km, and TOP45 of 10.4 km) are higher than those observed in Estonia (TOP15 of 12 km and TOP45 of 5.5 km) [14] (placed at higher latitudes). Additionally, they are like the estimates for Romania (between 4 and 16 km) [16,17] (in similar latitudes) and lower than in the United States of America (15 km) [20] with special conditions caused by the Great Plains. However, the values are higher than for South Africa (mean values of TOP23 of 9.1 km, TOP40 of 6.3 km, and TOP50 of 5.0 km) [13], located at lower latitudes. These important values of the vertical developments of the thunderstorms, not far in most of the cases of the tropical ones, can be caused by the combination of the proximity of the warm sea of the Mediterranean and the topographical influence.

5. Conclusions

As has been shown in this study, the spatial distribution of the TOP12, TOP35, and TOP45 display some notable differences for the period 2013–2021. First, TOP12 has the best correlation with the topography. This is a consequence of the influence of natural barriers in the initial convection triggering in most the cases. On the opposite, TOP35 and TOP45 have better correlations with the large hail observations and the lightning jump occurrences, that is, with severe weather. From the correlations between the different TOP products and topography and severe weather, we can deduce that topography is more necessary as a triggering element of the convection in those environments that are not prone to developing deep convection. The other main result is the quasi-simultaneity in space of the hail occurrence, the largest lightning activity (based on the LJ observation), and the largest values of TOP45. This implies that when the more intense nucleus of the cell acquires the highest altitude, it produces the lightning jump and severe weather phenomena. This result is coincident with other analyses based on the life cycle of severe thunderstorms [1,2,21], which compared lightning and radar parameters observing lead times of less than 1 h. Departing from the fact that thunderstorms move at averaged speeds of 16 m/s (~57 km/h), the highest distance between the highest vertical development and the severe weather occurrence should be less than 60 km. Furthermore, the topography will contribute to the rise in thunderstorms with important values of TOP12 in cases of low convective environments. However, it is highly probable that TOP35, and especially TOP45, present heights lower than in deep convection environments. This is a consequence of the combination of the triggering caused by the mountain range and the instability of the air mass. This triggering will be enough to lift and develop small to moderate precipitable particles. On the opposite, topography plays a secondary role in the generation of those thunderstorms with very high values of TOP45, which are the hail-bearing cases.

Another interesting point is the observation that the larger the value of the TOP and, also, the reflectivity threshold, the higher should be the diameter of the stones. However, it is important to notice that very elevated values of TOP12 are not necessarily associated with severe or big hail. For instance, there are values of TOP12 over the 16 km in the coastal region, but the probability of large hail is lower than in the internal areas. Then, TOP35, and especially TOP45, provide better diagnosis of hail probability, being coincident with [16,23]. The analysis has been focused only on the hail, but we suggest future research on other severe weather phenomena (tornadoes and straight-line winds) and heavy precipitation in heavy convection events.

Finally, TOP values in Catalonia are coincident with those observed in Romania, which has similar latitudes. The TOP values are higher than those estimated in regions at higher latitudes (Estonia), but lower than in the Great Plains of the USA (placed at lower latitudes). It seems that there exists a correlation between the vertical development of severe weather thunderstorms and the latitude of the region, but a more in-depth analysis should be made in the future because this correlation does not exist in the case of South Africa. However, topography and other meteorological factors can play an important role in some local particularities of the resulting regional climatology. In any case, it is possible to translate the proposed methodology into other regions because thunderstorms have the same life cycle. If it is needed to adapt the threshold values to the new areas of applicability, then a climatology of radar Echo Tops can help to identify the areas more prone to producing high electrical activity (lighting jumps) or of occurring severe weather on the ground (in this case, hail) easily.