The meteorological term atmospheric convection typically describes vertical transport of heat (and moisture) from a warmer lower layer to a cooler upper layer by the rising of warm (moist) air. When convection is triggered only in that way it is called 'free' convection. The well-known afternoon convection which leads to thunderstorms on hot summer afternoons is an example of free convection. Solar radiation heats the earth differently depending on soil composition and orientation to the sun. When the surface warms it heats the overlying air which gradually becomes less dense and, thereby, lighter. When near-surface air becomes lighter than its surroundings it begins to rise. If the lifting forces are strong enough an updraft channel can develop and eventually result in a convective cell.
Convection does not always become visible by cloud formation. But severe convective storms , the subject of this study, are defined as deep, moist convection  that is characterized by condensation processes and, thus, cloud formation. Those storms are of special interest since they can produce severe weather like intense showers, lightning, hail, downbursts, and even tornadoes, which pose a severe threat for humans, environment, and economy. The essential ingredient for severe convective storms is warm, moist, unstable air that is forced to rise either by free or forced convection. Forced convection is caused by dynamical convergence lines , orographic uplift, or weather fronts. Embedded in the atmospheric flow, convective storms develop an individual, distinctive dynamic which is mainly localized around the updraft channel(s). Nevertheless, individual storm development and intensity are strongly influenced by environmental conditions, atmospheric conditions and topographic aspects.
Convective storms, including severe convection, are mostly small-scaled and short-lived phenomena with lifetimes less than an hour. Convective processes are well observed by remote sensing (satellites, radar and lightning detection). Remote sensing data are available in such a high temporal and spatial resolution that convective processes can be adequately monitored. But those data are not available for the period of 30 years, the typical time basis of climatological studies. Nevertheless, data archives covering multiple years gradually become extensive enough to provide insights into the general appearance of convective storms as evidenced by the increasing number of studies on severe convection (Kaltenboeck and Steinheimer, 2014; Rudolph and Friedrich, 2013; Nisi et al., 2013; Puskeiler et al., 2013; Goudenhoofdt and Delobbe, 2013; Davini et al., 2012; Meyer and Schaffhauser, 2012, and more).
The generation of a convection climatology in mountainous regions such as the Alps in the research domain poses a true challenge since the main data source for a comprehensive monitoring of deep convection, the weather radar, suffers significantly from beam blockage. Although some radars are established on mountain tops, the visibility is limited due to beam blockage by surrounding mountains. In areas where the radar beam is blocked or shaded and also in areas further away from the radar station it is possible that low level precipitation is not captured. As severe convection generally has a pronounced vertical development, there is a good chance, however, that the upper part of severe storms is still captured by the radar, even when the lower parts are hidden due to beam blockage. Lightning data, in contrast, do not suffer from shading effects. Therefore they are used as a complementary information source to support continuous storm monitoring in areas with reduced radar visibility and to overcome occasional data gaps. But, as lightning discharges are not strictly bound to the main updraft region, but can also occur in attached, non-convective regions, it is more difficult to outline storm contours and, therefore, less reliable to continuously monitor storm developments by lightning activity only. However combined information from both data sources preserves the advantages of each and alleviates their respective disadvantages.
 Beside others, described in very detail in Doswell, 2001.
 A convergence line describes the confluence (convergence) of air near the ground. As the air masses converge, they are forced to rise. In case of sufficient humidity this leads to cloud formation and eventually precipitation. During the summer season thunderstorms are often triggered along convergence lines.
Both radar and lightning data are used to investigate convective processes in the study area. Radar data originate from three different C-band weather radar networks, the Austrian composite, the Veneto composite and the Bozen radar. In a composite the information from several radar stations are combined to a common map. This can be done by different methods. The following table shows the radar networks (radar domains):
Lightning data were provided by the European Cooperation for Lightning Detection (EUCLID), a collaboration among national lightning detection networks to detect lightning over the European Area. More information about EUCLID can be found at www.euclid.org. We want to thank EUCLID and especially the Austrian Lightning Detection and Information System (ALDIS) and the Centro Sperimentale Italiano Elettrotecnico (CESI SIRF) for providing the data which enabled the study in this form.
The analysis covers April through October of years 2009 to 2013. The period from April through October is considered the convective season since the majority of convective storms occur during this time. Data prior to 2009 are not included due to lack of data from the Valluga radar located near St. Anton. Since the Valluga radar contributes significantly to radar visibility in the study area, the availability of the Valluga data was a crucial factor in selecting the time span of the analysis.
An automated algorithm has been employed in order to consistently detect severe storms from the different datasets. The Austrian Thunderstorm Nowcasting Tool (A-TNT) monitors (moist) convective activity by identifying and tracking intensified (convective) precipitation regions and lightning activity based on the combined information of radar and lightning data (A-TNT builds thereby upon the principles of Meyer et al. 2013). The shortfalls of radar visibility in alpine regions are mitigated by the approach of combined storm tracking. Additionally, due to the lightning information thunderstorms and pure convective precipitation cells can be discriminated since the occurrence of lightning in convective cells is the common definition of a thunderstorm . For each radar domain a convective storm data base has been established and contains storm evolution in five minute time steps. The following table summarizes the criteria set in A-TNT for the detection of radar- and lightning-cells. Details on the criteria can be found in Meyer et al. 2013.
|accumulation time: 6 min
search radius: 9 km
min. cell area: 3 km²
min. Intensity: 38 dBZ (aprx. 9 mm/h)
min. cell area: 5 km²
 See AMS Glossary of Meteorology on http://glossary.ametsoc.org/wiki/Thunderstorm or the definition given by the German Weather Service on http://www.deutscher-wetterdienst.de/lexikon/index.htm?ID=G&DAT=Gewitter.
Although special care has been taken to overcome the shortages of radar measurements in mountainous terrain it is important to consider the variations in local radar visibility for the interpretation. A visibility map of the study area is provided and shows the calculated minimum height seen by a radar within the three radar composites. The respective radar adjustments are given by the providers and the topographic information is derived from SRTM-3 data, which has a resolution of 3 arc minutes (between 60 and 90 meter).
The storms included in the data analysis meet minimum area and duration thresholds of ≥ 50 km2 and ≥ 30 min. The size threshold can be met at any time during the storm's lifetime. The size and duration limits were applied to minimize false signals from residual radar clutter due to mountains and other non-precipitating objects. The disadvantage is that smaller, short-lived storms, such as those that typically form over mountain peaks, may be excluded from this study. Also, the radar signal decreases with the distance to the station because of geometrical broadening of the radar beam and because of signal attenuation when the beam passes through precipitation. As a consequence this results in higher reflectivity values and subsequently the detection of more storms in the vicinity of the radars. The lightning network data is not influenced by the radar locations and therefore dampens the effects of radar attenuation on identifying cell initiations. To avoid the bias introduced by the uncorrected radar signal attenuation, only convective cells detected by both radar and the lightning network (e.g. thunderstorms) were considered. All these compromises were made for the benefit of improving, as much as possible, the comparability of observation quality across the study region. Additionally, the selected storms also meet a minimum rain rate of approximately 9 mm/h. The reflectivity threshold is incorporated as part of the automated cell identification algorithm to identify convective cells at an early state by their intensifying precipitation.
clutter due to mountains and other non-precipitating objects. The disadvantage is that smaller, short-lived storms, such as those that typically form over mountain peaks, may be excluded from this study. Also, the radar signal decreases with the distance to the station because of geometrical broadening of the radar beam and because of signal attenuation when the beam passes through precipitation. As a consequence this results in higher reflectivity values and subsequently the detection of more storms in the vicinity of the radars. The lightning network data is not influenced by the radar locations and therefore dampens the effects of radar attenuation on identifying cell initiations. To avoid the bias introduced by the uncorrected radar signal attenuation, only convective cells detected by both radar and the lightning network (e.g. thunderstorms) were considered. All these compromises were made for the benefit of improving, as much as possible, the comparability of observation quality across the study region. Additionally, the selected storms also meet a minimum rain rate of approximately 9 mm/h. The reflectivity threshold is incorporated as part of the automated cell identification algorithm to identify convective cells at an early state by their intensifying precipitation.
The data from the three radar networks contain spatial overlap. The different radar networks are not time synchronized and differ in data processing, precipitation product, and data resolution. Therefore the convective storm analyses was used to generate a "composite of composites" in order to condense the information in regions which are covered by multiple radar networks into a single map. The process eliminates spatial overlap by selection of the appropriate radar data source for each location. The final composite was generated from the three storm data bases of the domains Austria, Veneto, and Bozen by selecting the maximum count of storm initiations at each location . This approach is based on the assumption that the radar providing the best visibility yields the maximum number of detected storms. This approach for storm selection is incorporated in all resulting analyses.
 As such, the composite initiation density map of this website represents the maximum count of storm initiations within each 10km × 10km area from among the three radar data sets.
Charles A. Doswell III, 2001. Severe Convective Storms—An Overview. Meteorological Monographs, 28, 1–26. doi: http://dx.doi.org/10.1175/0065-9401-28.50.1
Davini, P., Bechini, R., Cremonini, R., Cassardo, C., 2012. Radar-based analysis of convective storms over northwestern Italy. Atmosphere, 3, 33–58.
Fabry, F., Cazenave, Q., Basivi, R., 2013. Echo climatology, impact of cities, and initial convection studies: new horizons opened using 17 years of Conterminous US radar composites. AMS 36th Conference on Radar Meteorology, Breckenridge, CO.
Goudenhoofdt, E., Delobbe, L., 2013. Statistical characteristics of convective storms in Belgium derived from volumetric weather radar observations. J. Appl. Meteorol. Climatol. 52, 918–934
Kaltenboeck, R., Steinheimer, M. Radar-based severe storm climatology for Austrian complex orography related to vertical wind shear and atmospheric instability. Atmos. Res., 2014, doi:10.1016/j.atmosres.2014.08.006.
Meyer, V., Schaffhauser, A., 2012. Systematic investigations of intense convective precipitation events in Austria based on radar cell tracking. 7th Europ. Conference on Radar in Meteorology and Hydrology, Toulouse, France.
Meyer V., H. Höller, and H.-D. Betz, 2013. Automated thunderstorm tracking: utilization of three-dimensional lightning and radar data, Atmos. Chem. Phys., 13, 5137-5150, doi:10.5194/acp-13-5137-2013.
Nisi, L., Martius, O., Hering, A., Germann, U., 2013. Hail storms over Switzerland: spatial and temporal characteristics derived from radar-based hail products. 7th European Conference on Severe Storms, Helsinki, Finland.
O'Bannon, T., 1997; Using a "terrain-based" hybrid scan to improve WSR-88D precipitation estimates. Preprints, 28th Int. Conf. on Radar Meteorology, Austin, TX, Amer. Meteor. Soc., 506–507.
Puskeiler, M., Kunz, M., Schmidberger, M., 2013. Modelling of the hail hazard in Germany. 7th European Conference on Severe Storms, Helsinki, Finland.
Rudolph, J., Friedrich, K., 2013. Seasonality of vertical structure in radar-observed precipitation over southern Switzerland. J. Hydrometeorol. 14, 318–330.