Analysis of Sub-Daily Precipitation for the PannEx Region

The PannEx is a GEWEX-initiated, community driven research network in the Pannonian Basin. One of the main scientific issues to address in PannEx is the investigation of precipitation extremes. Meteorological Services in the PannEx area collected the hourly precipitation data and commonly used a computer program, which was developed in the INTENSE project, to produce a set of global hydro-climatic indices. Calculations are carried out on data aggregated 1-, 3- and 6-h intervals. Selected indices are analyzed in this paper to assess the general climatology of the short-term precipitation in the Pannonian basin. The following indices are illustrated on maps and graphs: the annual mean and maxima of 1-h, 3-h and 6-h sums, the count of 3-hr periods greater than 20 mm thresholds, the maximum length of wet hours, the timing of wettest hour and the 1-h precipitation intensity. The seasonal trends of the 1-h precipitation intensity were tested from 1998 to 2019. Analysis of sub-daily precipitation has been limited by the availability of data on a global or a regional scale. The international effort made in this work through collaboration in the PannEx initiative contributes to enlarging the data availability for regional and global analysis of sub-daily precipitation extremes.


Introduction
Heavy rainfall and storm events constantly trigger important damages and significant casualties in many regions, and enhanced understanding the variability of precipitation extremes would enable our communities to cope more efficiently with the challenges associated with the warming climate [1]. Understanding how and why precipitation extremes vary is vital for disaster preparedness and to enable adequate water resource management. This requires both understanding of how extremes have varied in the past and the ability to provide estimates for how they might change under warming climate. One of the key statements of the IPCC Fifth Assessment Report centres around changes in the extremes of observed precipitation, e.g., 'There are likely more land regions where the number of heavy precipitation events has increased than where it has decreased' [2]. Analyses of the station data that have contributed to available global datasets display increasing trends in annual maximum daily precipitation [3]. The observed more frequent heavy precipitation events are also consistent with increasing amounts of water vapor in the atmosphere due to global warming [4].
In recent decades, an increasing number of studies have reported the presence of significant positive trends in precipitation extremes in Europe e.g., [5][6][7][8]. Significant trends in precipitation extremes over Europe have been observed since the middle of the 20th century in regional studies for Northern Europe [9], the UK [10], the Mediterranean region [11], and western and eastern parts of the Czech Republic [12] and Poland [13]. [14] showed that despite the considerable decadal variability, 5-, 10-, and 20-year events of 1-day and 5-day precipitation amounts for the first 20 years in the period 1951-2010 became more common in the analyzed 60 years for the daily precipitation series from the European Climate Assessment and Dataset (ECA&D, http://www.ecad.eu (accessed on 5 May 2021)) project [15,16].
The area in focus of this study is the Pannonian basin and its surroundings. The Pannonian/Carpathian basin (both names are used) is part of the larger Carpathian region. The precipitation climatology for the Carpathian region together with other climate variables are discussed either at seasonal or monthly scale [17,18] ( based on the CARPATCLIM dataset. The CARPATCLIM data covers the Carpathian Region and provide a homogenized [19,20], harmonized and gridded [21] dataset for climate studies. The precipitation amounts showed no significant trend, though it increased slightly on an annual basis in the period 1961-2010. The changes in the number of very wet days (daily precipitation > 20 mm) range between -2 and +3 days in extended regions. The highest increase appears in the Northeast Carpathians and the Bihor Mountains i.e., 7 days [22]. Considering the precipitation changes in Hungary, in the centre of the Carpathian region, a decrease of the annual precipitation sum is not remarkable (3%, 1901-2019). From the beginning of the 1990s, precipitation has been increasing both on annual and seasonal scales; however, this rise is not significant. Recent years have been dominated by the extremes. The magnitude of the change in precipitation intensity (mm/day) is about 1.3 mm/day in the countrywide average. The number of days with a daily sum above 20 mm increased by 1 day in the period 1901-2017 [23]. In Croatia, the observed changes in annual and seasonal precipitation amounts are also generally weak, particularly when trends from the beginning of 20th century were considered [24]. However, in the eastern mainland (Pannonia plain) a significant positive trend in the autumn season was detected from the second half of the last century [25]. It was associated with more very wet days and with an increase of their contribution to the total precipitation as well as an increase in maximal 1-and 5-day precipitation and in daily precipitation intensity.
In general, the above-mentioned studies relied on daily or longer durations of precipitation, whereas studies documenting the characteristics of sub-daily precipitation are fewer. However, such events are of great importance for society and economy since extreme sub-daily precipitation may be associated with hazardous events such as floods, soil erosion, and debris. Moreover, mountains can play an important role by blocking further movement of cloud systems or even enhancing precipitation-related processes leading to possible flash flood events [26]. Due to a higher number of impervious surfaces, urban areas have a higher risk of pluvial flooding. Hence, cities in the vicinity of mountains can be additionally endangered since their drainage systems need to handle additional water amounts.
Fowler et al. [27] detected a notable minority of studies dealing with short-term rainfall, both on global or regional scales. Such studies are also sparse in the Pannonian region. Frequent occurrence, together with broad publicity for heavy rainfall events in recent years, has stimulated debates on the effects of climate change on the frequency and the magnitude of floods [28]. Due to spatial and temporal limitations of conventional climate models for Atmosphere 2021, 12, 838 3 of 18 resolving convective processes [29], measurements of short-term precipitation are of great importance for the analysis of such events.
For Hungary, Lakatos et al. [30] analyzed the return values of the 60-min extreme rainfall events using the measurements of the automatic weather station network. The return period of extremely high values of short-term rainfall has shortened in recent years in Hungary, as it is shown in a case study for the Pécs-Pogány meteorological station [31]. In Croatia, the analyses of a short-term precipitation were made for a few selected locations, mainly in urban areas [26,32,33] and over the region of Istria [34].
Extremes of sub-daily precipitation have been studied in the Czech Republic by Beranova et al. 2018. They compared the characteristics of observed sub-daily precipitation extremes with those simulated by several regional climate models driven by reanalyses and examined diurnal cycles of hourly precipitation and their dependence on intensity and surface temperature. Beranová et al. [35] showed that convection is not captured realistically. Hanel et al. [28] analysed sub-daily summer precipitation characteristics for 17 stations in the Czech Republic in the period 1961-2011 and found that the number of stations with significant positive trends is considerably larger than the number of stations with significant negative trends. Moreover, a significant trend in rain event characteristics is found only for a small number of stations, with the exception of the rain rate, which shows a systematic increase, and the consistent decrease in event duration.
According to the Clausius-Clapeyron relation, water vapor pressure in the atmosphere increases at a rate of 7% per Celsius degree (for constant relative humidity). Accordingly, it can be expected that the sub-daily precipitation will increase at the same rate. However, there is already evidence of the scaling of sub-daily precipitation with temperature that exceeds the expected 7% per degree e.g., [36,37]; thus, an increase in short-duration extremes as a consequence of the warming atmosphere prompts concern. In particular, on small temporal and spatial scales, rates of extreme precipitation are influenced by atmospheric circulation and vertical stability in addition to local moisture availability [38]. However, analysis of sub-daily precipitation has been limited by the availability of data on a global scale [5]. The situation is worse for sub-daily data, although efforts have been made recently to construct the first Global Sub-Daily Rainfall (GSDR) dataset, comprising over 23,000 hourly gauges [39], though with relatively short records in many locations. The first major international effort to focus on global sub-daily rainfall extremes is the INTENSE (INTElligent use of climate models for adaptatioN to non-Stationary hydrological Extremes) project. The INTENSE project is using a novel and fully-integrated data-modelling approach to enhance our understanding of the nature and drivers of global precipitation extremes and change on societally relevant timescales, leading to improved high-resolution climate model representation of extreme rainfall processes [40]. The INTENSE project is in conjunction with the World Climate Research Programme (WCRP)'s Grand Challenge on 'Understanding and Predicting Weather and Climate Extremes' and the Global Water and Energy Exchanges Project (GEWEX) Science questions. The first step towards achieving the project goals is to construct a new global sub-daily precipitation dataset. Comprehensive, open source quality control software is being developed to set a new standard for verifying sub-daily precipitation data and a set of global hydro-climatic indices to be produced. The derived indices, e.g., monthly/annual maxima, will be freely available to the wider scientific community.
The PannEx (Pannonian Basin Experiment) is also a GEWEX-related initiative. It is a Regional Hydroclimate Project (RHP) of the GEWEX GHP (Global Hydrology Panel). It is a GEWEX-initiated, community driven research network in the Pannonian/Carpathian Basin. It aims to achieve a better understanding of the Earth System components and their interactions in the Pannonian Basin (PannEx-https://sites.google.com/site/projectpannex/ (accessed on 5 May 2020)). The international efforts in PannEx involve research institutions, universities and national meteorological and hydrological services in an integrated approach towards identifying and increasing the capacity to face climate change in the Pannonian Basin. The PannEx research community has prepared a white book with the objective of identifying the main scientific issues to address, with the analysis of the precipitation extremes amongst them [41,42]. Thus, the PannEx community is providing an excellent framework for the GEWEX RHP and GHP joint work. The purpose of this paper is to contribute to INTENSE initiative with an international effort through collaboration of Meteorological Services in the Pannonian/Carpathian basin by collecting and analysing the hourly data and providing initial analyses of the sub-daily precipitation climatology in the PannEx region. The commonly used software Indices for quality control and derivation of several sub-daily precipitation indices was implemented in the INTENSE project at Newcastle University. We start by introducing the area and the data used in this study. We introduce the computer program we used. Then, we continue by analyzing and discussing the averages and trends of some selected indices with a focus on the climatology of the sub-daily precipitation. The last section is devoted to the main conclusions.

Area in Focus
The Pannonian Basin ( Figure 1) is enclosed by the Carpathians and the Transylvanian Plateau to the east and north [43,44]. This article focuses on the area of the countries located in the domain of PannEx but only partially covers the domain: Czech Republic (CZ) and Slovakia (SVK) to the north, Austria (AT), Hungary (HU) and Romania (RO) in the middle and Slovenia (SLO), Croatia (CRO) and Serbia (SRB) to the south. towards identifying and increasing the capacity to face climate change in the Pannonian Basin. The PannEx research community has prepared a white book with the objective of identifying the main scientific issues to address, with the analysis of the precipitation extremes amongst them [41,42]. Thus, the PannEx community is providing an excellent framework for the GEWEX RHP and GHP joint work. The purpose of this paper is to contribute to INTENSE initiative with an international effort through collaboration of Meteorological Services in the Pannonian/Carpathian basin by collecting and analysing the hourly data and providing initial analyses of the sub-daily precipitation climatology in the PannEx region. The commonly used software Indices for quality control and derivation of several sub-daily precipitation indices was implemented in the INTENSE project at Newcastle University. We start by introducing the area and the data used in this study. We introduce the computer program we used. Then, we continue by analyzing and discussing the averages and trends of some selected indices with a focus on the climatology of the sub-daily precipitation. The last section is devoted to the main conclusions.

Area in Focus
The Pannonian Basin ( Figure 1) is enclosed by the Carpathians and the Transylvanian Plateau to the east and north [43,44]. This article focuses on the area of the countries located in the domain of PannEx but only partially covers the domain: Czech Republic (CZ) and Slovakia (SVK) to the north, Austria (AT), Hungary (HU) and Romania (RO) in the middle and Slovenia (SLO), Croatia (CRO) and Serbia (SRB) to the south. The Carpathians are the longest mountain range and the geographic barrier between Central Europe, Eastern Europe and the Balkans. Carpathian/Pannonian Basin is bordered by the Alps in the northwest, the Dinaric Alps in the southwest and by the Carpathians in northeastern direction, and divided by the Danube River. The Carpathians constitute a geographic barrier between the cold and dry continental climate of Eastern Europe, the temperate climate of Central Europe and the warm and wet Mediterranean climate of the Balkans [17]. Precipitation is a climate parameter with high temporal and spatial variability in the region. Its spatial distribution is determined by distance from the seas, continentality and topography, as well as passing weather systems. Large-scale drivers such as North-Atlantic Oscillation (NAO) can influence the tracks of these systems [45]. However, the Carpathian chain is very long (more than 1500 km, but only the Tatra Mountains are above 2000 m) and it has a curved shape; consequently the Carpathians are not always able to stop the humid oceanic air masses coming from north-west. Topography also plays a role in the modification of the air flows in the large central plain, the Carpathian basin [23]. In winter, precipitation occurs because of large-scale, mainly The Carpathians are the longest mountain range and the geographic barrier between Central Europe, Eastern Europe and the Balkans. Carpathian/Pannonian Basin is bordered by the Alps in the northwest, the Dinaric Alps in the southwest and by the Carpathians in northeastern direction, and divided by the Danube River. The Carpathians constitute a geographic barrier between the cold and dry continental climate of Eastern Europe, the temperate climate of Central Europe and the warm and wet Mediterranean climate of the Balkans [17]. Precipitation is a climate parameter with high temporal and spatial variability in the region. Its spatial distribution is determined by distance from the seas, continentality and topography, as well as passing weather systems. Large-scale drivers such as North-Atlantic Oscillation (NAO) can influence the tracks of these systems [45]. However, the Carpathian chain is very long (more than 1500 km, but only the Tatra Mountains are above 2000 m) and it has a curved shape; consequently the Carpathians are not always able to stop the humid oceanic air masses coming from north-west. Topography also plays a role in the modification of the air flows in the large central plain, the Carpathian basin [23]. In winter, precipitation occurs because of large-scale, mainly stratiform precipitation systems, whereas convective precipitation dominates in late spring, summer and early autumn and develops many times locally, in unstable air masses. In the Mediterranean climate, most of the precipitation is obtained in autumn or winter, which is typically convective due to the unstable air masses above the sea.
The climate features of the sub-regions are described in Kocsis [46]. In the following, we outlined the foremost determinant features of the precipitating climate of the sub regions I. to IX.The Pannonian Great Plain (I): The lowland situated in the inner part of the Basin with scarce precipitation. In the middle of the Great Plain the flow is more moderate, typically north-westerly, while in the southern part south wind blows; Drava-Sava region (II): its climate is determined by Mediterranean character and by the varied topography, such as the precipitation-increasing effect of the mountains; Transdanubian hills with the Burgenland and the mediMurje (III): with favourable precipitation, growing Mediterranean character in the south; Little Plain (VI): the oceanic effect is stronger, there is more precipitation compared to the Great Plain and this is one of the windiest areas of the Pannonian basin; Vienna basin together with the upper Moravian hills (V): right up to the mountain frame, more affected by the ocean; Upper-Danube region with the medium and high mountains (VI): variation of deep valleys and closed basins, the local climate reflects the variation of the topography; Southern and Eastern Carpathians (VII) and Transylvanian Basin (VIII) and the Bihor-mountains (IX): the climate has more continental character there, which appears most intensively at higher elevation in the closed basins. The precipitation of cyclones from the west and from the Mediterranean region fall primarily due to the elevating effect of the topography. The flow here-due to the basin effect-comes as eddy-wind from all directions.

Data
Before automation, pluviographs were used to measure the precipitation at National Meteorological and Hydrological Services in the PannEx region. The pluviographs were also suitable for registering even at minute intensity; however, the evaluation of the registering paper was different from automatic sampling. In Hungary, for example, the most extreme partial sums were entered into the database, instead of the 10-min sums, which were stored from the late 90s onwards. Since the most intense 60-min sums are slightly different from the 1-h sums, to keep the homogeneity of the data sampling, the Hungarian data series from 30 stations used in this study only covers the period from 1998 to 2019, typically a period of automatic measurements. The Croatian mainland, as a part of the PannEx region, was covered here by 10 meteorological stations from the regular network of Croatian Meteorological and Hydrological Service (DHMZ). Short−term precipitation registers (with a discretization of 5 min) from the classical Hellman pluviograph were digitized, and hourly data from a common 1981−2019 period was selected for this study. Preliminary quality control compared the daily precipitation amounts registered by the ombrograph and a rain gauge at the same location. Initial data used rainfall measurements from 29 Slovenian pluviographic stations with at least 15 years of data, measured in 5-min time intervals. Although incoming station measurements undergo regular quality control, not all errors in the data can be filtered out. Additional filters at different stages of the process were therefore applied. Homogenized data sets of daily rainfall from rain gauges were used as a reference data in correcting the initial data. However, only two of the Slovenian stations are situated in the PannEx area in focus, and all derived Slovenian subdaily precipitation indices were offered to integrate to the global sub-daily precipitation dataset. In Romania, hourly data were extracted from paper pluviogram charts until 2018, and after automatic measurements using the tipping bucket and weighing precipitation gauges started in 2008. The pluviogram charts are available only for the warm season, namely from April to October. For the Czech Republic, heated rain gauges with tipping buckets were used for automated precipitation measurements. In the last few years, the professional station network introduced weighing rain gauges that replaced the original rain gauges. The period of these measurements covers about the last 20 years. For Slovakia, the situation is very similar, but weighing rain gauges started to replace gauges with tipping Although the data sharing agreements between WMO nations have moved in the direction of further sharing, the data was not collected to a common database. The participating researchers run the software to derive the sub-daily precipitation indices themselves. While there are obstacles to the exchange of long-term high temporal resolution precipitation data, there have been fewer barriers to the exchange of the derived indices. The strong restrictions on sharing daily and sub-daily data in some countries (despite a WMO resolution to the contrary-WMO 2016) typically do not apply to sharing indices. As a result of this joint effort of the Meteorological Services in the PannEx region, all the derived indices were offered to enlarge the global sub-daily rainfall dataset.
The input of the commonly used software was either the hourly or the 10-min data. The longest records go back to 1950 and earlier for some Romanian stations ( Figure 2). The series from the two Slovenian stations started in 1970 and 1975. The Croatian series are longer than 35 years starting from 1981. Overall the data density is the highest from the 1990s. The series for two stations are available for Serbia from early 1990s, but most of the data used for the analysis begin in 1998, as for Austria, Czech Republic, Slovakia and Hungary. However, many stations only have relatively short records, with a lot of missing data. The data coverage in time and the percent of missing data is varying from country to country. Beforehand, we underline that results for stations only located in the focus area are shown here. We included to the analysis gauges with less than 50 percent of missing values. The data availability by country for the longest records can be seen in Figure 2. urements covers about the last 20 years. For Slovakia, the situation is very similar, but weighing rain gauges started to replace gauges with tipping buckets earlier, since 2004, and from 2015 they became the only automated instrument in the station network.
Although the data sharing agreements between WMO nations have moved in the direction of further sharing, the data was not collected to a common database. The participating researchers run the software to derive the sub-daily precipitation indices themselves. While there are obstacles to the exchange of long-term high temporal resolution precipitation data, there have been fewer barriers to the exchange of the derived indices. The strong restrictions on sharing daily and sub-daily data in some countries (despite a WMO resolution to the contrary-WMO 2016) typically do not apply to sharing indices. As a result of this joint effort of the Meteorological Services in the PannEx region, all the derived indices were offered to enlarge the global sub-daily rainfall dataset.
The input of the commonly used software was either the hourly or the 10-min data. The longest records go back to 1950 and earlier for some Romanian stations ( Figure 2). The series from the two Slovenian stations started in 1970 and 1975. The Croatian series are longer than 35 years starting from 1981. Overall the data density is the highest from the 1990s. The series for two stations are available for Serbia from early 1990s, but most of the data used for the analysis begin in 1998, as for Austria, Czech Republic, Slovakia and Hungary. However, many stations only have relatively short records, with a lot of missing data. The data coverage in time and the percent of missing data is varying from country to country. Beforehand, we underline that results for stations only located in the focus area are shown here. We included to the analysis gauges with less than 50 percent of missing values. The data availability by country for the longest records can be seen in Figure 2.

The Commonly Used Software: INDICES
A team at Newcastle University is working on a global dataset of observed sub-daily precipitation indices associated with the INTENSE (INTElligent use of climate models for adaptatioN to non-Stationary hydrological Extremes) project. They have developed a Python library, INDICES, for calculating indices for sub-daily precipitation data. INDI-CES performs quality control of the data before deriving the sub-daily indices [47]. It also  45 50 1945 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995

The Commonly Used Software: INDICES
A team at Newcastle University is working on a global dataset of observed sub-daily precipitation indices associated with the INTENSE (INTElligent use of climate models for adaptatioN to non-Stationary hydrological Extremes) project. They have developed a Python library, INDICES, for calculating indices for sub-daily precipitation data. INDICES performs quality control of the data before deriving the sub-daily indices [47]. It also provides the summary statistics of the indices. All contributors to this article relied on this library, with the assistance of the INTENSE team. The indices themselves are adapted and extended versions of the ETCCDI indices. These indices were identified following discussions with the climate observations and modelling communities [48] and correspond, where possible, to the existing daily ETCCDI indices [49,50]. These indices condense information from the distribution of daily or sub-daily data to provide measures of intensity, frequency, and duration usually on monthly, seasonal, or annual timescales. The team is planning to release the station-level indices when possible (according to licence restrictions), and they are also producing and publishing a gridded version of the indices dataset with as much global coverage as possible.
Calculations are generally carried out on data aggregated in 1-, 3-and 6-h intervals, although some indices are only calculated using 1-h data. Each index is calculated on both a monthly and annual basis. Most indices are computed as time series, thus allowing us to calculate the trends. The calculation details for the relevant indices are given in Table 1.

The Selected Indices and Applied Statistics
We have chosen several key indices (Table 1) to visualize on maps and analyse for the entire region. For a few selected meteorological stations some of the indices are presented on graphs.
The seasonal trends for SPII1hr indices were tested with the Mann-Kendall test at the 0.05 significance level and the Sen-slope was computed [51]. The decadal change and its significance is illustrated on maps. We used R statistical language and environment for trend analysis and for mapping [52].

Spatial and Temporal Pattern of Sub-Daily Precipitation Indices
After processing the code of the INDICES software numerous results are available for 1-h, 3-h and 6-h sub-totals to analyse. Here, an initial analysis of the sub-daily precipitation climatology of the larger Carpathian region (Figure 1) with the main focus for the Pannonian basin are discussed.
The annual average of 1-h maxima (Rx1hr) and the maximum 1-h precipitation in the recording period are presented in Figure 3. The size of the markers represents the lengths of the available data series, whereas different colours present the range of indices' values. Large spatial variability can be seen on both the mean and the maximum of Rx1hr. The highest mean values, 24-26 mm, characterize the central part of the region for the most recent period (Figure 3, left, smallest dots).
outside the southern Carpathians. The highest annual average of 1-h precipitation exceeds 26 mm there. No clear spatial pattern or differences between the periods examined can be observed in the map of maxima of the 1-h totals (Figure 3, right). Large values, above 60 mm occur in all three examined periods, less in Czechia, but mainly in the middle of the region. Regarding Rx3hr and Rx6hr, the average annual maxima mainly occur in the central part of the region (Figures 4 and 5 left, HU and RO) while the absolute maxima exceeding 85 mm in 3 h and 100 mm in 6 h occur at the western and northern part of the region (Figures 4 and 5 right).  for 1-h, 3-h and 6-h sub-totals to analyse. Here, an initial analysis of the sub-daily precipitation climatology of the larger Carpathian region (Figure 1.) with the main focus for the Pannonian basin are discussed.
The annual average of 1-h maxima (Rx1hr) and the maximum 1-h precipitation in the recording period are presented in Figure 3. The size of the markers represents the lengths of the available data series, whereas different colours present the range of indices' values. Large spatial variability can be seen on both the mean and the maximum of Rx1hr. The highest mean values, 24-26 mm, characterize the central part of the region for the most recent period (Figure 3, left, smallest dots). This could be due to the openness of the northwestern part of the domain to prevailing weather systems coming from the Atlantic. Additionally, flat continental terrain warms up more compared to surrounding mountainous and coastal areas which may additionally enhance convection processes. Similar or slightly smaller values appear for the mid-long data (31)(32)(33)(34)(35)(36)(37)(38)(39)(40)(41)(42)(43)(44)(45)(46)(47)(48)(49)(50) year) series in Slovenia, Croatia and Romania. The datasets longer than 50 years (largest dots) indicate lower mean 1-h maxima in Transylvania, than outside the southern Carpathians. The highest annual average of 1-h precipitation exceeds 26 mm there. No clear spatial pattern or differences between the periods examined can be observed in the map of maxima of the 1-h totals (Figure 3, right). Large values, above 60 mm occur in all three examined periods, less in Czechia, but mainly in the middle of the region. Regarding Rx3hr and Rx6hr, the average annual maxima mainly occur in the central part of the region (Figures 4 and 5 left, HU and RO) while the absolute maxima exceeding 85 mm in 3 h and 100 mm in 6 h occur at the western and northern part of the region (Figures 4 and 5 right).    The maps of the seasonal averages of 1-h precipitation intensity (SPII1hr, Figure 6) reveal that the hourly precipitation is most intense in summer (JJA) due to the increased convective activity. The winter (DJF) intensity hardly reaches 1 mm/h; the spring (MAM) and autumn (SON) values are around 1 mm or slightly above; and the summer means  The maps of the seasonal averages of 1-h precipitation intensity (SPII1hr, Figure 6) reveal that the hourly precipitation is most intense in summer (JJA) due to the increased convective activity. The winter (DJF) intensity hardly reaches 1 mm/h; the spring (MAM) and autumn (SON) values are around 1 mm or slightly above; and the summer means exceed the 1.6 mm/h at a third of the gauges in the region. The results are consistent across the region irrespective to the period of the analysis. Comparing the corresponding mean 1-, 3-and 6-h intensities, the largest values in summer and the lowest in winter are obvious for each sub-total (Figures 7 and 8). In DJF and SON season somewhat-higher values are obtained in the south-western part of the domain (CRO) than in the rest of the area. This might be because of moist Adriatic air advection from southern directions, while air coming from northern directions over the continent is relatively dry. Similar can be seen for southern Romanian area as a result of Black Sea in the vicinity as a source of moisture and complex terrain, higher SPII1hr values are found on the windward side of the mountain compared to the lee. The maps of the seasonal averages of 1-h precipitation intensity (SPII1hr, Figure 6) reveal that the hourly precipitation is most intense in summer (JJA) due to the increased convective activity. The winter (DJF) intensity hardly reaches 1 mm/h; the spring (MAM) and autumn (SON) values are around 1 mm or slightly above; and the summer means exceed the 1.6 mm/h at a third of the gauges in the region. The results are consistent across the region irrespective to the period of the analysis. Comparing the corresponding mean 1-, 3-and 6-h intensities, the largest values in summer and the lowest in winter are obvious for each sub-total (Figures 7 and 8). In DJF and SON season somewhat-higher values are obtained in the south-western part of the domain (CRO) than in the rest of the area. This might be because of moist Adriatic air advection from southern directions, while air coming from northern directions over the continent is relatively dry. Similar can be seen for southern Romanian area as a result of Black Sea in the vicinity as a source of moisture and complex terrain, higher SPII1hr values are found on the windward side of the mountain compared to the lee.        The result of the trend analysis for the mean 1-h precipitation intensity (SPII1hr) can be seen in Figure 9. For the sake of consistency, the estimates of trend and the test of statistical significance was performed for the common period available from 1998. Due to the lack of DJF observations in Romania and in Serbia the trend estimates for winter are not available for these countries. The results revealed an increasing trend in the winter 1-h intensity concentrated to the western bound of the domain with significant increase on five stations (circled triangles). The summer SPII1hr increased in 70% of the analysed series. The largest rise exceeds 0.2 mm/decade, although only 13% of the gauges show significant change. Decreasing values appear as well, rather but not exclusively in the southwestern part of the domain and out of the southeastern Carpathians. Intensification of the autumn 1-h rainfall depth (0.15 mm/decade or more) is found in 67% of the stations (16% of them statistically significant), except in the eastern regions of Romania. A mixture of decreasing and increasing change was found in the spring trend pattern. A positive trend for some gauges (5 stations in total) with significant change appears in Austria and in Slovakia, ranging from 0.15 to 0.2 mm/decade, and the same rate but negative appears in Croatia and in Serbia (8 stations in total). A dipole pattern with an increase in northern and decrease in the southern part of the domain may be related to findings from other studies where a northward shift in storm track trajectories is detected, especially for The result of the trend analysis for the mean 1-h precipitation intensity (SPII1hr) can be seen in Figure 9. For the sake of consistency, the estimates of trend and the test of statistical significance was performed for the common period available from 1998. Due to the lack of DJF observations in Romania and in Serbia the trend estimates for winter are not available for these countries. The results revealed an increasing trend in the winter 1-h intensity concentrated to the western bound of the domain with significant increase on five stations (circled triangles). The summer SPII1hr increased in 70% of the analysed series. The largest rise exceeds 0.2 mm/decade, although only 13% of the gauges show significant change. Decreasing values appear as well, rather but not exclusively in the southwestern part of the domain and out of the southeastern Carpathians. Intensification of the autumn 1-h rainfall depth (0.15 mm/decade or more) is found in 67% of the stations (16% of them statistically significant), except in the eastern regions of Romania. A mixture of decreasing and increasing change was found in the spring trend pattern. A positive trend for some gauges (5 stations in total) with significant change appears in Austria and in Slovakia, ranging from 0.15 to 0.2 mm/decade, and the same rate but negative appears in Croatia and in Serbia (8 stations in total). A dipole pattern with an increase in northern and decrease in the southern part of the domain may be related to findings from other studies where a northward shift in storm track trajectories is detected, especially for Vb situations which are related to heavy precipitation events [53]. The result of the trend analysis for the mean 1-h precipitation intensity (SPII1hr) can be seen in Figure 9. For the sake of consistency, the trend estimates and the test of statistical significance was performed for the common period available from 1998. Due to the lack of DJF observations in Romania and in Serbia the trend estimates for winter are not available for these countries. In the last 20 years, the summer North Atlantic Oscillation (NAO) index has generally negative and has a decreasing trend; these results support findings that during such conditions the southern European area tends to be drier, while northern is wetter than normal (e.g., [45,54]). Still, it can be seen that the rate of the northern positive trend is generally stronger compared to the southern decreasing one which may indicate that there is an additional contribution to these changes. Besides changes in North Atlantic storm tracks, climate warming-related feedback processes are also one of the contributors to variations in extreme precipitation [55,56]. An increase in extreme precipitation due to general increase in air temperature, would result in a pattern observed here; more positive trend in southern area and less negative in northern part. However, this assumption needs to be investigated in detail. For example, Cheval et al. [32] showed that thermodynamic factors, such as air temperature and humidity are the major drivers for summer precipitation in the southeastern Carpathians, so that one can assume that local conditions have a major role in intense precipitation over the area, while NAO is an important triggering factor for cold season precipitation. We acknowledge that the period available for trend analysis is relatively short, so it will be interesting to update these results in the coming years to see if the results reflect say decadal variation or a long-term underlying trend. Important to note here is that not only the anthropogenic contributions can induce trends in the sub-daily rainfall depths but also the natural variability in the climate system.

Some Indices for Selected Stations
To complete the picture drawn in the previous section some additional indices are illustrated for 12 selected stations. The principle of the station selection was to cover climatic sub-regions of the Pannonian basin (Figure 1 right panel) on one hand and to include stations from all the countries contributed to this paper on the other hand. We selected one index from each group of the main types of indices to this initial study. One We acknowledge that the period available for trend analysis is relatively short, so it will be interesting to update these results in the coming years to see if the results reflect say decadal variation or a long-term underlying trend. Important to note here is that not only the anthropogenic contributions can induce trends in the sub-daily rainfall depths but also the natural variability in the climate system.

Some Indices for Selected Stations
To complete the picture drawn in the previous section some additional indices are illustrated for 12 selected stations. The principle of the station selection was to cover climatic sub-regions of the Pannonian basin (Figure 1 right panel) on one hand and to include stations from all the countries contributed to this paper on the other hand. We selected one index from each group of the main types of indices to this initial study. One from the "frequency/threshold" type indices: R3hr20mm, one from the "duration" type indices: MxLWS and the MoWH from the "Diurnal cycle" type indices. We have chosen the 3-h rainfall depth to analyse in this section instead of 1-h sum, as the 1-h sampling frequently cuts the intense rainfall events into two parts. The 20 mm threshold was chosen as it specifies a heavy rainfall event which is not extremely rare to analyse. In the half of the examined years at least one event was detected when the 3-h subtotal exceeded the 20-mm threshold. The indices R3hr20 mm and MxLWS are similar to the daily precipitation indices [50], while the MoWH is useful for the modelling community [48]. Wet hours are defined as ≥0.1 mm, acknowledging that not all instruments measure to this precision.
The graphs in Figure 10 are ordered by the number of the climatic sub-regions. We note that the time periods are not necessarily equal. The complete data series for the chosen stations were analysed to exploit most of the properties of the hourly or multi-hourly precipitation indices series.

Conclusions
One of the purposes of this paper was to explore the sub-daily precipitation data availability in the data archives of the National Meteorological and Hydrological Services (NMSs) in the PannEx region (Pannonian Basin Experiment), which is a GEWEX-related Regional Hydroclimate Project in the Pannonian/Carpathian basin. Through the collaboration of eight NMSs in the PannEx region and the INTENSE project software developed by the team at Newcastle University was applied for quality control of the hourly  The indices R3hr20mm represented by columns ( Figure 10 to the left axis on the left panel) show high variability on the selected stations. Rather variability than change can be explored in the frequency of 3-h aggregation above 20 mm. The most cases were registered in Beograd (Serbia) and Daruvar (Croatia): eight and seven respectively, which can be explained by the influence of the Mediterranean climate. Years with zero cases occurred at all stations in the examined period. The fewest cases occurred in Little Plain in Hungary (Mosonmagyaróvár) and at the Romanian stations, the latter in accordance with the growing continental character there. The shortest MxLWS can be seen in the Romanian stations; between 18-20 h on average for the examined periods ( Figure 10. left panel, line with marker). The maxima crossed 50 h on the stations representing the south-western part of the region in Slovenia and Croatia. The wettest hours for each month are indicated in the right panel of the Figure 10. The wettest hours appear from late afternoon to night-time in the warm season (May-September), as expected. Despite the relatively short time series of Szeged, the wettest hours separated well from early evening to dawn during the year. The wettest hours are grouped from early to late afternoon for Beograd from March to August, and around midnight during the fall months. We note that measuring was interrupted at Beograd in wintertime. The first insight to the wettest hours on the selected stations show that the range of the wettest hours extends from late afternoon to midnight during the year. The wettest hours happen typically not earlier than 2 pm at Miercurea Ciuc station in Romania, the most eastern station of the domain that we have included in this study.

Conclusions
One of the purposes of this paper was to explore the sub-daily precipitation data availability in the data archives of the National Meteorological and Hydrological Services (NMSs) in the PannEx region (Pannonian Basin Experiment), which is a GEWEX-related Regional Hydroclimate Project in the Pannonian/Carpathian basin. Through the collaboration of eight NMSs in the PannEx region and the INTENSE project software developed by the team at Newcastle University was applied for quality control of the hourly data and for derivation of sub-daily precipitation indices. As a result of this common effort, numerous derived statistics for 1-hourly, 3-hourly and 6-hourly rainfall depths have been produced for further analysis. Our aim was to assess some of the basic climatological properties of the sub-daily precipitation dataset collected in this initial analysis. Large spatial variability of the mean of the annual maxima and the absolute maxima of 1-h precipitation was observed in the region in the examined period. The highest mean annual maxima values around 25 mm appear in the middle part of the Pannonian basin. The maxima of the 1-h precipitation period exceeding 60 mm falls in the middle part of the region. The spatial distribution of the aggregated 3-h and 6-h maxima (annual mean and maximum) is similar to the 1-h sum, while the absolute maxima exceeds the 85 mm in 3 h and the 100 mm in 6 h at the western and northern part of the region. The summer (JJA) hourly and multi hourly precipitation is the most intense in the PannEx region, as expected. The influence of the Adriatic Sea in the southwestern part of the domain and the effect of the Black Sea in the southeastern part of the domain resulted in higher winter (DJF) and autumn (SON) hourly precipitation intensity.
The seasonal trends of the 1-h precipitation intensity were analyzed from 1998. The summer hourly precipitation intensity increased in the largest extent, at 70% of the analysed series in the region with above 0.2 mm/decade, although only 13% of the gauges show statistically significant change. The autumn 1-h rainfall depth intensified substantially, but less than the summer, overall at 67% of the stations there is an increase and 16% of them are statistically significant. It was noted that not only anthropogenic contributions can induce trends in the sub-daily rainfall depths but also natural variability in the climate system. Some of the indices for selected stations from different climatic regions of the Pannonian Basin are illustrated on graphs. Regarding the yearly occurrence of 3-h periods greater than the 20 mm threshold, a large variability in time was found. The influence of the Mediterranean climate was evident in more common occurrences (7-8 cases per year) at southern stations and, contrarily, a growing continental character was found by fewer occurrences at stations located in the eastern part of the domain. The annual maxima of the maximum length of wet hours are the fewest at the Romanian stations, between 18-20 h on average. The maximum crossing 50 h appears on the stations located at the southwestern part of the PannEx region. The wettest hours in the region range from the late afternoon to midnight. However, further study, including more stations, is required to conclude the main features of this index. Overall, in this study, we provided an initial climate analysis of sub-daily precipitation through collaboration within the PannEx initiative. One of the main aims was to contribute to enlarging the data availability for regional and global analysis of sub-daily precipitation extremes. It is important to note that sub-daily precipitation indices were analyzed here for the PannEx region, but the participating NMHSs offered more indices series to the global indices dataset.
Our plan is to involve the missing part of the PannEx region, namely the Transcarpathia, to this activity to cover the Pannonian basin and its surroundings as a whole. Including longer time series to the analysis of sub-daily precipitation indices is necessary to obtain reliable conclusions regarding the changes. The ongoing data rescue activity at NMHSs could support the lengthening the derived indices and including more stations to the investigations. Overcoming limitations imposed by short or non-existent records of sub-daily precipitation regional frequency analysis can be performed to identify homogenous extreme precipitation regions. Regionalization requires involving more climate variables and circulation types into analyses. The collaboration of the NMHSs evolved during this work provide an excellent framework for further analysis of the behavior of short-term extreme precipitation in the PannEx region.

Data Availability Statement:
The hourly precipitation data used in this study are stored locally at the National Meteorological and Hydrological Services contributed to this paper.