In Romania, the first S-band WSR-98D Doppler radar was installed in the second half of 2003. Since September 2003, the Bobohalma radar (RDBB) has performed observations in an operational regime. The fine spatial and temporal resolution permits a detailed statistical analysis of convective echoes development.

The main product of the meteorological radar is the spatial distribution of the reflectivity field, which is proportional with the density of cloud
particles and with precipitation intensity. Due to this fact the radar reflectivity field is used to detect convective storms. In this context, a statistical climatology of radar reflectivity can provide precious information to locate areas with enhanced convective potential.

There are numerous studies focused on statistical characteristics of radar reflectivity over different time intervals. The most known of them is the work of Bellon and Zawadski (2003) which analyses the reflectivity and wind data provided by the S-band Doppler radar

from Marshall Radar Observatory, aiming to realize a radar climatology of severe weather. The analysis was extended over a period of nine summer seasons between 1993 and 2001, the results providing a valuable insight into the spatial and temporal distribution and frequency of severe convection in the vicinity of Montreal. Also, extending the study over a nine-year period permits to reveal some inter-annual variability in occurrence and severity of convective storms.

Using reflectivity data on a hourly basis, recorded in May and June over the period 1975-1978, Marshall and Peterson (1980) show the evidence of topographic influences on radar echo climatology. They use a mesh in polar coordinates with a spatial resolution of $8^{\circ}\times 25$ nautical miles, and the domain has a radius equal to 100 nautical miles. The data in each cell of the mesh was treated independently by computing the mean frequency of convective echo occurrences and deviations from the mean. The results locate those cells of the circular mesh in which the frequency of convective echo occurrences are relatively high or low.

To generate a regional climatology of severe thunderstorms, Falconer (1984) uses hourly observational data of radar reflectivity for a 4 -year period, and quantifies the annual frequency of stormy days in New York state.

MacKeen at al. (1998) analyze the reflectivity data of 879 storms observed during 15 days of late spring and summer in Memphis, and using statistical analysis examine the relationships between parameters derived from reflectivity field and storm longevity.

Potts at al. (2000) process radar data recorded on 12 selected days with enhanced convective activity from the
summer of 1995/1996, and try to determine the radar characteristics of storms in the vicinity of Sydney, Australia. They analyze the relative frequency distribution of storms with reflectivity greater than or equal to 30 dBZ , and also the height and volume of these storms. Only storms situated in the range of $20-100\mathrm{~km}$ were considered.

To study the diurnal variations in relative frequencies of echoes greater than or equal to 40 dBZ , MacKeen and Zhang (2000) use reflectivity data of two radars from central Arizona, recorded in July-August 1999 and available at every 10 minutes. The results show the existence of a clear diurnal cycle in convective activity, and permit to locate the more favorable areas for convective initiation after sunrise.

Using a Cartesian mesh with 1 km x 1 km spatial resolution, and a time step of 5 minutes, Gourley and Dotzek (2004) examine the spatial and temporal structures of radar-estimated rain rates for both convective and stratiform clouds. They use data recorded in July-August 2003 from 6 Doppler radars, and by computing the frequency of maximum rain rates determine the areas with enhanced potential of occurrence of heavy rainfall and the hour of the day at which the maximum precipitation can occur.

To study the characteristics of severe convection in central Alberta, Brimelow et al. (2004) use radar parameters as VIL (Vertically Integrated Liquid), VIL from levels above 5 km and maximum reflectivity from levels above 7 km observed in July 2000. All of these parameters present a diurnal cycle with maximum intensity between 16 and 18 hours LT (Local Time).

Pascual et al. (2004) analyze the reflectivity data of a Doppler radar located in the vicinity of Barcelona recorded during the period June-September 2003, aiming to determine the location and timing of storm initiation over central Catalonia. They use two methods to identify preferred regions of convective initiation: an indirect one, by computing diurnal relative frequencies of radar reflectivity, and a direct one, by identifying and tracking the convective cells with an operational automated algorithm. The study has shown the importance of the orographic component and the diurnal forcing in triggering convection.

This study is the first in Romania that uses Doppler radar reflectivity fields to examine the spatial and temporal distribution of convection, making the first steps to create a local radar climatology.

Base reflectivity fields from the first 4 elevation angles (i.e. $0.5^{\circ},1.5^{\circ},2.4^{\circ}$, and $3.4^{\circ}$ respectively) provided by RDBB from 14 September 2003 to 31 August 2005 are examined. These fields are available at every 6 minutes in each cell, named pixel, of a polar coordinate mesh centered on radar site. The angular resolution of the mesh was $1^{\circ}$, while the radial resolution was 1 km with a maximum radius of 230 km , so the number of pixels in mesh was 82800. These pixels cover an area of 166190 $\mathrm{km}^{2}$. The observed reflectivity values corresponding to each pixel are coded and archived in binary files for each elevation level. The dataset used in this study contains 646968 such binary files, and this represents approximately $93.9\%$ of
the total number of observations that could be realized in the aforementioned period.

Romania has significant terrain variations, which influence the airflow dynamics (Fig. 1). The hilly and mountainous areas are often affected by dangerous meteorological phenomena such as heavy rain associated with flash flooding, hail and damaging wind in summer, and intense snowstorms and blizzards in winter. The ascendant motions on the windward mountain slopes, the convergence zones on the leeward sides and the elevated heat source generated by slopes exposed to the sun have strong influences on convective activity dynamics, and a major role in making areas favorable to convective initiation and later to convective cells evolution.

Although this dataset benefits by exceptional resolution and reliability some data are lost in the mountainous region where partial beam blockage occurs. Circular patterns of high frequencies of echoes greater than or equal to 40 dBZ centered on radar site are obvious in Fig. 2, especially in July and

September. This structure results from radar geometry (Ahijevych et al., 2003): when the beam is above the terrain the echoes are detected on conical surfaces. This makes the detection efficiency a function of range.

The data files are grouped by hours. For example, all files, which have the observational time between 00 and 01 UTC, were considered representative for the median moment of the interval, $00^{30}$ UTC. Using a computer code the files were decoded and the reflectivity values for each pixel identified. The number of echoes greater than or equal to 40 dBZ , $f_{40}(a,l,x,j,k,n)$, are computed and the same quantity is determined for echoes greater than or equal to $50\mathrm{dBZ},f_{50}(a,l,x,j,k,n)$. Here $a=\overline{1,3}$ stands for 2003, 2004 and 2005, $l=\overline{1,12}$ for months, $x=\overline{0,23}$ for hours, $j=\overline{1,360}$ for angles in polar coordinate system, $k=\overline{1,230}$ for the distance in $km$ from the radar site, and $n=\overline{1,4}$ for the first four elevation levels. These values represent the row data of this study and were recorded in separate files by months and hours, being 4608 such files. Then the monthly relative frequencies are calculated for both reflectivity thresholds. For echoes greater than or equal to 40 dBZ

where nrobs ( $a,l,x,j,k,n$ ) is the number of radar observations in year $a$, month $l$, hour $x$, in pixel ( $j,k$ ) from elevation level $n$. Similar expressions were obtained for $frl_{50}(l,j,k,n)$.

Over the summer months (June, July, and August) the diurnal variations in relative frequency distribution are analyzed by Fourier decomposition applied to each pixel. The discrete Fourier decomposition theoretically produces 12 harmonics,

where $x$ is the time in hours, $x\in\{0,1,\ldots,23\},\quad j=\overline{1,360},\quad k=\overline{1,230}\quad$ and $n=\overline{1,4}$, but in the present study only the zero, first and second components are calculated, which correspond to daily mean, and to the diurnal and semi-diurnal cycles. These are:

with $j=\overline{1,360},k=\overline{1,230}$ and $n=\overline{1,4}$.
Then the hourly amplitudes of first and second harmonics were determined, and for each of those 82800 pixels the maximum of these amplitudes and the time when this maximum is reached were calculated.

The dataset analysis was performed in several steps. First we have identified those pixels in which false echoes are suspected due to the radar beam return from ground surface or from other

objects. Then for both, 40 dBZ and 50 dBZ reflectivity thresholds, the monthly main statistical characteristics are determined. The analysis was performed separately for each of the 4 elevation levels. Afterwards the composite values of monthly relative frequencies are computed using reflectivity fields from the 4 elevations by

The diurnal and semi-diurnal variations of relative frequencies in June-July-August were investigated by computing the maximum amplitude and the hour of day when it occurs.

In summer the 40 dBZ reflectivity thresholds can be used to identify areas where deep convection may develop (MacKeen and Zhang, 2000). Echoes with reflectivity values greater than or equal to 50 dBZ indicate mature convective storms that can produce severe weather phenomena.

The main statistical characteristics of convection in RDBB radar coverage area are presented in Table 1. CS is the statistical property of convective activity; $\mathrm{S}_{0}$ is the total area, in $\mathrm{km}^{2}$, of pixels in
minutes, in pixels that belong to area $\mathrm{S}_{1}$; $T_{2}$ - the monthly mean duration of echoes in pixels that belong to area $\mathrm{S}_{2};T_{3}$ - the monthly mean duration of echoes in pixels that belong to area $\mathrm{S}_{3};$ Frm - the mean of monthly relative frequencies of pixels in which $f(l,j,k)>0;Az/R$ - the polar coordinates, relative to the radar site, of the pixel in which the duration of $T_{3}$ has the maximum value. Table 1 shows that the area $S_{0}$ varies between $78.2\mathrm{~km}^{2}$ in July and $144123.2\mathrm{~km}^{2}$ in January. Greater values, over $100000\mathrm{~km}^{2}$, were found in November, December, February

which the occurrence of echoes greater than or equal to 40 dBZ was zero; $\mathrm{S}_{1}$ - the total area of pixels in which the relative frequency of echoes is below normal; $\mathrm{S}_{2}$ - the total area of pixels in which the relative frequency of echoes is defined as normal; $S_{3}$ - the total area of pixels in which the relative frequency of echoes is above normal; $T_{1}$ - the monthly mean duration of echoes, in
and March, while low values, below $10000\mathrm{~km}^{2}$, in May and August. The area $S_{1}$ of below normal relative frequencies has values between $8465.1\mathrm{~km}^{2}$ in January and $56014.2\mathrm{~km}^{2}$ in June, while the area $S_{3}$ of above normal relative frequencies varies between $399.3\mathrm{~km}^{2}$ in February and $7607.4\mathrm{~km}^{2}$ in July.

The monthly mean duration of echoes greater than or equal to 40 dBZ in area $S_{3}$ has the greatest value of $21088.5^{\mathrm{m}}$ in June, followed by August with $20761.3^{\mathrm{m}}$. The lowest values were found in January with $407.4^{\mathrm{m}}$ and February with $844.2^{\mathrm{m}}$.

The mean of monthly relative frequencies, Frm, has the greatest values, over $0.1\%$, from April to August, with a maximum value equal to $0.326\%$ in July. The lowest values, below $0.05\%$, were found over the January-March period.

Figure 2 presents the distribution of monthly composite relative frequencies of echoes greater than or equal to 40 dBZ . The white areas in RDBB coverage represents pixels in which the montly relative frequency is equal to zero. Pixels colored in blue mark areas with relative frequencies below normal, while those colored in red, areas with relative frequencies above normal, according to the frequency thresholds defined above.

The yellow areas indicate normal frequency values for the month in which they are located.

One can see that the maximum monthly relative frequencies have values over $25\%$ from May to November, with the greatest value equal to $53.834\%$ in June. Over December-April period these maximum values are situated below $5\%$, having the lowest value equal to $1.084\%$ in January.

The values presented in Table 2 and the images in Figure 3 highlight the main statistical characteristics and the spatial distribution of monthly composite relative frequencies of echoes with reflectivity greater than or equal to 50 dBZ .

The area $\mathrm{S}_{0}$, in which the relative frequency is equal to zero, varies between $17748.1\mathrm{~km}^{2}$ in July and $165090.3\mathrm{~km}^{2}$ in January. Relative low values, below $81000\mathrm{~km}^{2}$, were found in months within the May-August period while highest values, over $148000\mathrm{~km}^{2}$, from October to April. This indicates, that the area of convective activity is reduced over the September-April period, when it is determined mainly by dynamical factors which appear during frontal passages, and during airflows over hilly and mountainous regions due to the interactions of them with these forms of relief. In the warm season, thermal convections acting on large domains are predominant. The mean of the monthly relative frequency, Frm, has the maximum value equal to $0.044\%$ in July. The lowest value of Frm is equal to $0.008\%$ and has been recorded in March.

Area $S_{1}$ contains pixels which have a reduced convective potential due to the low monthly relative frequencies. Over the April-October period, this area has values between $12137.0\mathrm{~km}^{2}$ in April and $25148.9\mathrm{~km}^{2}$ in August. Over the rest of the year, $S_{1}$ has values below $6500\mathrm{~km}^{2}$.
$S_{3}$ is the region in which the monthly relative frequencies of mature storms are above normal, and thereafter the frequency of the associated dangerous meteorological phenomena is also high. It has low values in cold seasons and relative high values in warm seasons. The lowest observed value was $14.6\mathrm{~km}^{2}$ in March, and the highest was equal to $11057.4\mathrm{~km}^{2}$ in July.

The mean monthly interval of time in which echoes greater than or equal to 50 dBZ were observed in area $S_{1}$ does not exceed $2.8^{\mathrm{m}}$ in January, and $5.5^{\mathrm{m}}$ in June. The low values of $T_{1}$ in the summer
months represent an additional argument to characterize area $S_{l}$ as a region with
reduced risk of dangerous meteorological phenomena.

The mean monthly duration of echoes accuring in $S_{3}$ area, $T_{3}$, has values between $9.8^{\mathrm{m}}$ and $71.6^{\mathrm{m}}$ in January, February and March. In April-October, $T_{3}$ rises and has values between $12.1^{\mathrm{m}}$ and $568.4^{\mathrm{m}}$. The highest $T_{3}$ values were observed in November and December, when they exceed 830 minutes. In these months the area of $S_{0}$ is high and the area of $S_{3}$ is low relative to their summer values suggesting a concentration of these high echoes on quite reduced regions of enhanced risk of severe weather phenomena, but with longer duration in their activity in November and December.

The maximum monthly relative frequencies has values below $0.2\%$ in the first three months of the year, with the lowest value equal to $0.04\%$ in January. In April-December, these values range between 0.683% in August and 2.137% in November.

In area $S_{3}$ of above normal monthly relative frequencies one can distinguish some semi-permanent zones, meaning that during the year they constantly appear in most months, on both images for monthly relative frequencies of echoes greater than or equal to 40 and 50 dBZ . In this light one can remark such a zone (Fig. 2 and 3) in the northern part of the RDBB radar coverage area, which traverses the central region of Maramureş county, having a belt aspect and being parallel with the Gutâiului and T, ibleş mountain ranges. In the north-west of Bistrița Năsăud county this belt presents a bifurcation, with one ramification extended toward east over the southern part of the Rodnei mountain, and another extended to the west over the northern part of Cluj county, and farther to the south-west near the boundary zones

C. PAVAI, C. VAMOŞ

between Cluj and Sălaj counties. Hereafter this zone will be referred to as "Zone A". Analyzing Fig. 2 and 3, it is easy to see that over the April-December period Zone A appears in every month, and it is more contoured over the SeptemberDecember period. Also, the pixel with maximum values of monthly relative frequency in April and December, for echoes greater than or equal to 40 dBZ (Table 1), and in July, August, October, November and December, for echoes greater than or equal to 50 dBZ (Table 2) lies in Zone A.

Another semi-permanent zone which can be observed in Fig. 2 and 3 is the "Zone B", which lies approximately 90 km south of RDBB, having a band aspect extending from west to east parallel to the Cândrel and Făgăraşului mountain ranges, over the northern slopes of the Southern Carpathians It is easiest to locate in June for echoes greater than or equal to 40 dBZ , but it appears in all images for echoes greater than or equal to 50 dBZ . The pixel with maximum monthly relative frequency in January and February for echoes greater than or equal to 50 dBZ (Table 2), and in MayNovember period for echoes greater than or equal to 40 dBZ (Table 1) lies in Zone B.

A third semi-permanent zone with high monthly relative frequencies, "Zone C", is situated near the Curvature Carpathians in the boundary regions between Braşov and Prahova counties, and also between Covasna and Buzău. Sometimes, for example in September, this zone is extended to the north-east following the frontier between Covasna and Vrancea counties. Similarly to Zone A and B, it appears as a belt of high monthly relative frequencies over the north-western slopes of the Curvature Mountains.

From April to September, and also in November and December, Zone C appears on images of monthly relative frequencies for echoes greater than or equal to 40 dBZ (Fig. 2). In the case of echoes greater than or equal to 50 dBZ , Zone C appears on images over the AprilNovember period in $S_{3}$ area (Fig. 3). The pixel with maximum monthly relative frequency for echoes greater than or equal to 50 dBZ in April and May lies in Zone C.

In Fig. 2 and 3 one can locate zones with high monthly relative frequencies belonging to $S_{3}$ appearing just in a few months of the year. For example in summer season, in July and August, in areas belonging to Cluj and Sălaj counties one can recognize a circular band of high relative frequencies, well contoured for echoes greater than or equal to 50 dBZ , which indicates an enhanced convective potential for that region and for these months.

The composite relative frequencies for zero harmonic were calculated using

relative frequencies from the 4 elevation levels by $A_{0}(j,k)=\max_{n=\overline{1,4}}\left\{A_{0}(j,k,n)\right\}$, $j=\overline{1,360},\quad k=\overline{1,230}.\quad$ In the same manner we have calculated the composite relative frequencies for the first and second harmonics. If I is the greatest cluster with the properties that $\mathrm{I}\subseteq\{1,2,..,360\}\times\{1,2,..,230\}$ and $\mathrm{A}_{0}(\mathrm{j},\mathrm{k})>0\forall(j,k)\in\mathrm{I}$, then the mean value of the zero composite harmonic will be $\overline{A_{0}}=\frac{1}{\operatorname{Card}(I)}\sum_{(j,k)\in I}A_{0}(j,k)$. In Figure 4 the spatial distribution of composite
rapidly. A more compact zone of high relative frequencies can be located in the southeastern half of Sălaj county, in the Meses mountain region. In the southern part of Cluj county, one can distinguish another band of high relative frequencies extending from west to east, flanked at the southern side by the Muntele Mare mountain range. At the eastern frontier of Cluj county this band is continued to the north with some isolated maximum of relative frequencies.
There are some other clusters of high relative frequencies south and north of the Făgăraşului mountain, over Vâlcea and

relative frequencies for the zero, the first and second harmonics are presented. The relative frequency values for $\mathrm{A}_{0}(\mathrm{j},\mathrm{k})>0,(j,k)\in\mathrm{I}$, are situated between $0.002\%$ and $0.943\%$, with the mean value $\overline{A_{0}}=0.029\%$. The high relative frequencies, over $0.07\%$, are more numerous in the northwestern part of the RDBB radar coverage area. There exists a local maximum extending as a band, which traverses the central region of Maramureş county along a line from NW to SE, parallel to the T,ibleş mountain. To the southwest and northeast from this band the relative frequency decreases

Sibiu counties, and also west and east of the Gurghiului and Harghitei mountains. All these zones can be found on images of monthly composite relative frequencies over the June-August period.

The spatial distribution of the zero, first and second harmonics at the 4 elevation levels are presented in Fig. 5. The predominance of the first elevation level in composite images is clear if we compare Fig. 4 with Fig. 5. At the higher levels, the radar detects echoes over more reduced ranges due to the overshooting effect. For example, at the second elevation level, pixels with relative

found also at the third and fourth elevation levels, while in any other part of the RDBB radar coverage area pixels with relative frequency greater the 0.07% do not exist.

The maximum composite amplitude of the first harmonic, $A_{1}(j,k),(j,k)\in\mathrm{I}$, varies between $0.001\%$ and $0.835\%$ (Fig. 4 ), with the mean value $\overline{A_{1}}=0.033\%$. One can see that the high amplitudes, with relative frequencies over $0.07\%$, are grouped predominantly in zones where the zero harmonic exceeds the 0.045% threshold. Again, the values from the first elevation level (Fig. 5) dominate the composite field of the first harmonic. At all levels, the areas of pixels with high diurnal amplitudes are greater than areas occupied by pixels with high values of the zero harmonic at the same level.

The maximum composite amplitude of the second harmonic has values between $0.001\%$ and $0.508\%$, with a mean value equal to $0.031\%$. As in the case for fields of $A_{0}(j,k)$ and $A_{1}(j,k)$, the composite image of $A_{2}(j,k)$ is dominated by values from the first elevation level, and also the values over $0.07\%$ are situated in areas where $A_{0}(j,k)$ has values greater than $0.045\%$, but the areas of these regions are reduced in comparison with the regions occupied by the same values of $A_{1}(j,k)$.

In general, the intensity and extension of fields $A_{0},A_{1}$ and $A_{2}$ decrease when the elevation angle increases. Some exceptions exist however, for example over Argeş county where an increase in intensities of these fields can be observed when the elevation level changes from 1 to 2 (Fig. 5), probably due to the partial beam blockage at the first elevation level caused by the Southern Carpathians (Carpații Meridionali mountain range).

It is important to observe that the fields of $A_{0},A_{1}$ and $A_{2}$ at all levels contain low relative frequencies over the central region of Cluj county.

In order to examine the magnitude of the diurnal and semi-diurnal variations the following ratios were calculated: $r_{10}(j,k)=A_{1}(j,k)/A_{0}(j,k),\quad r_{20}(j,k)=A_{2}(j,k)/A_{0}(j,k),\quad$ and $\quad r_{21}(j,k)=A_{2}(j,k)/A_{1}(j,k),\quad\forall(j,k)\in\mathrm{I}$ using the composite values of $A_{0},A_{1}$, and $A_{2}$. The ratio $r_{10}(j,k)$ varies between 0.067 and 2.5, having the mean $\mathrm{r}_{10}=1.291$. The number of pixels with $r_{10}(j,k)>1$ is equal to 62465 , in other words for $77.5\%$ of the total number of pixels the maximum diurnal amplitude exceeds the daily mean. $r_{10}(j,k)>2$ for a number of 2736 pixels, and this means $3.4\%$. The ratio $r_{20}(j,k)$ has values between 0.033 and 2.5, with the mean $\mathrm{r}_{20}=1.051$. Values greater than 1 were obtained for 43042 pixels $(53.4\%)$, and greater than 2 for 1789 pixels ( $2.2\%$ ). This means that the amplitude of semi-diurnal variations exceeds the daily mean on $53.4\%$ of pixels. The third ratio, $r_{21}(j,k)$ has values between 0.024 and 26.0, with the mean $\overline{\mathrm{r}_{21}}=0.892$. The number of pixels on which this ratio exceeds 1 is equal to 22324 ( $27.7\%$ ), and on which exceeds 2 is equal to $2788(3.5\%)$. So, on more than one quarter of pixels the semi-diurnal amplitude exceeds the diurnal amplitude.

Fig. 6 presents the spatial distribution of ratios $r_{10}(j,k),r_{20}(j,k)$ and $r_{21}(j,k)$. Fig. 6a shows clearly the predominance of zones where the maximum amplitude of diurnal variations exceeds the daily mean. In Zone A the diurnal maximum amplitudes are reduced in comparison to the daily mean, and the semi-diurnal maximum amplitudes are

more than twice of the diurnal maximum amplitudes (Fig. 6c).

Frequently, on extra-Carpathian plains the maximum diurnal and semi-diurnal

RDBB radar coverage area does not exist see/land breeze effects, but there are mountain breezes that affect large areas.

Fig. 7a shows the hours at which the

amplitudes exceed $1.5-2.5$ times the daily mean (Fig. $6a$ and $6b$ ). In the hilly regions between the mountainous and plain regions the semi-diurnal maximum amplitude exceeds over large areas the diurnal maximum amplitude (Fig. 6c).

Semi-diurnal variations are caused probably by mountain breezes. After Carbone et al. (2003), the semi-diurnal variations in precipitation fields has long been associated with atmospheric tidal lifting. But, using data from 7 warm seasons, they demonstrate that the semi-diurnal signals in precipitation fields are essentially unrelated to the atmospheric tides. Also, they show that in eastern and southeastern regions of the United States the local effects of sea/land breezes are responsible for the semi-diurnal variations, especially along the Gulf of Mexico and the Atlantic Ocean coasts. Hudak et al. (2004) found semi-diurnal signals in convective cell properties in the Great Lakes region, and they showed that these signals were related to the lake breeze/land breeze mechanism that worked in the area. In the
diurnal amplitude of composite relative frequencies rises to its maximum. One can see that in intra-Carpathian zones and in the eastern sector of the RDBB radar coverage area, the first harmonic has the maximum amplitude predominantly in the afternoon between 15 and 18 LT . In the mountainous region of the western part of Cluj and Alba counties, and in areas belonging to the Cândrelu and Lotrului mountains, and also in some areas of the Eastern Carpathians there are pixels in which the maximum diurnal amplitudes are reached early, between 11 and 14 LT . In the extra-Carpathian regions in the western and southern parts of the radar coverage area, the maximum of diurnal amplitude was reached in the evening and over the first half of the night (Fig. 7a). These aspects confirm the known fact that over the higher terrain the convective activity is most frequently near the time of maximum cumulative heating. This elevated heat source produces ascending motions over the mountainous regions during the day, and descending motions over the adjacent plains. During the

evening and night hours, subsidence and convective inhibition diminish over the plains, and thunderstorms propagate away from the mountains (Ahijevych et al., 2003).

A typical case of that type can be found in the eastern part of the RDBB radar coverage area. Analyzing in parallel Fig. $4b$, Fig. $7a$ and Fig. 1 one can see

values decreased below $0.025\%$, and in some parts of the band below $0.01\%$. The convective activity dies out here between 23 and 02 LT, excepting the northern extremity of the band, which persists until the first hours of the next morning and dies out between 03 and 06 LT. This peak motion to the east from high to low elevations with increasing intensity, and

Fig. 7. The time when the diurnal and semi-diurnal amplitude reaches its maximum in case of echoes greater than or equal to 50 dBZ .
that at noon and early in the afternoon over the Eastern Carpathians there are zones where the convection reaches its diurnal maximum. In late afternoon, between 15 and 18 LT , the peaks in maximum diurnal amplitudes appear moved to east and more intensified, with values greater than $0.07\%$, along a northsouth oriented line, parallel to the mountain range. Between 19 and 22 LT this band of peaks continues to move toward east, but the intensity of diurnal maximum amplitudes presents a decrease below $0.045\%$. Between 23 and 02 LT, this band can be found a little to the east, over the hilly and plain regions, with
then the weakening associated with deceleration of the propagation speed suggests that the mountain-plains dynamics is responsible for the nocturnal convective activity in this region.

Fig. $7b$ presents the spatial distribution of hours at which the semidiurnal amplitude reaches its maximum. The distribution is dominated by the morning hours between 03 and 06 LT, and the late afternoon hours between 15 and 18 LT . In the mountainous region, for example in the Western Carpathians over Cluj and Alba counties, the maximum in semi-diurnal variations is reached between 11 and 14 LT, and between

23 and 02 LT, while the minimum between 17 and 20 LT , respectively 05 and 08 LT. The maximum between 11 and 14 LT is superposed with the maximum in the diurnal variations (Fig. 7a), and this enhances the convective activity in this region and this time interval. In the same way, the minimum of semi-diurnal amplitude between 17 and 20 LT determines a weakening in convective activity.

Over the extra-Carpathian plains there are large areas where between 19 and 02 LT the semi-diurnal maximums are superposed with the diurnal maximum, favoring an extension of convective activity until late in the evening and first half of the night.

As in the case of echoes greater than or equal to 50 dBZ , the relative frequencies of reflectivity fields from the 4 elevation levels and of their composite field were analyzed.

Fig. 8 shows the spatial distribution of the composite relative frequencies for
$A_{0}(j,k),A_{1}(j,k)$ and $A_{2}(j,k)$ with $(j,k)\in\mathrm{I}$.

In the field of zero harmonic, relative frequency varies between $0.002\%$ and $47.031\%$, having the mean value of $A_{0}=0.238\%$ (Fig. 8a).

The maximum amplitude of diurnal variations has values between 0.001 and 5.212 %, with the mean value of $\overline{A_{1}}=0.198\%$ (Fig. 8b).

The maximum semi-diurnal amplitude has values between 0.001% and $2.499\%$, with a mean value $\overline{A_{2}}=0.170\%$. The lowest values were found in regions of the extra-Carpathian plains

The ratio $r_{10}(j,k)$ has values between 0.014 and 2.5, with the mean $\overline{\mathrm{r}_{10}}=0.775$. The number of pixels with $r_{10}(j,k)>1$ is $17995(22.3\%)$. The ratio $r_{20}(j,k)$ varies between 0.016 şi 2.5 , having the mean value $\mathrm{r}_{20}=0.558$. The number of pixels with $r_{20}(j,k)>1$ is 6310 (7.8%). Ratio $r_{21}(j,k)$ varies between 0.018 and 46.4, with the mean value $\mathrm{r}_{21}=0.826$.

Values greater than 1 in fields of ratios $r_{10}(j,k)$ and $r_{20}(j,k)$ can be found over extra-Carpathian regions. The number of pixels with values greater than

2 is very low, equal to 174 for $r_{10}(j,k)$ and 109 for $r_{20}(j,k)$. For $r_{21}(j,k)$ the number of pixels greater than 2 is 2904 , higher than
single radar can provide useful quantitative information related to the characteristics of local convective

in the case of the first two ratios.
Fig. 9 shows the hours of the day at which the diurnal and semi-diurnal amplitude reaches its maximum. The hours between 15 and 22 LT are dominant for the peak in diurnal amplitude, while for the semi-diurnal peak the hours between 03 and 06 LT, respectively 15 and 18 LT are more frequent. In these areas, between 15 and 18 LT the diurnal maximums are superposed with the semi-diurnal maximums, while between 03 and 06 LT the diurnal minimums are superposed with the semidiurnal maximums.

The results show that the analysis of radar reflectivity temporal series - even for such a short period of two years - from a
circulations. It was shown that in the RDBB radar coverage area the local influences over the convective circulations are due to the very diverse topographic aspects. There are numerous orographic forcings such as mountainplain dynamics, mountain-valley breeze, up slope and down slope winds and elevated heat source effects. The dynamical effects of the mountain ranges have a major impact over the local airflow and influence the climate of adjacent regions.

Convergence zones and convergence lines linked to topographies were located, which can activate in some specific synoptic and mesoscale conditions. These conditions are not explored in this study, but even knowing only the storm-prone zones and times when convective activity

reaches its maximum can be beneficial in operational nowcasting activity.

The results show clear mesoscale variability in storm occurrence and locate areas with high/low convective potential, highlighting their distinct statistical characteristics. In the summer, the monthly maximum relative frequencies vary between $0.008\%$ and $53.834\%$, $0.008\%$ and $49.333\%$, respectively $0.008\%$ and $52.664\%$ for June, July and August in the case of echoes greater than or equal to 40 dBZ , and between $0.008\%$ and $1.451\%$, $0.008\%$ and $1.129\%$, respectively $0.008\%$ and $0.683\%$ in the case of echoes greater than or equal to 50 dBZ. This large intra-seasonal variability shows the complexity of the meteorological control - monitoring and forecasting - over the convection in this area.

The study has found strong diurnal and semi-diurnal signals in reflectivity fields for echoes greater than or equal to 40 dBZ , and respectively 50 dBZ observed in the June-August summer season. In general, the mountainous regions present a maximum convective activity in the first afternoon hours. But, the fine structures of pattern of hours with the maximum diurnal amplitude are correlated with the local topography. These variations in phase are caused by the local mountain-valley circulations that influence the time of storm initiation.

It was confirmed the known result that the diurnal cycle of convective phenomena varies with elevation and
frequently convective cells appear at high elevations and then propagate to low elevations determining a nocturnal maximum in convective activity over the plain regions. So, in mountainous regions the convective activities are more enhanced in the first hours of the afternoon, while over the plains the late evening hours and the early hours of the next morning predominate.

The high reflectivity thresholds, 40 dBZ and 50 dBZ , minimize the problems associated with false echoes and with dependence of radar sensitivity on range (Bellon şi Zawadzki, 2003). However, in the analyzed domain there are some pixels with very high relative frequency suggesting the presence of persistent false echoes, especially on the first elevation level. Although according to statistical evaluations they have little influence on the results, it is unclear whether these pixels with very high relative frequency correspond to stationary orographic storms or to false echoes.

The study shows again that the radar reflectivity data and the statistical analysis methods can provide useful information to recognize and quantify the risks of dangerous meteorological phenomena associated with convective storms.

Acknowledgements: The authors are grateful to the two anonymous reviewers for their constructive remarks.

Ahijevych, D. A., Carbone, R. E. and Davis, C. A. (2003)
Regional-scale aspects of the diurnal precipitation cycle, Proceedings of the 31st International Conference on Radar Meteorology, American Meteorological Society, Boston, 349-352.
Bellon, A. and Zawadzki, I. (2003)
A 9-year summary of radar characteristics of mesocyclonic storms and of deep convection in Southern Quebec, Atmosphere-Ocean, 41, 99-120.
Brimelow, J. C., Reuter, G. W., Bellon, A. and Hudak, D. (2004)
A Radar-Based Methodology for Preparing a Severe Thunderstorm Climatology in Central Alberta, Atmosphere-Ocean, 42, 1, 13-22.
Carbone, R. E., Tuttle, J. D. and Ahijevych, D. A. (2003)
Inter-annual and semi-diurnal variations in summertime precipitation, Proceedings of the 31st International Conference on radar Meteorology, American Meteorological Society, Boston, 347-348.
Falconer, P. D. (1984)
A radar based climatology of thunderstorm days across New York state, Journal of Applied Meteorology, 23, 7, 1115-1120.
Gourley, J. J. and Dotzek, N. (2004)
The effects of vertical air motions on radar estimates of rainfall, Proceedings of ERAD 2004, 374-378.
Hudak, D., Boodoo, S. and Donaldson, N. (2004)
Time varying properties of convective systems in Great Lakes region of North America, Proceedings of ERAD 2004, 537-540.
MacKeen, P. L., Brooks, H. E. and Elmore, K. L. (1998)
Radar reflectivity derived thunderstorm parameters applied to storm longevity forecasting, Weather and Forecasting, 14, 289-295.
MacKeen, P. L. and Zhang, J. (2000)
Convective Climatology for Central Arizona During the 1999 Monsoon, Proceedings of the $2^{\text{nd }}$ Southwest Weather Symposium, Tucson, Arizona, 64-67.
Marshall, T. P. and Peterson, R. E. (1980)
Topographic influences on Amarillo radar echo climatology, Proceedings of the Eighth Conference on Weather Forecasting and Analysis, Denver, Colorado, June 10-13.
Pascual, R., Callado, A. and Berenguer, M. (2004)
Convective storm initiation in central Catalonia, Proceedings of ERAD 2004, 464468.

Potts, R. J., Keenan, T. D. and May, P. T. (2000)
Radar characteristics of storms in the Sydney area, Monthly Weather Review, 128, 3308-3319.