Assessment of QPF deficiencies in Romania

1. Report on Task 1.2 ofCOSMO Priority Project “Tackle deficiencies in QPF” Rodica DUMITRACHE, I.V.PESCARU, Liliana VELEA, Florinela POPA, Cosmin BARBU National Meteorological Administration, Bucharest, Romania

2. LM Integration characteristics Horizontal resolution: 14km (81x73, grid points) Vertical resolution: 35 levels Time step: 80s IC & LBC: GME 00, every 3h Grid-scale pp: 2-ice category scheme, prognostic Convection scheme: Tiedtke Data Assimilation:No Time of integration: 54 h

3. Methods and Criteria • Period of analysis:March-September 2005 • Method:subjective comparison(Simulated/Observed fields) with respect to: • locationof areas with significant amounts • over/under-estimationof observed amounts (at least twice larger/smaller than observed) • Comparison restricted to Romanian territory • Qualitative comparison of 48h and 24h anticipation for the same day • 10 cases selected • In choosing the significant cases, we didn’t start with the analysis of seasonal verification, since our data set used here covers only 8 months. Instead, daily subjective analysis of model results, using also the input from forecasters, showed that May-July and September were periods characterized by heavy precipitations, thus meaningful for the aim of this report. ’24 h anticipation’is used for the interval (30h-06h) of the“today”forecast and observation ’48 h anticipation’denotes the interval between (54-30) of“yesterday”forecast

4. Analyzed and presented cases • MAY 2005 • 4 - overestimation - well location • 9 - over/under-estimation • 31 - over/under-estimation • JUNE 2005 • 23 - overestimation • 30 - over/under-estimation • July 2005 • 2 - over/under-estimation • 11- overestimation • 12 - overestimation • 19 - overestimation • September 2005 • 19 – well estimation • Apart from the overestimation of observed precipitations, there are also cases when high amounts are found only in the model, in regions where maybe no precipitation was observed at all. • In some cases both values and area extension of heavy precipitation were underestimated.

5. 4 May 2005 2. • The frontal activity is manifested especially over the western part of Romania 06 UTC 12 UTC 3. • In altitude, in the south-eastern part of the continent, there is a ridge structure, dislocated by the advance of the trough from the Central Europe. 18 UTC 24 UTC => Low amounts of precipitation on large areas 1. • At surface, the pressure distribution shows the presence of Azores ridge over the western part of the continent and the expansion of the trough associated with the Icelandic Low in the center and south-east of Europe. • During the analyzed time interval, a channel is formed between this trough and the Mediterranean region. 4.This structure assures vertical instability, but the moisture content is relatively low, the air mass having continental characteristics.

6. The results of the model for 4.05.2005-24h anticipation and for 3.05.2005 -48h anticipation 4.05.2005+24h 3.05.2005+48h observed precipitation • 1. For 24 h anticipation there can be seen: • some well simulated maxima in Northern and central part of Eastern Carpathians. • area from NW part of the country is well located, but overestimated (about 40-80 l/m2 in the model, 25 l/m2 in observations). • overestimation for West of the country, SW, andEast-SE, including littoral area (5-20 lm2 in the model, no observed precipitation): 3. At 48 h anticipation presents in general the same spatial distribution , but with smaller values for the entire field. 2. => for this case in the model the amounts are overestimated mainly in south and south-east, possibly due to a faster large-scale circulation in the middle troposphere simulated in the model.

7. 19 May 2005 1. In the first part of the day, the atmospheric pressure was low over the SE of the continent, Romania being crossed by a frontal system which made the connection with the low from North of Europe. 12 UTC 06 UTC 18 UTC 24 UTC 2. In altitude, a trough with a N-S axis, toward the central basin of Mediterranean Sea, corresponded to the colder mass behind the surface front. During the entire interval, the circulation was mostly southerly over Romania, thus supporting the advection of moist air mass.

8. The results of the model for 19.05.2005- 24h anticipation and for 18.05.2005 - 48h anticipation 19.05.2005+ 24h observed precipitation 18.05.2005+48h 1. For the 24h anticipation: • the amounts are overestimated for the NE-Sband of intense precipitation, which does not appear in the observed field. • an underestimation of the amounts is found for the central part of the Eastern Carpathians (about 10 l/m2 in the model and 25-50 l/m2 in the observations) • 3. For the anticipation of 48h the area in the center of EasternCarpathians is somehow better simulated. • the area in the South is smaller and more realistic and so are the values. • The SE of the country is underestimated(no precipitationat all), but for the littoral area this is in agreement with the observation • Again the results at this anticipation seem to be characterized by lower amounts, in general, suggesting a faster large-scale circulation in the model. 2.The overestimation of forecasted quantities of precipitation may be associated with the excesive deepening of geopotential nucleus in middle troposphere (with about 8 damgp more than in the analysis)

9. 31 May 2005 1. The structure and evolution of pressure and geopotential fields indicatesthe dislocation of tropical air mass pre-existing over southern and south-eastern part of the continent and its replacement with a polar air mass, with continental characteristics, extending from north-west 06 UTC 12 UTC 3. • Quite significant was also the additional contribution of moist air in altitude, coming from Black Sea, in the first part of the considered time interval • The configuration assured a high degree of instability, manifested through intense electric activity and showers, which were generallymoderate quantitatively. 18 UTC 24 UTC • 2. This dislocation was faster at lower levels, cold air penetrating on two components: • one from North-East, advancing toward Romanian Plain, • the second from North-West, through southern part of the country. In the end, the cold air mass penetrated to the central part of the country, expelling the warm air mass in altitude.

10. The results of the model for 31.05.2005-24h anticipation and for 30.05.2005 -48h anticipation 31.05.2005+24h 30.05.2005+48h observed precipitation 1. The simulated precipitation field, at 24 h anticipation shows a false maxima ( 40—120 l/m2 in the model, no precipitation observed)inE of the country(central Moldova) An overestimation of the observed amounts is found in theSE(the modeled amounts are about 10 l/m2, while no precipitation were observed) and also in theWestern part of the country. 2. On the contrary, the intensification in the North ofEastern Carpathians is underestimated(about 5 –10l/m2 in the model, 40 l/m2 in observations). Well simulated is the maxima in the Carpathians Bend. 3. For the anticipation of 48 h the spatial distribution is similar, but the amounts are reduced compared with the 24 h anticipation for the same date.

11. 23 June 2005 • In the middle troposphere, the development of the trough toward central and southern part of the continent, associated to the baric low over Russian Plain, favored the penetration of cold air mass over the Romanian region. 12 UTC 06 UTC • In the second part of the day and during night the axis of the trough was located over Romania. • The closed nucleus accentuates the ascending motion in the eastern and southeastern regions. 18 UTC 24 UTC • In the lower troposphere, the mass separation was manifested through a cold atmospheric front, which crossed Romania from NWto SE

12. The results of the model for 23.06.2005- 24h anticipation and for 22.05.2005 - 48hanticipation 23.06.2005+ 24h 22.06.2005+48h observed precipitation • For the anticipation of 24h, the model shows a false maxima in the SWof the country (about 40 l/m2 in the model, no observed precipitation). • The eastern part of the country in generally overestimated especially in SE(Dobrogea region)(about 80-120l/m2 in the model, around 15-20l/m2 in observations). Here the convection leads to about 55 l/m2 in 24h. • The simulated field at 48 h anticipation is very similar with respect to spatial distribution, but with smaller values, especially in the East, thus being morerealistic for this area.

13. 30 June 2005 The surface and altitude fields structures suggest the extension of moist-tropical air massover the southern Europe. 12 UTC 06 UTC Surface pressure is relatively low and over Romania a frontal system evolves. 18 UTC 24 UTC The air mass with tropical characteristics determines an accentuated instability manifested in Romania through precipitation on almost the entire territory. The model overestimates the precipitation amounts in South-West, suggesting either a faster penetration of the altitude trough or an overestimation of moisture in the column on that region.

14. The results of the model for30.06.2005- 24h anticipation and for29.06.2005 - 48h anticipation 30.06.2005 + 24h observed precipitation 29.06.2005 + 48h 1.For the anticipation of 24 h, there are some regions where the simulated precipitation field presents more important errors: 3. At 48 h anticipation, the total amounts are much lower especially in SE, but the region in the Eastern Carpathians is overestimated. 2. • a false maxima in the SW of the country (in the model 40—80 l/m2, no observed precipitation); • a region with significant overstimation in Southern Carpathians/ Carpathians Bend (about 40-80l/m2 in the model; around 10 l/m2 in the observation); • the intensification in the extreme south of the littoralis drasticallyunderestimated ( 24-34 l/m2 in the observation, 2 l/m2 in the model );

15. 02 July 2005 1.The tropical air mass, from the previous case, still persists over the south of the continent, but it was slowly replaced by a mass with polar characteristics, associated to troughs in Nordic regions. 06 UTC 12 UTC 2.One of these troughs extends significantly toward south, forming a closed nucleus over south-western part of Romania. 18 UTC 24 UTC 4. • In these conditions, there was precipitation in the most parts of the country. • The model overestimated the quantities in East, where, in reality, the altitude ridge blocking persisted. 3.The pressure field at surface presents a low in the eastern basin of Mediterranean Sea, in extension toward Black Sea, having an energetic support from the altitude nucleus.

16. The results of the model for02.07.2005-24h anticipation and for01.07.2005 - 48hanticipation 02.07.2005 + 24h 01.07.2005 + 48h observed precipitation • 1. For the 24 h anticipation : • overestimation in NE–E of the country(no observed precipitation). • some centers of significant high amounts, not present in the observations: • extreme West, • intra - Carpathianregion • south-eastern part of South Carpathians. 2. The area with strong intensification in the South ofthe country is significantly underestimated(about 20 l/m2 in the model, 66-110 l/m2 in observations). • 3. For the anticipation of 48h: • the eastern partof thecountry is better simulated • overestimation on a small region in the W-NW; • in the rest of the country the amounts are underestimated, like in the previous cases.

17. 11 July 2005 1. Between 11-13 July over North-East of Europe the geopotential had very low values, looking like a large trough, with a NE-SW axis. 06 UTC 2. • This structure assures the transport of polar air mass from the extreme north of the continent and Polar Seas toward center and South of Europe. • Over the Mediterranean basin, the air mass increased its moisture content and then it was drawn on a south-western component of the flow toward our country • It should be noticed, too, the isolated cold nucleus in the Romanian region. 12 UTC 18 UTC 24 UTC 3. At surface, the gradual decrease of pressure in the western part of the country, together with the extension and enforcement of the anticyclone in the NW, suggest the organization of a frontal system. 4. In these conditions, in Romania there were precipitation in all regions, more important quantitatively in the first day (11 July) in SE, centerof the country and in the mountains, and in the second day in East of the country.

18. The results of the model for11.07.2005-24h anticipation and for10.07.2005 - 48h anticipation 11.07.2005 + 24h observed precipitation 10.07.2005 + 48h • For the anticipation of 48h: • the amounts in the eastern part of thecountry are more realistic • in S, the region with very heavy precipitation is moved toward SW. • At 24hanticipation, there are some regions where the model overestimated: • the maximum in the Eastern part of the country, also slightly moved to the north( 80-120 l/m2 in the model, about 70-90 l/m2 in the observations); • the band of intense precipitation from the S and SW(up to 120 l/m2 in the model, 31 l/m2 in the observations). • The spatial distribution is in general in good agreement with the observed field.

19. The results of the model for12.07.2005-24h anticipation and for11.07.2005 - 48h anticipation 12.07.2005 + 24h 11.07.2005 + 48h observed precipitation • The simulated precipitation field for24hanticipation is overestimatedfor: • the SE and littoral area (not in a very high degree). • for the NW of the country( up to 40 l/m2 in the model, about10 l/m2 in observations) • The general spatial distribution and the center of maximum near the Carpathians Bend are well reproduced. For the anticipation of 48 h there are more significant errors : • there are no precipitation in East ofthe country(in observations there are 10-60 l/m2); • the band of intense precipitation is located in the SW, in the model, while in reality it is on E-SE; • West of the country is strongly overestimated.

20. 19 July 2005 1. The warm weather in this interval suggests the existence of a tropical air mass in the SE of the continent, characterized by a relatively low pressure at surface and high values of geopotential in middle troposphere and of 1000-500 hPa layer thickness. 06 UTC 12 UTC 2. The instability was determined by the advance, to the east, of the trough of Icelandic Low. This trough had a correspondent in the middle troposphere. 18 UTC 24 UTC 3. The altitude structure, with the ascending part of the trough over Romania, explains the most important precipitation in this interval, in the mountainous regions. 4. Precipitation in Eastern Carpathians are explained by the additional instability caused by the moist tropical air mass, which was suddenly dislocated in the second part of the day, when the convective activity had a maximum of intensity.

21. The results of the model for19.07.2005 - 24hanticipation and for18.07.2005 - 48hanticipation 19.07.2005 + 24h 18.07.2005 + 48h observed precipitation For the 24 h anticipation there are some mismatches errors in in the simulated field of precipitation in comparison with recorded one: • in the East part of the country the area with moderate precipitations is overestimated • in N-NW part of the country the precipitation amount are overestimated too • the center of maximum in the W of Eastern Carpathians (Transilvania region) (values over 80 l/m2) is present in the observations field, but is overestimated; this center seems to be the consequence of strong convection • in SW of Romanian territory the model estimation are closed to observations. • in the S-SE light precipitation observed are not well estimatedin the model For the anticipation of 48h, the spatial distribution of precipitation is well estimated. The intensity of precipitations is slightly underestimated, and in S-SE part of the country, light precipitation observed are not present in the model.

22. 19 SEPTEMBER 1.In the middle troposphere, the geopotential field displays a large trough structure, with a NE-SW axis and the region of maximum ascendance in the Western part of Romania. 06 UTC 12 UTC 18 UTC 24 UTC 3. The precipitation in the Romanian region were generated at the contact area between acycloneformed in theAdriaticSeaand theridgeofAzoresHigh, very much extended to the east. 2. Over the eastern part there still persists the ridge associated to the warm and dry air mass from Asia.

23. The results of the model for19.09.2005 - 24hanticipation and for18.09.2005 - 48hanticipation 19.09.2005 + 24h 18.09.2005 + 48h observed precipitation 1. For the anticipation of 24h, the spatial distribution field is very realistically simulated and also the values forecasted are in good agreement with the observation 2. On the contrary, at 48h anticipation, the model doesn’t ‘see’anything– there are almost no precipitation over Romania. The spatial pattern of precipitation , compared with the anticipation of 24h, suggests a faster movement of the front toward W-SW.

24. SUMMARY I A subjective comparison of observed and simulated precipitation field for the period February – September2005, pointed out a number of 10 cases with poor QPF forecast of LM model over Romanian territory, inMay (3 cases), June(2 cases),July (4 cases) and September(1 case). The choice of these cases was done using two main criteria: • errors in location of important amounts of precipitation and with regard to the extension of the area with intense precipitation. • significant over/underestimation, especially of the heavy precipitation Also, a subjective comparison was done between the model forecast with 24 h and 48 h anticipation, for the same day.

25. SUMMARY II • In all cases, the precipitation was determined mainly by frontal systems, being accentuated by convection in some areas for some situations (31 May, 19 July). • In most of the cases the precipitations were quantitatively overestimated • in 5 cases in the model are found regions with high amounts which are not present in observations ( for 31.05, 23.06, 30.06, 2.07, 19.07), some determined by strong convection in the model (e.g. 23.06). • there was one case with significant underestimation of observed precipitation, namely 19 July. • in few cases, errors in location of areas with important precipitations are found in the model. • Comparing the results from the 24h and 48 h anticipation, it seems that possible causes for these errors in QPF forecast, for these cases, may be: • a simulated large-scale circulation faster than in reality (e.g. 4.05, 19.05, 02.07, 11.07); • lows of geopotential at 500 mb are deeper in the model than in the analysis (e.g. 23.06); • or improper estimation/simulation of specific humidity in the considered area (e.g. 31.05, 19.07).

26. THANK YOU