sign in sign up
<<
Home , Breznik, Cherni Vrah, Kulata, Marino Pole, Radomir (town), Sapareva Banya

B U L G A R I A N R E P O R T

Contents

No. Item Work Package Page 1. Bulgarian data supplied to EFFS 2,3,4 1.1. Descriptive data, GIS layers 4 1.2. Hydrological data 3 1.3. Meteorological data 2 2. HRLA meteorological forecasts 6 2.1. Partners’ forecasts 6 2.2. ALADIN forecasts 6 2.3. Comparison of observed data and ALADIN forecasts 6 4 2.3.1. Location of automatic observation stations 64 2.3.2. Methodological remarks 6 6 2.3.3. Analysis of the comparison results 6 7 3. HBV model schematization and calibration 8 8 3.1. Model schematization, sub-basins 8 8 3.2. Model calibration, first results 8 9 4. Simulations with the ISBA surface scheme 8 11 4.1. Modeling structure 8 11 4.1.1. ISBA (Interface Soil Atmosphere Biosphere) Land 811 surface scheme 4.1.2. Distributed regional scale hydrological and hydro- 812 geological model 4.2. Analysis of the project feasibility 8 13 4.3. Adaptation of ISBA-MODCOU coupled system for 814 using ALADIN precipitation fields 5. Bulgarian EFFS WEB page 9 16

Remark:

The involvement of the project team members in the implementation of the above listed tasks is as follows:

Dobri Dimitrov – overall coordination and participation in all tasks Snejana Balabanova – 1.1 Kamelia Kroumova – 1.2 Yordanka Berova – technical support 1.2, 1.3 Valery Spiridonov – 2 Georgy Koshinchanov – 2.3 Krasimir Stanev – 3,(1) Eram Artenian – 4,(2) Minka Stoyanova – 5 1. Bulgarian data supplied to EFFS

The test basin selected for the EFFS is one of the biggest at South - river, flowing from North to South up to the . Considerable part of the river basin is situated in northwest Bulgariais, heaving an area of more than 10000 km2 and average elevation about 900 m a.s.l. It is a mountain basin with steep slopes and relatively small concentration time, generating flush floods of snowmelt – rainfall type mainly in the late spring. Increased lead-time of the flood warnings is needed to allow the stakeholders to organize effective measures decreasing the damages and losses. The basin is transboundary one, shared with EU member state () with a clear need for increased lead time flood forecasts needed for effective management of Kerkini reservoir and selection of effective mitigation measures during floods. The basin is the closest one to the NIMH headquarters with well-developed infrastructure requiring reasonable costs to organize real time data transmission for river discharges. The available historical data series comprise meteorological and hydrological daily values at several reasonable locations of the river basin.

1.1. Descriptive data, GIS layers:

The supplied GIS data are in geographical coordinates WGS 84. They are uploaded to the EFFS FTP server as ArcView “.shp” files as follows: - DTM - cut for the Struma basin (300 x 300m); The hipsograph (area – elevation) curves were established for all sub-basins using the above DTM. Example of that calculation is given in item 3.1. - basin/sub-basin contours (only for the discharge calibration stations); - Stations for model calibration Discharge – Rajdavitza, , Dupnitza, Krupnik/Kres.hanche, Strumeshnitza, Air temperature and precipitation – , , Kustendil, , Precipitation – h.Osogovo, h.BAN, Yaz.Kalin, Fish.farm, Turishka cherkva, v.StDimitrovo, v.Rajdavitza, Kresnensko hanche, v.Zlatarevo, v.Marino pole, v., t., Boboshevo, Dolna Melna, , Divlya, Dolno selo, Bobovdol, Dren, Nevestino, , Dobro pole, Gradevo, Dolene, Levunovo, - River network (simplified); - big lakes - big reservoirs - Main roads and railways - Towns - State boundary (cut for Struma basin region)

1.2. Hydrological data

Data series with daily mean discharges from six hydrologic stations are provided for the model calibration over the period 1990 to 2002. Stations list is provided in Table 3 below. The number of sub-basin, which they are closing and some basic characteristics are given as well. The division of the pilot basin on sub-basins and the label (number) of sub- basin are given in fig. 1 as well. It should be noted that one of the stations is closing a basin located outside the Bulgarian territory. 1.3. Meteorological data

Daily precipitation totals are provided for 19 stations over the pilot basin. Six of them are of the so called “climatic” type. For those six stations daily averages of air temperature are also provided. Monthly evaporation totals averaged over the period 1990 – 2002 are provided for two of the climatic stations located at the middle part of the pilot basin. The list of precipitation and climatic stations is given in Table 4 below.

КЪ Lakes.shp r Places.shp Rivers.shp T R E S Roads.shp K V PERNIK L E 1 S# Hydrological_station.shp Y %U T %U A L Med-hycos station Y $Z N A S КЪ $T Wa te r lev el gauge K A Meteor ol ogical_station.shp S# automatic A UM %U climatic RO ST V& climatic and automatic $T # Med-Hycost station S# r %U r precipitation r КЪ precipitation and automatic %[ synoptic r 3 і# synoptic and automatic і# %U Watershed51700.shp N$T r A V& M Watershed51750.shp R E Watershed51430.shp H Z D Watershed51800.shp %[ NITSA $Z# KA LESH RILS Watershed51560.shp E S# Watershed51880.shp V& r Struma_grid GRADSKA) BISTRITSA (BLAGOEV 83 - 387 %[ 388 - 692 693 - 997 BLAGOEVGRAD 998 - 1302 1303 - 1606 r 1607 - 1911 1912 - 2216 2217 - 2521 $T 2522 - 2826 No Data S# S# A) SK AN D AN (S A S # IT S TR IS B %[

r

r S#$T $T S# A%U HNITS PETRICH UMES STRO

Fig. 1. Example of the GIS layers visualization

2. High Resolution Limited Area meteorological forecasts 2.1. Partners’ forecasts

After the review of the forecasts available in the EFFS FTP server, a conclusion was reached that the data of the numerical weather forecast models are inconvenient for the territory of Bulgaria as follows: • The data from DMI are on the domain: North: 75.07, South: 29.5, West –25.2 and East: 21.7, i.e. Bulgaria is outside this area. • The data from ECMWF and GME model of DWD are on coarse mesh for our purposes 0.75 x 0.75 degrees in Lat/Lon coordinates (approximately 70-80 km). • The fine resolution LM model of DWD covers areas apart of our domain of interest.

2.2. ALADIN (Aire Limitée Adaptation dynamique Développement Inter National) forecasts

Aladin short-range NWP model has been operating at the NIMH since May 1999. The numerical weather prediction model ALADIN (http://www.cnrm.meteo.fr/aladin/) - has been used as operational model in Bulgaria since June 1999. The weather forecast for 48 hours over the Balkan Peninsula is computed twice a day using as initial conditions the predictions for 12 and 00 UTC of the French global model ARPEGE (Action de Recherche Petite Echelle Grande Echelle). The horizontal resolution of ALADIN is approximately 12 km, with 31 levels vertically. The model is widely used in Europe. It has three main implementations in Western Europe – France, Belgium, and Portugal, more than 7 in central and Eastern Europe and two in North Africa.

2.3. Comparison of observed data and ALADIN forecasts

The Struma river is collecting waters from four countries (Yugoslavia, Bulgaria, Macedonia and Greece), flowing from North to South up to the Aegean Sea. Considerable part of the river basin is situated in southwest Bulgaria (fig. 1.), having an area of more than 10000 km2 and average elevation of about 900 m a.s.l. The Struma’s spring is situated in the southern slopes of the mountain, just below the Cherni Vrah peak, and its geographical coordinates are as follows: 42º33’40” north latitude and 23º16’40’’ east longitude at 2180 m altitude. The river leaves our territory at village and its coordinates there are as follows: 41º22’00’’ north latitude and 23º20’40’’ east longitude at 62 m altitude. The length of the Struma River (from its spring till the place, where it leaves our country) is 290 km. The Struma river catchment area is very elongated: the average length of the area is round 250 km, while the average width is only 40 km. The basin has a pronounced mountain character.

2.3.1. Location of automatic observation stations:

Network of 15 automatic consists of: • 6 meteorological stations (having average hourly data series for air temperature, rain/snowfall amount); • 5 hydrological stations (having hourly aggregated sums for rain/snowfall amount); • 4 rain/snow gauges (having hourly aggregated sums for rain/snowfall amount). Fig. 2. The pilot basin, location of automatic observation stations

Table 1. List of automatic observation stations used Station location Type Co-ordinates, elevation town Dupnitza), river Djermanska Level, rain / snow N: 42o15’43”, E: 23o07’27” v. Rajdavitza, river Struma Level, rain / snow N: 42o23’01”, E: 22o42’41” Kresnensko hanche, river Struma Level, rain / snow N: 41o46’51”, E: 23o09’11” v. Strumeshnitza, r. Strumeshtnitza Level, rain / snow N: 41o23’39”, E: 23o02’13” v. Marino pole, river Struma Level, rain / snow N: 41o25’06”, E: 23o18’57” h. Osogovo Met. station N: 42o23’01”, E: 22o42’41”; 1650 m. h. BAN Met. station N: 42o37’07”, E: 23o14’14”; 1480 m Yaz. Kalin Met. station N: 42o08’03”, E: 23o14’07”; 1500 m Fishing farm, v. Suhostrel Met. station N: 41o44’40”, E: 23o05’24”; 650 m Locality Turishka Cherkva Met. station N: 41o39’45”, E: 23o24’15”; 1560 m v. Rila Met. station N: 42o07’03”, E: 23o07’11”; 504 m v. Blagoevgrad Rain / snow N: 42o00’10”, E: 23o05’54” v. Breznik Rain / snow N: 42o44’51”, E: 22o54’34” t. Radomir Rain / snow N: 42o33’05”, E: 22o57’31” v. Kustendil Rain / snow N: 42o17’06”, E: 22o42’47” 2.3.2. Methodological remarks

The following comparison is for the temperatures and accumulated amount of rain(snow)fall for every 3 hours and for every 24 hours. The comparison is between the observed by the telemetric stations in the Struma river basin elements (temperatures and rain(snow)fall) amount and equivalent elements given by the forecasting model “ALADIN” for max 48 hours ahead of the time when issued. For the 3 hours comparison the measured elements are taken for each measured hour for which the forecast is issued (for the temperature) and the sum of rain(snow)fall for the preceding 3 hours before the term of the forecast. For example if the forecast is issued on 1.II.2003 at 00 UTC it will have the following terms: 1.II.2003 at 03 UTC, 1.II.2003 at 06 UTC , …. ,3.II.2003 at 00 UTC (which is the forecast issued for 48 hours ahead). We take the measured temperature for the corresponding terms of the forecast (having in mind the time difference zone-the data in the telemetric stations are stored in local time) and the amount of the rainfall for the last 3 hours before the according term. Example of the data is shown in table 2 below.

Table 2.a. Table 2.b. 29/09/02 00:00:00 4.37 3.40 4.20 5.00 30/09/02 00:00:00 7.0 3.1 4.9 0.0 29/09/02 03:00:00 4.96 2.60 3.70 30/09/02 03:00:00 0.6 3.7 0.7 29/09/02 06:00:00 5.56 2.60 4.70 30/09/02 06:00:00 0.2 0.9 0.6 29/09/02 09:00:00 6.13 4.20 6.00 30/09/02 09:00:00 0.2 0.8 0.4 29/09/02 12:00:00 5.73 5.00 5.50 30/09/02 12:00:00 0.0 1.2 0.6 29/09/02 15:00:00 5.61 4.60 5.20 30/09/02 15:00:00 0.2 1.1 4.9 29/09/02 18:00:00 5.62 3.70 5.50 30/09/02 18:00:00 0.2 0.3 9.6 29/09/02 21:00:00 5.71 2.70 4.30 30/09/02 21:00:00 3.8 2.2 4.1 30/09/02 00:00:00 5.59 1.70 3.10 2.40 01/10/02 00:00:00 3.6 7.2 13.60.0 30/09/02 03:00:00 5.30 1.20 2.20 01/10/02 03:00:00 3.2 8.0 2.0 30/09/02 06:00:00 6.08 1.10 2.30 01/10/02 06:00:00 6.8 1.3 0.7 30/09/02 09:00:00 7.49 4.20 5.70 01/10/02 09:00:00 3.2 1.4 0.8 30/09/02 12:00:00 5.76 5.60 6.40 01/10/02 12:00:00 2.4 2.0 1.4 30/09/02 15:00:00 6.25 5.10 5.10 01/10/02 15:00:00 1.2 1.9 1.2 30/09/02 18:00:00 3.88 2.80 3.40 01/10/02 18:00:00 0.0 1.6 1.4 30/09/02 21:00:00 3.73 1.50 3.00 01/10/02 21:00:00 0.0 1.1 1.0 01/10/02 00:00:00 2.82 0.70 2.30 3.00 02/10/02 00:00:00 0.2 0.0 0.9 1.8 01/10/02 03:00:00 1.10 1.30 2.40 02/10/02 03:00:00 0.0 0.5 0.6 01/10/02 06:00:00 0.69 1.40 2.10 02/10/02 06:00:00 0.0 0.2 0.7 01/10/02 09:00:00 0.37 3.80 4.30 02/10/02 09:00:00 0.0 0.1 0.1 01/10/02 12:00:00 0.73 4.90 5.70 02/10/02 12:00:00 0.0 0.0 0.2 01/10/02 15:00:00 0.58 3.90 4.80 02/10/02 15:00:00 0.0 0.1 0.2 01/10/02 18:00:00 0.56 2.60 3.10 02/10/02 18:00:00 0.0 0.0 0.0 01/10/02 21:00:00 0.62 2.70 2.10 02/10/02 21:00:00 0.0 0.0 0.0 02/10/02 00:00:00 0.52 0.30 2.60 1.60 03/10/02 00:00:00 0.0 0.0 0.0 0.0 02/10/02 03:00:00 0.26 0.20 1.30 03/10/02 03:00:00 0.0 0.0 0.0

The dates and terms for which the forecast is issued are in the first column (in the format dd/mm/yy hh:mm:ss), in the second column - the corresponding values measured by the telemetric stations and in the next 4 columns are inputted the predicted values issued by the model and it is taken into account the starting row of inputted forecast to correspond to time when the forecast is issued. As explained above the values for the temperatures (measured by the telemetric stations) correspond to the terms of the issued forecast, and the values for the rain 3 hourly sums for the hours preceding the issued forecast. The produced tables for each station both rain and temperature shows the number of events when the difference of the kind (predictedR-ObservedR), respectively (predictedT-ObservedT) are in given intervals. These tables are made for each term as well as a table summing all the events, which are in these intervals. It must be mentioned that for the precipitation distribution are excluded those cases in which both the observed and the forecasted sums are 0.

2.3.3. Analysis of the comparison results

• Spatial analysis for rain(12 hours sums) According the results it can be concluded that model gives good results for the stations Radomir and Breznik. These stations are situated in the north part of the catchment area. Still for station Breznik is observed slightly increase of the cases with difference over +5 mm between the predicted and the observed 3 hours sums of the precipitation. This increase is mainly for the predicted values for the period between the 12 hours and 48 hours after the issued forecast.

Normal Distribution for forecast +24 hours(Osogovo) 35 30 25 20 15

Events 10 5 0 < -5 -5 - - -4 - - -3 - - -2 - - -1 - 0 0.01 1.01 2.01 3.01 4.01 > 5 4.01 3.01 2.01 1.01 - 1 - 2 - 3 - 4 - 5 Difference intervals, degC

Normal distribution for forecast +36 hours(Osogovo) 80 70 60 50 40

Events 30 20 10 0 < -5 -5 - - -4 - - -3 - - -2 - - -1 - 0 0.01 1.01 2.01 3.01 4.01 > 5 4.01 3.01 2.01 1.01 - 1 - 2 - 3 - 4 - 5 Diffeerence intervals, degC

3. HBV model schematization and calibration 3.1. Model schematization, sub-basins The Struma river basin was divided on four sub basins. It was made according to the topography of the basin, area of sub-basins and the availability of historical information from the existing hydrological measurement stations. The sub basins are three for the main river and one for the river Djerman. The cross-sections are Rajdavitza, and Marino pole for the Struma river and Dupnitza for Djerman river (fig. 1). Data for 18 precipitation stations and 5 temperature stations were prepared according to required input format for the HBV model as well as the discharge data for 5 stations. The hypsographic are also available. The discharge observation stations to be used for the calibration of HBV, as well as some additional characteristics and sub-basins numbers are given in Table 3, while the meteorological stations are given in table 4.

Table 3. River discharge stations River Village Station Mean eleva- Area Sub-basin No. code tion [m] [km2] Struma Rajdavitza 51700 884 2171 1 Struma Boboshevo 51750 974 4320 Not used Dzermanska Dupnitza 51430 1001 396 3 Struma Kresna 51800 973 6777 2 Strumeshnitza Strumeshnitza 51560 - 1170 Outside inflow Struma Marino pole 51880 899 10243 4

Table 4. Meteorological stations Village Station code Sub-basin No Elevation [m] Precipitation stations (daily totals) Radomir 63415 1 691 Nevestin 62510 2 440 Sapbania 62535 3 746 Dolene 61690 4 720 Boboshev 62500 2 371 Bobovdol 62520 2 640 Breznik 63500 1 749 Dob_pole 61410 2 1209 Dol_selo 62450 2 770 Dolmelna 63440 1 928 Dren 63490 3 750 Gradevo 61490 2 466 Levunovo 61680 4 146 Climatic stations (air temperature, precipitation) Blagoevgrad 61010 2 417 Kustendil 62010 1 521 Sandanski 61100 4 206 Musala 64215 3 2925 Pernik 63010 1 693 Rila 62060 2 504 Station 51800 H [m] Area [km2] % Area 286-300 3.69 0 301-400 102.96 2 401-500 246.06 4 Hipsograph curve Station 51800 501-600 479.25 8 601-700 905.31 15 16 701-800 920.07 15 14 801-900 827.82 14 12 901-1000 705.69 12 10 8 % Area 1001-1100 524.7 9 6 % Area 1101-1200 293.67 5 4 2 1201-1300 209.79 3 0 1301-1400 138.6 2 1401-1500 127.71 2

1501-1600 91.98 2 286-300 401-500 601-700 801-900 1001-1100 1201-1300 1401-1500 1601-1700 1801-1900 2001-2100 2201-2300 2401-2500 2601-2662 1601-1700 74.79 1 H [m] 1701-1800 64.44 1 1801-1900 54.81 1 1901-2000 51.12 1 2001-2100 54.36 1 2101-2200 51.93 1 2201-2300 44.46 1 2301-2400 46.89 1 2401-2500 26.64 0 2501-2600 11.52 0 2601-2662 2.25 0

3.2. Model calibration, first results

Simulations for two sub basins (cross-sections Rajdavitza and Kresna) were made separately. The linking of Rajdavitza and Kresna sub basins was successfully done considering the measured discharge of Djerman river. The same measured discharges (station Mitino, fig. 1) done on Bulgarian territory will be used for Strumeshnitza river which flows from Makedonian territory.

The HBV model was roughly calibrated for the last hydrological station on Bulgarian territory. The achieved model parameters are given on table below. Those parameters will be further used for the separate calibration of all mentioned above sub- basins 1 – 4. After the separate calibration of the sub-basins they will be linked and the whole basin will be calibrated again. Those results are expected in the nearest months. Table 5. HBV optimization results Parameter Value Parameter Value PCORR 0.95 GMELT 4 PCALT 0.1 CFR 0.05 TCALT 0.6 WHC 0.1 RFCF 1.1 SFDISTFO 0.9 SFCF 1.1 SFDISTFI 0.6 FOSFCF 0.6 SCLASS 3 CFMAX 3.2 FC 200 TT -0.8 LP 0.7 DTTM -0.75 BETA 2.0 TTINT 1.44

Fig. 3. Example of the HBV model simulation 4. Simulations with the ISBA surface scheme

The region under consideration takes one-third part of the territory of Bulgaria – 34000 km2. It covers the central part of South Bulgaria, between the chain of Stara Planina Mountain and the borders with Greece and Turkey at the south (Figure 4). The climate is continental to Mediterranean in the valleys depending on the dominant atmospheric circulation. It has pronounced altitude variability as the elevation is going from 50 m up to 2925 m at the pick of Mussala in Rila Mountain.

Figure 4: Map of the Maritza, Arda and Tundja River basins in Bulgaria

4.1. Modelling structure

The distributed hydrologic modelling is able to reproduce the hydrologic events over a large surface, however it needs precise geographic, geomorphic and meteorological data at the earth surface level. The needed relatively static data, as vegetation cover, soil properties can be prepared with a great degree of accuracy and they remain unchanged for long period. In opposite the meteorological data are dynamic and the process of their interpolation for the 2D is quite complicated. This problem is already solved in the NWP models, however they do not take into account a big amount of local data, which are not distributed via GTS. The solution is to use NWP data where the degree of accuracy is good enough compared to the station data, but use station data for the variables that are statistically not well reproduced by the NWP models.

4.1.1. ISBA (Interface Soil Atmosphere Biosphere) Land surface scheme The ISBA surface scheme was developed for the climate, mesoscale and prediction atmospheric models used at Météo-France (http://www.cnrm.meteo.fr/mc2/). It aims to represent the main surface processes in a relatively simple way: it solves one energy budget for the soil and vegetation continuum, and uses the force-restore method (Deardorff, 1978) to compute energy and water transfers in the soil. Evapotranspiration is computed through four components: interception by the foliage, bare soil evaporation, transpiration of the vegetation (with a stress function computed using the method proposed by Jarvis (1976) and sublimation of the snowpack. Two fluxes of water in the soil are computed: a surface runoff (Qr) and drainage (D) (Figure 5)

Figure 5: ISBA land surface scheme coupled with MODCOU macroscale hydrological model.

4.2.2. Distributed regional scale hydrological and hydro-geological model

The macro-scale hydrological model MODCOU was used in several applications (Ledoux et al., 1989; Ambroise et al., 1995; Violette et al., 1997). MODCOU takes into account a surface and the underground layers. The surface routing network is computed from the topography, using a geographical information system (Golaz, 1995). The surface and underground domains are divided into grid cells of size 1, 2, and 4 km. The transfer between two grid cells is estimated from the topography. The surface runoff computed by ISBA is routed to the river network (Figure 5) and then to the gauging stations using isochronous zones with a daily time step. The drainage computed from ISBA contributes to the evolution of the groundwater table, which evolves according to the diffusivity equation. The exchanges of water between the groundwater table and the river are computed according to simple relations (Ledoux, 1980). At the end, the flows from the surface layer and from the groundwater table form the riverflow at the gauging stations.

4.2. Analysis of the project feasibility

• Comparison of point scale temperature data of a synoptic station to the nearest Aladin grid point value: This analysis showed good approximation of the surface air temperature when using valley stations and less good quality of the simulation for the mountain stations. The reason could be the large difference between the average grid mesh altitude and the real elevation at the station location (Figure 6). • Comparison of the point scale temperature data of a climatological station to the nearest Aladin grid point value: Because the climatological stations had not been taken into account by the Aladin model, it was expected to receive poor fitting results at these locations. It was found out that Aladin approximated well the daily minimum but very often underestimated the daily maximum temperature for the mountain stations. • Comparison of the point scale temperature data of a climatological station to the nearest Aladin grid point value, interpolated as a potential temperature: When the Aladin temperature field is re-interpolated into an 8 km grid (that is used by ISBA) the dependency of the potential temperature on the atmospheric pressure could be used to restore the temperature at the known elevation of the 8 km grid. This method gives higher bias and lower correlation coefficients (R2) compared to the simple interpolation (Table 6).

Table 6. Statistical results of the comparison between the measured air temperatures at the location of the synoptic and climatological stations and interpolated Aladin air temperatures of the adjacent grid point. RMS is the root mean square; STDE is the standard error; BIAS is the average bias and R2 is the squared correlation coefficient.

Station SIMPLE INTERPOLATION POTENTIAL TEMPERATURE INTERPOLATION Elevation Station RMS STDE BIAS R2 Station RMS STDE BIAS R2 160 0.30 3.12 0.10 0.84 Plovdiv 0.32 3.31 3.31 0.75 SYNOPTIC

STATIONS 2376 vr.Botev 0.36 3.45 2.37 0.71 vr.Botev 0.37 3.58 3.58 0.77 173 0.28 2.65 -0.62 0.89 Chirpan 0.28 2.82 2.82 0.88 138 0.28 2.61 -0.47 0.90 Elhovo 0.28 2.78 2.78 0.86 1750 Rojen 0.28 2.44 -0.21 0.86 Rojen 0.30 2.56 2.56 0.78 2925 vr.Musala 0.33 2.78 2.56 0.78 vr.Musala 0.35 2.91 2.91 0.78 392 0.32 2.44 -2.06 0.87 Kazanlak 0.30 2.71 2.71 0.82 AVERAGE 0.31 2.79 0.24 0.83 AVERAGE 0.31 2.95 2.95 0.81 CLIMATOLOGI 723 Devin 0.49 2.97 -0.87 0.80 Devin 0.42 3.23 3.23 0.76 STATIONS 743 0.45 2.64 -0.42 0.85 Velingrad 0.44 2.97 2.97 0.82 CAL 556 Panaguirisht 0.35 2.58 2.55 0.87 Panaguirisht 0.37 2.70 2.70 0.79 400 0.51 2.68 -0.85 0.84 Karlovo 0.44 3.30 3.30 0.74 1150 0.46 3.10 -0.43 0.79 Chepelare 0.42 3.08 3.08 0.77 AVERAGE 0.45 2.79 0.00 0.83 AVERAGE 0.42 3.06 3.06 0.79

• Comparison of the measured to the predicted precipitations at 3 h and 24 h step: The squared correlation between the 3h sums of Aladin precipitations and the measured precipitations at the point scale were below 0.35. The best results were achieved when using 24 h sums. As an average estimation, Aladin precipitations were usually higher than the measured values. That depended on the altitude of the grid cell, but also on the surrounding elevation patterns. • Comparison of the total (accumulated for the period 16/09/2002 to 20/02/2003) precipitation field produced by Aladin NWP and the point scale observations’ interpolation: This check showed the larger Aladin precipitation field variability compared to the observations. The model produced higher values over the Stara Planina Mountain that were not observed in reality.

The overall analysis showed that except the precipitation field estimation, which contains the higher error, all other fields could be used without special transformation. To correct the effect of the precipitation field on the water balance the measured precipitations could be used through sequential update of the soil moisture and other ISBA variables with that ones computed by using measured precipitations.

Comparison of Aladin 12 h forecast with data from a synoptic station Rojen Synoptic Station Rojen Forecast 25 20 15 10 5 0 -5 -10 -15 -20 2002/09/162002/09/19 002002/09/23 122002/09/26 002002/09/30 122002/10/03 002002/10/07 122002/10/10 002002/10/14 122002/10/17 002002/10/21 122002/10/24 002002/10/28 122002/10/31 002002/11/04 122002/11/07 002002/11/11 122002/11/14 002002/11/18 122002/11/21 002002/11/25 122002/11/28 002002/12/02 122002/12/05 002002/12/09 122002/12/12 002002/12/16 122002/12/19 002002/12/23 122002/12/26 002002/12/30 122003/01/02 002003/01/06 122003/01/09 002003/01/13 122003/01/16 002003/01/20 122003/01/23 002003/01/27 122003/01/30 002003/02/03 122003/02/06 002003/02/10 12 00

Figure 6 : Comparison between air temperature measured at a synoptic station and from the nearest Aladin grid point

4.3. Adaptation of ISBA-MODCOU coupled system for using ALADIN precipitation fields

As mentioned, the only use of Aladin precipitations was expected to produce largely overestimated runoff, caused by the overestimated precipitation. That is because the initial conditions (soil moisture, soil temperature, snow density and height) had to be corrected after one or two days of surface scheme integration. The values for the re- initialisation were taken from another ISBA integration having a delay of one or two daily steps (Noilhan, 2002), which used measured precipitations instead of Aladin outputs.

As a result the following four simulations were performed:

• (C) Control simulation. 2D fields of measured precipitations were used for the control simulation. All the other input data were taken from Aladin 12 h forecast. The last 48 h were also included as 48 h forecast. • (S0) Simulation using full set of Aladin produced 12 h forecast. The last 48 h were included as 48 h forecast. • (S1) Simulation during which, ISBA prognostic variables were updated with the values taken from the control simulation at 24 h time step. • (S2) Simulation during which, ISBA prognostic variables were updated with the values taken from the control simulation at 48 h time step.

By this way three set of results could be compared to the control simulation Two time steps were tested for the re-initialisation of ISBA prognostic variables: 24 h and 48 h. The Table 7 shows the linear trend squared correlation coefficients (R2) between the measured streamflow, the control and the corrected simulations for two anthropogenized and two natural river streamflows.

Table 7: Statistical results - squared correlation of the comparison between the four performed simulations and the measured daily streamflows. Catchment Measured Control S0 S1 S2 R2 Surface Streamflow Simulation Simulation Simulation Simulation [km2] Maritza 20840 1.0 0.64 0.27 0.43 0.39 (anthropogenized) Tundja Elhovo 5551 1.0 0.36 0 0.10 0.07 (anthropogenized) Chepelarska Bachkovo 825 1.0 0.65 0.09 0.44 0.44 (natural) Arda Vehtino 857 1.0 0.82 0.44 0.77 0.75 (natural)

Daily precipitation at Kirkovo station location compared with Aladin precipitation Station Aladin 100 80 60

mm 40 20 0 1/7/03 2/4/03 1/14/03 1/21/03 1/28/03 2/11/03 9/17/02 9/24/02 10/1/02 10/8/02 11/5/02 12/3/02 10/15/02 10/22/02 10/29/02 11/12/02 11/19/02 11/26/02 12/10/02 12/17/02 12/24/02 12/31/02

Figure 7 : Comparison between daily sum of precipitation at a rain gauge station and the predicted precipitation at the nearest Aladin grid point Measured Streamflow Control S(0) S(1) S(2) 350

300

250

200

m3/sec 150

100

50

0 09/16/02 09/23/02 09/30/02 10/07/02 10/14/02 10/21/02 10/28/02 11/04/02 11/11/02 11/18/02 11/25/02 12/02/02 12/09/02 12/16/02 12/23/02 12/30/02 01/06/03 01/13/03 01/20/03 01/27/03 02/03/03 02/10/03 02/17/03

Figure 8 : Results of daily streamflow simulation and forecast for the Arda River catchment at Vehtino location in the South part of the region. One can see that the S1 and S2 simulations have diminished the effect of a false precipitation event (see Figure 7) occurred on the 30.09.02

5. Bulgarian EFFS WEB page

According to the work plan WEB page of the project was created in http://hydro.meteo.bg. Some of the most important materials and basic implementation steps are published there “http://hydro.meteo.bg/proj/effs/ProgresReportBG.pdf”.