Hydrologic Reconstruction of the Bjelica River Flood of March 2016 using Radar Precipitation Measurements

Nikola Zlatanović1, Vanja Damjanović1, Stevan Prohaska1


1 Institute for the Development of Water Resources "Jaroslav Černi", Belgrade, Serbia; E-mail: This e-mail address is being protected from spambots. You need JavaScript enabled to view it



On March 7th 2016, the town of Lučani, together with numerous villages and towns along the Bjelica River Valley, was severely flooded. In order to reconstruct the flood event in the entire river basin, a semi-distributed hydrologic model was used, with radar precipitation measurements as input. The radar precipitation data were compared to traditionally measured data, and the hydrologic model was validated with the observed hydrograph at the Guča gauging station. Downstream, at the town of Lučani, a hydraulic model of the river was developed to validate the downstream hydrograph. The results exhibit great agreement, showing that radar precipitation can be successfully implemented in rainfall-runoff modelling, although care should be taken with certain issues such as radar shadowing.



The Bjelica River, located in the Dragačevo region in western Serbia, is a right tributary of the Zapadna Morava River. The source of the river emerges at an altitude of 930 m.a.s.l. on the northern slopes of Čemerno mountain, and flows northwestward through the Lučani municipality for 41 km until it enters the Zapadna Morava River in the village of Dljin, near the town of Lučani. Most of the tributaries of the Bjelica River are right-bank tributaries, flowing down from Jelica mountain.

The location of the town of Lučani in the Bjelica River Valley, coupled with the torrential nature of the river and the backwaters of the Zapadna Morava River, result in frequent flooding and extensive flood-related damages. One such flood occurred on March 7th, 2016, when the entire Bjelica Valley was flooded, severely affecting villages and towns, including Guča and Lučani. Several hundred hectares of agricultural land were submerged, about 80 houses were damaged in the villages of Turica, Puhovo, Đerađ, Dljin, Lis and Viča, as well as residential areas in the town of Guča. A total of 52 persons were evacuated and over 120 auxiliary facilities and outbuildings were flooded (Flood in Lučani, 2017). According to preliminary estimates, the March 2016 flood caused at least several hundred thousand Euros in damages.


Figure 1: Location of the municipality of Lučani (shown in red) within Serbia.


In this paper, the authors developed a rainfall-runoff hydrologic model to perform a hydrologic reconstruction of the 2016 flood using available radar precipitation measurements as input, to identify the formation and propagation of the flood wave and to explore advantages and shortcomings of such applications in this example.


Methodology and Datasets

The rainfall-runoff hydrologic model was developed using the HEC-HMS (Hydrologic Engineering Center – Hydrologic Modeling System) software package (USACE, 2000). The model structure was created based on physical characteristics of the basin and river system, while the model parameters were estimated based on available topographic, land use and soil data.

The Bjelica River Basin was delineated into a total of 18 subbasins, based on topographic data. The SRTM (Shuttle Radar Topography Mission) digital elevation model (DEM) (Rodriguez et al., 2005) was used for catchment and subcatchment delineation, as well as for estimation of topographic-based model parameters (flow length, channel slopes, etc.). Figure 2 shows the delineated subbasins with the digital terrain model as well as the model structure.


Figure 2: Bjelica River Basin: delineated subbasins with digital terrain model (left) and hydrologic model structure (right).


Direct runoff of excess rainfall was simulated using the Clark Unit Hydrograph Model (Clark, 1945). The parameters of the Clark Unit Hydrograph for each subbasin were estimated using the digital terrain model to obtain the longest flow path and slope for the modified Kirpich formula (Kirpich, 1940) proposed by Zelenhasic (1970). Initial values for the storage coefficient R were estimated as a linear relation to Tc, as proposed by Russel et al. (1979).

The SCS-CN loss method (SCS, 1985) was used in total runoff estimation to specify soil infiltration rates, or losses. The method uses an integration of landuse and soil data to determine Curve Number (CN) values of the watershed. Land use data for the basin were obtained from the CORINE (Coordination of Information on the Environment) Land Cover dataset (EAA, 2007), differentiating between a total of 44 land cover classes grouped in 3 levels - artificial surfaces, agricultural areas, and forest and semi-natural areas (Figure 3). Soil data were obtained from soil maps on a scale of 1:50.000, developed by the Soil Institute of Serbia, and soil types were classified into hydrologic soil groups according to their physical properties. Each subbasin was assigned a Curve Number (CN) based on the land use data and hydrologic soil groups.

Flow routing from upstream to downstream through river reaches was estimated using the Muskingum-Cunge model (Miller and Cunge, 1975)(Ponce, 1986). The parameters of the Muskingum-Cunge model, including reach length, average slope, Manning's roughness coefficient, and cross section shape and size, were estimated based on the digital terrain model as well as field visits to the sites.

Radar images, showing cloud structure, orientation and velocity, and rainfall intensity rates, were provided by the Republic Hydrometeorological Service of Serbia (RHMSS), which are uploaded at regular 15 minute intervals on the official RHMSS website. The radar images are provided as composite images, merging data from available radars. In the case of the 2016 Bjelica flood, the Jastrebac and Fruška Gora radar centers were operational, as can be seen on one radar image from March 6th 2016, shown in Figure 4.


Figure 3: CORINE 2015 Land Cover (left) and soil map (right) of the Bjelica River Basin.


Figure 4: Example radar image during the 2016 Bjelica flood, showing the storm cloud (source: Republic Hydrometeorological Service of Serbia).


The rain rates were calculated from the composite radar mosaic data via the standard Z-R formula (Marshall and Palmer, 1948), which were then used for precipitation accumulation calculations. The reflectivity parameter Z (in mm6m−3), measured by the radar, is converted to precipitation intensity R (in mm/h) using a pre-selected Z-R equation of the type:

Z = A · Rb          (1)

where A and b are empirical factors describing the shape and size distribution of the hydrometeors.

Parameters of the Marshall-Palmer formula were calibrated based on observed rain rates at three main meteorological stations, Valjevo, Požega and Kraljevo, for the duration of the event period in March 2016. First, series of observed reflectivity values were extracted at the locations of the meteorological stations (Figure 5).


Figure 5: Extracted series of radar reflectivity for the main meteorological stations, Valjevo, Kraljevo and Požega, during the Bjelica flood event of 2016.


The parameters A and b were calibrated so that the rain rates calculated from the obtained series using the Marshall-Palmer formula best fit the observed rain rates. Figure 6 shows this fit for the main meteorological station Požega, which is nearest to the Bjelica River Basin.


Figure 6: Calculated rainfall intensity based on observed radar reflectivity (shown in blue) compared to observed rainfall intensity at the main meteorological station Požega during the 2016 Bjelica river flood event.


Radar reflectivity was then converted to rainfall intensities using the calibrated parameters. Each subbasin of the hydrological model was assigned a single rainfall time series based on the calculated spatial rainfall intensities, by applying spatial averaging for each subbasin.

The calculated hydrograph was validated at the Guča gauging station, where the flood hydrograph was observed at hourly intervals. After successful validation, the flood hydrograph at Lučani was calculated. Flood stage at Lučani was estimated using a stage-discharge rating curve, and compared to results of the hydraulic reconstruction of the flood in Lučani.

The hydraulic reconstruction of the Bjelica flood in Lučani was performed on the most downstream river reach, from the mouth of the Bjelica River where it enters the Zapadna Morava River, and upstream for 4.5 km up to the bridge in the village of Krstac. The 3D geometric model of the Bjelica River was created based on a total of 33 surveyed cross sections encompassing the river channel and inundation, averaging a 500-1000 m cross-section.

The roughness of the river channel was estimated based on the granulometric composition of the river bed as inspected in field visits. The roughness coefficient of the inundation was estimated based on the land cover and other conditions (field visits and orthophoto imagery). Along the Bjelica River, the inundation/inundated areas that serve as detention basins were identified, and these areas were excluded from the active river flow cross section during flooding.

The downstream boundary condition for hydraulic modelling of the Bjelica River was the highest recorded stage on the Zapadna Morava River at the confluence of the Bjelica River during the 2017 flood, which was 300.25 meters above sea level.



The results of the hydrologic modelling are shown visually as hydrographs. At the Guča hydrologic station, the computed hydrograph is compared to the observed hydrograph (official data from RHMSS). As shown in Figure 7, the computed hydrograph shows good agreement with the observed values, especially peak discharge values. This successful comparison served as validation of the hydrologic model.


Figure 7: Calculated hydrograph (blue line) compared to observed discharge values (red dots) at the Guča gauging station on the Bjelica River; precipitation (shown in green) is spatially averaged for the upstream basin.


Near the town of Lučani, the resulting computed hydrograph was compared to results of the hydraulic reconstruction of the flood event. The hydraulic reconstruction at Lučani yielded a maximum calculated discharge of 230 m3/s, corresponding to the value associated with the 20-year flood, or an exceedance probability of 5%, which equals Q5%=240 m3/s. As seen on Figure 8, the results are in agreement, with less than 3% difference in the peak discharge between the hydrograph (hydrologic model) and maximum discharge (hydraulic model).


Figure 8: Calculated hydrograph of the Bjelica River in Lučani (blue line) and peak discharge calculated by hydraulic analysis (red dashed line).



While the results of the hydrologic reconstruction of the March 2016 Bjelica River flood using radar precipitations provided by the RHMSS were quite satisfactory beyond expectation, some issues were raised when calculating the total sum of rainfall during the flood event. Figure 9 shows the spatial distribution of total precipitation during the event, as observed by radar. What can immediately be seen are very strong radar shadows - regions shielded from radar illumination by an intervening reflecting or absorbing medium such as a hill or mountain. In this case, the radar is located near Jastrebac mountain and its radar waves are obstructed by the surrounding mountain peaks of Željin, Kopaonik and Jastrebac. Incidentally, the Bjelica River Basin is not affected by these radar shadows, as can be seen in Figure 9.


Figure 9: Total event precipitation (6.3.-8.3.2016.) as observed by radar; note the mountain peaks of Željin, Kopaonik and Jastrebac obstructing the radar and resulting in strong radar shadow



As shown in the results, radar observations of precipitation can be a very useful input for rainfall-runoff modelling of the Bjelica River Basin. Applications of such hydrologic and/or hydraulic modelling are numerous, including the development of flood early warning systems based on radar measurements and an improved spatial distribution of rainfall and consequently runoff, due to a much higher resolution of radar compared to traditional point rain gauges. However, for a more reliable application of such data, the reconstruction of flood events, as described in this paper for the March 2016 flood, should be performed in a series of events to properly calibrate the hydrologic and hydraulic model and reduce parameter uncertainty.



Clark, C.O. (1945), Storage and the unit hydrograph, Transactions, ASCE, 110, 1419-1446.

European Environment Agency (EEA) (2007), CLC2006 Technical Guidelines - EEA Technical Report No 17/2007

Flood in Lučani municipality (2017), Lučani municipality, Retrieved from www.lucani.rs (in Serbian)

Kirpich, Z.P. (1940),. Time of concentration of small watersheds. J. of Civil Engineering 10(6). ASCE. New York, NY. pp. 362.

Marshall, J. S. and Palmer, W. M. (1948), The Distribution of raindrops with size, J. Meteorol., 5, 165–166,

Marshall, J.S., R.C., Langille, and W. McK. Palmer (1947), Measurement of rainfall by radar. J. Meteor., 4, 186-192

Miller, W.A., and Cunge, J.A. (1975), Simplified equations of unsteady flow, K.Mahmood and V.Yevjevich eds, Unsteady flow in open channels, Vol. I, Water Resources Publications, Ft. Collins, CO.

Ponce, V.M. (1986), Diffusion wave modeling of catchment dynamics, Journal of the Hydraulics Division, ASCE, 112(8), 716-727

Rodriguez, E. et al., (2005), An assessment of the SRTM topographic products. (Technical Report JPL D-31639). Jet Propulsion Laboratory, Pasadena, California.

Russell S., Kenning B. and Sunnell G. (1979), Estimating Design Flows for Urban Drainage, Journal Hydraulics Division, American Society of Civil Engineers, Vol. 105, pp 43-52.

Soil Conservation Service (1985), National engineering handbook, Supplement A, Sect. 4, Chapter 10. US Department of Agriculture (USDA), Washington, D.C.

U. S. Army Corps of Engineers (USACE) (2000), Hydrologic Modeling System: Technical Reference Manual. Davis, CA: U.S. Army Corps of Engineers, Hydrologic Engineering Center.

Zelenhasić E. (1970), Theoretical probability distributions for flood peaks. Hydrol. Paper 42, Colorado State University, Fort Collins, CO.