Exploitation of Documented Historical Floods for Achieving Better Flood Defense

2.1. Overview of Historical Floods in the Catchment of the Tisza River

The basic parameters of a flood wave are its volume, peak discharge, and duration. Actually, the hydrograph of the flood wave provides the most information. The peak water level caused by the flood wave—depending on the location and time of interest—is rather consequence than basic parameter of the flood. Table 1 shows the peak water levels of the observed historical floods at four river stations along the Tisza River together with the corresponding intermittent periods.

The 1885 flood is of exceptional importance, not only for its volume, but for the fact it was the last extreme flood before regulation works along the Tisza River commenced. Meander cutoffs and restriction of the river by levees increased the peaks of successive floods. The 1879 flood destroyed more than 93% of buildings in Szeged and claimed 151 lives. The flood peak in 1888 was the absolute maximum until 1919, while it is still not overpassed at Dombrad. Following the accomplishment of the regulation works, the 1895 flood—even though moderate compared to the earlier ones—caused significant damage. The 1919 flood was distinctive in sense of threatening public security all along the river. Following the 1932 flood the practice of watercourse-restricting measures in flood control was abandoned. The flood in 1940 was associated with ice. In 1941 three floods occurred, one in January-February, the second in March-April, and the third in September-October. More moderate floods were characteristics of 1942. The following significant flood happened in 1964. The biggest flood ever—in terms of both peak discharge and duration—happened in 1970. It lasted 125 days, 65 settlements have been evacuated, and 43000 people have been engaged in flood defense. The following flood in 1979 produced peak level at Szolnok just 5 cm below the earlier maximum. In 1991 and in 2000 the earlier highest level (detected in 1970) was exceeded by 1.5 m, an increment not experienced for about a century. Proceeded by an early spring flood significant flood happened in fall 1998. The flood in 1999 was triggered by snowmelt and by the changes in the river bed during the last three decades. This flood was extraordinary in terms of peak level, velocity, and duration, exceeding the corresponding values of the 1970 flood. Leaking of dikes in 1100 locations was discovered (and treated), intervention in cases of 127 boils was needed, and creeping of embankment face in 26 cases has been registered. 2001 was not easier at all. 2005 and 2006 produced a flood lasting through the winter to the spring, supported by snowmelt and intensive rainfall in April. The 2010 flood was triggered by extraordinary rains, bringing six times more precipitation than average to the watershed.

Not only are the main parameters of the flood events informative; looking at the details of their genesis is even more instructive. In case of the Tisza River, in general, floods occur mostly at spring in the period of March–May. Floods are least probable in September. In addition to the extreme meteorological conditions, excessive floods are triggered by late snowmelts accompanied with heavy spring rainfalls, or by coincidence of tributary peak discharges, and by restricting the flood plain by embankments.

2.2. The Modeled Stream Network of the Tisza River

2.3. The HEC-RAS Model

Computer aided analysis of well-documented, reliable, historical flood data is the most instructive and straightforward way of learning about floods and predicting the most likely consequences [6]. The River Analysis System software—known by acronym as HEC-RAS—developed in the Hydrologic Engineering Center by the US Army Corps of Engineers has been adopted for this study by the authors [10, 11]. The HEC-RAS software is capable of simulating 1D steady and unsteady flow in a system of natural and constructed channels, producing as a result the corresponding water surface profiles and the related data.

The general principles of modeling unsteady flow in systems of open channels in the HEC-RAS environment are given in [3, 5, 12, 13]. Unsteady flow routing using HEC-RAS is provided by numerical solution of the continuity and momentum equations. The derivation of the governing equations is presented in [10] by Liggett. The most successful and accepted procedure for solving the one-dimensional unsteady flow equations is the four-point implicit scheme, also known as the box scheme. The software package can handle hydraulic structures like bridges, barrages, culverts, overflow weirs, floodgates, bottom stages, bottom sills, side overflows and gates, static reservoirs, pump stations, and water intakes. Example of application in a complex flood control project is given by [7]. Extended description of the software is given in [11].

2.4. Calibration and Verification of the Model

Calibration is feasible by adjusting the global parameters of the model—in most of the cases Manning’s coefficient of friction, eventually the loss coefficients of expansion/contraction [2, 8, 9]—providing successful reproduction of a well-documented past real flood event. Initial values of friction coefficient were estimated by digital aerial orthophotography and in situ surveys, while the final values—separately for the main channel and for the flood plain—were determined by calibration using historical flow data. Calibration is a straightforward procedure in cases of rather frequent floods. However, detailed data for extreme floods are seldom available, making them valued.

Flow measurements during flood recession in 1998 were carried out from bridges and boats. Comparison of the 1998 measurement with an earlier one in 1979 produced results shown in Table 2.

The data show almost unchanged characteristics in case of the main channel; however, flow conditions in the flood plain are significantly deteriorated during the corresponding two decades causing reduction in flow capacity up to 300 m³/s.

Successive measurements in 1998, 1999, 2000, and 2001 produced results shown in Figures 2 and 3 for the main bed and for the flood plain, respectively. The graphs reveal change in friction depending on the time of observation and on the actual water level. Variation of friction coefficient from 0.026 to 0.032 is observed in the main bed and from 0.025 to 0.048 in the flood plain. Increase in friction is due to the development of plant cover in the watercourse. Passing a flood wave often triggers decrease in friction as cutting a passage through an upsilted river section cleans the waterway.

Calibration of the Tisza River model was achieved making use of almost 50 time series comprising of hourly detected water levels at standard measuring posts, and water level readings of dam keepers. In the calibration process default values, 0.3 and 0.1, have been adopted for the expansion and contraction coefficients, respectively. Manning’s coefficients for the main channel and for the flood plain were determined separately in each cross section. Figures 2 and 3 demonstrate an attempt to establish water level dependant Manning’s coefficient; however, it could be achieved for short river sections only.

The blue line in Figure 4 represents the calculated peak water levels, while the red dots are the observed maximum water levels corresponding to the 2006 flood. The maximum difference between the calculated and observed water levels in the river section between Tiszabecs and Titel was 5 cm, which may be considered as very good agreement.

3.1. Flood Plain Maintenance, Detention Storages

The results are shown in Figure 5, where Δz (cm) denotes water level difference due to a particular intervention, compared to the peak water levels expected for the current condition of the river bed; x is river station (km). The light green line represents case (a), the blue line corresponds to case (b), and the dark green line shows the joint effect of both interventions, case (c). The increase of peak water levels in the downstream section (x < 220 km) caused by flood plain maintenance is successfully compensated by the effect of detention ponds. With both measures applied, the maximum water levels could be reduced up to 160 cm. Further improvement in terms of maximum water levels might be achieved using more sophisticated techniques in activation of the eight gate-controlled detention ponds [14]. Calculation results are similar for the 2006 flood as well.

3.2. Application of the 1888 Flood to the Current Condition of the Watercourse

The 1888 flood was so severe that the corresponding maximum water level detected at Dombrad is still not overpassed. In addition, it has occurred in case of watercourse unrestricted by dikes. Even more, extreme water levels over 800 cm lasted more than 14 days in the Vasarosnameny region. For the sake of comparison it is interesting to note that none of the recent floods (1998, 2001) exhibited peak water levels lasting longer than 3.5 days.

This simulation is meant to check if the 1888 flood could pass nowadays without causing trouble. The observed stage hydrograph of the 1888 flood has been set as upstream boundary condition. Stage hydrograph of the 2000 flood has been applied to the outlets of the tributaries and to the most downstream cross section of the Tisza River. The results of simulation compared to the consequences of the 2000 flood are shown in Figure 6. It clearly demonstrates that if the 1888 flood happened nowadays, it would have caused up to 80 cm higher peak levels than the 2000 flood did. Between Tiszalok and Tiszaug peak water levels would over pass the BFE up to 180 cm and they would last over 25 days! It is interesting to notice that, with flood plain maintenance carried out and all planned detention storages accomplished, most of the Hungarian section of the Tisza River could pass the 1888 extreme historical flood.