Assessment of causal mechanisms on flood conveyance on the Tisza River, Hungary using one-dimensional retro- and scenario-modeling.

摘要

During the past decade, a series of record flood stages have occurred along the Tisza River, resulting in extensive damage and displacing the local floodplain population. Previous research on the Tisza River in Hungary showed increases in flood stages for fixed discharges (above bankfull). These results suggest that a loss of conveyance has taken place on the Tisza River, contributing to recent record flood levels.;In order to assess the potential causes of flood conveyance losses, new hydrological and geospatial data were obtained from the Hungarian Federal Hydrological Authority (VITUKI) in order to develop hydrodynamic models for two reference conditions, 1979 and 1999, along two study reaches on the Middle Tisza River. In addition to these models of actual conditions at each time step (i.e., "retro-models"), four "scenario models" were developed in order to test the individual impacts of each of the hypothesized causal mechanisms: 1) channel geometry, 2) levees, 3) floodplain roughness (land cover), and (4) channel roughness. Comparison of the models for actual conditions ("retro-model") and the scenario models allows for quantitative assessments of the impacts of each of these parameters on flood conveyance. In addition to these four scenario models, an additional land cover scenario-model was created to further assess the impact of floodplain roughness on stage.;Assessment of the two retro-models showed changes in flood stage along both study reaches over the 20-year time step. Along the upper study reach, changes in stage ranged between 0.64 and 1.64 m. On the lower study reach, changes in stage ranged from 0 to 1.75 m. These changes in stage were larger on average than changes in stage previously detected by Venzcel, 2008.;Contributions from the channel geometry scenario model were significant on the upper study reach where channel geometry decreased flood stage by up to 0.58 m at Kiskore over the 20-year time step. This was attributed to the dam that was constructed immediately upstream from the Kiskore gauge on the upper study reach. Construction of a dam would cause incision downstream, and therefore, a decrease in flood stage over time due to channel geometry. On the lower study reach, change in stage due to channel geometry was negligible.;Contributions from the levee scenario model were negligible on both the upper and lower study reach with a maximum change in stage over the 20-year time step at only 0.11 m. However, during the time step analyzed, no significant change in levee configuration occurred.;Changes in stage due to changes in land cover were also negligible along both study reaches. An additional land cover scenario model was created to assess the impact of land cover change on modern flood levels. This scenario model set Manning's n values for floodplain roughness to 0.04, which approximates roughness for pasture and cropland. This scenario model showed that flood stages will decrease by 0.34-0.40 m, but only if the entire floodplain is converted from woodland to pasture/cropland.;Channel roughness accounted for the majority of change in stage during the 20-year time step. Along the upper study reach, channel roughness accounted for an average of 1.51 m of change over the 20-year time step. On the lower study reach, channel roughness accounted for 1.03 m of change. Change in the stage/discharge relationship is the best explanation for changes in stage due to channel roughness. Most likely, the channel did not experience such a large change in roughness during the 20-year time step. Instead, the two flood years used in this analysis were different "types" of floods (with differing degrees of hysteresis), and the stage/discharge relationships could not be easily compared. Comparing these two floods may have exaggerated the total change in stage between the two retro-models. Backwater effects are a likely explanation for the change in hysteresis between the two flood years.