| dc.description.abstract | Global warming and climate change lead to more frequent and extreme weather events. These include sudden and intense rainfall and rising temperatures which cause flash floods. In steep terrains, flash floods with high-flow velocities lead to erosion and sedimentation with potential disastrous changes of flood path. Flash floods caused by heavy rainfall with snowmelt contribution due to sudden rise in temperature (rain-on-snow events) have become common in autumn and winter in snow-covered Nordic catchments. These events have caused widespread damage, closure of roads and bridges and landslides leading to evacuations in affected areas. Therefore, the analysis of such flood types becomes more important in terms of inundation area, water depths and flow-velocities to identify critical locations in a catchment.
Hydrological and hydraulic models are usually used to simulate flash floods. But most of the traditional hydrological models only give output as a hydrograph but do not represent the consequences such as velocity, water depth, sheer stress etc. at any point or region in the catchment. So, the flood hydrograph from a traditional hydrologic model must be combined with a hydraulic model for downstream consequences. In the traditional method of manual coupling, the output from the hydrologic model is used as input and set as input boundary condition in the hydraulic models. This method of manual coupling requires the separate calibration of two models which makes it a time-consuming process. Sometimes due to many tributaries, more than one boundary condition is required and it is difficult to decide where to set the input boundary conditions in the hydraulic model. In addition to this, there is always some residual flow along the river from the catchment or the water from small tributaries, which is difficult to estimate and add in the hydraulicmodel calculations. In small and steep catchments, the inflow contribution from every section of the water course can be important to determine where critical conditions may arise.
To overcome these challenges and the hassle of manual coupling of the two models, the direct rainfall method (DRM) also known as rain-on-grid (RoG) technique was tested in this research work. Primarily, TELEMAC-2D, a Hydrodynamic Rainfall-Runoff (HDRR) model, with Curve Number (CN) infiltration method, was used for this purpose in a study site of 10.5 km2 steep catchment located in western Norway. Spatially distributed precipitation data with a resolution of 1km by 1km was used as input instead of point precipitation data to reduce the uncertainties related to precipitation distribution over the catchment.
Since TELEMAC-2D is an open source toolbox, it was possible to make changes in the source code ourselves and implement spatially distributed precipitation as input. Since TELEMAC-2D doesn’t have a snow routine, initially only those peak flow events were simulated which were induced by rainfall without any contribution from snowmelt. Seven such events were simulated and a sensitivity analysis was conducted for the parameters such as the CN values, size of mesh elements, roughness and antecedent moisture conditions (AMC) in the catchment. In addition, a 200-year design flood was simulated to show the potential damages in the catchment. The study explored the benefits and limitations of the approach through a comprehensive description of model construction, calibration, and sensitivity analysis. The results showed that calibrated models can satisfactorily reproduce peak flows and produce relevant information about water velocities and inundation. Since, floods can reach even more extreme levels when snowmelt combines with the surface runoff generated by rainfall events. When rain falls on an existing snowpack in addition to the sudden increase in temperature, it is known as a rainon-snow (RoS) event. These events can result into destructive flash floods due to the sudden melting of snow combined with the extreme rainfall.
Hence, in the next part of the thesis work, the contribution and effect of snowmelt in flash floods were analysed. The hydrological model HBV was used to find out the portions of snow and rain from the raw precipitation data and to calculate the snowmelt. The sum of snowmelt and rain calculated from HBV, which eventually contributes to flash floods, was used as the input precipitation in TELEMAC-2D for HDRR modelling. The results showed the importance of including snowmelt for distributed runoff generation and how the combination of hydrological and hydraulic models allows to extract flow hydrographs anywhere in the catchment. It is also possible to extract the flow velocities and water depth at each time-step showing the critical points in the catchment in terms of flooding. The RoG technique works particularly good for single peak events, but not for floods with longduration sustained flow and which are generated by multiple rainfall storms. The results indicated a need for implementation of time-varying CN values or another infiltration model with a time-varying infiltration rate for such multi-storm floods.
Therefore, another HDRR model HEC-RAS 2D with the Green-Ampt Redistribution (GAR) infiltration method was tested and compared with TELEMAC-2D for its RoG technique. CN method was applied in both the models to simulate two single storm events up to 20 hours duration. NSE and R2 for the models ranged from 0.70 to 0.90 and from 0.93 to 0.95. Moreover, the two models were compared for the calibration process, computational time, mesh size and shape, model availability in general, as well as for the results including inundated areas, water depths and velocity of water after a flood event. In addition, The GAR method was applied in HEC-RAS 2D for a multi-peak flood event with sustained flow between the peaks, but the results showed that even this method was unable to reproduce all the peaks of the flood event. Therefore a sensitivity analysis of the GAR parameters was done to understand why GAR method could not reproduce the multi-storm flood. The sensitivity analysis showed that the results are not very sensitive to the two GAR parameters which are responsible to influence the flow of the later peaks.
Neither of the HDRR models could reproduce such multi-storm long duration floods because of the fact that both the HDRR models permanently lose the infiltrated water out of the model domain which usually contribute as the return flow to the river which is mainly the reason for the sustained flow between the multiple storms and for a gradual recession limb of a flow hydrograph. However, neither of the models incorporate a return flow algorithm. Hence, the HDRR models should only be used if it is sure that the infiltrated water goes to the deep base flow where there is no chance of subsurface return flow, or they should be used only for shortduration single storm floods.
Potential follow up to this research work can be to implement a subsurface flow module or the contribution of return flow in the fully integrated hydrologic- hydrodynamic RoG models. This enhancement would enable these models to effectively handle both short and long duration floods. Moreover, a snow routine can also be implemented in the HDRR models which eliminates the need of a separate model to calculate snow storage and snowmelt in the catchment. | en_US |
