Event Based Flood Mapping
The previous article explained the constituent parts of an InfoWorks RS river model, namely a hydrodynamic model of the river system, incorporating ground models that extend across the catchment, and a hydrological model, representing the impact of wetting events on the river system by modeling run off of rain into the river system. The behavior of the river in terms of levels, rates of flow, and other parameters can then be simulated against various rainfall events. For flood mapping these are usually extreme events. Statistical analysis of historical rainfall patterns is used to determine design rainstorms of specific return periods. These design rainstorms are used as inputs to the model, hydrological and hydraulic analysis is undertaken to determine the resultant flows and levels, and finally the flood mapping results are calculated. By following this process, the model can be used to investigate the likely outcomes of specific design rainfall events. Options can be explored to understand their effect in minimizing any adverse effects, and a catalogue of flood inundation maps created for the differing return period rainfalls to provide an indication of flood risk for planning, insurance and development control purposes. The process may be summarized by the following flow chart.
The rainfall data input to the model must be geographically specific, covering the whole basin. The data used is usually from specific storms or extended rain events. For example, data from a rainfall event of a known return period can be used – say a 1 in 150 rainstorm – and the levels and flow of the river, or the extent of flooding, is calculated by the model. The conclusion from such a simulation could be that a 1 in 150 rainstorm would result in specific levels and flows at specific locations.
The following figure shows an example of an InfoWorks RS flood inundation map for a 1 in 150 year design rain storm.
However, such conclusions need careful examination. The results are based on a very specific rainfall pattern, with a very specific geographic distribution and duration. If the same amount of rain fell with a different geographic distribution, what would be the results? They may be similar, but could well be quite different. Moving rainfall a short distance can affect the amount of run-off into the river because it now falls on ground with different run-off characteristics, or may move it across a watershed into another river or tributary basin. River models are sensitive to the location of the rainfall. The result is that the worst storm for the whole basin will not necessarily be the worst for specific locations. A heavy local storm, not ranking in major storms across the whole basin, may result in higher levels at a specific location that the “more severe” 1 in 150 storm.
Flood Frequency Mapping
To address this complication, many agencies and river authorities have started to consider continuous simulation. Aminal, part of the Flemish Government in Belgium, in association with Wallingford Software, has pioneered an evaluation method that gives a more accurate indication of the flood risks at each specific location in a river system. For simplicity, the approach is named Flood Frequency Mapping and the process is summarized in the flow chart below.
In essence, the approach looks at each node of the model – each specific location on the river – and examines how a large number of storms have affected that particular node in terms of height and flow, and then builds a statistical distribution of the individual results.
Achieving this greater accuracy requires many more model runs than the simpler method, and many more rain events. For example, in the Aminal study of the Yser basin, 79 rain events were simulated, representing the extreme events over 45 years, using a model of 2300 nodes and 102 storage areas. The mechanics of undertaking this would be onerous if InfoWorks RS did not offer a fast and stable simulation engine and did not automate many of the procedures.
The way in which InfoWorks RS supports each step of the Flood Frequency Mapping (FFM) methodology is as follows:
Selection of the rain events to be modeled
The FFM methodology involves selecting a number of the most critical rain events - perhaps 50 or 100 events - along the course of the river over the chosen time span. Events are selected for input into the model if their magnitude caused river levels to exceed designated thresholds at each specific location being examined. This selection is undertaken by InfoWorks RS, using two inbuilt analysis tools.
Running the selected events through the model
The next step of the methodology is to run each of these selected rain events through the model. InfoWorks RS handles this as a single task, using a ‘Multirun’ tool that executes all the selected events. On completion of the runs and the production of results for every event, the frequency analysis on every node in the network can be undertaken.
Statistical analysis of the results
The results for every node, a node being a particular river section or storage areas for example, can be examined for every simulation run in the Multirun. The Frequency Analysis tool in InfoWorks RS produces frequency (probability) curves of the maximum stages (water depths) or flows, for every node. The continuous frequency curve is fitted onto the results points curve by InfoWorks, using a pre-defined statistical function.
Flood Frequency Mapping
For each return period the frequency results can immediately be represented. The results can be reproduced in the form of flood frequency maps, tables and figures for every designated location. The following figure shows and example of a flood frequency mapping for the 1 in 150 year water levels rather than 1 in 150 rainfall.
For most situations, creating an accurate model of the river basin and simulating the flood effects of rain patterns is a great step forward in risk analysis, opening up the way to flood mitigation and disaster planning. However, when this level of analysis has been completed and the model proven, Flood Frequency Mapping provides, at the cost of more detailed work, very accurate probabilities of the occurrence of severe events at every location within the basin.