Stormwater flood hazard assessments in urban areas comprising the identification of overland flood paths and flood levels form an important part of territorial authorities responsibilities. In the past such data have been compiled on “best available” data such as field observations and resident reports. More recently 1D stormwater network models such as MIKE URBAN (formerly MOUSE) or 1D-2D (MIKE FLOOD URBAN) which include descriptions of both the piped and above ground flow network (ie primary channels and secondary overland flow paths) have provided a more robust basis for such assessments. This has been aided by the availability of high resolution and accurate land level data obtained from airborne laser surveys (LiDAR). Nevertheless the definition of overland flow paths remains problematic in a pure 1D model due to the need to pre-define the flow path in the 1D network. With roads, buildings, culverts, fences etc all being common features of the urban landscape, the construction of the 1D model which is capable of accurately defining the surface flood hazard needs to be undertaken with a lot of attention to detail, which may not always be economic. Similarly, MIKE FLOOD URBAN models where the 1D pipe network is linked to a 2D surface description, provide a good description of the flow processes but require time to develop and test.
DHI have trialled a number of alternative approaches for assessing flood extents for low probability flows where the sub-surface drainage network is essentially ignored. These so called rapid flood hazard mapping (RFHM) assessments can provide a cost effective way in which to quickly and efficiently map the worst case flood extent and identify the major overland flow paths. The sub-surface drainage system is assumed to be at capacity in the assessment, which allows for a rapid assessment of the overland flow paths to be undertaken by making direct use of the LiDAR data in a MIKE 21 2D hydrodynamic model. By introducing flows onto the LiDAR surface, the MIKE 21 model is able to simulate the accumulation and routing of the rainfall runoff, simultaneously generating flood extents, depths and velocities for the entire catchment.
Two main approaches have been developed and applied in the Auckland region. The first approach, developed with NSCC, utilizes the TP108 rainfall runoff model to generate a series of runoff hydrographs for the 100 year rainfall event for sub-catchments up to 10ha in area. The generated hydrographs are introduced as “source points” on the MIKE 21 grid, which is then routed by the hydrodynamic model through the catchment. A very fine (1m) grid has been used in the NSCC modeling so that fine features such as roads and rapid slope changes can be resolved in the model. Smaller catchment thresholds may also be applied. For example DHI have tested and compared a 10ha threshold to a 2ha threshold. The latter provides a greater coverage of the flood extent but does not significantly affect flood levels in the parts of the catchment common to both approaches.
An alternative method is to apply the net rainfall (rainfall minus initial and continuing losses) directly to the model grid. In this method all routing is undertaken by the MIKE 21 model but the rainfall must first be processed to take into account the effects of local storage losses and surface imperviousness. This method has been successfully applied by DHI in Australia.
Decisions need to be made on how to handle surface features such as cross drainage structures (bridge and culverts) and large buildings. Drainage structures may either be “cut out” of the model grid, which assumes minimal headlosses, or at the other extreme “blocked out”, which assumes full blockage so that flows are forced over the road crest. Similarly buildings may be filtered from the model grid (“bare earth”) or alternatively if building outlines are available as part of the LiDAR survey these may be extruded back into the model grid, which assumes complete loss of conveyance and storage due to the buildings. An example of this approach is shown in the figure below.
View of 100 year flood extent of Wairau Park in Auckland’s North Shore computed by MIKE 21
Both methods have been compared to more detailed flood mapping studies and despite the coarseness of the model assumptions the results have compared very favourably. In the case of the NSCC modelling, the 2D RFHM modelling has highlighted flood prone areas that were not previously identified in more detailed 1D modelling. The rainfall on grid approach has similarly been compared to detailed 1D-2D modelling in Australia and maximum water levels were within 200mm of the more detailed study.
The advantage of RFHM is the speed at which the assessments can be carried out, which typically is many times faster than traditional detailed model assessments, with consequent cost savings. DHI have so far undertaken assessments for 24 catchments for NSCC, ranging in size from 170ha to 1300ha. It is important to note that the resulting product does not necessarily provide a definitive flood hazard extent. However it can very quickly provide information of worst case flood extent and possible overland flow paths which could be used to prioritise areas where additional more detailed analyses (eg MIKE FLOOD URBAN) are required.
(First published in the NZWWA journal September 2008)