Melbourne Water Managing Director Rob Skinner today reported that he was satisfied that the proposed flood protection works at the Flemington Racecourse could proceed without increasing the flood risk to surrounding areas, including the Maribyrnong Village.
Mr Skinner said his decision was based on an assessment of the processes undertaken by Melbourne Water.
“Some members of the local community raised concerns about the works with me and proposed alternative methodologies for assessing the flood risk. In response, I requested that a comprehensive re-examination be undertaken of the modelling we had commissioned. This review addressed each of the concerns raised by some local residents.
“In each case it was determined that the conclusions drawn would not have been materially different had other modelling methods been used. As a consequence, I can confirm that Melbourne Water’s advice to the Minister for Planning about the flood risks of the wall is well founded,” Mr Skinner said.
Part of the review involved an independent review of the hydraulic modelling by Associate Professor Bob Keller of Monash University, a renowned expert in the theory and practice of flood level modelling, who also has 20 years experience in analysing flows in the Maribyrnong River.
As the Floodplain Management Authority for the metropolitan area, Melbourne Water is responsible for the assessment of any proposal involving development on floodplains to ensure that they do not increase the risk of flooding.
Mr Skinner said that works on other sites downstream of the proposed VRC wall are required prior to the walls construction to ensure that future flood levels are not increased.
A detailed response to the issues regarding modelling raised by the Maribyrnong Floodplain Committee and consultants is attached.
| Issue |
Description |
Advice |
| 1 |
Total Energy Line (TEL) versus Water Surface Elevation (WSE) when calibrating to debris lines |
- When calibrating models, it is a case by case decision on whether to calibrate to WSE or TEL (This is a view also held by the US Army Corps of Engineers).
- Need to weigh up what information is the model being calibrated against (are debris levels representative of TEL or WSE?) .
- Dr Keller has contacted a European colleague who specialises in compound channel flow who commented that the WSE approaches the TEL where the velocity approaches zero (ie at the edge of a flood plain). Dr Keller also relates his experience that, in practice, there is clearly a difference in the WSE between channel and flood plain and, in line with the thinking of his colleagues such as Dr Eric Laurenson, the difference is assumed to be of the order of the difference in velocity heads. He is not persuaded by the paper subsequently provided by Dr McCowan, which gives results which appear to justify his assumption, but which are based on a two-dimensional model prediction.
- Dr Keller reaffirms his belief that the approach undertaken is correct - ie that the WSE approaches the TEL near the edge of a flood plain where the velocity approaches zero.
- The concept of different flood levels between the boundary of the floodplain and the main channel is also documented in the MMBW's Report on the 1974 Flood Event.
|
| Table 8.3 is Flood Levels not TEL |
- This matter has been further reviewed (particularly in light of Ron Sutherland's letter to Rob Skinner of 17 May 2005.) GHD and Dr Keller still believe it is still a reasonable conclusion that values in table 8.3 are Total Energy Levels. It is agreed that the tables in question in 1986 report have a number of typographical errors.
- Comments that Monash work was not used in 1986 results is also difficult to substantiate. (However we have found a hand written calibration profile which is believed to be source of Fig 8.2 of 1986 report and this actually labels lines as "Monash Calibration Run").
|
| Impact on base flood levels. |
- There would be minimal impact as after accepting GHD's report, Melbourne Water adopted flood levels for the purposes of Encumbrances and Development Referrals that are the TEL levels from the HEC-RAS output. Melbourne Water also adjusted the in-channel Manning's "n" to a minimum of 0.02 which caused the TEL's to increase above those within the GHD Report. At the village, this resulted in lower flood levels than applied in 1980's but with a reasonable level of conservatism.
|
| Impact on compensatory works and sensitivity analysis. |
- By way of sensitivity analysis, Melbourne Water has re-calibrated the models assuming that the values are WSE (for the purposes of this review even though the assumption is not accepted) which has resulted in Manning's "n" values of 0.018 to 0.025. Re-running the model that assesses the impact of the wall and compensatory works generally resulted in an approximate increase of 10 mm in flood levels at the Maribyrnong Village when flood levels rise some 4 metres in a flood event.
|
| 2 |
Appropriateness of HEC-RAS model |
|
| Age of model/technology (1995 versus 2003) |
- The HEC-RAS model is still an industry standard.
- It continues to be developed and US Army Corps of Engineers are releasing updates on a regular basis.
|
| 1D versus 2D |
- It is disputed that a 2D model would produce any better result.
- We do not believe that it will model scour appropriately (either at the bridges or along the channel)
- We do not believe that it will resolve uncertainties in calibration data or design flow hydrographs.
- Where the level of available data is comparable (scour, geometry and flow) a 2D model cannot be expected to predict the Water Surface Level profiles to any greater accuracy than a 1-D model
- 1D models include bridge relationships as well as expansion and contraction of flow coefficients that have been based on empirical data from research.
- It also needs to be recognised that all models require engineering judgement to ensure that they appropriately reflect the unique river system it is attempting to replicate.
- 2D models need to be of very fine resolution at bridge openings in order to actually reflect the impact that piers and abutments have on flows.
|
| Use of 2D Model by GHD to assess impact of wall. |
- The modelling carried out by GHD primarily used a 1D model, created in HEC-RAS. This model is generally well-suited for determining flood levels on the Maribyrnong River.
- However, in 1D models, flood levels are based on representative cross-sections and cross-section averaged flow characteristics. Typically, water surface levels and flow velocities vary across a wide flood plain. A 1D model cannot resolve these variations, which may be significant for determining the impact of modifications to the flood plain, such as that of the VRC flood wall.
- That is, although a 1D model is generally appropriate for setting flood levels, the detailed assessment of the flow distribution and the effect of particular flood plain works is best carried out using a 2D model.
- To this end, a 2D model of the Flemington Racecourse site was developed using the FLS model. This model was based on the same bathymetry, photogrammetric survey, and Melbourne Water 1m contours as used for the 1D modelling.
- The FLS model was then used to determine the hydraulic conveyance of floods through the Flemington site under existing conditions and the effects of the flood wall on upstream flood levels. It was also used to obtain the elevation-volume relationship for the site, which was used to model the effect of the loss of flood plain storage.
|
| Localised swirling flows |
- Where swirling is an issue, 1D models have been proven to adequately reflect the associated energy loss with a localised adjustment to the cross-section data and an appropriate Manning's n.
|
| 3 |
Low Mannings "n" values used for calibration |
- To simply change Manning's "n" values to "more realistic" values would be inappropriate as it would ignore the calibration and result in unreasonably higher flood level. These higher flood levels although more conservative would be very difficult to defend in setting development conditions as they would be generated by a model that doesn't replicate actual recorded catchment and river behaviour.
- It is also important to be aware that traditionally at times, Manning's "n" was the factor used to apply levels of conservatism to studies to make up for any shortfalls in the input data (flow or topography etc).
- Recent (1998) published evidence of roughness coefficients derived through measurements for rivers with similar hydraulic characteristics to the Maribyrnong show values of 0.015 to 0.018 for similar orders of flow (Roughness Characteristics of New Zealand Rivers - NIWA Taihoro Nukurangi)
- Furthermore, Dr Keller and GHD advise that when cross-sections along a waterway are "sparse" the Mannings "n" value compensates for the flood flows expanding and contracting as the floodplain/river meanders within the reach between cross sections. Therefore, the recent modelling with a greater number of cross-sections does not rely on Mannings to take this into account and relatively smoother values would be expected with more cross-sections.
|
| Scour at bridges or all along river |
- 1986 report states - ".. it became evident that general scouring of the river bed would occur throughout the study area at high flows." It is agreed that scour is a larger issue at constrictions (ie bridges etc)
- The levels of scour are not quantifiable and therefore the Mannings "n" adopted through calibration is believed to possibly reflect this scour issue. The alternative would be to have dynamic river geometry that changes the cross sectional area as velocity increase and decreases, however this would be introducing a further engineering judgement.
|
| Affect of coffer dam on scour in 1974 |
- The coffer dam is implicitly allowed for in any calibration as the recorded levels would have taken in any afflux created by the temporary works.
- Adjusting the expansion and contraction co-efficients to the maximum allowable values indicates that the coffer dam would have had minimal impact.
- It would be reasonable to assume that the dam and partly completed bridge pile in 1974 are comparable to the fully constructed bridge pile today rather than assuming any conservatism in the Mannings "n" values.
|
| 4 |
Others advice on use of HEC-RAS |
|
| ASCE Manual and 2ft/mile slope issue |
- The ASCE Manual states that "Unsteady flow analysis should be used for all streams where the slope is less than 2 ft per mile." It is also true that, in parts of the Maribyrnong study reach, the gradient is less than this.
- It is important to note that the slope issue disappears if the flow hydrograph is completely flat (ie no change in flow with time) The design hydrograph for the present study is relatively flat, compared with flood hydrographs in other rivers. Dr Keller remains of the view that engineering judgement justifies the use of steady-state modelling, based on peak flows, as used by GHD.
|
| MFPC advice from US Army Corps of Engineers (USACE) and that its not available to others |
- With regard to this issue, Melbourne Water has not been privileged to see any comment that the authors of the software have made and we are not aware that the USACE have actually seen the modelling undertaken on the Maribyrnong by GHD.
|
| 5 |
What impact does the wall have on smaller events |
|
| Is storage a bigger issue at lower flows?Has any assessment been done. |
- For events of a similar ARI, any loss in storage will have greater impact on peaky events (small flood volume) and less on longer duration (bigger volume events). A sensitivity run with a "worst case" scenario of the actual 1974 event (which is a peaky 50 year event) saw slight increase in flows downstream, which for the most part were contained with the channel with the exception of small area upstream of Dynon Road where the industrial areas already within the floodplain had slightly increased flood levels.
- For the 50 year ARI design event the increase in flows due to the removal of the floodplain storage was equivalent to an increase (2cumec) in flows which has a negligible impact on downstream flood levels for this event. In an 100 year event the impact of Flemington is 8 cumecs.
|
| 6 |
If so many culverts were required for Kensington Banks loss of storage, how can the access path lowering compensate for Flemington |
|
| 200m2 area of culverts were required for Kensington Banks, why only scraping several hundred millimetres are required for a larger storage? |
- The railway culverts were constructed with the primary objective of lowering upstream flood level, not to offset the "relatively small" increase in downstream levels due to loss of floodplain storage at Kensington Banks. By comparison the Flemington Racecourse mitigation works are primarily designed to address the loss of floodplain storage.
|