Botany Bay flood risk management study, Australia
The Botany Bay project aims to update the existing flood study and to provide a
framework to reduce the identified flood risk in the study area by way of structural works to physically reduce the impacts of flooding on development, policy and planning provisions. This is to ensure future development does not increase flood risk. It also includes emergency planning and public education to reduce the potential harm to people presented by the residual flood risk in future flood events.
Software run with Flood Cloud: TUFLOW
Time to run 70 simulations with normal approach: 490 hours
Time to run 70 simulations with Flood Cloud: 7 hours
Time saved using Flood Cloud: 98.5% (20 days)
High performance computers cost: £0
Extra licence costs: £0
Additional labour costs: £0
The study area is located within the Bayside Council Local Government Area (LGA) in New South Wales, Australia. With a surface of approximately 5.4 km2, the area is fully developed with residential, commercial and industrial buildings. Due to the relative flat nature of the topography in the catchment and the low-lying areas, the flooding issues are significantly impacted by the performance of the pit and pipe drainage networks and the tidal levels at the outlets.
The previous flood study was conducted in 2015 and was undertaken to define the nature and extent of flooding within the area. The study represented catchment conditions in 2015 and utilised Australian Rainfall and Runoff (ARR) 1987 methodologies to estimate rainfall runoff. However, based on the data and model review report, it was recommended that the 2015 flood study be updated in accordance with the ARR 2016 guidelines, and recent topographic modifications and developments.
A full range of design flood events were simulated for the updated flood study. Design flood events includes the 20%, 5%, 2%, 1%, 0.5% and 0.2% Annual Exceedance Probability (AEP) events and the Probable Maximum Flood (PMF) event. As per new Australian Rainfall and Runoff guidelines (ARR2016), each design storm AEP and duration consists of an ensemble of 10 storm temporal patterns which define the timing and intensity of rainfall throughout a given storm event. Each storm in the 10 temporal pattern ensemble has an equal probability of occurring.
The 20% AEP, 5% AEP and 1% AEP events were simulated for the 10 temporal patterns for the full range of storm durations from 15 minutes up to 9 hours. The 2% AEP, 0.5% AEP and 0.2% AEP events were simulated for the identified critical storm and temporal patterns.
The large number of scenarios, more than 400 simulations needed to be run within TUFLOW, would require thousands of hours of computing time. Each of the TUFLOW simulations would have taken about 7.5 hours to complete using one of their local computers. Due to both hardware and software (TUFLOW licensing) restrictions, it meant that all the simulations could not be run concurrently to obtain results faster.
To overcome the issue and run more simulations concurrently, Jacobs’ Flood Cloud service (which is compatible with TUFLOW models), was used. Flood Cloud enables users to deliver their flood modelling projects quicker than ever before. Once a user has set up their models locally, they can send their simulations for processing on high performance hardware in the cloud directly from within the Flood Cloud interface. This innovative service removes the need for TUFLOW licences as well as high-performance computers. It can be scaled-up and down again at the click of the button and is charged on a pay-as-you-go basis, allowing users to plan and manage their projects more effectively. Hosted by Amazon Web Services’ state-of-the-art offering, it makes it less prone to in-house IT infrastructure challenges that can greatly impact project deadlines.
Due to existing licence availability, only 70 simulations were run using Flood Cloud, however this still provided significant savings when compared to running them locally. Not only was each Flood Cloud simulation faster (by 0.5 hour) when compared to those run locally, the ability to seamlessly scale-up both the software and hardware resources achieved speed increases of between 75% (based on four concurrent simulations) and 98% (based on one concurrent simulation).
This was achieved by running all 70 simulations at once within Flood Cloud, each allocated its own TUFLOW licence and machine, allowing all 70 simulations to be completed in just 7 hours at a cost of approximately 1,559 AUD (£828 / 1,093 USD).
Flood Cloud offers very useful features which helped streamline the study. The test mode allowed the modelling team to make sure that their simulations were set up and packaged correctly before uploading them to the cloud. This saved time and effort as they had assurance that their simulations would run properly and not be interrupted or stopped. It also meant that they would not lose credits unnecessarily.
The real-time dashboard, which provides updates on flows, iterations, convergence and mass balance helped the team to monitor the progress of the simulation to make sure they were keeping in time with the deadline.
Finally, the fact that they had the ability to download the results to any machine and not only their modelling machine saved more time as they didn’t have to transfer the data from one machine to another. This meant they could start processing the data straight away.