When the result is a landslideTue 24th May 2011
A landslide can wipe an entire village off the map leaving a trail of destruction in its wake. Their effects are amplified in poorer countries by under-developed infrastructure and a lack of resources for disaster recovery. To combat this geographers based in Bristol have been working with colleagues in the Eastern Caribbean to model landslides and their after-effects, and to identify potential solutions. They are now porting the work on to the grid with support from GridPP.
Since 2005, the MoSSaiC project (Management of Slope Stability in Communities) has been studying and modelling landslides in the Eastern Caribbean to help locals and provide information to the World Bank for their development schemes. Their work looks at landslides from various angles including the risk of landslide on a specific slope, its potential size and the economic impact. However the workflow had some failings; managing data from different simulations was difficult and the available CPU was being overstretched.
Of course these issues aren’t new in the world of science so one of the researchers, Dr Liz Holcombe from Bristol’s School of Geographical Sciences, approached the university’s Advanced Computing Research Centre (ACRC) to help. The entire process was a collaboration, with Dr Holcombe being very involved in the design of the system “In developing the initial proof-of-concept, I worked together with Jon Wakelin at ACRC as much as we could on the user interface. This meant that I was able to give ongoing feedback on how users might approach the data entry part of the modelling process.” Work focussed on two popular applications; CHASM and QUESTA. CHASM models the stability and physical characteristics of the slope while QUESTA, developed by Dr Holcombe, looks at the economic impact of a potential landslide.
Alongside his work with ACRC, Jon Wakelin was also a member of the physics department at Bristol. His experience with the grid had persuaded him that it was suitable for the work, “We had already made a lightweight front-end, a simple web form of only 4 pages for entering the different kinds of data required. Once all the data was entered the user could view a drawing of the slope geometry they had created. To help users we incorporated on-the-fly error checking capability, the ability to view job progress once submitted, downloadable raw input/output files and reloading previous simulations. What was important in the design process was the fact that the end users would often have limited local computing power and rather unreliable internet connections. I felt that the grid would help overcome the limitations while still working within the obvious constraints.”
It was at this point Michael Wallace from the High Energy Physics group came on board to integrate the system with the grid, “When I first got involved I was happy to see that the application was written in C and ran on Scientific Linux which meant running the jobs on grid nodes was going to be relatively straightforward. Their workflow was also quite simple: a task defines the simulation required, this is split into individual simulation jobs and they are then submitted to the grid. So far I have been able to run the modelling software on the grid and am working on integrating the existing user interface with the grid infrastructure. However we are still in the early stages of development and the hard part of the work is still to come. One of our biggest obstacles will be developing methods for managing large volumes of input and output data and turning it into useful information. We want them to be able to answer questions like “Of the slopes that are likely to fail, which ones will have a significant impact?” or “How can the resources we have be best used to reduce that impact?”". To help with this the team have been awarded an Innovations Partnership Schemes grant by STFC. The grant is aimed at getting the project off the ground by funding dedicated effort to improve data management and grid access. The team have chosen the DIRAC framework, used by LHCb, as their starting point.
The application is currently pre-alpha and the team are using continuous deployment to release new features to users for testing. User testing is an important part of the development cycle and features will only be released to production once they are signed off by test users. Dr Holcombe will continue to be involved in the development and is very optimistic “The prototype has been tested by an engineer working for the Government of St Lucia. They were very pleased with the user-friendly interface to both model inputs and outputs and the potential power of having remote access to resource. They did raise some issues about user and data management as well as secondary processing and error checking and we have taken those comments into account as we continue development.”
The next step for the project is to increase access to grid resources. The VO is currently only supported by Bristol with a couple of pledges for future support but they would like more sites within GridPP and further afield to enable their VO (vo.landslides.mossaic.org). If you are interested please contact Michael Wallace at firstname.lastname@example.org