1 UNISDR - Global Risk Assessment: Towards a high-performance environment WCDRR, Sendai 15 March, 2015 Sahar Safaie, Julio Serje Global Assessment Report Team United Nations Office for Disaster Risk Reduction Geneva, Switzerland
2 GAR15 Probabilistic Global Risk Assessment Components: Hazard, Exposure And Vulnerability
3 GAR Global Risk Assessment Partners
4 GAR15 Global Risk Assessment 6 Hazards are covered: flood, earthquake, cyclone wind and storm surge, tsunami, volcanic eruption, drought Probabilistic Hazard modeling of 5 hazards at global scale Probabilistic risk modeling using CAPRA software Datasets and results including national risk profiles are available on the open platform
5 Earthquake Hazard Model:
6 Earthquake Hazard Model: 1. Identification of the principal seismic sources to characterize the tectonic regions and seismic provinces. 2. Calculation of the seismicity parameters of each seismic source 3. Generation of a set of stochastic set of scenarios (1 Million synthetic earthquakes) 4. Model the probability distributions of ground motion intensity associated with a given earthquake magnitude. 5. Generation of hazard maps for representative events (seismic intensity parameter and the probability of it in each 30x30km grid)
7 Exposure Model
8 Exposure Model 5x5km Grid with global coverage, 1x1 km grid for selected coastal areas The total number of cells of the grid is 9,008,829
9 Exposure Model Example of exposure data: urban stock in selected areas of Europe
10 Vulnerability Vulnerability is the susceptibility of exposed elements or assets to suffer damage and loss Vulnerability curves used in risk model relate hazard intensity with the potential loss of the element 620 Vulnerability curves used in GAR Global Risk Assessment
11 Risk model calculation With the simulated hazard scenarios, the exposed assets and their vulnerability it is possible to calculate the probabilistic loss. For earthquake: 1 million scenarios 9 million exposure cells 18 types of assets types (vulnerability curve)
12 Tsunami, Cyclone Wind and Surge, Riverine Flood, Drought, Volcano
13 Risk number crunching: current approach (most hazards) Windows Workstation based tools for Calculation (CapraGIS) Manual partitioning of data (by Country) Distributed Processing in a small Local Area Network TOTAL PROCESSING TIME: MONTHS!!!
14 Risk number crunching: current approach - Floods Linux Workstation based tools for Calculation (new CapraJava ) Manual partitioning of data (by Country) Distributed Processing in a small Local Area Network (2 machines, 24 CPUs) Brazil, Canada, Russia: 7-10 days each USA: 18 days; China: 5 weeks! Other countries: between 1 hour and 1-2 days. TOTAL PROCESSING TIME: 5 Weeks
15 Risk number crunching: future requirements Increase in Exposure resolution: Exposure at 1x1Km = 25 times Exposure at 100mts=2,500 times Increase in Hazard resolution: Hazard models at 1x1K =25-90 times, etc. Increase in number of Hazards: Increase in number of exposed assets (Agriculture, infrastructure, etc.) Increase in processing time: anything between a thousand to several million times current processing needs. Unfeasible Under The Current Approach
16 Risk number crunching: moving forward Platform Independent version to run under LINUX (initially to Java) NO Graphical user interface GUI Parallelization of Software (with MPI Message Passing Interface?) Move to a High Performance Environment (collaboration with LRZ) TOTAL PROCESSING TIME: Days? Hours? Minutes? Real Time? Linux Cluster Super MUC
17 Thank you