Proposed by: Australia - Indonesia Facility for Disaster Reduction and Geoscience Australia
Contact (name, email): David Robinson, firstname.lastname@example.org
Team David Robinson, Peb Ruswono Aryan, Mrigesh kshatrya
Best way and times to contact during RHoK 2.0 Dec 4/5 2010: I'll be at Rhok 2 Jakarta
Source Code http://www.aifdr.org/projects/eqrm_rhok/wiki
Indonesia has more earthquakes than any other country in the world. In recent times there have been huge advances in understanding Indonesia's earthquake hazard - we are starting to understand where earthquakes are most likely to strike and how big they are likely to be. Now our challenge is to understand how these earthquakes are likely to affect people.
Geoscience Australia's Earthquake Risk Model (EQRM) is designed to model the impact of earthquakes on buildings. The code basically simulates the ground shaking associated with hundreds or thousands of synthetic earthquakes (called an event set) and then uses engineering models to estimate the damage that each one of these earthquakes would cause to a dataset of buildings (called exposure or building inventory). While this is important, and without doubt the most rigorous way to understand risk, it requires large, complex and expensive datasets describing where buildings are located and what they are made of. We can't wait for these datasets - we need something now.
We want RHoK to hack the EQRM to allow it to model the fatalities from earthquakes directly from population data. Equations exist to do this and we have access to a number of global population datasets that can get us started. We need you to take these and implement them in the EQRM to lets us start understanding Indonesia's earthquake risk today.
There are two key use cases.
1) Single event fatalities
I should be able to simulate a single, user-defined earthquake and feed the EQRM a population dataset (typically defined as a grid of population density). The EQRM should then tell me how many people are likely to have died at each point in the grid.
2) Probabilistic event set fatalities
I should be able to do the above but for a suite of hundreds or thousands of earthquakes, each one with a defined probability of occurrence. It will be important that I keep track of the fatalities for each individual earthquake so that I can then estimate the probability of a region suffering fatalities
The EQRM tool is written in a combination of python and C.
The job at hand is to try and create a new data structure for population and the tools for EQRM to be able to import and use such data. We will then also need to hack the "vulnerability" functions to allow the EQRM to estimate fatalities based on levels of ground shaking.
Visualisation of Results The EQRM already has a suite of post-processing tools that will us to visualise and analyse impacts to buildings. Hack these to ensure that they work sensibly for the fatality estimates
Importing Population from Geoserver Teach the EQRM software to import the population data directly from AIFDR's geoserver
OpenGEM (http://www.globalquakemodel.org) is working on an open-source earthquake risk model. If the hacks from this Project are good enough they may well find themselves incorporated within this new global initiative.
The EQRM can be downloaded from http://sourceforge.net/projects/eqrm/
A global population dataset can be downloaded from http://sedac.ciesin.columbia.edu/gpw/
An example vulnerability model for fatalities is: F = 10**(a*H-b)*E (where F is fatalities, a=0.97429, b=11.037, H=hazard (ground shaking in MMI), E=exposure(number of people))
This work will be used to produce a rapid earthquake fatality risk assessment for Central Java and, if successful, a preliminary analysis of earthquake risk across all of Indonesia.
The EQRM is an open-source community model that is currently being developed by Geoscience Australia. Components of this model will likely be incorporated into the Global Earthquake Model (GEM). See also http://www.globalquakemodel.org
these results below are generated from randomly generated population data but basically we already implemented the fatality calculation to the eqrm. the total site location contains 64580 points that are simulated on site-by-site basis so to speed things up we simulate just the part of it (say like 65 or 650).