Open matamadio opened 3 years ago
This needs to be discussed with developers of MOVER - look at proposals together with them and understand implications of simplifying (noting it was already simplified somewhat)
Better to discuss over implementation examples, which we also need for demo showcase sooner or later. I remember you were already looking into function repos etc, it would be good to have V examples for at least 3 different hazards (FL, EQ, WI). On my side I have flood depth-damage functions (JRC ones or others) that could be put in form of step-value tables and/or mathematical functions in order to test implementation.
As examples of Vulnerability dataset that could be included:
Flood Damage Model catalogue is a rare example of V models collection with metadata and table data. Includes curves made by JRC, with macro-regional coverage.
RiskScape catalogue has several hazards covered, but only provides bibliography of resource.
However these are not the curves used to calculate loss in our showcases datasets. In the case of SWIO-RAFI, the AIR propetary model was used; it is not open, we only have pictures of the functions from the report.
In the case of Afghanistan, we have a number of different approaches depending on the hazard.
EQ has proper fragility functions only displayed as picture (no tables):
FLood uses open source Delft-FIAT model for which it provides picture of curves for residential buildings (no tables):
Other hazards (drought, landslide, avalanches) are assessed using a simpler approach, classes of vulnerabilities described by tables and geodata, i.e. risk is given as exposure to different classes of hazard, and a fixed damage ratio is applied.
Prahl et al. 2018 - Damage and protection cost curves for coastal floods within the 600 largest European cities https://www.nature.com/articles/sdata201834
using building-material-based vulnerability curves for an object-based approach in urban and rural areas
https://nhess.copernicus.org/articles/19/1703/2019/
I was not aware of this! Made by GFDRR in 2019, we should definetely try to gather the functions mentioned in the report.
By installing this software, it is possible to view details and compare vulnerability functions developed for the CAPRA model, which was used for assessing risk in GAR13 and GAR15.
This software uses a simple analytical methodology to obtain new vulnerability functions based on a few parameters. The methodology used to obtain new vulnerability functions are based on the methodology proposed by (Yamin et al. 2014) and (Council 1985) with a few modifications. The vulnerability functions are represented by three main parameters the hazard intensity parameter value, the expected and the variance value of the MDR for either physical or human damage.
https://ecapra.org/topics/funvul-simplified
was previously reviewed by WB: https://openknowledge.worldbank.org/handle/10986/22981
I expect similar additional material to pop out from a review of these reports: https://openknowledge.worldbank.org/handle/10986/7755
We will refer some of these report/PDF/image only curves to stats expert at UCL, to understand how they retrieve curve parameters from images, which we can then apply to further curves for inclusion in dataset.
@Mattia Amadio @.***> can you suggest 3-5 which would be most useful to put to UCL? I would suggest AIR seismic and wind (do we have building curves, or only the infrastructure curves?). GPSS data we can almost certainly get the curves - let me follow up with contacts in the program.
Proposing to use http://www.fdm.polimi.it/models as a source for flood V curves
Vulnerability functions used in GAR15 are discussed in Cardona et al 2015.
More about the CAPRA Vulnerability model of GAR. I think the way it describes the functions together with metadata looks neat, very clear and intuitive to use.
Example output file
A simple text file, both human readible for data/metadata and machine readable (by specific tool/module) to generate the actual function.
Proposing to consider this approach and adapt it for RDL V schema/package. It seems very fit for majority of fragility/vulnerability functions; may be inadequate to describe other types of models. Could rely on a simple python utility or webtool to plot the functions instead of standalone program.
We received the Flood and Cyclone vulnerability curves from the OIA-SEA study. Good candidate for schema proofing.
Flood depth/velocity functions for different exposure types https://knowledge.aidr.org.au/media/5667/updating-national-guidance-mcluckie.pdf
GEM archive of earthquake fragility, vulnerability, damge-to-loss functions https://platform.openquake.org/vulnerability/list?type_of_assessment=2
The V schema as it is implemented in JKAN has been simplified and made interlinked with the other schema components, compared to the original sql version (had 97 fields, most of which are quite specific and optional. In comparison, other schemas had 1/4 the number of fields). It still has a high level of complexity, and it would benefit from being tested and adjusted on real datasets (bottom-up approach). Suggesting to start fitting it on a selection of vulnerability models to use as showcase data.