Open johcarter opened 2 years ago
@johcarter including the common hazard types and intensity measures in OED would definitely be a good start. Could the more unique/less common ones then be captured in the model settings file in parallel to the ones in OED or could this pose an issue with duplication? I only ask as a hybrid approach may accommodate more users initially.
yes I think so @MattDonovan82 . as long as the meta data is well structured so that it would not matter whether an OED identifier was present or not.
GFDRR, in developing the Risk Data Library standard looked at this question, with a consortium of hazard experts.
An example of metadata shows fields to give specific information related to hazards, added to ISO geospatial metadata - added with metadata_type RDL_DATA_SPECIFIC. hazard_maps_flu_def_updated.meta.csv Data with these metadata will be stored on the World Bank data catalogue. This page of documentation shows how RDL stores hazard data, with associated data fields and values: https://docs.riskdatalibrary.org/hazard.html
Fields 'Type' and 'Process' use the hazard taxonomy defined at: https://docs.riskdatalibrary.org/taxonomy.html
We defined a long list of intensity measures per hazard process, which you can use. Requesting @MamadioCMCC to provide that list here.
In GFDRR RDL, metric and unit are tied together. The following is the list we put together, knowing it will likely be expanded:
HAZARD | UNIT | DESCRIPTION |
---|---|---|
EQ | PGA:g | Peak ground acceleration in g |
EQ | PGA:m/s2 | Peak ground acceleration in m/s2 (meters per second squared) |
EQ | PGV:m/s | Peak ground velocity in m/s |
EQ | AvgSa:m/s2 | Average spectral acceleration |
EQ | Sd(T1):m | Spectral displacement |
EQ | Sv(T1):m/s | Spectral velocity |
EQ | PGDf:m | Permanent ground deformation |
EQ | D:s | Significant duration |
EQ | IA:m/s | Arias intensity (Iα) or (IA) or (Ia) |
EQ | Neq:- | Effective number of cycles |
EQ | EMS:- | European macroseismic scale |
EQ | MMI:- | Modified Mercalli Intensity |
EQ | CAV:m/s | Cumulative absolute velocity |
EQ | D_B:s | Bracketed duration |
FL, CF | fl_wd:m | Flood water depth |
FL, CF | fl_wv:m/s | Flood flow velocity |
WI | v_ect(3s):km/h | 3-sec at 10m sustained wind speed (kph) |
WI | v_ect(1m):km/h | 1-min at 10m sustained wind speed (kph) |
WI | v_etc(10m):km/h | 10-min sustained wind speed (kph) |
WI | PGWS_tcy:km/h | Peak gust wind speed |
LS | ls_fd:m | Landslide flow depth |
LS | I_DF:m3/s2 | Debris-flow intensity index |
LS | v_lsl:m/s2 | Landslide flow velocity |
LS | ls_mfd:m | Maximum foundation displacement |
LS | SD_lsl:m | Landslide displacement |
TS | Rh_tsi:m | Tsunami wave runup height |
TS | d_tsi:m | Tsunami inundation depth |
TS | MMF:m4/s2 | Modified momentum flux |
TS | F_drag:kN | Drag force |
TS | Fr:- | Froude number |
TS | v_tsi:m/s | Tsunami velocity |
TS | F_QS:kN | Quasi-steady force |
TS | MF:m3/s2 | Momentum flux |
TS | h_tsi:m | Tsunami wave height |
TS | Fh_tsi:m | Tsunami Horizontal Force |
VO | h_vaf:m | Ash fall thickness |
VO | L_vaf:kg/m2 | Ash loading |
DR | CMI:- | Crop Moisture Index |
DR | PDSI:- | Palmer Drought Severity Index |
DR | SPI:- | Standard Precipitation Index |
This looks ideal, thank you @stufraser1 @matamadio ! At first glance, it covers most of the common hazards in catastrophe modelling. The ones I can't see are hail, wildfire. fire following earthquake, sprinkler damage.
Separately I wonder to what extent we could/should bring ORD model meta data in line with RDL. Many of the attributes in the example hazard map are also appropriate for catastrophe models.
I think we'd be happy to extend the RDL code list to include hail, wildfire, fire following earthquake, sprinkler damage What intensity and units do you use for these?
We can share the metadata RDL uses for loss estimates too. Here is the data structure for loss datasets. @matamadio do you have a metadata file example too?
Would be great to align them, and a good outcome since GFDRR and Oasis had a joint workshop last year discussing how we can better align / support each other.
@pchrzanowski (now leading RDL) and @dickiewhit will be interested in this development.
Thanks @stufraser1. I think hail stone diameter in cm is common. I'm not sure about fire intensity and sprinkler damage, but we could ask the oasis modeller community for input on this.
happy to continue the conversation on data standard alignment, I'll review the documents you link to.
The metadata file (json) is still a work in progress.
Some comments on the proposed additional metrics:
To clarify on sprinkler damage: rather than damage to sprinklers, industry models can include loss incurred due to the additional damage caused by the water sprayed from sprinklers in a fire. In that sense it is coded as a peril in OED or AIR CEDE datasets, for locations/buildings which have that included in its insurance coverage. This effect is modelled as a factor (in my recollection) which increases the loss incurred from the main hazard. Its something the development sector would be modelling very rarely, if at all. Given RDL is designed for development sector use, it may be something we choose not to support (per the decisions on simplifying GED4GEM exposure taxonomy when developing the simplified but hazard-expanded GED4ALL).
Hi, May have missed them, but adding some more parameters here:
For Fire Following & Sprinkler Leakage, often these sub-perils include a 'trigger' component, as well as severity (typically captured by the 'hazard') component. We wouldn't look to support hazard parameter outputs for these sub-perils at this time.
Hi @johcarter bringing this back on the radar as the OED/RDL works picks up again.
@johcarter for the missing units from some perils (wildfire, etc) is best to get in touch with the model vendors who have models for them and then include what they are using. Great idea to standardise the units!
Rather than using the list pasted above in this issue, please use the latest dev codelist from RDLS: https://github.com/GFDRR/rdl-standard/blob/0.3-codelist-updates/schema/codelists/open/IMT.csv?plain=1
And the production version: https://github.com/GFDRR/rdl-standard/blob/main/schema/codelists/open/IMT.csv?plain=1
@stufraser1 is there a way to associate all the units with the respective perils they cover? In some cases it's pretty obvious, but the further down the list you go, the more obscure they get.
We've been discussing adding some ORD reports for hazard, such as hazard EP curves and hazard value by location and event. However there is no clear structure for hazard meta data in ORD. Here is a table of the current relevant fields in ORD:
If we support reports which contain values such as windspeeds and flood depths, we should have clearly structured meta data which references exactly what those values mean in terms of hazard intensity measure and units.
Peril codes are used as a proxy for a hazard intensity measure, but in practice this is not well specified in model settings.
e.g WTC means Tropical Cyclone, but is normally used to represented the hazard of wind speed. This is a 'peril type' or event type which is being used as a proxy for wind speed, but they are different concepts.
A meta data structure might look like this, for example;![image](https://user-images.githubusercontent.com/11421733/173834101-5aec3298-de76-4e0c-9360-33a8e41233c1.png)
We could also consider codifying common hazard intensity measures in ODS, e.g.
HazardIntensityMeasureType =1 name = 'windspeed' desc = '3 second maximum peak gust speed at 10 metre elevation' unit = 'metres per second'
We can't expect to codify all possible hazard intensity measures that modellers choose to use, but having a list of common ones would help newcoming modellers and facilitate pairing of hazard and vulnerability modules across providers.
Obviously a very important design question, so i welcome lots of comments and suggestions