[Docs update] Describe use of and mapping to UNDRR Hazard Taxonomy (HIPS)

pzwsk commented 3 years ago

If you are filing a proposed change, e.g. a new attribute, relationship, use of a new taxonomy, or a revised definition, then please fill in the below template.

What is the context or reason for the change?

Publication of the UNDRR Hazard definition and classification review https://www.undrr.org/publication/hazard-definition-and-classification-review

Why is this not covered by the existing model?

UNDRR taxonomy was not available when the RDL project started.

What is your proposed change?

To align RDL hazard taxonomy with UNDRR Hazard definition for hazard within the scope of RDL.

i.e. reviewing names, code and hierarchy of the common tables for hazard against the UNDRR taxonomy https://docs.riskdatalibrary.org/common.html

Rationale: UNDRR official source for hazard definition and classification.

Can you provide an example?

RDL

UNDRR

matamadio commented 3 years ago

I'll update the existing comparison matrixes we did with new UNDRR framework.

matamadio commented 3 years ago

UNDRR taxonomy for natural hazards

(excluded chemical, biological, environmental, extraterrestrial, technological, societal hazards)

Show huge table

matamadio commented 3 years ago

See comparative table of our hazard schema and UNDRR hazard list. Includes also taxonomies by DRMKC, OED, Inspire.

Let's focus on RDLxUNDRR hazard taxonomy comparison. UNDRR hazard list is copied from the Annex 6 of 2020 report.

There is matching for most primary hazards, but the UNDRR taxonomy is huge and extends far out of the common natural hazards, including phenomena not commonly assessed in risk analysis. Others would be nice to include but lack datasets: e.g. previously we evaluated the inclusion of "convective storms > hale | lightning | tornado" but in the end couldn't find anything for the first two hazards, and only kept "convective storms > tornado".
It goes into a lot of detail (and repetition..) for certain categories, eg distingush "gale" from "wind" and has 12 types of "cold weather" hazards:
- Hail
- Ice storm
- Snow / Ice
- Snow storm
- Cold wave
- Dzud
- Freeze
- Frost (hoar frost)
- Freezing rain
- Glaze
- Ground frost
- Icing (including ice)

but at the same time it has only one general "Drought" hazard category while RDL has 4 sub-categories (meteo, hydro, agri, socio-eco).

Some phenomena are just environmental conditions and not necessarily an hazard tied to an impact.
It uses explicit combinations of triggers and hazards as individual items (e.g. landslide triggered by EQ, landslide triggered by VO, etc), whereas the RDL can define any hazard/process as a trigger for another hazard/process.

Ultimately I don't think we should adopt the UNDRR list as a whole, considering that many of those "hazards" are never ever mentioned in risk assessment. Rather justify that our schema matches with the most relevant items in this list and other taxonomies.

pzwsk commented 3 years ago

Thanks a lot, Mattia, based on your evaluation I agree that a perfect (one to one) alignment with UNDRR hazard taxonomy might be problematic, especially due to those two reasons:

different hierarchies used : cluster>specificfor UNDRR versus hazard>processfor RDL
trigger event is embedded into specificfor UNDRR while trigger event is a relation between two process for RDL

Suggestion, include the mapping to UNDRR taxonomy in the documentation, but do not replace RDL Hazard Taxonomy with UNDRR one.

Having said that, I see one issue in the current Hazard taxonomy of RDL which is that we have hazard and process type that are similar and this can create confusion for users:

Landslide
Coast Flood

Should we open a specific issue on that?

matamadio commented 3 years ago

Yes, taxonomy could be improved based on UNDRR, but not fully mirroring it.

In agreement with UNDRR (combining Meteo-Hydro Terrestrial and Shallow geohazard), landslide could contain 5 processes:

Rockfall
Debris flow
Mud flow
Lahar
Avalanche

As a reference in the case of Afghanistan data, landslide hazard is classified as:

bedrock landslides in slow evolution
bedrock landslides in rapid evolution
cover material in rapid evolution
avalanche

Fitting these into the proposed process change would translate as 1 and 2 to be "rockfall" (not including the slow-rapid distinction in schema), 3 "debris flow" and 4 "avalanche".

Similar for coastal flood, my proposal would be to have it include:

Storm Surge
Tsunami (turning it into a process of CF instead of its own hazard type)
Extreme tide (component that is often accounted in coastal flood modelling)
Coastal erosion (have been proposed in the past)

Because often hazard layers can describe the combination of two or more processes, the choice of process shouldn't be univocal. E.g. a landslide layer may consider both rockfalls and debris flows. Then both processes should be included in the metadata.

matamadio commented 3 years ago

Please ignore my previous post. Better check alignment with existing hazard taxonomies instead.

This online sheet includes definitions from:

UNDRR 2020
EMDAT
Munich-RE
DRMKC
Inspire
OED/OASIS

The first tab compare shows the alignment/matching of these sources compared to current RDL hazard schema, and propose a new "common denominator" hazard taxonomy that is based on these criteria:

the best alignment with sources
split as "main" and "sub" (some taxonomies have 3 levels, some only 1)
relevance of phenomena for DRR

Below is the resulting hazard taxonomy:

Main	Sub
Earthquake	Ground shaking
Earthquake	Tsunami
Landslide	Rockfall
Landslide	Debris flow
Landslide	Mudslide
Landslide	Snow avalanche
Subsidence
Volcanic	Ashfall
Volcanic	Lahar
Volcanic	Pyroclastic Flow
Volcanic	Lava
Storm	Extra-tropical storm
Storm	Tropical storm
Storm	Convective Storm
Storm	Lightning
Storm	Tornado
Storm	Sand/dust storm
Storm	Winter storm/blizzard
Flood	Riverine flood
Flood	Flash flood
Flood	Coastal flood
Extreme temperature	Heat wave
Extreme temperature	Cold wave
Drought
Wild fire	Forest fire
Wild fire	Land fires (grass, scrub, bush)
Epidemic
Insect infestation

pzwsk commented 3 years ago

Thanks for doing this Mat, this is very good and actually something that could be featured within a blog post.

matamadio commented 3 years ago

Latest proposal for hazard taxonomy discussed with Stu, matching (at best compromise) with UNDRR and other sources while maintaining our current taxonomy structure (main hazard - process). Not using "clusters" as in UNDRR (repetition across categories, non intuitive term). New categories and types were added, other were reorganised, renamed or changed code. Not all taxonomy have (double) ID codes, at least disclosed, so that is from us. I added new codes for new items, please suggest if you have better alternative.

New main hazard types:

Mass movement (for landslide, avalanches and subsidence)
Storm (aggregating strong winds and convectice storm)
Environmental degradation (wildfire, coastal erosion, sea level rise)
Biologiocal hazard (for epidemic and infestation)

Notable differences with other taxonomies:

RDL has 4 processes of Drought whereas all other taxonomy just have "Drought" as individual item. We stick to our more in-depth classification, motivated by different types of water scarcity indicators: FAO-ASI for Agriculture, SPEI for Meteorological, WCI for socio-economic..)
RDL has Tsunami as independent main type and process instead of being a process of Earthquake like most taxonomies. many other taxonomy puts it as earthquake process but the very different approach in damage assessment for the two suggest to have a separate categories
No extra-terrestrial hazards
No technological hazards

Summary table:

Hazard	Process	Code	Comment
Earthquake	Ground shaking	QGS	Was QGM ("shaking" substitutes "motion")
Earthquake	Surface rupture	QSR	Was Q1R and Q2R (primary and secondary rupture)
Earthquake	Liquefaction	QLI	No change
Tsunami	Tsunami	TSI	No change
Volcanic	Ashfall	VAF	No change
Volcanic	Lahar	VLH	No change
Volcanic	Pyroclastic Flow	VPF	No change
Volcanic	Lava	VLV	No change
Mass movement	Landslide	MLS	Added new category "mass movement" including generic landslide and more specific
Mass movement	Rockfall	MRF	New
Mass movement	Snow avalanche	MSA	New
Mass movement	Subsidence	MSS	New
Storm	Tornado	STN	Changed from "Convective storm > Tornado (TOR)"
Storm	Lightning	SLN	New
Storm	Extra-tropical cyclone	SEC	Was under "Strong wind" category, now under "Storm", new code
Storm	Tropical cyclone	STC	Was under "Strong wind" category, now under "Storm", new code
Storm	Sand/dust storm	SSS	To be reviewed for data and imt
Storm	Winter storm/blizzard	SWB	To be reviewed for data and imt
Flood	Fluvial flood	FFF	No change
Flood	Flash flood	FPF	No change
Flood	Coastal flood/Storm surge	FCF	Previously separated category, now under Flood and aggregated with storm surge as in DRMKC
Extreme temperature	Heat wave	ETH	New code (was EHT)
Extreme temperature	Cold wave	ETC	New code (was ECD)
Drought	Agricultural	DTA	No change
Drought	Hydrological	DTH	No change
Drought	Meteorological	DTM	No change
Drought	Socio-economic	DTS	No change
Environmental degradation	Wildfire	EWF	Was main hazard type with identical process
Environmental degradation	Coastal erosion	ECO	New
Environmental degradation	Sea level rise	ESR	New
Biological hazard	Epidemic	BEP	New
Biological hazard	Insect infestation	BII	New

Next steps:

Team review of proposed taxonomy
Selection of a starting list of IMT for each process. IMT definitions are "open" for new items when necessary, but not publicly; RDL admin to evaluate case by case, checking to avoid duplications.
Taxonomy approval
Include hazard definitions (main source: INSPIRE definitions)
Update code lists https://github.com/GFDRR/rdl-standard/issues/22
Update docuentation (Taxonomy and Hazard sections)

ldodds commented 3 years ago

If you have mappings from the RDL codes to those other standards, then it might be useful to also publish those in a machine-readable form (e.g. a CSV file sitting alongside the code lists). This will help share the mappings and potentially allow people to translate/transform between datasets.

stufraser1 commented 3 years ago

Adding a note on tsunami - it shouldn't be a process in EQ given there are also landslide and volcanic generation mechanisms, so to place tsunami in earthquake hazard would ignore these. Equally, we did not want to replicate the process type as we do not need to link the process all of it's triggers (we have the link between trigger event and triggered event that does not rely on hazard classification.

stufraser1 commented 3 years ago

Noting that environmental degradation and biological were added to expand potential to relevant hazard and create alignment with UNDRR for these (also a reason to move wildfire under environmental)

Noting that for earthquake, primary rupture and secondary rupture provide a detailed separation which we do not experience in practice - therefore they are combined.

Note on combination of storm surge and coastal flood. The previous separation was because coastal flood includes storm surge but also technically includes wave set up, wave run-up which contributes to the hazard but are independent of cyclone/pressure (though generally storm surge provides the greatest wave heights), through they use the same IMTs and vulnerability curves. Combination will simplify the categorisation of very similar datasets and would use the dataset description to explain the components of hazard included.

stufraser1 commented 3 years ago

Requirement: IMTs to be defined for:

[ ] Environmental degradation | Coastal erosion | ECO | New
[ ] Environmental degradation | Sea level rise | ESR | New
[ ] Biological hazard | Epidemic | BEP | New
[ ] Biological hazard | Insect infestation | BII | New
[ ] Mass movement | Rockfall | MRF | New
[ ] Mass movement | Snow avalanche | MSA | New
[ ] Mass movement | Subsidence | MSS | New
[ ] Storm | Lightning | SLN | New
[ ] Storm | Sand/dust storm | SSS | To be reviewed for data and imt
[ ] Storm | Winter storm/blizzard | SWB | To be reviewed for data and imt

matamadio commented 3 years ago

Proposed IMT for new hazards

To be approved @stufraser1

Main Type	Process	Code	IMT (proposal)	Reference
Environmental degradation	Coastal erosion	ECE	Shoreline retreat (m); or Annual rate of shoreline retreat (m/y)	UNDRR Report
Environmental degradation	Sea level rise	ESR	Annual change in Mean Sea Level compared to historical trend (m/y)
Biological hazard	Epidemic	BEP	Number of people affected; Rt index
Biological hazard	Insect infestation	BII	Affected area (km^2); Severity Index
Mass movement	Rockfall	MRF	Susceptibility index	Approach for AFG and other risk projects
Mass movement	Snow avalanche	MSA	Avalanche pressure (kPa)	Approach for AFG
Mass movement	Subsidence	MSS	Annual change in land elevation compared to historical trend (m/y)
Storm	Lightning	SLN	15-Minute Lightning Strike Density (strikes/km^2/min * 10^3)	NOAA
Storm	Sand/dust storm	SSS	Annual number of events (n/year); Wind speed (km/h); aerosol optical depth (AOD); Coarse particulate matter (PM10)	https://earthobservatory.nasa.gov/global-maps/MODAL2_M_AER_OD; https://blog.breezometer.com/how-dust-storms-confuse-air-quality-reports/
Storm	Winter storm/blizzard	SWB	Annual number of events (n/year); Wind speed (km/h); Northeast Snowfall Impact Scale (NESIS); Affected area (km^2); Winter Storm Severity Index (WSSI)	https://www.ncei.noaa.gov/access/monitoring/rsi/nesis; https://www.weather.gov/gjt/WSSI_Tutorial

When hazard list and imt are approved I will update docs and data package specifics.

stufraser1 commented 1 year ago

@matamadio I updated proposed list above https://github.com/GFDRR/rdl-standard/issues/2#issuecomment-844287573 Go ahead and publish in docs

stufraser1 commented 1 year ago

Process_type.csv has been updated with descriptions here: https://github.com/GFDRR/rdl-standard/issues/112#issuecomment-1622524165

However, I didn't complete the table according to these proposals. I added rockfall, mudflow, rockslide while retaining a general landslide category, given we have lots of data which is not defined as the subtype of landslide, but landslide overall.

@matamadio When we have the codelists for process_type and IMT both published as csv, we should make changes to make them comprehensive according to the above proposal https://github.com/GFDRR/rdl-standard/issues/2#issuecomment-844287573

stufraser1 commented 1 year ago

IFRC suggested at SC mtg (July 12 2023) to create a mapping between RDLS hazard_type and process_type to UNDRR HIPS

RDLS 4 drought codes -> drought, for example Our tsunami code -> HIPS two tsunami (may use trigger information too - HIPS includes trigger explicitly)

stufraser1 commented 1 year ago

Proposing csv mapping file to be included somewhere in repo, to describe how RDL hazard_type and process_type map to UNDRR HIPS: Below updated from Mat's earlier sheet (https://github.com/GFDRR/rdl-standard/issues/2#issuecomment-818823181) (not updated for other taxonomies, but could do the same for those - next priority: Open Data Standards peril codes):

UNDRR Hazard cluster	UNDRR Specific hazard	RDL Hazard type	RDL Process type
Convective-related	Downburst
Convective-related	Lightning (Electrical storm)
Convective-related	Thunderstorm
Flood	Coastal ﬂood	Coastal Flood	Coastal Flood
Flood	Estuarine ﬂood	Coastal Flood	Coastal Flood
Flood	Flash ﬂood	Flood	Pluvial Flood
Flood	Fluvial (riverine ﬂood)	Flood	Fluvial Flood
Flood	Groundwater ﬂood	Flood	Groundwater Flood
Flood	Ice-jam ﬂood including debris
Flood	Ponding ﬂood	Flood	Pluvial Flood
Flood	Snowmelt ﬂood
Flood	Surface water ﬂooding	Flood	Pluvial Flood
Flood	Glacial lake outburst ﬂood
Lithometeors	Black carbon (Brown clouds)
Lithometeors	Dust storm or Sandstorm
Lithometeors	Fog
Lithometeors	Haze
Lithometeors	Polluted air
Lithometeors	Sand haze
Lithometeors	Smoke
Marine	Ocean acidifcation
Marine	Rogue wave
Marine	Sea water intrusion
Marine	Sea ice (ice bergs)
Marine	Ice ﬂow
Marine	Seiche
Marine	Storm surge	Coastal Flood	Storm Surge
Marine	Storm tides
Marine	Tsunami	Tsunami	Tsunami
Pressure-related	Depression or cyclone	Strong Wind	Tropical cyclone
Pressure-related	Extra-tropical cyclone	Strong Wind	Extratropical cyclone
Pressure-related	Sub-tropical cyclone	Strong Wind	Tropical cyclone
Precipitation-related	Acid rain
Precipitation-related	Blizzard
Precipitation-related	Drought	Drought	Agricultural Drought
Precipitation-related	Hail
Precipitation-related	Ice storm
Precipitation-related	Snow / Ice
Precipitation-related	Snow storm
Temperature-related	Cold wave	Extreme Temperature	Extreme cold
Temperature-related	Dzud
Temperature-related	Freeze	Extreme Temperature	Extreme cold
Temperature-related	Frost (hoar frost)	Extreme Temperature	Extreme cold
Temperature-related	Freezing rain
Temperature-related	Glaze
Temperature-related	Ground frost
Temperature-related	Heatwave	Extreme Temperature	Extreme heat
Temperature-related	Icing (including ice)
Temperature-related	Thaw
Terrestrial	Avalanche	Landslide	Snow Avalanche
Terrestrial	Mud ﬂow	Landslide	Mudlfow
Terrestrial	Rock slide	Landslide	Rockslide
Wind-related	Derecho
Wind-related	Gale (strong gale)
Wind-related	Squall
Wind-related	Subtropical storm	Strong Wind	Extratropical cyclone
Wind-related	Tropical cyclone (cyclonic wind, storm surge)	Strong Wind	Tropical cyclone
Wind-related	Tropical storm	Strong Wind	Tropical cyclone
Wind-related	Tornado	Convective Storm	Tornado
Wind-related	Wind	Strong Wind	Extratropical cyclone
Seismogenic	Earthquake	Earthquake
Seismogenic	Ground shaking (earthquake)	Earthquake	Ground Motion
Seismogenic	Liquefaction (earthquake trigger)	Earthquake	Liquefaction
Seismogenic	Surface rupture/fissuring	Earthquake	Primary Rupture and Secondary rupture
Seismogenic	Subsidence and uplift, including shoreline change (earthquake trigger)
Seismogenic	Tsunami (earthquake trigger)	Tsunami	Tsunami
Seismogenic	Landslide or debris ﬂow (earthquake trigger)	Landslide	Landslide
Seismogenic	Ground gases (earthquake trigger)
Volcanogenic	Lava ﬂow	Volcanic	Lava
Volcanogenic	Ash/ Tephra fall (physical and chemical)	Volcanic	Ashfall
Volcanogenic	Ballistics (volcanic)	Volcanic	Ballistics
Volcanogenic	Pyroclastic density current	Volcanic	Pyroclastic Flow
Volcanogenic	Debris ﬂow/Lahar/Floods	Volcanic	Lahar
Volcanogenic	Landslide (volcanic trigger)	Landslide	Landslide
Volcanogenic	Ground shaking (volcanic earthquake)	Earthquake	Ground Motion
Volcanogenic	Ground gases (slow release) and Volcanic gases and aerosols (CH4, CO2, H2S, SO2 etc.)
Volcanogenic	Tsunami (volcanic trigger)	Tsunami	Tsunami
Volcanogenic	Lightning (volcanic trigger)
Volcanogenic	Urban fire (during/following volcanic eruption)
Volcanogenic	Subsidence and uplift, including shoreline change (magmatic/volcanic trigger)
Shallow geohazard	Ground shaking (induced earthquake, reservoir fll, dams, cavity collapse, underground explosion, impact, hydrocarbon felds, shale exploration, etc.)	Earthquake	Ground Motion
Shallow geohazard	Liquefaction (groundwater trigger)	Flood	Liquefaction
Shallow geohazard	Ground fissuring
Shallow geohazard	Subsidence and uplift, including shoreline change
Shallow geohazard	Ͱ Localized subsidence (e.g. due to swelling clays)
Shallow geohazard	ͰSinkhole
Shallow geohazard	Ground gases (CH4, Rn, etc.)
Shallow geohazard	Riverbank / Coastal erosion
Shallow geohazard	Sand encroachment
Shallow geohazard	Aquifer recharge (systems failure/ outages)
Shallow geohazard	Landslide or Debris ﬂow, including submarine landslides	Landslide	Landslide
Shallow geohazard	Rockfall	Landslide	Rockfall
Shallow geohazard	Landscape creep
Shallow geohazard	Avalanche	Landslide	Snow Avalanche
Shallow geohazard	Tsunami (submarine landslide trigger)	Tsunami	Tsunami
Environmental degradation (Forestry)	Wildfires	Wildfire	Wildfire

duncandewhurst commented 1 year ago

Proposing csv mapping file to be included somewhere in repo, to describe how RDL hazard_type and process_type map to UNDRR HIPS: Below updated from Mat's earlier sheet (#2 (comment)) (not updated for other taxonomies, but could do the same for those - next priority: Open Data Standards peril codes):

Sounds good. Since there might be more than one mapping, I suggest that we:

[x] Create a new mapping directory to put the CSV files in
[x] Add a new 'Taxonomy mappings' subsection under https://rdl-standard.readthedocs.io/en/dev/guides/
[ ] Add a page for each mapping under the new subsection with a link to the CSV file and a HTML version of the mapping table
[x] Link to the taxonomy mappings from the guidance on how to publish RDLS metadata drafted in https://github.com/GFDRR/rdl-standard/issues/149, e.g. "If your risk datasets use terms from existing taxonomies or classifications, use the [taxonomy mappings]() to identify the equivalent codes in RDLS".

(@stufraser1 edited to add tick boxes)

stufraser1 commented 1 year ago

new folder: https://github.com/GFDRR/rdl-standard/tree/2-addHazardMappings/schema/mappings contains RDL - UNDRR mapping (csv, with corrected spelling/spacing errors from table above). Created mappings page, with csv and link to original source, but not html.

odscjen commented 1 year ago

@stufraser1 I've taken what you'd done in https://github.com/GFDRR/rdl-standard/commit/a074aa224ec672f9aff2f7036692bf0b7c2c3aa1 and added it to the larger PR for updating the guidance pages as some of the guidance needs to reference this so its easier if it's all together

duncandewhurst commented 1 year ago

Should this mapping include the hazard identifiers assigned in the UNDRR-ISC Hazard Information Profiles? e.g. MH0001 for 'Downburst':

duncandewhurst commented 1 year ago

@stufraser1 to add identifiers to the CSV file

stufraser1 commented 8 months ago

Noting here a consideration to map RDLS hazard process codes to those used by EMDAT (IRDR 2014). This is not being done because EMDAT have a roadmap from 2023 to align their codes with UNDRR HIPS, which we are already aiming to align with https://doc.emdat.be/docs/data-structure-and-content/disaster-classification-system/

stufraser1 commented 8 months ago

@stufraser1 to add identifiers to the CSV file

See PR #266

GFDRR / rdl-standard