[DATA] CCDR risk analytics

[matamadio]

CCDR risk analytics

Data produced for the CCDR reports by M. Amadio and A. Essenfelder.

Countries (9)

South & South-East Asia	West Africa & Sahel countries	Central & East Africa	Latin America & Caribbean
Pakistan Nepal Bangladesh Cambodia India Thailand	Ghana Burkina Faso Mali Niger Chad Mauritania Guinea-Bissau Nigeria Senegal	Ethiopia Rwanda	Dominican Republic OECS countries

Data types

• Hazard datasets (global)

  • Floods: Fathom 2019 (v2)
  • Coastal floods: Aqueduct, Deltares
  • Landslide: ARUP 2018
  • Drought: FAO ASI season 1 and 2 for >30% and 50% area affected
  • Cyclones: GAR15, STORM Wind speed 2020, IBTrACSv4
  • Heat stress: VITO WBGT 2016, UTCI 2020
  • Air pollution: ACAG PM2.5 2020

• Exposure datasets (global)

  • Population: Worldpop 2020, GHS 2020
  • Built-up: WSF 2019, WSF 3D, GHS-S
  • Land cover: ESA World cover 2020-2021

• Vulnerability models (global):

  • Floods: JRC depth-damage curves (built-up); Jonkman et al. 2008 (mortality function)
  • Heat stress: standard index classification (Blazejczyk et al. 2012)
  • Air pollution: WHO Air Quality guidelines (2021)
    Models and thresholds are included in the report annexes and in the open python notebooks released alongside.

• Loss (risk)

  • Expected Annual Impact (EAI) as combination of probabilistic hazard models, exposure data and vulnerability functions
  • Expected Annual Exposure (EAE) as combination of hazard models, exposure data and vulnerability classes

Type	Format	Description
Hazard data	Raster (tiff)	Multiple or individual layers packages on various resolutions - global scale, not clipped for the countries, except Fathom
Exposure data	Raster (tiff)	Individual raster layers
Vulnerability data	Excel file	Noted excel file including all models used in the assessment
Loss data	Excel file, vector (gpkg)	Output including impact for individual RPs and EAI/EAE calculations, indexing, charts, fields legend

All global data uses CRS: 4326

Data samples

Exposure

Hazard

Vulnerability

Loss

[matamadio]

While the discussion about loss schema is ongoing, the other CCDR components can be included.
The question is how to add them. @stufraser1 please check the proposal:

Global hazard data

The files with global coverage should have just one global entry, instead of multiplicating for each country/extent.

• Already on the catalog, need to add RDLS metadata:
  • Flood (FATHOM) (Note there is no direct linkage to the zip, we should link to the form request)
  • Landslide (ARUP)
  • Heat stress (VITO)
• To be added on the catalog:
  • Coastal floods:
    • Aqueduct
    • Deltares on NASADEM90m.
  • Drought: FAO ASI season 1 and 2 for >30% and 50% area affected
  • Cyclones:
    • GAR15
    • STORM Wind speed 2020
    • IBTrACSv4
  • Air pollution: ACAG PM2.5 2020

Global exposure data

In the case of exposure, the data are often downloaded as tiles and then processed (country mosaic and clip; resampling in some case). In some cases (WSF3D), we also have the global extent datasets, but in a very large file (>2 Gb).
Thus, differently from Hazard layers, I would provide the national (analysis) extent for these data, uploaded on our catalog and with reference to source models.

Global vulnerability data

These are global model with minimal regionalisation, so they should be just one global entry for each.

[matamadio]

Please check metadata sheet for global hazard layers:

• Fathom Flood (v2): needs to add license type (commercial or custom) as in [Schema] Commercial licenses? #183
• STORM Tropical cyclone wind: being added to DDH/RDL (under review)
• Deltares global coastal flood maps: being added to DDH/RDL (under review)

[stufraser1]

For Fathom Flood: Only one request: Please remove values from hazard/event_sets/0/event_count per this field description "You should only use this field when details of individual events are not included in the RDLS metadata."
In reality the no. of events in the modelled hazard numbers many more, so including the number of return period layers shown is misleading.

STORM:

• check attributions/0/entity/url for Nadia - includes fathom URL incorrectly
• Same comment as above wrt hazard/event_sets/0/event_count
• URL (https://data.4tu.nl/datasets/504c838e-2bd8-4d61-85a1-d495bdc560c3/3) is missing from resources/0/download_url
• There is an argument for including the different climate model projections as different event_sets. The suggested URL above refers to the general dataset page where there are multiple model runs and median dataset. We could go further to define all, but not necessary for this demonstration maybe -- or define that we're referring to the median with these resources (if thats what you want to refer to).

[matamadio]

Thanks Stu, see fixed:

• Fathom Flood (v2): event count deleted.
• STORM Tropical cyclone wind: awaiting publishing on DDH. Event count deleted and urls fixed. I have the original download page link in access url and the direct link to zip in DDH to add as download url. I specified it's the median of projections (cause I only added that to DDH) but yes, it would be good to include all climate models.

Related to this: I tried to upload a couple of datasets (deltares global flood, STORM) directly to DDH to produce the download link for metadata. However there seems to be issues with data upload from external networks (outside intranet); does it happen to you as well?

[matamadio]

Adding for review:

• ARUP Landslide hazard

[stufraser1]

For ARUP Landslide:

• I understand why you've added 'convective storm' as trigger, but this might be a limitation we need to address: rainfall is the trigger, but not necessarily only from convective storm.
• IMT is given as ls_hzd:-; We need to add a dimensionless landslide hazard index as maybe ls_HI:- but this data gave annual frequency so we also need to reference that too, can I suggest: ann_freq:-
• 'hazard/event_sets/0/events/0/occurrence/deterministic/thresholds': I think there is a copy-down error here - please confirm. Also remind me what the thresholds refer to please.

[matamadio]

* I understand why you've added 'convective storm' as trigger, but this might be a limitation we need to address: rainfall is the trigger, but not necessarily only from convective storm.

Yes, agree on principle. The list of triggers is however the same of hazards and processes; proposing to add "extreme rainfall" or "heavy precipitation" to process list. Usefull for floods anyway.

* IMT is given as `ls_hzd:-`; We need to add a dimensionless landslide hazard index as maybe `ls_HI:-` but this data gave annual frequency so we also need to reference that too, can I suggest: `ann_freq:-`

From the ARUP report:

It's a composite index that considers frequency and magnitude, but yes, I would add annual_freq; not just for LS, but applicable to all hazards. Please also see comment I left on imt.csv PR:

It's not always useful to specify the hazard type in the metric. In some cases it just create unnecessary duplication when the same metric:unit could be used, e.g.

• Why we need to differentiate FL from LS when speaking of flow depth, velocity?
• How tsunami inundation depth is different from FL or LS "flow depth"?
• Why we need to indicate "etc" (extra-tropical cyclones) for strong wind unit?

My proposal is to have a first section of general imts (meters, km/h, m/s, annual frequency, and others).
Following, a list of more hazard-specific imts (like PGA:g and others).

• 'hazard/event_sets/0/events/0/occurrence/deterministic/thresholds': I think there is a copy-down error here - please confirm. Also remind me what the thresholds refer to please.

Thanks, yes copy-down error occurred. Thresholds are the same for all three layers.

Index value	Hazard class
> 0.01	High
0.001 – 0.01	Medium
< 0.001	Low