Open stufraser1 opened 3 months ago
matamadio
commented
3 months ago Agree, we need to add clear distinction between historical records and projections. It's implicit that when a future period is specified, than it's a projection. ATM we have the period attributes, and we only suggest to mention it by naming of the resource.
That opens the way for a number of additional attributes related to future projections, most importantly climate scenario, model, index, source, ...
stufraser1
commented
3 months ago Potentially then creating an object 'projection'
From: Mattia Amadio @.> Sent: Thursday, April 11, 2024 10:03:15 AM To: GFDRR/rdl-standard @.> Cc: Stuart Fraser @.>; Author @.> Subject: Re: [GFDRR/rdl-standard] [Proposal] Forecast data - label H, E, L datasets as forecasts (Issue #267)
Agree, we need to add clear distinction between historical records and projections. It's implicit that when a future period is specified, than it's a projection. ATM we have the period attributes, and we only suggest to mention it by naming of the resourcehttps://docs.riskdatalibrary.org/en/latest/guides/datasets/packaging/#thematic-grouping.
That opens the way for a number of additional attributes related to future projections, most importantly climate scenario, model, index, source, ...
— Reply to this email directly, view it on GitHubhttps://github.com/GFDRR/rdl-standard/issues/267#issuecomment-2049246341, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AC7PNYRPBIVTTY3ANND6EDTY4ZGVHAVCNFSM6AAAAABE72FBQKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDANBZGI2DMMZUGE. You are receiving this because you authored the thread.Message ID: @.***>
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?
Risk data could include simulated forecast data and forecasts can be deterministic or probabilistic. Consider how to label a dataset as a forecast. Extend this to include 'projections'?
Why is this not covered by the existing model?
We have included data that records the timeframe of a dataset (temporal object) but not included an explicit label for this case.
What is your proposed change?
Consider whether the following codelists are the appropriate place(s) to include 'forecast' as a value to label data. https://docs.riskdatalibrary.org/en/latest/reference/codelists/#data-calculation-type https://docs.riskdatalibrary.org/en/latest/reference/codelists/#analysis-type Neither type is mutually exclusive with 'forecast' however (forecast can be inferred or simulated, and deterministic or probabilistic. Is another codelist needed to place a hazard dataset / exposure dataset / loss dataset in time using e.g., 'forecast', 'projection', 'snapshot', 'historical'? Could this be achieved with the existing temporal object? It describes the time of event / dataset, but may not be clear enough for filter/search.
Forecast data may be kept for longer than the event duration as a record of forecast activity, or comparison with later retrospective modelling, so this is valid a use case for medium-long term data storage and sharing.
Can you provide an example?
TBC - but typically most attributes will be the same as a modelled event footprint, but temporal information would indicate the lead time (time ahead of event at which the forecast is made)