Add "scenario projection"

[stap-m]

Description of the issue

Add a concept for projection / scenario projection which focusses on the data of a "scenario", not on the assumptions.

Ideas of solution

it is related to data set that includes endogenuous data

Workflow checklist

• [ ] I discussed the issue with someone else than me before working on a solution
• [ ] I already read the latest version of the workflow for this repository
• [ ] The goal of this ontology is clear to me

I am aware that

• [ ] every entry in the ontology should have a definition
• [ ] classes should arise from concepts rather than from words

[han-f]

Thinking out loudly: A scenariois an information content entity that contains statements about a possible future development based on a coherent and internally consistent set of assumptions and their motivation. https://openenergy-platform.org/ontology/oeo/OEO_00000364

If we kept that, could we then say something along the lines: a projection is the quantification of a scenario by means of a model calculation. A projection covers exogenous data and endogenous data where exogenous data expresses the assumptions of the scenario quantitatively (e.g as input to the model calculation)

Alternatively we could also use scenario projection instead of projection

[l-emele]

That is a good start for a definition. However, we don't have a concept of quantification in the OEO. To me there are two possible interpretations what you could have meant with:

• Either quantification is something like a data set (A data item data set[^1] that is an aggregate of other data items of the same type that have something in common. Averages and distributions can be determined for data sets.)
• Or quantification is the process of doing something in which the scenario is involved.

[^1]: IAO seems to have here a mistake in the definition.

[han-f]

Thanks for these thoughts and here come my initial follow-up thoughts (not as structured as yours though): Is a projection maybe the process of quantifying a scenario. As a result of this quantification we receive outputs (data sets, endogenous data).

Also the inputs to a projection (data set, exogenous data) may have undergone a quantification process to begin with? E.g. by translating assumptions that are qualitatively described in the scenario into quantitative data that is then used in the model calculation.

[l-emele]

Can we describe a projection as an umbrella process that covers these steps?

1. The translation of assumptions into data sets.
2. The data prepocessing
3. The model calulation(s)
4. The data postprocessing

Further steps that might be also part of the projection (process):

5. Plotting of results into diagrams
6. Describing the scenario in the study report.

So a projection would have these parts and would itself be a party of a study. (A study is a project with the goal to investigate something.) Am I right?

[han-f]

Sounds really good to me. What do the others think? @Ludee, @jannahastings, @stap-m ?

[stap-m]

I am hesitating with classifying projection as process. I'd interpret it rather as the output of a process. This would also fit better to @han-f s draft for a definition.

[stap-m]

I read again some sources about the wording:

• The acatech study Zur Interpretation von Energieszenarien says "Synonym zum Begriff „Szenario“ wird gelegentlich auch die Bezeichnung „Projektion“ verwendet, etwa im „World Energy Outlook 2013." p.9 They further distiguish between
  • "Ein Szenario ist ein sprachliches Gebilde, das als Aussagensystem rekonstruiert werden kann.[...]"
  • "In einer Szenarioanalyse werden ein oder mehrere solcher Szenarien generiert und analysiert. Idealtypisch lässt sich die Analyse isolierter Szenarien von der vergleichenden Szenarioanalyse unterscheiden.[...]"
  • "Als eine Szenariostudie wird im Folgenden die Textgattung bezeichnet, in der Szenarien und Szenarioanalysen präsentiert werden.[...]" p. 10
• The enargus ontology doesn't use "projection" at all. There's just the term "Szenariorechnung", with "Szenarien-Analyse" as synonym. Nothing else :(
• The IPCC uses the following distinction in the TAR chapter 13 on Climate Scenario Development: "A climate scenario is a plausible representation of future climate that has been constructed for explicit use in investigating the potential impacts of anthropogenic climate change. Climate scenarios often make use of climate projections (descriptions of the modelled response of the climate system to scenarios of greenhouse gas and aerosol concentrations), by manipulating model outputs and combining them with observed climate data. "
• For some scenarios, the term "outlook" is used, e.g. World Energy Outlook.

Probably, there're more (English) sources we could take into account. I think, we should define and add all of these terms, at least as synonyms of each other.

Finally, we should add (and contrast) the term "forecast".

[l-emele]

In my understanding, a forecast is a statement about the future predicting what will actually happen. In contrast, a scenario or projection is a statement about the future what will happen under certain assumptions.

[han-f]

I add another source on (climate) projection which may inspire how to generalise for our definition:

• Climate projection A projection of the response of the climate system to emissions or concentration scenarios of greenhouse gases and aerosols, or radiative forcing scenarios, often based upon simulations by climate models. Climate projections are distinguished from climate predictions in order to emphasize that climate projections depend upon the emission/ concentration/radiative-forcing scenario used, which are based on assumptions concerning, e.g., future socioeconomic and technological developments that may or may not be realized and are therefore subject to substantial uncertainty. Source: p.557: https://archive.ipcc.ch/pdf/special-reports/srex/SREX-Annex_Glossary.pdf

This also includes confirmation on what @l-emele wrote about forecast, here called prediction.

[l-emele]

This climate projection seems to a special kind (= subclass) of a more general projection.

[han-f]

Coming back to thinking whether a projection is a process or not, while writing on the typology paper. I thought the following (apologies not very structured), maybe take this aboard in the oeo dev meeting on projection: Doesn't a model calculation project a scenario (narrative) into quantitative values (output = endogenous data) So could we see projecting as a process that involves a scenario, model, exogenous data model calculation and produces some output (endogenous data)? And potentially all these elements together make up a projection?

And in contrast to a forecast projection includes exogenous data (the assumptions). So a forecastcould be made up of the same elements excluding however exogenous data? The corresponding process could then be forecasting?

[l-emele]

I cannot follow the reasoning that forecast do not use exogenous data. Examples of forecasts that come to my mind are weather forecast, oil price forecast and forecast of energy production from fluctuating renewable energy sources like wind and photovoltaic. All of these extensively use exogenous data.

[han-f]

Hm - this is correct, maybe we need to distinguish between different types of exogenous data? Thinking: Oil price forecast probably uses input oil prices and also outputs oil prices, right? So in this case the oil price is exogenous and endogenous at the same time.

May we need to distinguish exogenous data further, for example exogenous data that are empiric and exogenous data that are explicit assumptions made by the modellers (e.g. demand elasticities, future renovation rates, future CO2 prices ...)? Would this be a possibility to work out a distinction between what is used for forecasts and what is used for projections?

[l-emele]

I don't think, that you can distinguish forecast and scenario by the type of exogenous data that you use. The difference is more, what you do with the data. Furthermore, one way of producing forecasts is to do a lot of scenarios (i.e. model runs with slightly different assumptions on various parameters) and then statistically analysing the results of the scenarios. The median or mean of the scenario results typically then is the forecast. In the case of weather forecast: Some more nerdy weather websites not only show you the forecast itself but also the scenarios itself (or the range of the scenarios). Here is an example including description (scroll down to "14-Tage-Wetter").

[l-emele]

Btw, we already have a distinction between:

• assumption: An assumption is an information content entity about a property of a system or process. It determines a part of a scenario content.
• exogenous data: Exogenous data is a data item whose quantity value is determined outside of a model and is imposed on a model.

So the exogenous data can be about some assumption, but also about about different things like historic data. Do you think that we need a special class, something like exogenous assumption data?

[han-f]

Thanks. That is a good question. If we can say that exogenous data is about an assumption, maybe the current setup is sufficient already.

[han-f]

Can we describe a projection as an umbrella process that covers these steps?

1. The translation of assumptions into data sets.

2. The data prepocessing

3. The model calulation(s)

4. The data postprocessing

Further steps that might be also part of the projection (process): 5. Plotting of results into diagrams 6. Describing the scenario in the study report.

So a projection would have these parts and would itself be a party of a study. (A study is a project with the goal to investigate something.) Am I right?

Meeting @stap-m, @l-emele and @han-f May 09, 2022:

We think this is a good approach and would add a very first step to this process that is scenario-creation or scenario-selection that describes that before going through the process of projecting a scenario a scenario needs to be present - and thus either created or selected:

0. Scenario-creation / scenario-selection (input to the projection process)
1. The translation of assumptions into data sets.
2. The data prepocessing
3. The model calculation(s)
4. The data postprocessing

Further steps that might be also part of the projection (process): 5. Plotting of projection results into diagrams 6. Describing the scenario-creation / scenario-selection and the projection process in the study report.

[han-f]

Btw, we already have a distinction between:

* `assumption`: _An assumption is an information content entity about a property of a system or process. It determines a part of a scenario content._

* `exogenous data`: _Exogenous data is a data item whose quantity value is determined outside of a model and is imposed on a model._

So the exogenous data can be about some assumption, but also about about different things like historic data. Do you think that we need a special class, something like exogenous assumption data?

Meeting @stap-m, @l-emele and @han-f May 09, 2022:

Assumptions (which can be very general) are "translated" into exogenous data that are used in a model calculation. So the current setup is OK. We want to add the axiom: exogenous data is about some assumption.

[han-f]

Listening into oeo-dev on 2022-05- 19: of course one can project other things than scenarios, thus I suggest to change the concept discussed here to scenario-projection, changing the title of the issue accordingly.

[l-emele]

Assumptions (which can be very general) are "translated" into exogenous data that are used in a model calculation. So the current setup is OK. We want to add the axiom: exogenous data is about some assumption.

Maybe we can make it a subclass of exogenous data and define it as XYZ is exogenous data that is about some assumptions. XYZ EquivalentTo 'exogenous data' and 'is about' some assumption. We might label XYZ as input data, then we would have a nice parallel class to the already existing output data. But maybe there is an even better label for such a class.

[l-emele]

From OEO DEV meeting 39:

Scenario projection editor note / example: A scenario projection is an intentional process (with human participation): A scenario is either created or selected, its assumptions are translated into data sets. These datasets are quantified and serve as inputs to a model calculation which is applied to quantitatively project one or more a variables of interest into the future. (Intentional: a research question is basis for the projection to be done)

Suggested definition: A scenario projection is an intentional process in which output (endogenous) data of interest are quantified for future points in time using one or more model calculations based on a scenario.

• Discussion: should we broaden the definition to also include back-casting projections of the past ?
• Back-casting is a separate entity for the ontology, not the same one as scenario projection, could be added, but should then have a different definition. Scenario is strictly future defined.
• Is 'applied to a scenario' the best formulation for what is being done? We apply model calculations to assumptions that are expressed in a scenario.
  • used 'based on' rather than 'applied to'
• Scenario projections are tied to just a single scenario at a time. If you take multiple scenarios into consideration, you still have to first combine them to form an additional, separate, composite, single scenario, before making a scenario projection. We could add a separate class - scenario creation, where we describe this process.
  • Do we want the 'based on' relationship to extend from the specific scenario on which the projection is based, to the other scenarios that this scenario was based on?
  • Complicated, because different assumptions vary very specifically.
  • The additional relationships between scenarios that are combined is not 'based on' but maybe something like 'derived from'??? (derived from in this case would have at least two sub-properties - 'alteration of' and 'combination of')
  • If these properties between scenarios (perhaps for the knowledge graph) we should log a separate issue for them.