Skip to content

Latest commit

 

History

History
83 lines (68 loc) · 6 KB

tutorial.md

File metadata and controls

83 lines (68 loc) · 6 KB
Raw

Tutorial

  • To understand how to submit to the IIASA database, read this REMIND tutorial.
  • In the following, we differentiate templates (list of variables and corresponding units used in a project) and mappings (specifying which PIAM variable will be mapped to a project variable).

Mappings

Mappings found in the inst/mappings folder serve to map variables from the PIAM framework to variables needed for the submission to databases. The mappings are ;-separated files, using # as comment character, with the following mandatory columns:

  • variable: name of the variable in the project template
  • unit: unit corresponding to variable
  • piam_variable: name of the variable in REMIND / MAgPIE / EDGE-T etc. reporting
  • piam_unit: unit corresponding to piam_variable
  • piam_factor: factor with which the piam_variable has to be multiplied for units to match

Recommended column:

  • description: description text defining the variable. Never use " and ; in the text.
  • source: abbreviation of the PIAM part where the piam_variable comes from. Use B = Brick, C = MAGICC, M = MAgPIE, R = REMIND, S = SDP postprocessing, T = EDGE-Transport. This column is used to select the variables passed to remind2 and coupling tests. If the variable is not normally reported, add a small x after the model abbreviation for it to be skipped.

Additionally, some mappings use those columns:

  • idx: serial number of variable
  • Tier: importance of variable. 1 means most important
  • Comment: place for comments
  • weight: calculates a weighted average of multiple entries of piam_variable. Provide a different piam_variable in this column, and generateIIASASubmission() will split the data on the rows which contain weight pointers, resolve these weights into numerical values (via a join operation between the submission and the input data) and then modify the value based on the weighting. This takes place in the private .resolveWeights method.

To edit a mapping in R, use:

mappingdata <- getMapping("AR6")
...
write.csv2(mappingdata, "test.csv", na = "", row.names = FALSE, quote = FALSE)

Opening the csv files in Excel can be problematic, as it sometimes changes values and quotation marks. You can edit the files in LibreOffice Calc using these settings in the Text Import dialog:

  • Text Import with:
    • Character set: Unicode (UTF-8)
    • Separated by: Semicolon.
  • Save with:
    • Character set: Unicode (UTF-8)
    • Field Delimiter: ;
    • String Delimiter: (none)

The github diff on a large semicolon-separated file is often unreadable. For a human-readable output, save the old version of the mapping and run:

remind2::compareScenConf(fileList = c("oldfile.csv", "mappingfile.csv"), row.names = NULL)

Creating a new mapping

Since templates contain between several hundreds and a few thousand variables, relying on existing mappings can save substantial amounts of work compared to setting up a new mapping from scratch. Since the template itself is most likely built based on earlier templates from other projects, chances are good that existing mappings already provide parts of the required new mapping. Using R, we describe a simple way to create a new mapping mapping_NEW.csv based on existing mappings.

  1. Identify which existing mappings are most relevant for your new mapping. Criteria might include the time at which the existing mapping was created and the proximity of the templates (e.g. follow-up project). If you are unsure, ask your experienced colleagues for advice. This provides you with a list of existing mappings that is ordered by relevance, say mapping_OLD1.csv, ... , mapping_OLD9.csv.
  2. Use read.csv2 to get the template as a dataframe template.
  3. Looping over the existing mappings in descending order of relevance,
    • use getMapping to get the existing mapping mapping_OLDi as a dataframe,
    • filter mapping_OLDi for variables which are contained in template and which have not yet been added to mapping_NEW,
    • left_join (by variable) the filtered mapping_OLDi with template to add the information from the template (consider using str_to_lower for case-insensitive matching when filtering and joining),
    • select/mutate the columns of the joined dataframe to keep the desired columns for the new mapping (see above for description of mandatory and recommended columns),
    • bind_rows to mapping_NEW
  4. Use write.csv2 to export mapping_NEW as mapping_NEW.csv. It is recommended to make a few checks (e.g. by looking at all variables for which the description or the unit does not agree between the existing mapping and the template).

Model intercomparison

  • To compare the results of different models, pass as modeldata a quitte object or a csv/xlsx file. You get a PDF document for each scenario and each model with area plots for all the summation groups in AR6 (or NAVIGATE) summation files plus line plots for each variable in the lineplotVariables vector you supplied. It takes some time, better use a slurm job for:

    plotIntercomparison(modeldata, summationsFile = "AR6", lineplotVariables = c("Temperature|Global Mean", "Population"))

  • If your modeldata is not well filtered such that for example model regions are not too different, you can use interactive = TRUE which allows to select models, regions, scenarios and variables that you like in your PDF. As lineplotVariables, you can also specify mapping names.

    plotIntercomparison(modeldata, summationsFile = "AR6", lineplotVariables = c("AR6", "AR6_NGFS"), interactive = TRUE)