EcoSIM Output Variable List

Purpose

Use this skill to rebuild the catalog of EcoSIM model output variables registered in Fortran history files. The source of truth is call hist_addfld...(...) registration calls, especially in HistDataType.F90.

The catalog is meant to guide downstream EcoSIM output processing: variable selection, dimension handling, aggregation interpretation, unit checks, and NetCDF metadata review.

Workflow

1. Locate the selected .F90 source. If the user gives a path, use it. If no path is given, search data/ first, then the repository, for HistDataType.F90 or files containing hist_addfld.
2. Create an output folder under result/, for example:

mkdir -p result/ecosim_output_variables

3. Run the bundled extractor. Prefer JSON for automated processing and CSV for quick review:

python3 .agents/skills/ecosim-output-variable-list/scripts/extract_hist_output_vars.py /absolute/path/to/HistDataType.F90 --format json --output result/ecosim_output_variables/ecosim_output_variables.json
python3 .agents/skills/ecosim-output-variable-list/scripts/extract_hist_output_vars.py /absolute/path/to/HistDataType.F90 --format csv --output result/ecosim_output_variables/ecosim_output_variables.csv

4. Inspect the summary. Report total fields, counts by hist_addfld* routine, counts by rank, the number of dynamic fname expressions, and any duplicate literal variable names.
5. Preserve dynamic names as templates instead of inventing expansions. For example, names built from trim(micpar%cplxname(jj)) or trim(fieldname) should remain marked with is_dynamic_fname=true unless the loop values are explicitly available from source or runtime configuration.
6. Use the catalog to guide processing code. Preserve the registered units, avgflag, type2d, pointer_kind, pointer_target, long_name, condition_context, and loop_context fields.

Output Contract

Each extracted variable record includes:

• source_file, source_line, end_line: where the history registration call appears.
• addfld_subroutine: the registration routine, such as hist_addfld1d or hist_addfld2d.
• rank: derived from the routine name when possible (1d, 2d, etc.).
• fname: literal variable name when available; otherwise a readable template.
• fname_expr: raw Fortran expression for fname.
• is_dynamic_fname: true when fname is assembled from expressions.
• units, avgflag, type2d, default, long_name: named arguments from the registration call.
• pointer_arg, pointer_kind, pointer_expr, pointer_target: pointer metadata used to infer whether the output is column-, patch-, or other domain-oriented.
• condition_context, loop_context, enclosing_subroutine: lightweight source context for conditional and loop-registered variables.

Interpretation Rules

• Do not include commented-out hist_addfld calls in the active variable list.
• Treat avgflag as EcoSIM history metadata and preserve it verbatim. Do not change temporal units solely from avgflag; check whether units already encode rates (/hr, /s) or interval totals (gC/m2, mm H2O/m2).
• Treat type2d as the vertical or secondary dimension key for 2-D history variables, such as levsoi, levcan, node, rootaxs, elements, or pmorphunits.
• Flag unit inconsistencies for review, especially carbon, nitrogen, phosphorus, gas, water, and heat variables where g, mol, umol, m2, m3, hr, and s conversions can create mass-balance errors.
• Do not edit the Fortran source unless the user asks for code changes. This skill updates the output-variable catalog derived from the source.

Reporting

When reporting the update, include the catalog paths, record counts, dynamic-name examples, duplicate-name warnings if any, and the selected F90 source path. Keep the output focused on how the catalog should guide EcoSIM model-output processing.