FRIDA v3.1 Endogenous Model Behavior (EMB) 100000 member ensemble digest

# FRIDA v3.1 baseline (EMB) ensemble, 100 000 samples: run digest This record holds the digest of the FRIDA v3.1 baseline ensemble (EMB), a 100 000 sample runof the FRIDA uncertainty analysis. EMB is the reference run of the FRIDA v3.1 overview paper: William Schoenberg, Benjamin Blanz, Jefferson K. Rajah, Beniamino Callegari, ChristopherWells, Theresia B. Putranti, Francisco Mahu, Maria Molina, Alexandre Köberle, Catherine Li,Wanderson Costa, David Collste, Karel Zwetsloot, Jannes Breier, Lennart Ramme, Chris Smithand Cecilie Mauritzen: *An overview of FRIDA v3.1: An update to a feedback-based, fullycoupled, global integrated assessment model of climate and humans.* The same digest is part of the figure package of that paper. ## The run - Run name: `UA-v3-1-2026-09-14-S100000-policy_EMB-ClimateFeedback_On-ClimateSTAOverride_Off`- Scenario: the EMB policy file (`policy_EMB.csv`), climate feedbacks on, no surface temperature override.- Samples: 100 000 parameter draws. 85 312 (85.3 %) ran to completion; the others stopped early. `runStatus.csv` and the "run completion" section of `runMetadata.txt` give the failure year of each.- Model: [WorldTransFRIDA](https://github.com/metno/WorldTransFRIDA), branch `development`, commit 89b3979724e524f53909872fa160f3545ff1c015, run through the Stella Simulator.- Analysis scripts: [WorldTransFrida-Uncertainty](https://github.com/BenjaminBlanz/WorldTransFrida-Uncertainty), commit 947baab0ddcb54e90272b378f8c86c1efa113192 with local changes. The scripts and configuration exactly as they ran are in `runScriptsAndConfiguration/`.- Run on DKRZ Levante in September 2026. The representative samples and the plot data were redrawn on 2026-09-24 (`LOG.repSampleRepair*.log`). ## What a digest is The full run holds about 38 GB, nearly all of it the per variable output of every sample.The digest keeps everything else, about 540 MB: - `runMetadata.txt`: model and script versions, config settings, run completion, and checksums of the model files as they ran.- `samplePoints.csv.gz`: the parameter values of all 100 000 samples.- Parameter space and calibration: `sampleParmsParscale*.csv/.RDS`, `parscale.RDS`, `sigma*.RDS`, `calDat.RDS`, `frida_info_ranged.csv`, `Calibratio_Data_Cleaned_and_Transposed.csv`.- `repSample/`: the representative samples (11 and 101 members) drawn from the ensemble.- `figures/CI-plots/completeEquallyWeighted/plotData/`: the ensemble's per year uncertainty ranges and default run for 444 model variables, as RDS and CSV. These are what the paper's figures are drawn from.- `runScriptsAndConfiguration/`: the analysis scripts, config and input files of the run.- The run's logs and status files. Left out: `detectedParmSpace/`, the per variable output of every sample (36.4 GB), and thefigure images (221 MB). `digest.txt` names the source run, what was left out, and the md5 of every file. The digestwas written with `runMakeDigest.R` of WorldTransFrida-Uncertainty, commit19f135f8a4de3b62a984c6738fbdc7eade26c6d4, and every file in it was checked against thesource run. ## Using it Unpack with `tar -xzf <file>.tar.gz`. The folder can stand in for the full run wherever onlythe plot data, sample points or representative samples are read. To use it in place ofthe run, point the scripts' data location at the folder that holds it; the figure scripts of thepaper take the run where it exists and the `-digest` folder otherwise. ## Licence Creative Commons Attribution 4.0 International (CC BY 4.0). If you use this data, pleasecite the paper above.