# README - E3SM Genesis Shared Dataset This directory contains a subset of E3SM initial-condition files and historical model output prepared for the Genesis project. The current dataset is based on an E3SM S2D initialization at: ```text 2018-05-01 00:00 UTC ``` The shared files include atmosphere, land, ocean, sea-ice, river-routing, and coupler components. For the Genesis machine-learning workflow, the two main data groups have distinct roles: ```text initial_condition/ -> correction target -> input to the drift prediction model model_output/ -> historical E3SM prediction -> source used to construct drift labels ``` ## Directory Structure ```text genesis_share/ |-- initial_condition/ | |-- atm/ | |-- cpl/ | |-- ice/ | |-- lnd/ | |-- ocn/ | `-- rof/ `-- model_output/ |-- atm/ |-- ice/ |-- lnd/ |-- ocn/ `-- rof/ ``` # 1. Initial Conditions The `initial_condition/` directory contains the E3SM model states used to initialize the coupled prediction. In the Genesis machine-learning workflow, these initial-condition files serve two primary purposes: 1. **Correction target** - the initialized state is the state that the Genesis correction framework aims to adjust. 2. **Input to the drift prediction model** - the same initialized state is used as input to the machine-learning model to predict subsequent initialization drift. Conceptually: ```text Initial Condition | +------------------------------+ | | v v Correction Target Input to Drift Prediction Model | v Predicted Drift | v Corrected Initial Condition ``` The initialization time for the files in this dataset is: ```text 2018-05-01 00:00 UTC ``` ## 1.1 Atmosphere - EAM Directory: ```text initial_condition/atm/ ``` Atmospheric initial conditions are provided for 10 ensemble members: ```text v3.LR.amip_0101.HICCUP.atm_era5.EN00.eam.i.2018-05-01-00000.nc v3.LR.amip_0101.HICCUP.atm_era5.EN01.eam.i.2018-05-01-00000.nc v3.LR.amip_0101.HICCUP.atm_era5.EN02.eam.i.2018-05-01-00000.nc v3.LR.amip_0101.HICCUP.atm_era5.EN03.eam.i.2018-05-01-00000.nc v3.LR.amip_0101.HICCUP.atm_era5.EN04.eam.i.2018-05-01-00000.nc v3.LR.amip_0101.HICCUP.atm_era5.EN05.eam.i.2018-05-01-00000.nc v3.LR.amip_0101.HICCUP.atm_era5.EN06.eam.i.2018-05-01-00000.nc v3.LR.amip_0101.HICCUP.atm_era5.EN07.eam.i.2018-05-01-00000.nc v3.LR.amip_0101.HICCUP.atm_era5.EN08.eam.i.2018-05-01-00000.nc v3.LR.amip_0101.HICCUP.atm_era5.EN09.eam.i.2018-05-01-00000.nc ``` These files provide atmospheric initial conditions for ensemble members EN00 through EN09. The atmospheric initial conditions are based on ERA5-derived EAM initial states. ## 1.2 Coupler Directory and file: ```text initial_condition/cpl/ 20251024_s2d_spinup.cpl.r.2018-05-01-00000.nc ``` This file provides the coupler restart state for the 2018-05-01 initialization. ## 1.3 Land - ELM Directory and file: ```text initial_condition/lnd/ 20251024_s2d_spinup.elm.r.2018-05-01-00000.nc ``` This file provides the ELM land-model restart state. ## 1.4 Ocean - MPAS-Ocean Directory and file: ```text initial_condition/ocn/ 20251024_s2d_spinup.mpaso.rst.2018-05-01_00000.nc ``` This file provides the MPAS-Ocean restart state. ## 1.5 Sea Ice - MPAS-Seaice Directory and file: ```text initial_condition/ice/ 20251024_s2d_spinup.mpassi.rst.2018-05-01_00000.nc ``` This file provides the MPAS-Seaice restart state. ## 1.6 River Routing - MOSART Directory and file: ```text initial_condition/rof/ 20251024_s2d_spinup.mosart.r.2018-05-01-00000.nc ``` This file provides the MOSART river-routing restart state. The non-atmospheric components currently use a common coupled spin-up state for this initialization date, while the atmospheric component includes 10 ensemble members. # 2. Model Output The `model_output/` directory contains historical E3SM predictions generated after initialization. In the Genesis machine-learning workflow, these files serve two primary purposes: 1. **Historical model prediction output** - they provide the E3SM prediction trajectory following initialization. 2. **Source for constructing drift labels** - the historical model predictions are compared with the appropriate reference data to diagnose initialization drift and construct drift labels for machine-learning training and evaluation. Conceptually: ```text Initial Condition | v Historical E3SM Prediction | + Reference Data | v Diagnosed Initialization Drift | v Drift Labels | v Train and Evaluate Drift Prediction Model ``` The current model output is from the BruteForce/Reanalysis initialization experiment starting at: ```text 2018-05-01 00:00 UTC ``` The files currently provided are for ensemble member EN00 and contain monthly output for May 2018. ## 2.1 Atmosphere - EAM Directory and file: ```text model_output/atm/ WCYCL20TR_ne30pg2_r05_IcoswISC30E3r5_BruteForce_2018050100.EN00.eam.h0.2018-05.nc ``` This file contains monthly atmospheric history output from EAM. ## 2.2 Land - ELM Directory and file: ```text model_output/lnd/ WCYCL20TR_ne30pg2_r05_IcoswISC30E3r5_BruteForce_2018050100.EN00.elm.h0.2018-05.nc ``` This file contains monthly land-model history output from ELM. ## 2.3 Ocean - MPAS-Ocean Directory and file: ```text model_output/ocn/ WCYCL20TR_ne30pg2_r05_IcoswISC30E3r5_BruteForce_2018050100.EN00.mpaso.hist.am.timeSeriesStatsMonthly.2018-05-01.nc ``` This file contains monthly MPAS-Ocean time-series statistics. ## 2.4 Sea Ice - MPAS-Seaice Directory and file: ```text model_output/ice/ WCYCL20TR_ne30pg2_r05_IcoswISC30E3r5_BruteForce_2018050100.EN00.mpassi.hist.am.timeSeriesStatsMonthly.2018-05-01.nc ``` This file contains monthly MPAS-Seaice time-series statistics. ## 2.5 River Routing - MOSART Directory and file: ```text model_output/rof/ WCYCL20TR_ne30pg2_r05_IcoswISC30E3r5_BruteForce_2018050100.EN00.mosart.h0.2018-05.nc ``` This file contains monthly MOSART river-routing output. # 3. Data Roles in the Genesis Workflow The relationship between the shared initial-condition files and model output can be summarized as: ```text +---------------------------+ | Initial Condition | +---------------------------+ | | | | v v Correction Target Drift Model Input | v Predicted Drift Initial Condition | v Historical E3SM Prediction | + Reference Data | v Observed/Diagnosed Drift | v Drift Labels ``` In summary: ```text initial_condition/ - state to be corrected - input features for drift prediction model_output/ - historical E3SM prediction - source data for diagnosing drift - source data for constructing drift labels ``` # 4. Experiment Naming The model-output case name is: ```text WCYCL20TR_ne30pg2_r05_IcoswISC30E3r5_BruteForce_2018050100 ``` Important parts of the naming convention include: ```text 2018050100 Initialization at 2018-05-01 00 UTC EN00 Ensemble member 00 BruteForce Reanalysis/BruteForce initialization experiment ``` The simulation uses the coupled E3SM atmosphere-land-ocean-sea-ice-river system. # 5. Current Dataset Coverage The current shared dataset includes: ```text Initial conditions: Atmosphere EN00 through EN09 Coupler common restart state Land common restart state Ocean common restart state Sea ice common restart state River routing common restart state Model output: Atmosphere EN00 Land EN00 Ocean EN00 Sea ice EN00 River routing EN00 ``` The provided model-output files currently contain monthly output for May 2018. # 6. NetCDF Files All files are in NetCDF format. File metadata and variable information can be inspected using: ```bash ncdump -h FILE.nc ``` or NCO: ```bash ncks -m FILE.nc ``` For example, to inspect an atmospheric initial-condition file: ```bash ncdump -h \ initial_condition/atm/v3.LR.amip_0101.HICCUP.atm_era5.EN00.eam.i.2018-05-01-00000.nc ``` To inspect the atmospheric model output: ```bash ncdump -h \ model_output/atm/WCYCL20TR_ne30pg2_r05_IcoswISC30E3r5_BruteForce_2018050100.EN00.eam.h0.2018-05.nc ``` # 7. Intended Use These files are shared primarily for development and testing associated with the Genesis project. Potential uses include: * inspecting E3SM coupled initial states; * preparing machine-learning inputs from E3SM initial conditions; * developing initialization-correction methods; * predicting initialization drift from the initialized model state; * analyzing historical E3SM prediction trajectories; * diagnosing initialization adjustment and drift; * constructing drift labels using model predictions and reference data; * training and evaluating drift-aware machine-learning models; * comparing initialized simulations with reference datasets; * testing workflows before scaling to the larger E3SM S2D hindcast archive. Users should preserve the original shared files and create separate copies or derived datasets for processing whenever possible.