CoupledModelDriver

coupledmodeldriver generates an overlying job submission framework and configuration directories for NEMS-coupled coastal ocean model ensembles.

It utilizes nemspy to generate NEMS configuration files, shares common configurations between runs, and organizes spinup and mesh partition into separate jobs for dependant submission.

Supported models and platforms

• models
  • circulation models
    • ADCIRC (uses adcircpy)
  • forcings
    • ATMESH
    • WW3DATA
• platforms
  • local
  • Slurm
    • Hera
    • Stampede2

Usage

Example scripts can be found at examples/<platform>

1. write a configuration to a directory

The following code (examples/nems_adcirc/hera_shinnecock_ike_spinup_tidal_atmesh_ww3data.py) creates a configuration for coupling (ATMESH + WW3DATA) -> ADCIRC on Hera, over a small Shinnecock Inlet mesh:

#! /usr/bin/env python

from datetime import datetime, timedelta
from pathlib import Path

from adcircpy.forcing.tides import Tides
from adcircpy.forcing.tides.tides import TidalSource
from adcircpy.forcing.waves.ww3 import WaveWatch3DataForcing
from adcircpy.forcing.winds.atmesh import AtmosphericMeshForcing

from coupledmodeldriver import Platform
from coupledmodeldriver.generate import NEMSADCIRCGenerationScript, NEMSADCIRCRunConfiguration

# paths to compiled `NEMS.x` and `adcprep`
NEMS_EXECUTABLE = '/scratch2/COASTAL/coastal/save/shared/repositories/ADC-WW3-NWM-NEMS/NEMS/exe/NEMS.x'
ADCPREP_EXECUTABLE = '/scratch2/COASTAL/coastal/save/shared/repositories/ADC-WW3-NWM-NEMS/ADCIRC/work/adcprep'

MODULES_FILENAME = '/scratch2/COASTAL/coastal/save/shared/repositories/ADC-WW3-NWM-NEMS/modulefiles/envmodules_intel.hera'

# directory containing input ADCIRC mesh nodes (`fort.14`) and (optionally) mesh values (`fort.13`)
MESH_DIRECTORY = Path('/scratch2/COASTAL/coastal/save/shared/models') / 'meshes' / 'shinnecock' / 'v1.0'

# directory containing input atmospheric mesh forcings (`wind_atm_fin_ch_time_vec.nc`) and WaveWatch III forcings (`ww3.Constant.20151214_sxy_ike_date.nc`)
FORCINGS_DIRECTORY = Path('/scratch2/COASTAL/coastal/save/shared/models') / 'forcings' / 'shinnecock' / 'ike'

# directory to which to write configuration
OUTPUT_DIRECTORY = Path(__file__).parent / Path(__file__).stem

HAMTIDE_DIRECTORY = '/scratch2/COASTAL/coastal/save/shared/models/forcings/tides/hamtide'
TPXO_FILENAME = '/scratch2/COASTAL/coastal/save/shared/models/forcings/tides/h_tpxo9.v1.nc'

platform = Platform.HERA
adcirc_processors = 11
modeled_start_time = datetime(2008, 8, 23)
modeled_duration = timedelta(days=14.5)
modeled_timestep = timedelta(seconds=2)
tidal_spinup_duration = timedelta(days=12.5)
nems_interval = timedelta(hours=1)
job_duration = timedelta(hours=6)

# dictionary defining runs with ADCIRC value perturbations - in this case, a single run with no perturbation
runs = {f'test_case_1': None}

# describe connections between coupled components
nems_connections = ['ATM -> OCN', 'WAV -> OCN']
nems_mediations = None
nems_sequence = [
    'ATM -> OCN',
    'WAV -> OCN',
    'ATM',
    'WAV',
    'OCN',
]

slurm_email_address = 'example@email.gov'

# initialize `adcircpy` forcing objects
tidal_forcing = Tides(tidal_source=TidalSource.TPXO, resource=TPXO_FILENAME)
tidal_forcing.use_all()
wind_forcing = AtmosphericMeshForcing(
    filename=FORCINGS_DIRECTORY / 'wind_atm_fin_ch_time_vec.nc',
    nws=17,
    interval_seconds=3600,
)
wave_forcing = WaveWatch3DataForcing(
    filename=FORCINGS_DIRECTORY / 'ww3.Constant.20151214_sxy_ike_date.nc',
    nrs=5,
    interval_seconds=3600,
)
forcings = [tidal_forcing, wind_forcing, wave_forcing]

configuration = NEMSADCIRCRunConfiguration(
    fort13=MESH_DIRECTORY / 'fort.13',
    fort14=MESH_DIRECTORY / 'fort.14',
    modeled_start_time=modeled_start_time,
    modeled_end_time=modeled_start_time + modeled_duration,
    modeled_timestep=modeled_timestep,
    nems_interval=nems_interval,
    nems_connections=nems_connections,
    nems_mediations=nems_mediations,
    nems_sequence=nems_sequence,
    tidal_spinup_duration=tidal_spinup_duration,
    platform=platform,
    runs=runs,
    forcings=forcings,
    adcirc_processors=adcirc_processors,
    slurm_partition=None,
    slurm_job_duration=job_duration,
    slurm_email_address=slurm_email_address,
    nems_executable=NEMS_EXECUTABLE,
    adcprep_executable=ADCPREP_EXECUTABLE,
    source_filename=MODULES_FILENAME,
)

configuration.write_directory(OUTPUT_DIRECTORY, overwrite=False)

generation_script = NEMSADCIRCGenerationScript()
generation_script.write(OUTPUT_DIRECTORY / 'generate_nems_adcirc.py', overwrite=True)

This code will generate JSON configuration files in the directory hera_shinnecock_ike/, with the following structure:

📦 hera_shinnecock_ike/
┣ ✎ configure_modeldriver.json
┣ ✎ configure_adcirc.json
┣ ✎ configure_nems.json
┣ ✎ configure_slurm.json
┣ ✎ configure_tidal_forcing.json
┣ ✎ configure_atmesh.json
┣ ✎ configure_ww3data.json
┗  ▶ generate_nems_adcirc.py

These JSON configuration files contain the configuration values for an ADCIRC run. You may change these values to alter the resulting run configuration.

2. run generation script

Running generate_nems_adcirc.py will read the JSON configuration and generate an ADCIRC run configuration, as so:

📦 hera_shinnecock_ike/
┣ ✎ configure_modeldriver.json
┣ ✎ configure_adcirc.json
┣ ✎ configure_nems.json
┣ ✎ configure_slurm.json
┣ ✎ configure_tidal_forcing.json
┣ ✎ configure_atmesh.json
┣ ✎ configure_ww3data.json
┣  ▶ generate_nems_adcirc.py
┣ 📂 coldstart/
┃  ┣ 📜 fort.13
┃  ┣ 🔗 fort.14 -> ../fort.14
┃  ┣ 📜 fort.15
┃  ┣ 🔗 nems.configure -> ../nems.configure.coldstart
┃  ┣ 🔗 config.rc -> ../config.rc.coldstart
┃  ┣ 🔗 model_configure -> ../model_configure.coldstart
┃  ┣ 🔗 adcprep.job -> ../job_adcprep_hera.job
┃  ┣ 🔗 adcirc.job -> ../job_adcirc_hera.job.coldstart
┃  ┗ 🔗 setup.sh -> ../setup.sh.coldstart
┣ 📂 runs/
┃  ┗ 📂 test_case_1/
┃    ┣ 📜 fort.13
┃    ┣ 🔗 fort.14 -> ../../fort.14
┃    ┣ 📜 fort.15
┃    ┣ 🔗 fort.67.nc -> ../../coldstart/fort.67.nc
┃    ┣ 🔗 nems.configure -> ../../nems.configure.hotstart
┃    ┣ 🔗 config.rc -> ../../config.rc.hotstart
┃    ┣ 🔗 model_configure -> ../../model_configure.hotstart
┃    ┣ 🔗 adcprep.job -> ../../job_adcprep_hera.job
┃    ┣ 🔗 adcirc.job -> ../../job_adcirc_hera.job.hotstart
┃    ┗ 🔗 setup.sh -> ../../setup.sh.hotstart
┣ 📜 fort.14
┣ 📜 nems.configure.coldstart
┣ 📜 nems.configure.hotstart
┣ 📜 config.rc.coldstart
┣ 📜 config.rc.hotstart
┣ 📜 model_configure.coldstart
┣ 📜 model_configure.hotstart
┣ 📜 job_adcprep_hera.job
┣ 📜 job_adcirc_hera.job.coldstart
┣ 📜 job_adcirc_hera.job.hotstart
┣ 📜 setup.sh.coldstart
┣ 📜 setup.sh.hotstart
┣ 📜 cleanup.sh
┗  ▶ run_hera.sh

3. run job submission script

Running run_hera.sh will start the actual model run.

sh run_hera.sh

This will submit the requested jobs to the queue (or run the commands directly if the platform is set to LOCAL):

squeue -u $USER -o "%.8i %.21j %.4C %.4D %.31E %.20V %.20S %.20e"

   JOBID                  NAME CPUS NODE                      DEPENDENCY          SUBMIT_TIME           START_TIME             END_TIME
16368044 ADCIRC_MESH_PARTITION    1    1                          (null)  2021-02-18T19:29:17                  N/A                  N/A
16368045      ADCIRC_COLDSTART   11    1  afterany:16368044(unfulfilled)  2021-02-18T19:29:17                  N/A                  N/A
16368046 ADCIRC_MESH_PARTITION    1    1  afterany:16368045(unfulfilled)  2021-02-18T19:29:17                  N/A                  N/A
16368047       ADCIRC_HOTSTART   13    1  afterany:16368046(unfulfilled)  2021-02-18T19:29:17                  N/A                  N/A