ROMS-Rutgers

Regional Ocean Modeling System

The Regional Ocean Modeling System (ROMS) is a widely used, open-source, free-surface, terrain-following, primitive equations ocean model designed for studying coastal and regional ocean dynamics. It has been developed by an international community of scientists and modelers since the late 1990s. Please visit MyRoms and ROMSwiki for more information on the model.

ROMS Key Features

• Terrain-following vertical (sigma) coordinates: ROMS uses a curvilinear vertical coordinate system that adapts to the ocean bottom topography, allowing for accurate representation of coastal and shelf processes.

• Free surface and baroclinic dynamics: The model resolves both the surface elevation and stratified (density-driven) flows, capturing tides, currents, and thermohaline circulation.

• Modular and flexible framework: ROMS supports a variety of physical processes including advection, mixing, sediment transport, and biogeochemical cycles, and can be coupled with atmospheric and wave models.

Typical Applications

• Coastal circulation and upwelling studies

• Estuarine and shelf sea dynamics

• Harmful algal bloom modeling

• Sediment transport and morphodynamics

• Coupled physical-biogeochemical simulations

• Operational forecasting and climate research

Below we describe the ROMS interface to DART. Please note that this interface supports the Rutgers version of ROMS only. Other versions of the model are also supported. Please refere to ROMS_ucla for the UCLA interface. If you have any questions regarding this interface or other ROMS-DART related questions, please reach out to the DART team at dart@ucar.edu

Interface Overview

The interface between DART and ROMS supports multiple grid configurations, ranging from fully regular grids to curvilinear (fully irregular) grids. The choice of grid type affects both the compilation and the performance of the interpolation routines.

To compile the DART interface, run quickbuild.sh with the appropriate grid option:

• Regular grid: ./quickbuild.sh reg_grid

• Irregular grid: ./quickbuild.sh irreg_grid

After compilation, the interface module ROMS_rutgers/mod_mod.f90 is generated.

For grids that are logically rectangular with separable longitude and latitude coordinates (e.g., rectilinear grids with uniform or nonuniform spacing), the reg_grid option enables faster and more efficient interpolation by using optimized search routines based on 1D coordinate arrays.

The irreg_grid option is intended for fully curvilinear grids, where longitude and latitude vary in both horizontal directions. In this case, the interpolation requires access to the full 2D coordinate arrays, which increases both memory usage and computational cost.

The ROMS_rutgers/mod_mod.f90 interface provides the routines required for DART to interact with the model state and perform data assimilation tasks, including:

• Model initialization and state management

• Observation-space interpolation of model state variables

• Vertical coordinate transformation and interpolation

• Mapping between DART’s state vector and ROMS variables

Note

In many practical ROMS configurations, grids may be irregularly spaced but still rectilinear. These grids can take advantage of the more efficient interpolation pathway associated with the reg_grid option.

Observation Handling and Interpolation

Unlike the UCLA ROMS-DART interface, this version does not rely on precomputed observation-space values output by ROMS (e.g., from MODname in s4dvar.in). Instead, it computes observation-space equivalents directly via interpolation routines implemented using the quad_utils module. Interpolation to the observed location is done both horizontally and vertically. Observations outside the model domain or falling on masked grid cells are excluded from the assimilation process (see failure codes below). The interpolation further assumes the model depth to be ordered from deepest to shallowest. It also handles extrapolation gracefully for values above or below the model domain.

! Failure codes: model_interpolate
integer, parameter :: QUAD_LOCATE_FAILED   = 13
integer, parameter :: QUAD_EVALUATE_FAILED = 21
integer, parameter :: SSH_QUAD_EVAL_FAILED = 34
integer, parameter :: QUAD_MAYBE_ON_LAND   = 55
integer, parameter :: OBS_TOO_DEEP         = 89

The image below shows an example assimilating Temperature observations from different sources including: XBT, CTD, BOTTLE, MOORING, and FLOAT. The experiment, designed for illustration purposes, uses only 3 members within the Indonesian archipelago. The effective localization radius is set to 800 km. Temperature ensemble mean at a depth of ~100 m is shown in the left panel. Ensemble mean increment (i.e., posterior - prior) at the same depth is shown on the right.

Warning

This interface only supports single time-level variables in ROMS history files. Variables with multiple time levels (e.g., leapfrog schemes in restart files) are not supported and may lead to incorrect assimilation behavior.

Namelist Configuration: model_nml

The ROMS–DART interface is configured through the model_nml namelist. This namelist is read from the file input.nml. Namelists start with an ampersand ‘&’ and terminate with a slash ‘/’. Character strings that contain a ‘/’ must be enclosed in quotes to prevent them from prematurely terminating the namelist.

The table below describes the configurable variables in this namelist:

&model_nml Namelist Variables

Variable	Type	Default	Description
roms_filename	character(len=256)	‘roms_input.nc’	Path to a ROMS NetCDF template file used to extract static grid and mask information.
assimilation_period_days	integer	1	Number of days in each assimilation window.
assimilation_period_seconds	integer	0	Number of seconds (in addition to days) in each assimilation window.
perturbation_amplitude	real(r8)	0.02	Amplitude used to perturb ensemble members for ensemble generation.
use_mean_SSH_from_template	logical	.false.	Indicates whether to use mean SSH from the template (ensemble mean).
debug	integer	0	Debugging verbosity level. Set >0 for more detailed log output.
variables	character(len=vtablenamelength), dimension(MAX_STATE_VARIABLES * table_columns)	‘ ‘	Specifies the list of ROMS variables to be assimilated. The variable table is parsed as flat strings with metadata.

Variables Table Format

The variables field in the &model_nml namelist is used to declare each state variable to be included in the DART state vector. Each variable entry consists of five elements (columns), listed in a single Fortran character array:

Format of Each variables Entry

Field	Description	Example	Notes	Requirement
Variable Name	Name of the ROMS variable in the NetCDF file.	‘temp’	Must match the ROMS file variable name exactly.	Required
DART Quantity	DART internal quantity label.	‘QTY_TEMPERATURE’	Must be a valid DART quantity name.	Required
Minimum Value	Lower bound as a string, or ‘NA’ for none.	‘0.0’	Used for clamping or bounds checking in DART.	Optional
Maximum Value	Upper bound as a string, or ‘NA’ for none.	‘NA’	Same usage as above.	Optional
Update Rule	Whether DART should write back this variable to the ROMS restart file.	‘UPDATE’	‘UPDATE’ = write back; ‘NO_COPY_BACK’ = internal use only.	Required

Example namelist snippet:

variables = 'temp', 'QTY_TEMPERATURE'        , 'NA' , 'NA', 'UPDATE',
            'salt', 'QTY_SALINITY'           , '0.0', 'NA', 'UPDATE',
            'u'   , 'QTY_U_CURRENT_COMPONENT', 'NA' , 'NA', 'UPDATE',
            'v'   , 'QTY_V_CURRENT_COMPONENT', 'NA' , 'NA', 'UPDATE',
            'zeta', 'QTY_SEA_SURFACE_HEIGHT' , 'NA' , 'NA', 'UPDATE'

Each variable must appear as a consecutive 5-element group in the flat variables array. The interface supports up to MAX_STATE_VARIABLES, each with 5 fields.

Note

• Variables marked as ‘NO_COPY_BACK’ are updated within the DART filter but are not written back to the ROMS restart file. 1990s

• Only variables in restart files can be updated in ROMS. Ensure roms_filename points to a restart file (e.g., roms_input.nc) when using ‘UPDATE’.

• Observation times are assimilated if they fall within ±0.5 × assimilation_period_days from the model forecast time.

• It’s recommended to set use_mean_SSH_from_template to .true. and provide the ensemble mean as the template file roms_filename. This ensures the right SSH values are used in the localization routine.

Generating an Initial Ensemble

The ROMS interface provides the ability to create an ensemble of initial ROMS history files from an initial file by using the perturb_single_instance routine. You can specify an ensemble of any size in the perturb_single_instance namelist in input.nml and this program will randomly perturb the temperature and salinity fields of an initial ROMS history file to generate the ensemble. The size of the perturbation is set using the namelist parameter perturbation_amplitude and the resulting initial distribution is Gaussian.

References

• Shchepetkin, A.F. and McWilliams, J.C., 2005. The regional oceanic modeling system (ROMS): a split-explicit, free-surface, topography-following-coordinate oceanic model. Ocean Modelling, 9(4), pp.347-404.