Doppio (ROMS, Mid-Atlantic Bight and Gulf of Maine)

Doppio is a ROMS configuration for the Mid-Atlantic Bight and the Gulf of Maine, run by the Rutgers University Ocean Modeling Group for MARACOOS (López et al., 2020). Rutgers runs it in two forms, both with data assimilation and both served hourly from the Rutgers THREDDS server:

  • the reanalysis DopAnV3R3-ini2007 (4D-Var; Wilkin et al., 2022), 2007-01-02 01:00 to 2024-12-31 00:00 UTC, every hour with no gaps; and

  • the real-time system (2017_da), 2017-11-02 01:00 UTC to about two days ahead, as the “Best” time series of a forecast model run collection.

forcingkit.fetchers.doppio reads either as a full-depth subset in the same layout as the NECOFS volume (NECOFS: Northeast Coastal Ocean Forecast System (FVCOM), schema volume-v1), so a comparison between the two models, or between either and a downstream ocean model, can treat them alike. It is a library function: no service route or parent-ocean (z-v3) delivery uses it yet.

Underlying model

Checked against the aggregations’ OPeNDAP metadata on 2026-11-17.

  • Grid: curvilinear Arakawa C-grid, 106 x 242 rho points (eta_rho, xi_rho), about 7 km, from 80.5 W to 59.7 W and 32.2 N to 46.6 N. The grid’s xi axis is turned about 37 degrees from east (angle), so u and v in the files are along the grid, not east and north.

  • Vertical coordinate: 40 terrain-following s-levels (s_rho, bottom first), Vtransform 2, Vstretching 4, theta_s 7, theta_b 2, hc 250 m.

  • Fields: u (on u faces), v (on v faces), w (on the 41 s_w interfaces), temp (potential temperature, Celsius), salt, zeta, plus ubar, vbar, mixing coefficients and surface fluxes. The real-time aggregation also carries u_eastward and v_northward at rho points, but they are empty (fill values) for 2022 and populated for recent dates.

  • Datum: the free surface is relative to the model mean sea level; the Rutgers metadata states no geodetic datum, so do not compare it with NAVD88 water levels without an offset.

  • Time: ocean_time in seconds since 2006-01-01 (reanalysis); time in hours since 2017-11-01 (real time).

The full-depth subset

volume(bbox, start, end) returns every s-level for the index rectangle around the cells whose centers lie in the box, hourly from start to end inclusive, with no horizontal regridding and no vertical conversion. read_volume(bbox, start_time, hours, dap_url=None) reads one aggregation.

from forcingkit.fetchers.doppio import volume

bbox = [-69.89, 39.66, -69.26, 40.14]  # Veatch Canyon mouth and the shelf break above it
vol = volume(bbox, "2022-08-11T00Z", "2022-08-14T00Z")  # 73 hourly records

Name

Dimensions

Meaning

u, v

(time, siglay, eta_rho, xi_rho)

Eastward and northward velocity at the cell centers, m s-1: the two faces either side of each center averaged (a land face counts as zero flow, as in ROMS’s own u2rho), then rotated by angle. On 2026-11-10 this matched the real-time aggregation’s own u_eastward and v_northward to 6e-8 m s-1.

ww

(time, siglay, eta_rho, xi_rho)

ROMS w (upward velocity in z), averaged from the interfaces above and below each layer center; the name follows the NECOFS volume.

temp, salinity

(time, siglay, eta_rho, xi_rho)

Potential temperature (degree_Celsius) and salinity (1e-3).

zeta; h

(time, eta_rho, xi_rho); (eta_rho, xi_rho)

Surface elevation relative to the model mean sea level (m, positive up); depth (m, positive down).

siglay, siglev

(siglay, eta_rho, xi_rho), (siglev, eta_rho, xi_rho)

The fraction of the water column at layer centers and interfaces, surface first: (hc * s + h * Cs) / (hc + h), which for Vtransform 2 makes the depth of a layer center \(\sigma (h + \zeta) + \zeta\), the NECOFS formula.

lon, lat, mask, angle

(eta_rho, xi_rho)

Cell centers in degrees (lon and lat are 2D auxiliary coordinates, so CF tools and the cache viewer place the cells), water mask, and the grid angle in radians.

in_bbox

(eta_rho, xi_rho)

The cell center lies in the box. The rectangle’s corners may not, since the grid is rotated: for the box above, 59 of the 100 cells.

eta_rho, xi_rho (coordinates)

(eta_rho), (xi_rho)

Each row’s and column’s index in the full grid.

Field names and units match the NECOFS volume (u, v, ww, temp, salinity, zeta, h, siglay, siglev). The horizontal layout differs: NECOFS returns elements and nodes of an unstructured mesh, Doppio rows and columns of a curvilinear grid, with every field at the cell centers. Land cells are NaN in the time-varying fields.

variables and layers select as for NECOFS: variables among u, v, ww, temp, salinity, zeta; layers as zero-based indices, surface first, negative ones from the bottom, returned with a layer coordinate.

volume uses the reanalysis when it holds the whole window, else the real-time system, else raises; dap_url names an aggregation explicitly. It raises if any requested hour is missing. Results are cached as Zarr under $FORCINGKIT_CACHE_DIR/doppio/volume/, keyed by the schema, the aggregation, the box, the window and the selection. The 73 hours above (10 x 10 cells, 40 levels, 6 MB) took 45 s from Rutgers on 2026-11-17, and 0.1 s from the cache.

forcingkit.layered_mesh.doppio(bbox, start, end) returns the same subset as a LayeredMesh (see Layered mesh: 3D model output on its own mesh) for rendering and sections next to NECOFS.

Limits

  • Resolution. Cells are about 7 km across (6 to 7 km over the box above), so features under about 15 to 20 km, such as the flow inside a single shelf-break canyon, are not resolved. Use Doppio for the slope current, stratification and tides at shelf scale, and NECOFS or a finer model inside canyons.

  • Sea level datum. zeta is relative to the model’s own mean sea level; the Rutgers metadata names no geodetic datum.

  • Cache viewer. The cache viewer (Cache viewer) lists the stores under doppio/volume/, draws their footprints on its coverage map and previews them on their own curvilinear cells, every layer at its height from siglay and h.

Credit

The Doppio catalog describes the data as “Freely available”. Cite the model (López et al., 2020), the reanalysis (Wilkin et al., 2022) and, for the reanalysis output, the dataset as its catalog asks: Wilkin, J. and J. Levin (2022), SEANOE, doi:10.17882/89673. The observations it assimilated are on Rutgers ERDDAP. See Northeast Shelf Models and Reanalyses for the other models of the region.