This section provides directions to download the ECCO v4 r2 output (Section 2.1), the underlying model setup that can be used to re-run ECCO v4 r2 (Section 2.2), tools for manipulating and analyzing model output (Section 2.3), and a list of additional resources (Section 2.4).

## 2.1. The Release 2 Solution¶

The ECCO v4 r2 state estimate output is permanently archived within the Harvard Dataverse that provides citable identifiers for the various datasets as reported in this README.pdf. For direct download purposes, the ECCO v4 r2 output is also made available via this ftp server by the ECCO Consortium. The various directory contents are summarized in this README and specific details are provided in each subdirectory’s README. Under Linux or macOS for instance, a simple download method consists in using wget at the command line by typing

wget --recursive ftp://mit.ecco-group.org/ecco_for_las/version_4/release2/nctiles_grid
wget --recursive ftp://mit.ecco-group.org/ecco_for_las/version_4/release2/nctiles_climatology
wget --recursive ftp://mit.ecco-group.org/ecco_for_las/version_4/release2/nctiles_monthly


and similarly for the other directories. The nctiles_ directory prefix indicates that contents are provided on the native LLC90 grid in the nctiles format [FCH+15] which can be read in Matlab using the gcmfaces toolbox (see Section 2.3). Alternatively users can download interpolated fields, on a $$1/2\times1/2^\circ$$ grid in the netcdf format, from the interp_* directories. The input_* directories contain binary and netcdf input files that can be read by MITgcm (Section 3.1). The profiles/ directory additionally contains the MITprof, netcdf collections of collocated in situ and state estimate profiles [FCH+15].

## 2.2. The Release 2 Setup¶

Users can donwload the MITgcm from this github repository and the model setup there from that github repository by typing:

git clone https://github.com/MITgcm/MITgcm
git clone https://github.com/gaelforget/ECCOv4
mkdir MITgcm/mysetups
mv ECCOv4 MITgcm/mysetups/.


Re-running ECCO v4 r2 additionally requires downloading surface forcing input (96G of 6-hourly fields in ECCO v4 r2), initial condition, grid, etc. input (610M), and observational input (25G) either from the Harvard Dataverse permanent archive or directly from the ECCO ftp server as follows:

cd MITgcm/mysetups/ECCOv4
wget --recursive ftp://mit.ecco-group.org/ecco_for_las/version_4/release2/input_forcing/
wget --recursive ftp://mit.ecco-group.org/ecco_for_las/version_4/release2/input_init/
wget --recursive ftp://mit.ecco-group.org/ecco_for_las/version_4/release2/input_ecco/
mv mit.ecco-group.org/ecco_for_las/version_4/release2/input_forcing forcing_baseline2
mv mit.ecco-group.org/ecco_for_las/version_4/release2/input_ecco inputs_baseline2
mv mit.ecco-group.org/ecco_for_las/version_4/release2/input_init inputs_baseline2/.


The Recommended Directory Organization is shown below. While organizing the downloaded directories differently is certainly possible, the Section 3.1 instructions to Compile, Link, And Run the model and Verify Results Accuracy are based on this organization.

Recommended Directory Organization

MITgcm/
model/   (MITgcm core)
pkg/     (MITgcm modules)
tools/
genmake2        (shell script)
build_options   (compiler options)
mysetups/         (user created)
ECCOv4/
build/                (build directory)
code/                 (compile-time settings)
input/                (run-time settings)
results_itXX/         (reference results)
forcing_baseline2/    (user installed)
inputs_baseline2/     (user installed)


Note

Some subdirectories are omitted in this depiction.

## 2.3. The Gcmfaces Toolbox¶

The gcmfaces toolbox [FCH+15] can be used to analyze model output that has either been downloaded (Section 2.1) or reproduced (Section 3.1) by users. From the command line, you can install either the Matlab version by executing:

git clone https://github.com/gaelforget/gcmfaces


or the Octave version by executing:

git clone -b octave https://github.com/gaelforget/gcmfaces


The gcmfaces toolbox can be used, e.g., to reproduce the standard analysis (i.e., the plots in [FCH+16]) from released, nctiles model output (Section 2.1) or from plain, binary model output (Section 3.1). For more information, please consult the gcmfaces user guide.

## 2.4. Other Resources¶

• A series of three presentations given during the May 2016 ECCO meeting at MIT provides an overview of ECCO v4 data sets, capabilities, and applications (Overview; Processes; Tracers).
• Various Python tools are available to analyse model output (see, e.g., this tutorial).
• Any netcdf enabled software such as Panoply (available for MS-Windows, Linux, or macOS) can be used to plot the interpolated output (interp_* directories).
• The stand-alone eccov4_lonlat.m program can be used to extract the lat-lon sector, which spans the 69S to 56N latitude range, of native grid fields [FCH+15].
• ECCO v4 estimates can be plotted via the NASA Sea Level Change Portal tools (interpolated output) or downloaded from the Harvard Dataverse APIs (native grid input and output).