ESA Mediterranean Sea Datacube

The ESA Mediterranean Sea Datacube brings selected marine datasets onto a common 1/24° latitude/longitude grid (EPSG:4326) for the Mediterranean Sea.

The data are available in four online GeoZarr stores: monthly fields, hourly currents, daily surface fields, and daily biophysics. The stores share the same horizontal grid and keep the time and depth coordinates each dataset needs.

This shared grid lets you compare variables without repeating the spatial alignment.

Access and Visualization¶

The associated notebooks show how to:

1. Acces the ESA Mediterranean Sea Datacube

2. Build one local datacube for a small ROI selects salinity from the original 4DMED-SEA source dataset, resamples a small region, and saves and plots the local cube.

3. Do larger scale analysis (Time Series at Every Depth) .

The access examples open the online GeoZarr stores linked below directly from EarthCODE object storage.

A simple visualisation of the data cube is available at https://sunnydean.github.io/ocean_cubes_vis/:

Use the individual dataset notebooks to explore the source products.

Data structure¶

• All spatial layers are aligned to one EPSG:4326 lat/lon grid. Full-resolution level 0 uses 1/24° cells, with longitude centres from approximately -6.0625 to 36.0625 and latitude centres from 30.2708 to 45.9792, giving 1,012 columns by 378 rows.

• Each store is a multiscale GeoZarr with groups 0, 1, and 2. These represent 1/24°, 1/6°, and 1/3° spacing, with spatial shapes 378 × 1,012, 94 × 253, and 47 × 126.

• The same lat, lon position refers to the same place across every store at a given level. (The grid has different physical cell areas at different latitudes.)

• Data values are spatially interpolated using nearest neighbour from the source data.

• Monthly fields are averaged into calendar months and indexed by month start. Daily fields are aligned to daily timestamps.

• The biophysics cube preserves 18 common depth levels from 3 to 135 m, positive down. WOC has one depth level at 15 m.

• All four stores include water_mask: 1 retains water or unmapped offshore cells and 0 excludes mapped land. It comes from ESA WorldCover 2021 overviews and is applied across time and depth; The mask is not guaranteed permanent water, but rather an np.isin([0,80]) (i.e. non land and permanent water bodies) applied to the land classificaiton product.

Table of Cubes¶

The access links below point to the online GeoZarr stores.

Current Assumptions and Future Refinements¶

• The upsampling or downsampling used to create the cubes can make the variables unsuitable for some scientific analyses. The original datasets remain available through 1_Datasets for workflows that need their native grids, times, or depth levels.

• Monthly aggregation of 8-day products uses an unweighted mean of samples assigned to each calendar month; it is not a day-overlap-weighted integration. Reindexed dates may be entirely missing for some variables. Source-wide MITHO summaries repeated along time should not be treated as independent monthly observations. Likewise, a maximum category in a coarser GeoZarr group represents the most severe contributing cell, rather than a class at every location in that block.

• Open the store containing the variables you need, then select a region, time, and depth before computing. When comparing hourly, daily, and monthly values, aggregate them to a common time interval.