M-VRE: The MOSAiC - Virtual Research Environment
The M-VRE project is funded by the German BMBF initiative „MARE:N – Polarforschung/MOSAiC“ from July 2021 to June 2024.
This project is a collaboration of AWI Bremerhaven, DKRZ Hamburg and DLR Jena.
The main goal of the M-VRE project is to provide novel
methods and software tools for the
MOSAiC
consortium, the global
climate community, and the general public for efficiently exploring,
analysing, and visualizing MOSAiC data in an online and
user-friendly manner. This enables scientists to exploit large,
complex, and heterogeneous data sets to answer fundamental research
questions of the MOSAiC expedition, connected to the overarching
goal of the identification of reasons and consequences of changing
and shrinking sea ice cover. To achieve this, a web-based data
analysis and exploitation environment, a so-called "virtual research
environment (VRE)", will be developed as part of the AWI cloud
infrastructure. The collaborative and reproducible online analysis
of MOSAiC data opens up new possibilities in climate research and
increases the visibility and usability of the MOSAiC mission. The
VRE offers online data analysis (e.g. quality control,
interpolation), exploration (data cubes), and visualization (maps,
scatter plots, time series, section plots and more) of MOSAiC data
as well as automated data quality control through Deep Learning
methods.
Watch the M-VRE overview video below!

webODV, the online pendant of the desktop Ocean Data
View (ODV, https://odv.awi.de) software. ODV, with its diverse functionality,
is widely used in the marine and environmental sciences
for a large variety of applications. Currently we have
approx. 10,000 users worldwide. Its features include
creating maps, surface plots, section plots, scatter
plots, filtering data, extracting data subsets from large
collections, aggregating single data files into larger
datasets and collections, flagging data, and modifying
data, among others. webODV is a highly interactive
tool running in the browser, where the user can interact
directly with the data in the maps or plots.
Productive webODV services for the GEOTRACES and EMODnet Chemistry projects are operated at
https://webodv.awi.de.

Data Cubes. Raster data management systems, such
as Rasdaman or SciDB, and data cube frameworks, such as
openDataCube and EarthSystemDataCube, provide efficient
data access, query processing, and data analysis support
for large-scale scientific multi-dimensional data. As
such, it serves as a foundation and backend for complex
scientific data analyses and offers intuitive programming
interfaces (e.g., in scientific scripting languages, such
as Python or R) for scientific users, hence perfectly
suited in the M-VRE. We will develop a data cube service
that allows querying raster data from the MOSAiC
expedition and connect it as a backend service to other
data analysis and visualization services offered in the
M-VRE.

DIVAnd. DIVAnd, the Data-Interpolating
Variational Analysis in n dimensions, is a powerful
interpolation tool developed by the GeoHydrodynamics and
Environment Research group at the University of Liège. It
allows to create smooth and continuous fields in one, two,
three or n dimensions from a set of observations, while
handling constraints which are typical for earth-science
applications (e.g. non-regularly located data points,
physical constraints). This is a recurring task in a wide
range of scientific disciplines like physical oceanography
We are already working closely together with
the developers of DIVAnd (Dr. A. Barth and Dr. C. Troupin)
and will invite them to our regularly planned workshops
for teaching MOSAiC users. More infos at
https://github.com/gher-ulg/divand.jl.

Image by Florian Weihmann from Pexels.
AutoQC. Until today the manual and visual quality
control (QC) of ocean data is a time consuming effort done
by ocean experts. With increasing amounts of big data,
algorithmic help is urgently needed, where artificial
intelligence could play a dominant role. We develop deep
learning algorithms for automated QC of ocean data to
support scientists.
We will train the algorithms
with AWIs UDASH data to "learn" arctic QC. We will then
apply the algorithms to new MOSAiC data together with the
ocean experts to create a novel quality controlled arctic
hydrographic data set. More infos about our previous project at https://salacia-ml.awi.de.