Short_course_18: Gap Filling Ocean Data (statistical and machine learning methods)
مخطط الموضوع
-
Webinar Aida Alvera Azcarate; Moderator: Hayley Evers-King (EUMETSAT)
19 October 2021; 11 UTC;
The ocean is a complex system with processes at various scales interacting among them, from thousands of kilometers down to less than 1 km. Satellite data offer a unique amount of information about the ocean surface, thanks to the high spatial and temporal resolution they provide. However, satellite sensors measuring at the visible and infrared wavebands are affected by the presence of clouds and have therefore a large amount of missing data (clouds cover about 75% of the Earth at any given time). In order to study these multi-scale oceanic processes it is therefore necessary to deal with this missing information. Data interpolation techniques are often used for that, and various approaches have been developed over time.
The GHER (GeoHydrodynamics and Environment Research) of the University of Liege in Belgium works on the development of interpolation techniques for satellite data. In this seminar we will present two approaches. We’ll start with DINEOF - Data Interpolating Empirical Orthogonal Functions- which is a data-driven technique using EOFs to infer missing information in satellite datasets. DINEOF uses the information present in the dataset to infer the most probable state of the ocean at the missing points. We will follow with a more recent development, DINCAE - Data Interpolating Convolutional AutoEncoder. Training a neural network with incomplete data is problematic, and this is overcome in DINCAE by using the satellite data and its expected error variance as input. The autoencoder provides the reconstructed field along with its expected error variance as output.
We will provide examples of reconstructed satellite data for several variables, like sea surface temperature and chlorophyll concentration. We’ll show some recent developments with DINCAE to grid altimetry data to 2D complete fields.
The code for these techniques is openly available for any interested user. We will show where to find them and provide background information on their use, as well as some practical examples. We will also briefly point out to operational ocean satellite products using gap-filling techniques (DINEOF among others) and that are available through the CMEMS catalog.
-
-
Our training activities support users in the member states and among the World Meteorological Organization (WMO) members in the application of EUMETSAT data, products, and services, including Copernicus data provided by EUMETSAT. Find our more at our web page.
Subscribe to the Training Bulletin the joint EUMETSAT, EUMETCAL, EUMeTrain newsletter on new activities and resources in European meteorological training — published every quarter.
Contact us for questions about training events and professional development opportunities at training@eumetsat.int.
Follow us on Twitter @eumetsat_users.