The Sentinel-3 constellation of Earth Observation satellites is designed to provide accurate and timely information to better manage the marine environment, and to understand and mitigate the effects of climate change by utilizing systematic measurements and products of sea-surface topography, sea-state and ecosystem characteristics over the open ocean and the regional and shelf seas. The aim of the paper is twofold: (a) to provide a brief overview of the types of data available to users from the current Sentinel-3 satellites, and (b) to outline the new features, compared to the conventional radar altimeters, and the new capabilities provided by the Sentinel-3 SRAL (SAR mode) altimeters. The presentation will show representative results based on comparative analyses using previous (conventional) and current Sentinel-3 altimetry data, in an effort to identify critical data handling aspects (i.e. data access) and associated constraints (such as primary outputs, ground coverage, etc.) vis-à-vis typical user requirements for various marine applications, especially closer to the coastal areas where the SAR altimetry method is shown to be superior to conventional systems in terms of accuracy, spatial resolution of the observations etc.

Deep learning techniques are used for capturing intricate structures of large-scale data by employing computational models of multiple processing layers that can learn and represent data with multiple levels of abstraction [1]. Such methods can include Convolutional Neural Networks, stacked auto-encoders and Long-Short Term Memory (LSTM) architectures. LSTM networks are suitable for dealing with time-dependent data through mapping input sequences to output sequences as it is done, for instance, in language modeling and speech recognition. One application that has recently attracted considerable attention within the geodetic community is the possibility of applying these techniques to account for the adverse effects of the ionospheric delays on the GNSS satellite signals. LSTM architectures model long-range dependencies in time series, making them appropriate for ionospheric modeling in GNSS positioning. This paper deals with a modeling approach suitable for predicting the ionospheric delay at different locations of the IGS network stations using the LSTM networks. We also incorporate a Bayesian optimization method for selecting the best configuration parameters of the LSTM network, thus improving network’s performance.

Achieving high accuracy for Global Navigation Satellite Systems (GNSS) positioning at low cost, is the main reason for the wide acceptance that has received the technique of Precise Point Positioning (PPP) from the geodetic community. However, the long convergence time required to achieve centimeter-level accuracy in positioning, remains of PPP’s main disadvantage. Our proposed method is aiming to provide accurate prediction of ionosphere variations at regional level, in a form suitable to be applied in PPP processing as external ionospheric information, so that to successfully deal with the ionospheric error effects. Machine Learning regression-based approaches for sequence modeling are suitable for predicting the ionospheric variability. Gaussian Process Regression (GPR) and Support Vector Regression (SVR) are introduced for ionospheric variations modeling at different locations of the International GNSS Service (IGS) network. The proposed algorithms predict the future electronic content per satellite from a specific station, fitting a non-linear model. The evaluation of our proposed methods for vertical total electron content (VTEC) values prediction is compared against the traditional Autoregressive (AR) and Autoregressive Moving Average (ARMA) methods, per satellite. Additional, mean TEC timeseries are compared against classical Global Ionospheric Maps (GIM), NeQuick and International Reference Ionosphere (IRI) models.