Optical Feeder Links to GEO Satellites: Statistical Analysis of Link Availability Using Deep Learning-Based Cloud Segmentation Data

Abstract

Optical feeder links (OFLs) to geostationary orbit (GEO) satellites offer a promising solution for significantly increasing the throughput of satellite systems, especially in constellations with high data rate demands. However, cloud coverage substantially raises the likelihood of link outages, thereby reducing the availability of OFLs. This paper presents a statistical analysis of cloud data recorded by the German Aerospace Center (DLR) at the Plataforma Solar de Almería (PSA), a facility of the Spanish Centre for Energy, Environmental and Technological Research (CIEMAT), where an optical ground station (OGS) from DLR is currently under construction. Using deep learning-based cloud segmentation of whole-sky images, we estimate the a priori probabilities and distributions of clear-sky and cloud-covered states of the Alphasat and European Data Relay Satellite System (EDRS)-A satellites. The hourly and monthly cloud coverage probabilities are analyzed and visualized, suggesting time-zone and hemisphere distribution for future OGS networks to enhance system availability. Results also show that data links often remain operational only for short intervals before cloud obstruction, highlighting the need for rapid connection handovers. Finally, we propose a satellite diversity approach to complement a distributed OGS network, improving system availability and reducing the number of required OGS sites.