Weather-Based Link Prediction for LEO-Satellite Networks using the WetLinks Dataset

The performance of Low Earth Orbit (LEO) satellite networks such as Starlink is volatile and susceptible to weather conditions. In this demo, we explore the potential to predict Starlinks link parameters - precisely the upload and download throughputs and the RTT - based on historical weather data and weather-forecast data. An accurate prediction model would enable the development of weather-aware applications. We train different machine learning (ML) models, compare their performance, and demonstrate that the prediction accuracy for all network parameters is sufficient to predict general trends. Especially, the download throughput can be predicted fairly well.