In order to understand and mitigate the effects of space weather on the performance of geostationary (GEO) communications satellites, we analyze 16 years of archived telemetry data from Inmarsat, the UK-based telecommunications company. We compare 665,112 operational hours of housekeeping telemetry from two generations of satellites, designated as Fleet A and Fleet B. Each generation experienced 13 solid-state power amplifier (SSPA) anomalies for a total of 26 anomalies from 1996 to 2012. We compare telemetry from the Inmarsat anomalies with space weather observations, including data from the OMNI2 database, Geostationary Operational Environmental Satellites, the Advanced Composition Explorer Satellite, and Los Alamos National Laboratory (LANL) GEO observations; the evolution of the sunspot number; and the Kp index. Most SSPA anomalies for Fleet A occur as solar activity declines; Fleet B has not yet experienced a full solar cycle. For both fleets, the average value of Kp remained <2 over time periods of 2 days, 3 days, and 2 weeks around the time of anomaly, which suggests that the anomalies occurred at times of relatively quiet geomagnetic activity and that they were probably not solely caused by surface charging. From 1996 to 2009, the average of the 1.8-3.5MeV electron flux was 1.98 #/(cm(2)s st keV). Five of the 26 anomalies, unfortunately, do not have corresponding science observations (specifically, electron flux data in the LANL data set), so part of this study focuses on the 21 anomalies when science observations were available. Six out of 21 anomalies experienced a high-energy electron flux greater than 1.5 standard deviations above the mean of the log(10) of the flux between 7 and 14days prior to the anomaly. By contrast, a Monte Carlo simulation finds that on average, only 2.8 out of 21 (13%) of randomly assigned anomalies occur between 7 and 14days after an electron flux greater than 1.5 standard deviations above the mean. Our observations suggest that internal charging from either past elevated radiation belt fluxes or some conditions related to relativistic electron enhancements (either causally or accidentally) is most likely responsible for the SSPA anomalies. We next consider the timing of these anomalies with respect to the local time (LT) and season. Anomalies occur at all LT sectors with 46% (Fleet A) and 38.5% (Fleet B) in the midnight to dawn sector and 54% (Fleet A) and 46% (Fleet B) in the local noon to dusk sector. From the local time distribution, surface charging does not appear to be the sole causative agent of the anomalies. Understanding the connection between the space weather conditions and anomalies on subsystems and specific components on identical and similar geostationary communications satellites for periods of time longer than a solar cycle will help guide design improvements and provide insight on their operation during space weather events.
