Dependability is an important issue in LEO satellite networks. In this paper, we study the dependability of a proposed LEO satellite network for scientific data collection based on the Illinois Observing Nanosatellite (ION). In particular, we first use analytical techniques to determine the amount of radiation shielding and number of redundant components, such as batteries and regulators, needed to achieve particular reliability goals for an individual satellite. We then incorporate the satellite model into a larger satellite network model to analyze network performance and dependability using simulation. Both models are represented as stochastic activity networks (SANs) in M?obius, making it possible to combine them and choose the best solution technique (analytical vs. simulation). Our results show how accumulated radiation dosage affects the functioning of the proposed network, and how redundancy and shielding can be profitably used to increase dependability.