This paper addresses the sensor selection problem in a passive sensor network under dynamically varying environment. It is assumed that sensors in the same local area share similar dynamic environmental characteristics but only a few may have events within their respective sensing radii. The main challenge is to select sensors for separation in two categories: (i) those sensors whose outputs contain combined information of event and dynamic environment and (ii) those sensors whose outputs contain the environment information only. The criteria for sensor selection is made based on the entropy rate measurements of the standard Probabilistic Finite State Automata (PFSA) constructed from sensor time series and cross D-Markov machines constructed from combined time series of a sensor and the first (i.e., dominant) principal component representing the dominant linear mode in the ensemble of related sensor data within the network. The proposed method is applied to experimental data collected from a small local passive sensor network for target detection under dynamic environments, where characteristics of environmental signal is similar to the target signal in the sense of time scale and texture.