Abstract: In many sensor network applications, sensor collects correlated measurements of a physical field, for example: temperature field in a greenhouse. However, due to nodes’ inherent instability and the severe environment, sensors are prone to fail. The measurements of a faulty sensor node will incur confusions in global readings, while turning them into sleeping mode will degrade network connectivity and overload balance. Therefore, it is significant to exploit residual energy of those faulty sensor nodes so as to obtain accurate integrated readings as well as overload balance. In this paper, a cut-point set based faulty sensor node tolerance algorithm is proposed by introducing the concepts of spatial correlation model, strong correlation graph and cut-point set. The algorithm first finds out a cut-point set, which has strong spatial correlation with faulty sensor node. According to the observations of the cut-point set, the faulty sensor node is able to predict its missing sensor readings by using orthogonal intersection estimation method. Analytic results show that the algorithm not only can tolerate the faulty sensor node, but also accurately predicts miss-readings, and keeps network connectivity and overload balance. The results of miss-readings estimation, obtained from simulations and greenhouse monitoring experiments, show that the methodology presented can successfully predict the missing sensor readings.