This study describes several experiments to simulate and study temperature sensor networks. The ultimate goal of this study is to accurately predict a temperature sensor failure in a sensor network, using Network Neighborhood Watch theory. Simulated temperature measurements are used to compare the results of accuracy, and quickness in detecting a failure in a malfunctioning sensor node. This is done by changing the intervals between temperature sampling, by considering different rates of temperature raise, as well as varying the number of nodes in the network neighborhood (deployment density). The thesis studies the threshold for temperature sensor failure by computing a non-parametric hypothesis-testing statistical 'failure' criteria, computed over a range of temperature data for the network neighborhood. When the temperature sensor value crosses the threshold, a sensor node failure is signified. Sensor node failure prediction delay has been measured by trading off accuracy of data and quickness in failure prediction. For this study, an animal habitat monitoring environment (a research project for animal environment studies) is examined.