This paper proposes a localized recursive estimation scheme for parameter estimation in wireless sensor networks. Given any parameter occurred at some location and time, a number of sensors recursively estimates the parameter by using their local measurements of the parameter that is attenuated with the distance between a sensor and the target location and corrupted by noises. Compared with centralized estimation schemes that transmit all measurements to a sink (or a fusion center), the recursive scheme needs only to transmit the final estimate to a sink. When a sink is faraway from the sensors and multihop communications have to be used, using localized recursive estimation can help to reduce energy consumption and reduce network traffic load. Furthermore, the most efficient sequence of sensors for estimation is defined and the necessary condition for such a sequence is determined. Some numerical examples are also provided. By using some typical industrial sensor parameter values, it is shown that recursive scheme consumes much less energy when the sink is three hops or more faraway from the local sensors.