In wireless sensor networks, if the local sensor node transmits wrong decided data to a fusion center, the energy-efficiency of the network is degraded. In this paper, we propose wireless sensor networks with local fusion to reduce the degradation caused by observation noise, channel noise, and fading. In the proposed systems, first, information transmitted from local sensor nodes are decided by a local fusion center allocated in the neighborhood of local sensor nodes. Here, we consider two signal decision schemes (local fusion) for use at the local fusion center: majority decision and likelihood decision. By using local fusion, even if the local sensor node makes an error in the decision of the event, the probability that the local fusion center makes an error in the decision of the event becomes small. Thus, the probability that the local fusion center transmits wrong decided data becomes small. Then, from the local fusion center to a destination node (a global fusion center), error correcting coded bits are transmitted in multihop communication. We show that the proposed systems can achieve good energy-efficiency as compared with the system without local fusion. We also show that the total energy consumption of the system with like-lihood decision is smaller than that of the system with majority decision when the influence of observation noise is large. Moreover, we show that the optimal number of hops minimizing the total energy consumption of the proposed systems depends on the signal attenuation parameter K and a long-haul transmission distance between the local fusion center and the global fusion center.