A localization algorithm using arrays with special geography for wireless sensor networks is discussed in this paper. The algorithm is based on the assumption of small aperture. The sensors are divided into several groups. Each group has a special geography. Firstly, the bearing information such as azimuth and elevation are estimated according to the time delay of arrivals. The estimate result in lines of bearings that can be thought of as rays that emanate from each sensor. Then the source localization is conducted through the intersections of the rays. The algorithm has the advantage of relative low synchronization requirement. The bearing and localization ability of two types of array, the four-element-array and the five-element-array, are discussed. The bearing estimation equations are given in a different form. Simulation results show that the five-element-array has much better localization ability than the four-element-array. The experiment confirmed the validity of the localization algorithm. And it also shows that the root-mean-square (RMS) error for the source localization increased with the increase of the aperture.