Raptor surveys for environmental assessment and other purposes are conducted mainly by field surveys and observation. However, the surveys are labor-intensive, and there are also issues such as missing findings. In this study, as an efficient and effective raptor survey tool, we developed a system (Sound radar) that estimates the location of goshawks Accipiter gentilis by automatically identifying and extracting their calls from audio data recorded at multiple locations. The calls of goshawks were classified into three patterns (i.e., warning, begging, and juvenile), which were determined by decision tree analysis based on spectrogram (voice pattern) characteristics. The location of individuals was estimated by solving a ternary simultaneous equation based on the planar coordinates and sound pressure levels at four points using the distance decay equation for sound. The fit rate (correct response rate) of the call discrimination ranged from 0.8 to 0.9 for the training data and from 0.5 to 0.8 for the validation data. The estimated locations by Sound radar and the results of field confirmation were generally in agreement.