Many practical applications of signal processing require accurate determination of signal parameters from sensorarray measurements. Most estimation techniques are sensitive to errors in the array response model. Thus, reliable array calibration schemes are of great importance. A paradigm for generating an array model from noise corrupted calibration vectors is developed. The key idea is to use a local parametric model of the sensorresponses. The potential improvement using the suggested scheme is demonstrated on real data collected from a full-scale hydroacoustic array.