The problem of statistical inference in large-scale sensor networks observing spatially varying fields is addressed. A method based on multiple hypothesis testing and Bayesian clustering is proposed. The method identifies homogeneous regions in a field based on similarity in decision statistics and locations of the sensors. High detection power is achieved while keeping false positives at a tolerable level. A variant of the EM-algorithm is employed to associate sensors with clusters. The performance of the method is studied in simulation using different detection theoretic criteria.