This document discusses cooperative sensor network localization using Gaussian mixture modeling and expectation-conditional maximization (ECM) algorithms. It proposes using a Gaussian mixture to approximate the unknown non-Gaussian measurement error distribution. Centralized ECM algorithms are developed for parameter estimation with proofs of convergence properties. Distributed ECM algorithms are also created to improve scalability, using average consensus to update the Gaussian mixture model locally. Computer tests show the distributed algorithms can perform well even with model mismatch and unknown error statistics, outperforming alternatives.