Horizontal coherence length informs array aperture design and coherent processing performance assessment. This study presents a method for measuring horizontal coherence length along an array in the presence of interference and background noise. Subarray beamforming suppresses interference from non-target bearings. Effective degrees of freedom (EDOF) and eigenvalue difference entropy (EDE) are constructed from the eigenvalue difference spectrum of the cross-spectral density matrix to reduce the effect of an approximately common noise contribution to the eigenvalues. Numerical mappings to coherence length are established for three spatial coherence functions, accounting for subarray processing effects on signal and noise correlations. At the lower signal-to-noise ratios examined, simulations show that mean EDOF and EDE estimates are generally closer to the true values than power factor (PF) and eigenvalue ratio (ER) estimates, despite correlated noise introduced by subarray processing. Measurements from a convergence zone in the deep western Pacific show that, at higher signal-to-noise ratios, the estimates are comparable in magnitude to PF and ER estimates and exhibit broadly consistent frequency trends, supporting the method’s effectiveness. The estimates vary with time and frequency, and their scale depends on the coherence function model and decay parameter.