Subspace-based methods require a large number of sensors for localization of closely spaced sources since the spectral magnitude of Multiple Signal Classification (MUSIC) is used. However, the MUSIC-Group delay (MUSIC-GD) method has been used earlier to resolve closely spaced sources with a limited number of sensors. In this work, the MUSIC-GD method is used in high resolution azimuth and elevation estimation of spatially close sources under reverberant environments over a planar array. The efficiency of the MUSIC-GD method in effectively resolving closely spaced sources, even when the noise eigenvalues change considerably under reverberation, is described and illustrated. Localization error analysis is performed on the proposed method and its performance is illustrated using two dimensional scatter plots. Cramer-Rao lower bound (CRB) analysis is also performed and the CRB is compared with the Root Mean Square Error (RMSE) of the proposed method. Large vocabulary speaker dependent speech recognition experiments are conducted on sentences from the TIMIT database acquired over a planar microphone array. The proposed MUSIC-GD method indicates reasonable improvements in terms of localization and the Cramer-Rao lower bound error analysis. A reasonable reduction is also observed in terms of word error rate (WER) from the experiments conducted on distant speech recognition.