A six transceiver ocean acoustic tomography array was deployed to monitor ocean ventilation and circulation over the 1988-89 winter cooling season. A stochastic inverse method computer code which attains a solution by minimizing mean square error is used to perform inversions of the Greenland Sea tomography data. A computer simulated ocean is used to evaluate various aspects of system performance. We first consider the advantages and problems associated with using a ray theory based algorithm. Next, we made two adjustments to our inversion code and discuss the effects on system performance. The first adjustment allows for layers of different thicknesses in the inverse solution to increase the density of estimates in regions of interest. The second adjustment allows the estimator to expect variability of the unknown field to decrease exponentially with depth. The ray theory based algorithm is an adequate method of modeling ray paths in the Greenland Sea, but has limitations. Reliability of ray paths degrades as launch angles become shallower and if strong gradients and rapidly changing gradients in sound speed are present in the vicinity of the transceiver elements. A set of Greenland Sea data between one transceiver pair was processed. Although only three groups of eigenrays are involved, initial inversion results indicate the estimator detects seasonal variations and synoptic scale events occurring at time scales greater than 20 days, however, solutions show wide fluctuations at shorter times.