Cold tube banks were subjected to a crossflow of water. Internal cooling of the tubes below freezing resulted in gradual ice deposits to occur on the exterior of the tubes; this moving ice-water interface was then allowed to stabilize. The tube bank heat transfer rate with ice formations was examined in a global sense in Parr I of this study and showed trends significantly different from the convective heat transfer rates applicable to non-icing tube bank designs. These differences could not be resolved on a global scale. Thus, the resulting ice shapes are analyzed in greater depth in the second half of this investigation, with the focus redirected locally with an initial objective to determine the distribution of the local convective heat transfer coefficient h(local) along the ice-water interface. While this provided some new and interesting information, it nevertheless was still not able to explain the reasons behind the disparity. Thus, the problem was re-analyzed from a totally different perspective, as a conjugate problem. It was found that the thermal resistance of the ice layer, due to its relatively low thermal conductivity, cannot be neglected and becomes a controlling parameter in the problem, particularly as the ice thickness increases. In the limit, as the thickness of the ice layer vanishes, the problem is seen to reduce to one of pure convection, but, as the ice layer thickens, it quickly becomes one of conduction dominated heat transfer. Finally, the non-uniform shape (taper effect) of the ice deposits on the tubes was found to drive up the heat transfer from the tubes as compared to the heat transfer which would occur if the same amount of ice were uniformly distributed around the tubes’ perimeter.