We use Monte Carlo simulations to identify the mechanism that allows for phase transitions in dipolar spin ice to occur and survive for applied magnetic field, H, much larger in strength than that of the spin−spin interactions. In the most generic and highest symmetry case, the spins on one out of four sublattices of the pyrochlore decouple from the total local exchange+dipolar+applied field. In the special case where H is aligned perfectly along the [110] crystallographic direction, spin chains perpendicular to H show a transition to q = X long range order, which proceeds via a one to three dimensional crossover. We propose that these transitions are relevant to the origin of specific heat features observed in powder samples of the Dy2Ti2O7 spin ice material for H above 1 Tesla. [5]. For a field H larger than 1 Tesla (T), a sharp peak suggesting a phase transition occurs at a field-independent temperature of 0.35 K, surviving up to the largest field considered (6 T). Another rounded peak at 1.2 K first appears at 0.75 T and also remains observable up to 6 T, although it is less sharp than the 0.35 K feature. Finally, a third peak at 0.5 K first appears at 1 T, where it is quite sharp, but disappears for H > ∼ 3 T. Over the past six years, much theoretical effort has been devoted to the search for a microscopic explanation of the field-induced spin correlations responsible for these specific heat fea- suggested that some of these features might be due to phase transitions arising in crystallites that accidently happen to be oriented such that some of the Ising spins in the unit cell are perpendicular to H. These "fielddecoupled" spins would then be free to interact among themselves, giving transitions at temperatures that are field-independent up to a very large field [5]. A similar proposal had been made earlier for the case of garnet systems [9]. Experimental evidence [10] suggests that the 0.5 K feature in powder samples arises due to a multicritical point at a temperature of 0.5 K and field of 1 T along the [111] crystallographic direction, which is related to the interesting phenomenology of kagomé ice [11]. However, to date, there is no concensus as to whether or not the specific heat features at 0.35 K and 1.2 K are caused by the magnetic field directed on crystallites of particular orientation [12,13,14]. In this paper we address this question by using the dipolar spin ice model, where Ising spins