Since the original formulation of the positive-degree-day (PDD) method, different PDD calibrations have been proposed in the literature in response to the increasing number of observations. Although these formulations generally provide a satisfactory description of the present-day Greenland geometry, they have not all been tested for paleo ice sheets. Using the climate-ice sheet model CLIMBER-GRISLI coupled with different PDD models, we evaluate how the parameterisation of the ablation may affect the evolution of Northern Hemisphere ice sheets in the transient simulations of the last glacial cycle. Results from fully coupled simulations are compared to time-slice experiments carried out at different key periods of the last glacial period. We find large differences in the simulated ice sheets according to the chosen PDD model. These differences occur as soon as the onset of glaciation, therefore affecting the subsequent evolution of the ice system. To further investigate how the PDD method controls this evolution, special attention is given to the role of each PDD parameter. We show that glacial inception is critically dependent on the representation of the impact of the temperature variability from the daily to the inter-annual time scale, whose effect is modulated by the refreezing scheme. Finally, an additional set of sensitivity experiments has been carried out to assess the relative importance of melt processes with respect to initial ice sheet configuration in the construction and the evolution of past Northern Hemisphere ice sheets. Our analysis reveals that the impacts of the initial ice sheet condition may range from quite negligible to explaining about half of the LGM ice volume depending on the representation of stochastic temperature variations which remain the main driver of the evolution of the ice system. The main findings of this paper underline the need for conducting studies with high resolution climate models coupled to detailed snow models to better constrain the temporal and spatial variations of the PDD parameters. The development of such approaches could improve the calibration of the PDD formulation which is still widely used in climate-ice sheet studies.