Knowing the conditions on an emergency environment is useful for rescuers to take critical decisions in saving a life. A specific case is an ice-covered lake where accidents of people falling through ice occur. It is desirable that trained rescuers reach the victim by going over the ice within three minutes from the moment that they arrive to the scene . However, it is difficult to achieve this efficiency since they start making a plan after arriving and evaluating the situation. Therefore, we propose an ice rescue support system for estimating dangerous zones. Rescuers are provided with the application to visualize the ice condition by computer graphics (CG) and plan candidate routes before arriving. A model  has obtained the average of ice thickness for the whole lake. We consider a detailed thickness resolution by a physics based approach because it varies between small distances. Lake morphology is needed to predict ice variations although most lakes have not been surveyed. Our method estimates the morphology based on a model  from hydrology, and constructs geometrical representations that contribute to predict the ice condition over time.