Adaptive systems are required to address variabilities in their dynamic execution environments and continue to provide their services with the expected qualities in spite of variabilities. Currently, the available approaches to specify requirements for adaptive systems focus on what the system should do and what performances should be ensured in any condition. They usually do not specify explicitly the variabilities a system may cope with. When they are specified, it is through a natural language and in an ah-hoc manner for a particular system. However, variabilities are key aspects in adaptive systems, and they trigger the adaptation process. Therefore, they should be explicitly considered and modelled in order to be properly managed through the entire life-cycle of a system. In this paper, we propose a systematic way to describe explicitly variabilities together with meaningful details such as their context, impact, and mitigation strategies, Our approach is based on the available taxonomy for the description of the sources of uncertainty and on RELAX, a requirements specification language for self-adaptive systems, We illustrate the application of our approach through the variabilities description for TAS - Tele Assistence System, a service based solution in the healthcare domain.