Abstract : The quality of communication in a wireless network is primarily determined by the wireless link signal quality expressed in term of signal-to-interference-plus-noise ratio. The fact that better signal quality enhances the communication quality incites to look for states where each receiver connects to a transmitter providing it with the best signal quality. Using stochastic geometry and then extreme value theory, we obtain the distribution of the best signal quality, of the interference, and of the maximum signal strength in both bounded and unbounded path loss conditions. We then investigate temporal variations of wireless links, which are also essential to wireless networking, in terms of level crossings of a stationary process $X(t)$. We prove that the length of an excursion of $X(t)$ above a level $\gamma \to -\infty$ has an exponential distribution, and obtain results associated with the crossings of several levels. These results are then applied to mobility management in cellular networks. We focus on the handover measurement function, which differs from the handover decision-execution by identifying the best neighbouring cell to which a connection switching is to be decided and executed. This function has an important influence on the user's experience, though its operation has been questionable due to the complexity of combining control mechanisms. We firstly address this topic with an analytical approach for emerging macro cell and small cell networks, and then with a self-optimisation approach for neighbour cell lists used in today's cellular networks.