The Tapas project investigates novel approaches for accurately modeling and analyzing the behavior of various non-stationary characteristics of network links (e.g., error, latency, etc.). Simple application of traditional modeling approaches such as Discrete Time Markov Models are limited in their ability to model non-stationary or time-varying characteristics. We focus on the wireless domain where the problem is exacerbated by fading events which create extreme burstiness in wireless data. We show that when these traditional approaches are used for simulation during the development of new application or network protocols, they can lead to incorrect conclusions and decisions. We have developed modeling methodologies based on data preconditioning for the analysis and modeling of non-stationary datasets.