The paper describes a qualitative reasoning approach aimed at representing and interpreting a dynamic process evolution. The problem specifically addressed is the presence of multiple timescales in complex systems. Definitions of a temporal granularity as well as related concepts are provided. For the representation of a single process, a segmentation and abstraction method is described. The identification of dynamic features at any level of abstraction then supplies a help to better choose relevant sampling frequencies of the simulated process and helps in interpreting its outputs. An example is given for a complex crop growth model, whose interpretation is tested against expert knowledge.