Quality of models is paid more attention, when models are reused during software development. Model libraries such as MoCCa support model-driven development by offering reusable models. However, an approach for improving model quality during evolution in model libraries is missing. In model libraries models have to fulfill different quality aspects of distinct modelers in order to be reused in software. These quality dimensions include social aspects such as appeal, which are hard to measure. During model evolution in model libraries, model quality has to be retained to keep reusability of models or it has to be improved to make models reusable. In this thesis, an approach to describe model evolution in model libraries is proposed, which partitions model evolution into four stages: vague, elaborating, and productive. Each stage characterizes a different modeling goals and the model?s reusability during model evolution. In addition, four measurable model quality dimensions for models in model libraries are introduced: syntactic, semantic, pragmatic, and emotional quality. Based on the staged model evolution approach and measurable model quality dimensions an approach is proposed to improve model quality during model evolution.