Software-intensive systems increasingly pervade our society and economy, and their application in safety-critical contexts can even decide about life or death (e.g., driver assistance systems). Such systems are typically developed in a multidisciplinary manner, are often subject to real-time requirements, and are executed on distributed and concurrent platforms influencing their timing behavior. A high quality of the requirements on these systems' software is crucial, because the software requirements are the basis for the software design and development. The application of models in Requirements Engineering (RE) is considered beneficial, because they foster automatic analysis techniques that aim at ensuring high-quality requirements. However, existing model-based RE approaches take neither the transition from multidisciplinary to discipline-specific RE phases nor platform-induced timing effects during real-time requirements validation sufficiently into account. This results in potential software requirements defects introduced during the transition as well as costly development iterations due to timing analyses conducted in late engineering phases. This thesis proposes and evaluates a model-based RE approach that addresses these problems by means of two techniques. First, it presents a semi-automatic technique for the transition from multidisciplinary system models to software RE models. Second, it presents a technique for the semi-automatic verification of timing-relevant platform properties against real-time requirements as part of the software RE models. These contributions improve the quality of software requirements by reducing the likelihood to introduce requirements defects during the transition from multidisciplinary system to software RE models and by early revealing platform-induced real-time requirement violations.