In this paper we discuss the generation of models for emissions of a Diesel engine, produced by genetic programming based evolutionary system identification: Models for the formation of NOx and particulate matter emissions are identified and analyzed. We compare these models to models designed by experts applying variables section and the identification of local polynomial models; analyzing the results summarized in the empirical part of this paper we see that the use of enhanced genetic programming yields models for emissions that are valid not only in certain parts of the parameter space but can be used as global virtual sensors.