Designing urban drainage elements can be done by setting criteria for four components: 1) proper planning horizons, 2) indicators to limit the solution space, and 3) an unambiguous calculation method with 4) corresponding input data that together encompass the solution space. Design is then a relatively simple exercise of optimizing against a set of criteria such as economy and technical feasibility. Numerous UDM conferences have discussed these criteria and in particular which modelling principles could best be used to address a given problem. We have over the past years observed this paradigm being challenged. We discuss the challenges by considering the simplest possible urban design problem: designing a retention pond for storm water for an acceptable overflow frequency. Given that we in the small country of Denmark have more than 4000 retention ponds this should be easy. However, we find that we have reached a point where it is very difficult to be able to develop and apply design tools primarily because development of a training data set with corresponding objective function is impossible.
Arnbjerg-Nielsen et al. (Fri,) studied this question.