Urban road transportation performance is the result of a complex interplay between the network supply and the travel demand. Fortunately, the framework around the macroscopic fundamental diagram (MFD) provides an efficient description of network wide traffic performance. In this paper, we show how temporal patterns of vehicle traffic define the performance of urban road networks. We present two high-resolution traffic data sets covering a year each. We introduce a methodology to quantify the similarity of macroscopic traffic patterns. We do so by using the concepts of the MFD and a dynamic time warping (DTW) based algorithm for time series. This allows us to derive a few representative MFD clusters that capture the essential macroscopic traffic patterns. We then provide an in-depth analysis of traffic heterogeneity in the network which is indicative of the previously found clusters. Thereupon, we define a parsimonious classification approach to predict the expected MFD clusters early in the morning with high accuracy.