An approach to reconstruct the macroscopic fundamental diagram (MFD) was proposed based on probe data in order to detect and evaluate the congestion of urban traffic network in real time, and analyze the inherent characteristic of urban traffic flow on road network. The characteristics and variation patterns of MFD were analyzed. Given the non-homogeneous distribution of probe penetration rates, procedures were presented to magnify MFD variables with varying penetration rates estimated by the GPS data combined with the fixed detections. The accuracy and the characteristic of MFD were tested with trajectories of 7 109 taxis in Hangzhou, China. An estimation method of expansion coefficients was presented for the MFD reconstruction, using piecewise mean according to the ranges of the flow rate and density, respectively. The method generates a more accurate MFD that can be useful to inform real-time traffic monitoring required to implement control strategies in the field compared with the historical mean as the uniform expansion coefficient.