The merger of a supermassive binary black hole (SBBH) is one of the most extreme events in the universe with a huge amount of energy released by gravitational radiation. Although the characteristic gravitational wave (GW) frequency around the merger event is far higher than the nHz regime optimal for pulsar timing arrays (PTAs), non-linear GW memory might be a critical smoking gun of the merger event detectable with PTAs. In this Letter, basic aspects of this interesting observation are discussed for SBBHs, and the detection numbers of their memory and inspiral GWs are estimated for ongoing and planned PTAs. We find that the expected detection number would be smaller than unity for the two types of signals even with the Square Kilometre Array. We also provide various scaling relations that would be useful to study detection probabilities of GWs from individual SBBHs with PTAs.