This paper investigates practical predefined-time trajectory tracking control problem with guaranteed performance for unmanned surface vehicles (USVs) suffering from dynamic uncertainties and unknown disturbances. A observer-based practical predefined-time integral sliding mode control scheme is proposed. Salient features are as follows. First, by devising a practical predefined-time extended state observer, lumped disturbances arising from uncertainties and external disturbances can be exactly identified within the expected predefined-time, in conjunction with the fuzzy logic systems, and thereby improving the disturbance rejection property of USVs. Then, a nonsingular predefined-time integral sliding mode (NPTISM) surface is proposed using a class of time-dependent scaling function, and the NPTISM surface has only the sliding phase, while the reaching phase is eliminated. Besides, a novel-type prescribed performance function is introduced into the controller to characterize transient and steady-state performances, simultaneously. A rigorous stability analysis guarantees that practical predefined-time stability of the system, and the upper bound of settling time can be precisely set by only a single parameter in advance. Finally, the simulation results and comparisons are carried out to validate the proposal.