We propose a new dynamically feasible trajectory generation algorithm that incorporates sternward motion for unmanned surface vehicles. This work is motivated by riverine applications where the operating environment is large and poorly known. We extend a navigation approach for forward path planning into a more versatile framework that includes safe and dynamically feasible backward trajectories. We pose the backward trajectory generation problem as a finite-horizon optimal control problem and transform it into a nonlinear programming problem by utilizing the direct shooting method. The nonlinear programming problem is solved using the Hooke-Jeeves numerical algorithm. We provide successful simulation and field-trial results that demonstrate the performance of backward path planning algorithm.