Tabu search design for difficult forest management optimization problems

A series of tabu search (TS) methods for solving the stand harvesting and road access optimization problem was developed and evaluated. This challenging forest management problem includes spatial constraints for maximum opening size, adjacency delay (green up), as well as timber-flow targets derived exogenously from a strategic planning process. The base harvest decision unit is the stand, and harvest blocks are created dynamically as adjacent stands are scheduled for treatments. The road network subproblem is solved using a fast heuristic method to select a minimum discounted cost set of road construction projects so that scheduled stands are accessible. The TS methods range from simple ones with fixed tabu tenure to an adaptive search with feedback mechanisms to control tabu tenure and to direct the search near constraint boundaries. It was found that while simple TS algorithms can find feasible solutions to the problem, these may be far from optimal. A good short-term memory strategy, constraint boundaries smoothed using penalty functions, and customized diversification moves were important design elements in the most successful TS algorithm for this problem. This paper points out the necessity to design the TS method carefully, since there are many possible TS designs and the design choices matter.