Forest Multi-objective Planning Based On Simulated Annealing Algorithm

The objectives of forest management planning are to maximize the comprehensive benefits of forest ecosystems when the forest resources are very limited,however forest planning policies have generally focused on the traditional timber production function in many countries over the last decades,ignoring the other ecological and social benefits of forest ecosystem,such as wildlife habitat protection,water conservation and entertainment.Today,carbon sinks of forest ecosystem are received much more attentions due to the dramatically changes of global climate and environment.Therefore,the Paris Convention that combats with global climate changes has recently been accepted by 195 countries across the worldwide,in which the roles of forest ecosystems in mitigating global climate changes were further strengthened.To achieve this goal,the Chinese government has put forward some objectives that the amount of forest volume in 2030 should be increased approximately 450 billions cubic meters when compared that in 2005.Since the seriously shortcoming of afforestation areas in our country,adaptive forest management treatments have become the unique strategy to improve the amount of forest volume and carbon consequestion.Thus,establishing the environmental friendly forest planning models has become the hotspot in the current forest management researches.The overall aims of this paper were theref’ore to establish the forest multi-objective planning models for the Pangu forest farm in the northeastern part of-China by which one could assess either a carbon stocks objective,a timber production objective,or a spatial objective related to the arrangement of forest management activities.DLle to the performance of heuristics are usually sensitive to the parameters used,the paraleters sensitivity of different neighborhood search technieques of simulated annealing were sy’stematicly evaluated when applied to forest spatial harvest scheduling problems.In order to improve the quality of heuristic solutions,the neighborhood search,hybrid search and reversion search processes of simulated annealing were also evaluated in a set of difficultly forest planning problems.Finally,the optimal forest management plans for the research area were produced based on a set of quatificationally simulations and e-valuations.The conclusions and contributions of this paper can be summarized as:1)Parameter sensitivity(Part IV):The effects of the number of initial solution(N).initial temperature(T),cooling rate(r)and the number of-iterations for each temperature(nrep)of threee different neighborhood search techniques of simulated annealing on the objectivc feunction values were systematically analyzed when they xverc applied to five hypothetical grid datasets,which encompassed 400 to 10000 management units.The results indicated that the objective function values were not necessarily increased as the parameter value increases,in which the parameter N had no significant effects on the objective function values,however the parameter T,r and nrep all have significhant effects on the quality of solutions,while the relations might be restricted by the number of management units within a forest landscape.The performance of the three neighborhood search techniques of simulated annealing were all significant related to problem size,in which the number of satifactory solutions significantly increased with the increases of problem size,however the resolution efficiency of maximum solution for each problem all decreased-with the increases of problem size.2)Neighborhood search technique(Part V):The performance of two different neighborhood search techniques of simulated annealing were systematically evaluated when applied to three increasing diflficult planning problems(i.e.,non-spatial,URM and ARM)for one actual forest dataset and four hypothetical datasets with different age-class distributions.The results showed that the 2-opt technique not only produced consistently better solutions than the 1-opt technique in terms of the mean and maximum solutions values,but also significantly decreased the standard deviations of the sets of solutions.The validation results indicated that the maximum solution values were usually more than 95%of the estimated optimal values.Therefore,the neighborhood search technique was more effective to improve the quality of heuristic solutions.3)Reversion search technique(Part VI):The performance of three enhanced techniques(i.e.,neighborhood,hybrid and reversion search techniques)of simulated annealing algorithm were systematically assessed when applied to the forest spatial harvest scheduling problem for the datasets of Pangu forest farm in the part of northeastern China.The results showed that the number of exchange between 1-opt and 2-opt moves did have no significant effects on the maxinum objective function values.but significantly affected the mean objective function values.The three search strategies can significantly improve the qualities of solution,however the optimization time was also significant increased.For a minimization problem,Method 6 produced the best(lowest)mean objective function values,which were less than 1%of the means developed using conventional methods.Although Method 6 performed very well,the other methods should not be neglected because they also have the potential to produce high-quality solutions.4)Multi-objective planning(Part Ⅶ):The effects of age-class distribution and economic parameters on the carbon management objectives were systematically evaluated using the theory of utility function when they were applied to the forest spatial harvest scheduling problems for the dataset of Pangu forest farm in the part of northeastern China.The results showed that carbon density across the forest landscape for the ending of planning horizon were as large as 43.26 t/ha.which was very similar to the current carbon level(45.06 t/ha).The results may have some important meanfuls on forest management practices when considered the large amount of harvest timber(3 ×106 n3)during the whole horizon.The age-class distributions not only affected the objective function values,but also have significant effects on the optimization time.The coefficients of variation for all the planning scenarios were only varied from 0.20%to 1.34%,indicating adequate solution stability.5)FMPS v1.0 software(Part Ⅷ):Since the complexity,integration and special-regional characteristics of forest planning models,thus these planning techniques might be difficult to use in forest management practices.Therefore,the Forest Multi-objective Planning System(FMPS v1.0),including the forest multi-objective planning models,forest growth and yield model systems and simulated annealing algorithm,were developed within the Microsoft Visual Basic 6.0 environment.The software has a friendly interactive interface which is easily to use for forest manager and planner.The running instances also validated the stability of the independent solutions for difficult planning problems,which further indicated that FCSMP can be widely used in the similar planning problems or research area.