Studied site is located in Jingou forestry farm Wangqing forestry bureau Jilin province,one representative of over-logged forest areas in Northeastern China.The purpose of this study is to seek a spatial way for forest sustainable management, forest landscape and stand are selected for the two important scales. The main methodologies include nonlinear multi-objective integer programming, GIS technology application, TWINSPAN quantitative classification, mingling analysis, competition analysis, spatial pattern analysis, greedy algorithm, Monte Carlo algorithm and object-oriented programming(OOP),etc. The main conclusions are as follows: 1. In theory(l)The basic frame of spatial way for forest sustainable management includes two important scales: forest landscape and stand. In landscape scale, the objective of forest management is to keep integrity and stability of forest landscape. In stand scale, the objective of forest management is to maintain health and energy of stand. No matter what scale, forest protection is more important than forest utilization.(2)The theory of point structure analysis includes tree point structure and stand integer structure. Tree point structure is part spatial structure of stand. Stand integer structure is the average structure of tree point structure. It is possible that the theory of point structure analysis unify many aspects of forest management.(3)Based on the modeling theory and method of spatial structure optimization, one spatial structure optimizing model of stand selective thinning is built. The model integrates modern forest management theory and information technology, and is successfully connected with control method. The model belongs to nonlinear multi-objective integer programming, and can be solved by Monte Carlo algorithm.(4)The maximal covering model is introduced in reserve site selection of forest type diversity, greedy algorithm is an available algorightm.(5) The method of TWINSPAN quantitative classification is applied to classify mixed forest and therefore a succession serial and law of mixed forest has been gotten. The result shows that TWINSPAN method is also suitable for mixed forest classification and can indicate gradient of time and space in the process of mixed forest succession to some extent.(6) Tree species diversity mingling, an improved spatial structure index, can indicates the real mingling degree of mixed forest.(7)It is proved that if distance is given, competition of trees in the same DBH class isintenser than that of those in different DBH class. Dissymmetric competition of trees is probably one of the reasons that uneven-aged forest has high stability.(8)In comparison with refined nearest neighbor analysis, only Ripley's K(d) can analyse spatial pattern in each distance scale and it is a better method.(9)It is found that there is the fourth case in the edge correction of Ripley's K(d) function, and an improved method is brought forward.(10) Among three kinds of Monte Carlo test methods for spatial pattern, the one based on largest DBH is the best. 2. In application(1) Forest of study area is classified into three groups of forest type and twenty-two forest types in qualitative and quantitative method..(2)The maximal covering pattern includes seven compartments, 31,37,39,48,51,55 and 64 compartment, can be selected for the reserve sites of forest type diversity.(3)Based on 20 plot of natural forest, tree competition index has exponential descending relation to DBH. Cultivation cutting is available just before tree's DBH reaches 30.9 cm. Intrespecific competition is intenser than that of intraspecific, but bass tree's intraspecific competition is the intensest. Conifer species have weaker competition ability than broadleaves. So intrespecific competition should be the main regulative relation in thinning, removing of broadleaf trees which influence conifer trees will advantage the evolutional succession of ecology system.(4)The theory of tree point structure analysis is applied in stand spatial str