Research On The Application Of SOM Neural Network Models In Ecological Sensitivity Analysis For Ecological Planning

Ecological sensitivity is the adaptation ability of a certain area against outside stress or disturbances without damage to or declination of the environmental quality. Evaluation of ecological sensitivity is to evaluate this ability and is an important content and method for analysis and evaluation of ecological system. Evaluation of ecological sensitivity analyzes the adaptation ability of each system in the area against outside disturbances from the natural resources point of view. It is the basis of ecological planning and is instructive in urban planning, environmental protection planning, etc.Our research reviewed the concept, evolution, development and trend of ecological planning, with emphasis on ecological factors relevant to the single factor sensitivity, including ecological factors relevant to soil erosion sensitivity, acid rain sensitivity, etc. It also introduces the systematic ecological sensitivity and two common methods: ecological factor rating and iterative evaluation of single factor ecological sensitivity rating. And then introduced other method of ecological system evaluation in brief, including: analysis of ecological process, analysis of ecological potential, analysis of ecological pattern, etc. Our research also investigated the production and development of neural networks models, and analyzed in particular the characteristic of structure, composition, learning algorithms of SOM neural networks.After clarifying the concept of relative systematic ecological sensitivity and considering the actual situations in Hangzhou, main ecological problems affecting Hangzhou, soil erosion and acid rain, are analyzed, as well as other factors (e.g. habitat sensitivity, etc.). According to the principle of choosing factors scientifically, representatively, operably, concisely, convergently, etc., 12 ecological factors are adopted in the analysis of relative systematic ecological sensitivity in this paper, including: annual average precipitation, annual average pH value of precipitation, annual average wind speed (1961-1999), the average times of successive 3 days with total precipitation exceeding 150mm in the Mei-yu rainy season (1961-1999), vegetation distribution, soil types, physiognomy, soil erosion, etc. Criteria in evaluating the unit systematic ecological sensitivity are given. An evaluation unit is a grid of 6000(?)6000 square meters, which is then numbered to provide a numbered graph of evaluation units, and.then table of input data are obtained through evaluation criteria. An appropriate SOM network is constructed via experiences, prediction, adjustment, etc., and convergent results of systematic ecological sensitivity for each evaluation unit with training steps of 5, 10, and 20, respectively, are obtained with neural networks toolbox in matlab 7.0. The result after 20training steps is deemed as acceptable, and the evaluation graph of ecological sensitivity in Hangzhou is produced in three ranks, 1-33, 34-67, and 68-100. With the result of single factor sensitivity analysis, the significance of ecological planning in Hangzhou is: the types and possible scales of ecological problems in Hangzhou are clarified, and the regional diversity rules of ecological sensitivity are analyzed according to its developing mechanism;the extent and possible impact of specific ecological environmental problems in Hangzhou is defined through the analysis of single factor sensitivity. And it finally plays a directional role in ecological and other planning for Hangzhou. Besides, it also discussed the advantage of SOM models in ecological sensitivity analysis.In the end, the innovative ideas in the research is discussed: compared with traditional ecological sensitivity analysis, it uses more basic materials, SOM models are used in ecological sensitivity analysis for the first time;the evaluation process is more objective with the support of SOM networks;the concept of relative systematic ecological sensitivity is clarified. Its shortcomings include: its precision is not sufficient and the ranking standard still needs improving. Perspective for forthcoming research is: how to apply 3S technology in ecological sensitivity evaluation and how to establish a scientific evaluation system for absolute ecological sensitivity.