摘要
生态系统的脆弱性评价对于生态系统的管理具有重要作用。在分析生态系统脆弱性特征和影响因素的基础上 ,构建了针对森林和草地生态系统的脆弱性评价指标体系 ,涵盖了生态系统的结构、功能和生境 3个方面 ,评价指标分别是物种多样性、群落覆盖度、NPP、建群种年生长量、地表干燥度以及土壤有机碳等。评价系统将生态系统的脆弱性划分为轻微脆弱、中度脆弱、重度脆弱以及系统崩溃 4级。作为案例研究 ,构建了结构和性能优化的多层感知器 ,评价了温带落叶阔叶林生态系统的脆弱性。结果表明 。
Assessing the fragility of ecosystem has an important role in sustainable ecosystem management. Based on a critical review of current research, we developed an indictor system for assessing the fragility of natural ecosystems, which includes NPP, species biodiversity, community coverage, annual growth of dominant species, surface aridity, and density of soil carbon. These variables affect the structure, function, and habitats of ecosystems. The proposed indicator system classifies the fragility of natural ecosystem into four levels, i.e., slight, medium, severe and collapse.Defining the ecological baseline of an ecosystem is a precondition for establishing assessment standard. Here, two approaches were employed to build ecological baselines. One is the fundamental niche of dominant species obtained from their ecological ranges and the other is the average value of the structural or functional indices of the ecosystem in its distribution region.As a case study, we developed the specific assessing indicators for temperate deciduous broad-leaved forests according to the defined ecological baseline and general grading standard of assessment system. By using intelligent solver of artificial neural network (ANN), a Multi-layer Perceptron (MLP) which is a supervised learning ANN, was developed to assess the fragility of ecosystem in the case study area. The topological structure of MLP is 6×6×1, namely six neurons in the input layer, six neurons in the hidden layer, and one neuron in the output layer. The input variables of MLP are above-mentioned six indicators, and the output variable is the fragility index of the ecosystem. In order to train and test the MLP model, the training data of the six indicators were linearly interpolated.Based on the ANN assessing results, during 1961-2000, the fragile index of the temperate deciduous broad-leaved forest ranged from 0.64 to 1.80, which indicated slight to medium fragility. The fragility index showed slight increase with a 0.003/40a tendency rate. The tendency rates of fragility index, however, were all negative and exhibited decreasing but fluctuating trends during the first 30 years of the whole assessing period and then increased during last decade.Artificial neural network, originally inspired by their biological namesakes, are composed of many simple intercommunicating elements, or neurons, working in parallel to solve a complex problem. In this study, we tested the feasibility of ANN in assessing the ecosystem fragility. The results indicated that ANN can be an alternative assessment approach, and it shows advantage of fault tolerance, robustness in modeling process.Like other approaches, ANN-based methods also face some difficulties due to lack of data for choosing indicators. Further improvements should include developing appropriate ecological baseline and grading standard for each indictor, and considering other factors such as topography in assessing ecosystem.
出处
《生态学报》
CAS
CSCD
北大核心
2005年第3期621-626,共6页
Acta Ecologica Sinica
基金
国家科技攻关资助项目 (2 0 0 1-BA611B-0 2 -0 5 -0 3 A)
中国科学院知识创新工程重要方向资助项目 (KZCX3 -SW-3 3 9-0 5 )~~
