旱区生态环境质量的综合定量评价模型

Comprehensive quantitative assessment models for ecological environment in arid area

生态环境质量是社会经济可持续发展的基础。因此 ,建立旱区生态环境质量的定量评价是区域可持续发展的主要依据。现有的评价方法大多是通过根据评价区评价指标量化值与评价等级标准来建立评价模型 ,评价区不同 ,评价模型也不相同 ,计算工作量较大。根据给定的生态环境质量评价等级标准 ,采用随机技术模拟生成足够数量的评价指标序列 ,应用人工神经网络模型 (artificial neural network,ANN) ,以评价指标生成序列与其所属的评价等级值进行网络训练。网络训练后 ,以评价区的评价指标为网络的输入 ,通过计算 ,可获得相应的生态环境质量评价等级值。并以甘肃省石羊河流域生态环境脆弱的民勤县为研究对象 ,应用 1975～ 2 0 0 0年资料进行了实例研究。结果表明 ,民勤县 1975～ 2 0 0 0年生态环境质量效应评价值分别为 2 .95 0 1,4 .0 0 90 ,4 .1342 ,4 .16 37,4 .9736 ,5 .0 12 8,说明该地区的生态环境质量是持续下降的 ,与以往采用的模糊综合评价等级值 3,4 ,4 ,4 ,5 ,5一致。文中 ANN模型建立后 ,对于不同评价区 ,只要给定相应的评价指标值 ,通过 ANN模型计算 ,可直接得出生态环境质量评价等级值。因此 ,模型具有实用、可操作性强的特点 ,大大减少了评价区的计算工作量 。

The quantitative assessment of local environmental conditions constitutes one of the most important issues for researchers in the area of sustainable development in arid areas. During the past two decades, a number of methods have been used to assess environmental quality. The methods include: comprehensive assessment methods, fuzzy assessment methods, gray system clustering, principle component analysis, factor analysis and projection pursuit algorithms. These models usually establish functions based on assessment indicators and their assessment grade values. Variation in these models is high and the calculations are cumbersome. Environmental quality, economics, and society constitute a very complicated system. The system is influenced by multiple factors and constraints, such as weather, geography, political systems and world affairs. The factors in this complex system have mutual effects and different forms of uncertainties, therefore the relationship between indictors and their assessment grade value is nonlinear. An artificial neural network (ANN) is a mathematical model structure that is capable of representing the complex nonlinear processes related to the inputs and outputs of any system. ANNs also have excellent nonlinear approximation capabilities. Due to these factors ANN techniques have drawn a lot of interest from researchers studying environmental quality. In this case study, we applied ANN techniques to analyze data from Minqin County, Gansu Province, China. The area, which is in the Shi Yanghe basin, represents one of the most environmentally degraded areas in all of China. Our purpose was to establish a general-purpose model for the quantitative assessment of environmental quality in arid areas. The ecological environment quality indicator system in the study area had 32 indicator and assessment criteria. The criteria were classified into 5 grade values, i.e. 1, 2, 3, 4 and 5, where 1 = excellent, 2 = good, 3 = average, 4 = bad, and 5 = worst. Our work consisted of three steps. First, according to assessment criteria, we used stochastic simulation methods to generate 21 training samples (#1～#21) containing 10, 20, 30, 40, 50, 60, 70, 80, 90, 100, 200, 300, 400, 500, 600, 700, 800, 900, 1000, 1500 and 2000 indicator groups for each grade value respectively. Their sample sizes for training samples #1～#21 were 50, 100, 150, 200, 250, 300, 350, 400, 450, 500, 1000, 1500, 2000, 2500, 3000, 3500, 4000, 4500, 5000, 7500, 10000 respectively. Secondly, a 3-layer Back Propagation (BP) network was built to assess the environmental quality. It had one input layer, one hidden middle layer and one output layer. The stochastic simulation indicators were used as inputs and their corresponding grade values as target outputs to train the network. The input and output data were normalized to fall in the range [0.05,0.90]. The network was trained until the mean squared error (MSE) was less than 0.001. After that, weights and biases were obtained for the 21 ANNs. Finally, environmental indicators collected in Minqin County from 1975 to 2000 were input into each network and the results assessed. According to the assessment results, the assessment values were stable after the sixth sample (training sample size = 600). This indicated that the model can be used to assess environmental conditions in arid areas. The assessment results for 1975, 1980, 1985, 1990, 1995, and 2000 were 2.9501, 4.0090, 4.1342, 4.1637, 4.9736, 5.0128 respectively. The results are in agreement with results derived using fuzzy assessment methods (Shi Y W, 2003), and indicate that environmental quality in the study area is decreasing. This decline can be attributed to an overall shortage of water in the region as well as inefficient use of the water that is available. We suggest that the situation could be improved by constructing water saving systems. The model in this paper requires iterative calculations, but with the ANN calculation toolbox conta