【作者】 陈波宇；

【导师】 赵劲松；

【作者基本信息】 华中农业大学 ， 环境科学， 2011， 硕士

【摘要】 物种敏感度分布(species sensitivity distribution,SSD)是一种基于单物种测试的外推方法,其在污染物环境质量标准的制定与污染生态系统的生态风险评价中的应用近年来已成为研究热点。本研究针对水环境中日益严重的有机磷农药污染问题,选择广泛使用的毒死蜱作为研究对象,利用其对水生生态系统中不同营养层次生物物种的半数效应浓度(median effective concentration, EC50),建立基于对数-逻辑斯蒂、对数-正态、对数-三角形、韦伯分布的水生生物SSD模型,采用概率图和拟合优度检验方法对模型进行了检验和评价,并应用参数Bootstrap对模型参数及模型输出的5%危害浓度(hazardous concentration for 5% of the species, HC5)和潜在影响比例(potentially affected fraction, PAF)进行不确定性分析。论文还从模型选择角度探讨了降低SSD的不确定性的方法,同时研究了小样本数据构建SSD的方法,并确定出构建SSD所需的最小样本容量。本研究可为环境管理机构对受污染水环境进行快速、科学的生态风险评价,以及制定合理、安全的水质标准提供理论依据和方法支持。主要结果如下：(1)对四种确定概率SSD模型进行概率图和拟合优度检验,结果表明选用的四种累积函数分布均可以较好地拟合物种敏感度数据。依据Anderson-Darling (A-D)检验统计量的比较,确定对数-三角形分布对数据拟合得更好,其参数为a=-2.2260±0.4016,b=3.8306±0.1588,c=2.9032±0.0401。(2)根据建立的对数-三角形SSD模型,预测毒死蜱的HC5值为0.1048μg/L,据此获得毒死蜱的最大浓度标准值(criteria maximum concentration, CMC)为0.0524μg/L。依据毒死蜱的监测数据及对数-三角形SSD模型,预测渤海莱州湾海域、伊波兰加河、普雷图河及波隆伊河中毒死蜱PAF分别为0.01%、4.84%、11.01%、12.35%。简单地表征了渤海莱州湾海域生态系统的生态风险很小；伊波兰加河遭受的风险较小,但需加强对该水域的监测,防止有较高风险危害发生；普雷图和波隆伊这两个水域生态系统遭受的风险较大,应禁止毒死蜱的使用。(3)模型参数的不确定性分析表明,四种SSD模型均具有良好的稳健性。但是四种SSD模型输出的HC5和PAF的估算结果存在明显的差异,说明模型选择造成SSD的不确定性。(4)为了降低模型选择带来的不确定性,采用了非参数Boostrap方法构建毒死蜱的SSD,预测毒死蜱的HC5值为0.0700μg/L,CMC值为0.0350μg/L。预测伊波兰加河、普雷图河及波隆伊河中毒死蜱PAF值分别为7.00%、16.00%、16.00%。渤海莱州湾海域中毒死蜱PAF值在分布下限之外,生态系统的生态风险很小。与对数-三角形SSD模型相比,非参数Bootstrap方法获得的结果较为保守。(5)依据参数Bootstrap方法,确定至少需要9个不同物种急性毒性数据才能构建SSD。而依据非参数Bootstrap方法,确定出使用非参数Bootstrap方法构建SSD所需的最小样本容量为10。更多还原

【Abstract】 Species sensitivity distribution (SSD) is an extrapolation method, which is based on the single-species toxicity test. In recent years, the applications of SSD in the derivation of environmental quality criteria and ecological risk assessment for contaminated ecosystems have become the research hotspots. The increasing pollutions of organophosphorus pesticides (OP) in water have been of concerns. Taking the widely used chlorpyrifos as the object, the SSD models were developed on the median effective concentrations (EC50) of chlorpyrifos to aquatic species at various trophic levels. Four different distributions, i.e., log-logistic, log-normal, log-triangular and Weibull, were chosen to fit the SSD. Then, all models were tested and evaluated using probability plots and good-of-fit tests. The uncertainty analysis for model parameters, hazardous concentration for 5% of the species (HC5), and potentially affected fraction (PAF) were performed through parametric bootstrap simulations. The ways of reducing the uncertainty of SSD through the model selection were discussed in the paper. Besides, the development of SSD based on the small data sample and the minimum sample size for constructing a valid SSD were also studied. The present study will facilitate the environmental regulatory agency to perform ecological risk assessment to a polluted aquatic environment, and will offer the scientific evidences and technical supports in derivation of water quality criteria. The main results are as follows:(1) The results of probability plots and good-of-fit tests showed that SSDs for aquatic biota exposed to chlorpyrifos were well fitted with all the specific probability distributions. Based on the Anderson-Darling test, the fitted log-triangular distribution was preferred, which was determined by three parameters, a=-2.2260, b=3.8306, and c =2.9032, with standard error 0.4016,0.1588 and 0.0401,respectively.(2) According to the built log-triangular SSD model, the HC5 and the criteria maximum concentration (CMC) of chlorpyrifos were 0.1048μg/L and 0.0524μg/L, respectively. In addition, according to the reported data and log-triangular SSD model, the PAF of species exposed to chlorpyrifos in the Laizhou Bay (Bohai Sea, China) area, Iporanga River, Preto River and Piloes River were 0.01%,4.84%,11.01% and 12.35%, respectively. These PAF values could simply characterize that the ecological risk of chlorpyrifos in Laizhou Bay area was very low; the ecological risk of chlorpyrifos in Iporanga River was lower, but the strengthening monitoring was a necessary measure to prevent a higher risk harm happening; the ecological risks of chlorpyrifos Preto River and Piloes River were higher, and chlorpyrifos should be banned.(3) The results of uncertainty analysis for model parameters indicated the high robustness of each model. Obvious differences of the estimation results of HC5 and PAF among all SSD models showed that the model selection led to the uncertainty of SSD.(4) In view of reducing the uncertainties of SSD through the model selection, the nonparametric bootstrap method was introduced. The median estimation of the HC5 and CMC of chlorpyrifos were 0.0700μg/L and 0.0350μg/L, respectively. The PAF of species exposed to chlorpyrifos in Iporanga River, Preto River and Piloes River were 7.00%,16.00% and 16.00%, respectively. The PAF of species exposed to chlorpyrifos in the Laizhou Bay areawas below the lower tail of Bootstrap distribution, and the corresponding risk might not be significant. The nonparametric bootstrap method gave the more conservative results when compared to the log-triangular SSD model.(5) Based on the parametric bootstrap method, a sample with size 9 was minimum requirement for a valid and valuable SSD,while the nonparametric bootstrap method gave a sample size with 10 that was the minimum for constructing a valid SSD.更多还原