我国土壤受试植物筛选与毒性预测

Screening of soil test plants and developing of their toxicity prediction models in China

生态毒性数据缺乏是我国土壤基准与生态风险评估研究中一直存在的问题,开展本土受试植物的筛选可提供更多的生态毒理试验选材,从而获得不同物种的生态毒性数据.鉴于植物对土壤污染物的敏感性,从被子植物中依据分布范围、代表性和易于获得性等原则,对我国潜在的受试植物进行筛选,结果发现,13科53种被子植物分布广泛且易于获取,可作为本土受试植物;结合生态毒性数据的搜集与分析,其中12种受试植物的生态毒性数据相对丰富,并对部分典型污染物表现高敏感.此外在受试植物生态毒性预测模型研究方面,对12种受试植物两两进行建模预测,共得到132个物种种间关系估算模型(Interspecies Correlation Estimation,ICE),其中88个为显著性模型(F检验P<0.05);此处,回归分析了已构建ICE模型的评价参数,得出预测效果较好的ICE模型应满足交叉验证成功率≥80.00%、MSE≤0.62、R^(2)≥0.76和分类学距离≤4的标准.最终筛选出25个符合上述标准的ICE模型,涉及禾本科-禾本科、十字花科-十字花科的相互预测,其中当燕麦Avena sativa、芜青Brasrapa、普通小麦Triticurn aestivum、玉蜀黍Zea mays和黑麦草Lolium perenne等作为替代物种时,预测物种的实际生态毒性值与预测值较为接近.受试植物的筛选与生态毒性预测模型的建立有助于生态毒性数据的产生,并为土壤污染管理和生态风险评估提供科学依据.

The lack of ecotoxicity data has always been a problem in the research of soil criteria and ecological risk assessment in China.The screening of native test plants can provide more ecotoxicity test materials,so as to obtain the ecotoxicity data of different species.In view of the sensitivity of plants to soil pollutants,potential test plants in China were screened from angiosperms according to the principles of distribution range,representativeness and accessibility.The results showed that 53 species of angiosperms in 13 families were widely distributed and easy to obtain,and could be used as native test plants.The result showed that ecotoxicity data of 12 plant species were relatively abundant.Therefore,the ecotoxicity prediction models were developed in this study.A total of 132 Interspecies Correlation Estimation(ICE)models were obtained for the 12 plant species,of which 88 were significant models(F-test P<0.05).Moreover,the selection and evaluation principle of the constructed ICE models were analyzed,and it was concluded that the ICE models with better prediction effects should meet the principle of cross-validation success rate≥80.00%,MSE≤0.62,R^(2)≥0.76 and taxonomic distance≤4.Finally,25ICE models were screened meeting the above principle,involving the mutual prediction of Gramineae-Gramineae and Cruciferae-Cruciferae.Among which when Avena sativa,Brassica rapa,Triticum aestivum,Zea mays and Lolium perenne were used as the surrogate species,the estimated toxicity values of predicted species were close to the actual tested values.The screening of the test plants and the establishment of the ecotoxicity prediction model can help generate ecotoxicity data and provide a scientific basis for soil pollution management and ecological risk assessment.