回归模型法推导油菜田土壤Cd限值的不确定性

为研究土壤-作物迁移模型推导农田土壤环境基准的不确定性,以油菜为例,通过贵州省4种典型微酸性土壤——红壤、黄壤、石灰土和黄棕壤进行Cd盐添加的盆栽试验,除全量Cd外,选择5种有效态提取剂,分别与油菜籽w(Cd)建立土壤-作物迁移模型. 参照GB 13078—2001《饲料卫生标准》w(Cd)限值(0.5 mg/kg),采用模型预测中位值和95%预测上限进行土壤Cd限值推导并分析模型的不确定性. 结果表明:全量和各提取态Cd质量分数与油菜籽w(Cd)均能建立极显著的一元(R2为0.907~0.946)或多元(考虑pH)回归模型(R2为0.875~0.962). 5.550(以多元模型中位值推导的农田安全种植油菜的Cd限值)随pH的升高从0.93 mg/kg增至1.45 mg/kg,HNO₃提取态、HCl提取态、DTPA提取态的Cd的质量分数也分别从0.81、0.97、0.73 mg/kg升至1.39、1.79、1.66 mg/kg. 基于CaCl₂提取态Cd的一元模型对pH不敏感,SEQS₅₀不受其影响. 对所有模型而言,SEQS₅₀是SEQS₉₅(基于模型95%的预测上限推导的限值)的1.1倍左右. 其中,CaCl₂提取态Cd的SEQS₅₀和SEQS₉₅分别为0.052和0.049 mg/kg,小于福建省CaCl₂提取态Cd标准限值(0.15 mg/kg). 为降低不确定性,需要选择合适的提取剂,使用全量Cd,HNO₃、HCl或DTPA 3种提取态Cd建模时须兼顾pH的影响. 研究显示,SEQS₉₅能够降低土壤Cd限值达标而作物Cd含量超标的可能性. 

Uncertainties of Soil-Crop Transfer Models for Deriving Critical Values of Cd in Rape Planting Soils

Abstract: This study aimed to investigate the uncertainties of soil-rapeseed transfer models of Cd on deriving soil environmental quality standards (SEQS), and to examine whether it is safe togrowrape on Cd contaminated soils. Due to natural pedogenic processes, with a narrow range of pH (5.54 for pot experiment. Besides the total Cd, five extractants, including 0.43 mol/L HNO₃,0.1 mol/L HCl, 0.05 mol/L DTPA, 0.01 mol/L CaCl₂ and 1 mol/L NH₄OAc, were selected to evaluate Cd phytoavailability and were correlated to Cd concentrations in rapeseed by either linear regression or multi-linear regression (log transformed). Sequentially, soil Cd critical values were calculated by derived models in view of the food quality standard. The results indicated that weak adsorbed Cd in soils with a high Cd background accounted for 16.6% of total Cd content and had a high Cd phytoavailability. Very statistically efficientsoil-rapeseed transfer models were established by both linear (R²:0.907-0.946) and multi-linear (R²:0.875-0.962) regression. Model uncertainties on deriving SEQS values were significant and varied among different models. Within pH range(5.550) (0.93-1.45 mg/kg) increased rapidly with the increase of pH. It was the same case with the derived SEQS₅₀ values based on HNO₃ (0.81-1.39 mg/kg), HCl(0.97-1.79 mg/kg) and DTPA(0.73-1.66 mg/kg), respectively. Only the CaCl₂-based model was not affected by pH.Meanwhile, SEQS₅₀ levels were 1.1 times SEQS₉₅ levels, which were derived from 95% upper prediction limits. Highly predictive models are needed, and pH has to be considered for the application of the total-, HNO₃-, HCl-, or DTPA-based model, during which the 95% upper limit is preferred for SEQS derivation.