电脑散热风扇灰尘中多溴二苯醚的污染特征和环境健康风险评价

不同模型对土壤污染物空间分布预测精度具有重要影响，针对现有方法不能较好模拟土壤污染物较强的空间变异特征以及缺乏对影响污染物空间分布的关键环境因子识别，本研究基于随机森林（RF）模型，通过融合多源环境要素，开展了某冶炼厂周边农田土壤砷含量空间分布预测研究，并与反距离加权（IDW）和逐步线性回归模型（STEPREG）相比较.结果表明，研究区农田土壤砷污染范围较广，污染严重区域主要分布在研究区南部，3种模型模拟的砷污染空间分布虽总体趋势相似，但局部区域差异明显，IDW和STEPREG模型不能很好地反映研究区土壤污染的强空间变异特征，RF模型模拟结果较好的表达局部高污染区域的细部变化.不同环境要素对农田土壤砷含量空间分布影响的重要性不同，研究区环境变量和地形变量是影响土壤砷含量空间分布的关键环境因子.交叉验证结果表明，RF模型相对IDW和STEPREG模型具有最小的均方根误差（RMSE）、平均绝对误差（MAE）、平均误差（ME）和最大的R²，RF模型的RMSE、MAE、ME较IDW模型分别降低了10.8%、5.5%和88.1%，较STEPREG模型分别降低了17.8%、18.4%和94.7%，表明采用RF模型对研究区农田土壤砷含量预测精度最高，取得了最优的预测效果.本研究结果能够为土壤重金属污染空间分布制图提供方法学参考.

Contamination characteristics and environmental health risk assessment of polybrominated diphenyl ethers in dust from cooling fans in computers

Different models have important effects on the prediction accuracy of spatial distribution of soil pollutants. The existed methods have limitations in simulating spatial variation and distribution of soil pollutants and identifying the key environmental factors. With the random forest (RF) model of integrating multi-source environmental factors, the spatial distribution of arsenic content was studied in farmland soil around a smelter, and compared with the inverse distance weighted (IDW) and stepwise regression (STEPREG) models. The results showed that arsenic pollution was widely distributed and the south areas were seriously polluted in farmland soil. The overall spatial distribution of arsenic pollution simulated by the three models was similar, but the IDW model and STEPREG model showed less spatial variation of soil pollution, while RF model simulated better spatial variability of heavy arsenic polluted area. The environmental factors were of different influences on the spatial distribution of arsenic content. Environmental and topographic variables in the study area were the key environmental factors influencing the spatial distribution of arsenic in soil. The cross-over studies have shown that the RF model has the smallest RMSE, MAE, ME and the largest R². Compared to the IDW and STEPREG models, the RMSE, MAE and ME of the RF model decreased by 10.8%~94.7%, respectively. Thus, the RF model has more accuracy for predicting spatial distribution of arsenic content in this study, providing the reference of cartographic methodology.