嗪吡嘧磺隆在土壤中的吸附特性及其吸附过程模型的建立

磺酰脲类除草剂是应用较为广泛的农药之一，其在土壤中迁移、降解、转化和滞留等多个过程受其吸附、解吸行为的影响。本文以嗪吡嘧磺隆为研究对象，采用批量平衡法研究了其在8种不同类型土壤中的吸附、解吸附行为。结果表明:嗪吡嘧磺隆与土壤溶液接触4 h内为快速吸附阶段。Freundlich模型可较好地拟合嗪吡嘧磺隆在土壤中的等温吸附解吸过程，相关系数 (r) 值在0.9584~0.9973之间。8种土壤对嗪吡嘧磺隆的吸附能力均为弱，吸附常数 (K_f-ads) 在0.281~3.515之间。其中，以黑龙江白浆土对嗪吡嘧磺隆的吸附能力最强，且远高于其他土壤。除广西赤红壤外，嗪吡嘧磺隆在其他7种类型土壤中的滞后系数 (H) 均小于1，解吸过程存在滞后现象，存在潜在环境风险。单因素试验结果表明，嗪吡嘧磺隆在土壤中的吸附行为受腐殖酸的影响极显著 (P＜0.01)，受pH值和Mn2 + 的影响显著 (P＜0.05)，受高岭土和稻壳生物炭的影响不显著 (P＞0.05)。采用中心复合试验设计，建立了具有一定预测功能的嗪吡嘧磺隆在土壤中的吸附过程BP神经网络模型，并进行了验证，拟合结果较好。

Adsorption characteristics and simulation of metazosulfuron in soils

Sulfonylurea herbicide is one of the widely used pesticides. Its migration, degradation, transformation, and retention in the soil system are affected by its adsorption and desorption behaviors in soil. In this study, 8 types of soil from different provinces were selected to investigate the adsorption and successive desorption of metazosulfuron using equilibrium batch experiments. The results showed a very rapid adsorption process in the initial stage (0-4 h), then followed by a slow desorption (4-24 h) process. The Freundlich model fits the isothermal adsorption and desorption process of metazosulfuron in soil well, with r values ranged from 0.9584 to 0.9973. All the selected soils had a low adsorption capacity of metazosulfuron, with K_f-ads ranging from 0.281 to 3.515. Among them, the albic soil from Heilongjiang has the strongest adsorption capacity of metazosulfuron, which was much higher than others. Except for latosolic red soil from Guangxi, the desorption hysteresis values in the other 7 selected soils were below 1, indicating that hysteresis existed in the process of desorption, which means potential environmental risks. The single factor experiment result showed that the effect of humic acid on the adsorption capacity of soil to metazosulfuron was extremely significant (P < 0.01), and the effects of pH and Mn2+ were significant (P < 0.05), while the effects of kaolin and biochar were not significant (P > 0.05). Using the results of the central composite design experiment, a BP neural network model with certain prediction functions, was established and verified.