基于近红外光谱技术的农药土壤吸附系数预测及其流域尺度空间分布规律研究

Sorption is one of the major processes that determine the fate, effects, and ecological risks of pesticides in terrestrial and aquatic environments. Within a watershed, soil properties are varied greatly because of landscape and management practices, which led to spatial variation of pesticide sorption coefficients (Kd) and organic carbon adsorption coefficients (Koc). It needs a mountain of work for accurately quantifying the spatial variability of sorption coefficients using conventional batch equilibrium technique. In this study, it was proposed that near-infrared spectroscopy (NIRS) could be used to more rapidly quantify the sorption variability of pesticides in soils of Baima river catchment. The herbicides atrazine and acetochlor were selected because of their widely application in these areas. And the study area was an agricultural catchment with hummocky terrain landscapes, which was adaptable to our research target. Several chemometrics methods were used to optimize the prediction model, such as continuous wavelet transform (CWT), standard normal variant transform (SNV), multiplicative scatter correction (MSC), Monte Carlo uninformative variable elimination (MC-UVE), randomization test (RT), consensus strategy PLS and so on. Ultimately, through the optimal prediction model, it was attained to predict the spatial distribution of target pesticide sorption coefficients in soil of the catchment, rapidly and accurately. The objectives of this study were to determine the feasibility of NIRS as a rapid tool to quantify the spatial variability of pesticide sorption coefficients in catmint scale. Compared to the methods of batch equilibrium techniques and Koc evaluation, this study was expected to be more cost-effective, rapid, and reliable.

农药在土壤中的吸附是决定其在陆地和水环境命运、效应和生态风险的主要过程之一。农药的吸附系数（Kd）和有机碳吸附系数（Koc）因流域土壤性质空间差异而具有较大变化，详细、准确地获取流域农药吸附特征分布需要大量样品的测定分析，传统分析方法费时费力，具有一定难度。近红外光谱技术具有无需前处理、快速、绿色、廉价等特点，为大规模分析土壤对农药吸附系数的空间分布规律提供了新的可能。本研究拟以我国广泛使用的除草剂阿特拉津和乙草胺为目标化合物，以典型农业小流域内不同区域表层土壤作为供试土壤，运用近红外光谱技术，结合化学计量学方法，通过多种光谱预处理和变量选择方法消除光谱中非目标因素的影响，然后运用多模型共识建模方法，建立适合本流域的土壤吸附系数最优预测模型，并应用此模型预测白马河流域阿特拉津和乙草胺的土壤吸附系数的空间分布规律。相较于传统批量吸附和Koc估算方法，本研究具有快速、廉价、准确等特点。

农药在土壤中的吸附是决定其在陆地和水环境命运、效应和生态风险的主要过程之一。农药的吸附系数（Kd）因流域土壤性质空间差异而具有较大变化，详细、准确地获取流域农药吸附特征分布需要大量样品的测定分析，传统分析方法费时费力，具有一定难度。近红外光谱技术具有无需前处理、快速、绿色、廉价等特点，为大规模分析土壤对农药吸附系数的空间分布规律提供了新的可能。本研究以我国广泛使用的除草剂阿特拉津为目标化合物，以典型农业小流域内不同区域土壤作为供试土壤，运用近红外光谱技术，结合化学计量学方法，通过多种光谱预处理和变量选择方法消除光谱中非目标因素的影响，然后运用偏最小二乘建模方法，建立了适合本流域的土壤吸附系数模型，其预测值和参考值的相关系数为0.8090，并应用此模型考察了白马河流域阿特拉津的土壤吸附系数的空间分布特征，发现表层土壤的吸附系数大于下层土壤，上、中、下游的分布规律不明显。另外，土壤中的C、N、H的含量会影响土壤吸附系数的大小，通过光谱预处理和波长选择方法采用偏最小二乘建立了土壤中C、N、H元素的最优预测模型，其预测值和参考值的相关系数分别为0.9285、0.6599和0.8245。并分析了吸附系数与有机碳含量的关系，发现阿特拉津土壤吸附系数与土壤中有机碳的含量成正比关系。相较于传统批量吸附和定量分析方法，本研究具有快速、廉价、准确等特点。