Spatial And Seasonal Patterns Of Dissolved Organic Matter Hydrophobicity In Lake Taihu And Organic Sorption Model

Dissolved organic matter（DOM）widely exists in natural water and is an important adsorbent for organic pollutants in the environment.The partition process of organic pollutants to dissolved organic matter significantly influences their environmental fate and bioavailability.The hydrophobicity of dissolved organic matter which is reflected by the percentage of non-polar components is a key property influencing the partition process,which is usually quantified by organic carbon-water partition coefficient(K_OC).The study of the spatial and seasonal pattern of DOM hydrophobicity and its major controlling factors and the accurate and high throughput K_OC prediction model is still limited.In this study,high-resolution DOM samples were collected from northern Lake Taihu and characterized with water quality parameters,optical indices and hydrophobicity(K_ATPS).The high-resolution spatial patterns and seasonal variations of DOM hydrophobicity as well as its major controlling factors were assessed.Furtherly,log K_OC prediction model was established based on the normalized UV-visible spectra of DOM and the octanol-water partition coefficient(K_OW)of organic pollutants.The applicability,interpretability and optimization of the model were analyzed.The main findings achieved are summarized as follows:（1）K_ATPSis a simple and quantitative measure of DOM hydrophobicity,which is suitable for spatiotemporal investigations.K_ATPS of DOM from northern Lake Taihu varied from 0.44 to 3.00 in March and 0.15 to 3.17 in July,showing strong spatial heterogeneity and DOM from Zhushan Bay in July was statistically more hydrophobic than that in March.The hydrophobicity of DOM in northern Lake Taihu was significantly affected by the riverine input,algae activity and photodegradation.In terms of DOM hydrophobicity,the west part of the lake strongly affected by the riverine input in March and by the riverine input and the algae activity in July.The central area and the east part of the lake were affected by the algae activity and photodegradation respectively.（2）log K_OC can be accurately assessed with the K_ATPSand log K_OW dataset by the two-phase system model.Taking phenanthrene for demonstration,the spatial and seasonal pattern of log K_OC in northern Lake Taihu varied significantly.log K_OCvaried from 2.81 to 3.88 in March and 1.86 to 3.90 in July.The K_OCvalues in different regions of northern Lake Taihu varied more than one magnitude and log K_OCof Zhushan Bay and East coast had significant difference between March and July.The result indicates that the hydrophobicity of DOM should be considered into the risk assessment of pollutants.（3）log K_OC prediction model was established using random forest algorithm based on the normalized UV-visible spectra of DOM and the log K_OW of organic pollutants,which can be used on aquatic,sediment,soil and peat DOM.The out-of-fold root mean squared error（RMSE）of all the samples in the dataset was 0.259 and the out-of-fold RMSE of soil and peat DOM was 0.291.The features important for the model were related to the molecular size,the extent of humification and the substituents on aromatic rings of DOM,which were significant for organic sorption.（4）With feature selection or dimension reduction,RMSE of log K_OC prediction model can be reduced by 0.007 and 0.011.Feature selection can also reduce the measuring cost.With the high throughput of UV-visible spectra measurement and log K_OW,random forest model can be applied to study of sorption behavior of DOM in large scale region.High throughput log K_OC prediction can facilitate our understanding of the environmental fate and bioavailibility of organic pollutants.This study analyzed the controlling factor of the spatial and seasonal pattern of DOM hydrophobicity and establish log K_OC prediction model,which can facilitate our understanding of the fate of organic pollutants and help realize accurate risk assessment and management.