| Tracking the sources of main non-point source pollutants is the first step to carry out effective agricultural watershed management. From the perspective of the pollution characteristics of receiving river waters, the source apportionment method has been widely used. This simple method can generate the reliable analytical conclusions without many considerations of the output loads of pollutants and their complex migration processes. Herein, the main agricultural non-point source pollutants were tracked in a typical mountain-plain watershed of Tiaoxi River, which is located in southern Lake Taihu. In the study area, the sources of the high amount of nitrogen in surface water body were still unclear, and the knowledge of dissolved organic matter (DOM) in rivers was extremely lack. Therefore, we have adopted the above strategy to address these problems. The routine quality monitoring, 15N-18O dual stable isotope and fluorescent fingerprint were introduced as the basic research tools to reveal the pollution status of carbon and nitrogen in Tiaoxi River. The data mining methods of parallel factor analysis (PARAFAC) and self-organizing maps (SOM) were also used. Consequently, the sources and spatial occurrence of NO; (the dominating N form) and DOM were fully explored. Base on our study results, we presented some scientific and effective watershed management strategies. Meanwhile, some innovations to the analytical technique of three-dimensional fluorescence data were made. The primary results obtained in this study were as follows:(1) A large amount of water samples were collected in winter from two upstream subcatchments of East and West Tiaoxi River Watershed (ETRW and WTRW). Then, the spatial variations of surface water quality were evaluated by the integrated approaches of traditional and SOM analysis. We found that the overall water quality of WTRW was better than that of ETRW, but they could both meet the Grade III standard of surface water quality in China. The average pollution level of total nitrogen in ponds and ditches was lighter than that in rivers, lakes and reservoirs, but it was opposite for the dissolved organic carbon (DOC), turbidity and NH4-N indexes. Roughly, the water quality of headwater streams, primary tributaries, main streams, reservoirs and small-scale lakes remained good. However, the pollution problems in the secondary tributaries (especially in ETRW), ponds and ditches were obvious. Most of water samples were rich in nitrate, and its contribution to total dissolved nitrogen (TDN) was over 50%. Therefore, the ecological stoichiometric relations using the ratios of nitrogen to phosphorous (N/P) and carbon to nitrogen (C/N) were further explored. Combined with the historical water quality data, we confirmed the high nitrogen pollution problem and the nitrate accumulation mechanism in the study area. In conclusion, we should focus the water pollution prevention and control works on ETRW at the watershed scale. Meanwhile, the water quality management for the secondary tributaries, ponds and ditches should be strengthened. Finally, because of the more serious non-point pollution in summer, the source apportionments of NO3 and DOM in ETRW were suggested as the next works.(2) To reveal the sources of water and NO3-N in the main stream and the tributaries within the ETRW in summer, the D-18O and 15N-18O dual stable isotope methods were applied, respectively. Similarly, the high total N and NO3--N content were presented in all water samples, and NO3--N was one of the most important N forms in water. The δD and δ18O-H2O of all water samples could be well fitted according to the local meteoric water line (LMWL), indicating that the rain was a key driving factor for water resource of ETR. Analysis of isotopic compositions (δ18O, δD) of water suggested that the river water of ETR mainly originated from three tributaries (except for Middle Tiaoxi River) during the sampling period. Thus, the measures of controlling nitrate pollution should firstly focus on these three rivers. There was a wide range of δ15N-NO3- and a narrow range of δ18O-NO3- in the mainstream waters. This phenomenon could be well explained. The high δ15N and δ18O-NO3- values in some tributaries of South, Middle and North Tiaoxi River revealed that the nitrate might mainly originate from the rural domestic sewage or livestock and poultry breeding wastewater. Otherwise, NO3--N in most of the samples with relatively low δ15N and δ18O-NO3- values were mainly coming from soil organic N and fertilizer N. Therefore, the sources of NO3--N in diverse rivers could be verified qualitatively. Our results also suggested that the development of the best management practice for relieving N pollution in rivers should consider the sources of NO3--N and its pollution characteristic carefully at the tributary scale.(3) In a typical combined polluted sub-catchment of ETRW (diffuse agricultural pollutants and domestic sewage), the sources and spatial occurrence of fluorescent DOM (FDOM) in three orders of receiving rivers were studied. Then the FDOM fluorescent fingerprint information of water samples was determined by the excitation-emission matrices coupled with parallel factor analysis (EEMs-PARAFAC). Four typical PARAFAC components including the humic-like (Cl), marine humic-like (C2), tryptophan-like (C3) and tyrosine-like (C4) fluorescent materials were separated. Their fluorescence peaks were similar to the traditional defined peaks of A+C, A+M, T and B, respectively. Cl and C2 accounted for the dominant contributions to FDOM (> 60%). Except for the autochthonous produced C4, the allochthonous components (C1 and C2) had the same terrestrial origins, but C3 might possess the separate anthropogenic and biological sources. In addition, the allochthonous FDOM was gradually homogenized along the migration directions in multistage rivers. Interestingly, the average content of C1～C3 in secondary tributaries and source pollutants had significantly higher levels than that in subsequent receiving rivers, thus suggesting that the supervision and remediation for secondary tributaries would play a prominent role in watershed management works as well as the direct pollution sources control.(4) The traditional EEMs-SOM and the newly proposed PARAFAC-SOM neural network models were built and compared to resolve the complex fluorescent data of EEMs. We found the EEMs-SOM model was simple to use and less time-consuming. It could provide us with the intuitive contour plots of reference EEMs in interesting neurons. However, the identification of different fluorescence peaks by the "peak-picking" method only depended on the people’s subjective judgment, so this method might not effectively identify the overlapping fluorescent peaks and cause that the connections between multiple peaks of one fluorophore at similar emission wavelengths were artificially dissevered. Thus, the reliability and environment significance of the results were affected. The PARAFAC-SOM model could address the above-mentioned problem. This method could effectively remove the noise signals or useless fluorescence information, and then greatly reduce the number of input variables. Moreover, this model could not only save the running time, but also generate the nondestructive and easily interpretable results. However, the requirements of PARAFAC pretreatment were much rigid, thus the workload was much higher. The results calculated by these two methods were consistent, which made the new and complementary explanations to the findings by conventional approach. Therefore, we strongly recommended these coupling methods to handle the three-way fluorescence data (EEMs) and make a more vivid and robust description for the FDOM’s sources and spatial occurrence in multistage rivers. |
PDF Full Text Request