Abstract
Abstract: The objectives of this study were to characterize heterogeneities in agricultural non-point source (NPS) pollution loads, and to identify the uncertainty of time series of water flow and NPS pollution loadings at different scales. The Daily flow rates, concentrations of ammonium nitrogen (NH3), total nitrogen (TN), total phosphorus (TP), and the permanganate index (IMn) and the degradation coefficients of pollutants were measured in a river basin during the period from 2013 to 2015. The dynamic time warping distance (DTW) was used to characterize the differences of runoff flow rates and NPS pollution (i.e., NH3, TN, TP, and IMn) loadings under different underlying surfaces at the sub-basin and basin scales. Information entropy and complexity were used to measure the uncertainty and predictability of different variables at the 2 scales. At both sub-basin and basin scales, NPS pollution loadings increased nonlinearly with rainfall until the rainfall reached 40 mm, and then the NPS pollution loadings maintained stable. The coefficients of variation for NH3, TN, TP, and IMn at the basin scale were 69.1%, 47.0%, 14.2%, and 85.8% higher than those at the sub-basin scale, respectively, with the rainfall higher than the threshold amount. During the dry seasons, the differences in NPS pollution loads between the sub-basins were higher than those between the sub-basins and basin, which were mainly affected by the different pollution sources in the regions. The differences in NPS pollution loadings were significantly higher during the wet seasons than during the dry seasons at both the sub-basin and basin scales. Compared with the NPS pollution loads during the dry seasons, the differences in the NPS pollution loads between the 2 sub-basins and between the sub-basins and the basin increased by 3.18 times and 2.44 times, respectively, during the wet seasons. The differences in fluxes of NH3, TN, and TP between the sub-basins and between the sub-basins and basin were mainly attributed to the underlying surface conditions and the pollutions sources. Compared with the results at the sub-basin scale, the information entropy values for the time series of the water flux per unit area, TN, TP and IMn loads at the basin scale were decreased by 4.8%, 9.3%, 31.9% and 10.7%, respectively, but NH3 were increased by 15.3%. The efficient measure complexity increased with the spatial scale. Compared with the sub-basin scale, the efficient measure complexity values at the basin scale for the water flux per unit area, NH3, TN, TP and IMn loads increased 4.6%, 15.4%, 17.4%, 49.5% and 19.8%, respectively. The result indicated that more parameters were required at the basin scale than at the sub-basin scale for the efficient prediction. The NPS pollutions load and flow processes were not synchronous at different spatial scales. For the flow rates and the NPS pollution loads, the information entropy decreased and effective measure complexity increased with the basin scales, indicating that the uncertainty decreased and the predictability decreased with the scales. The investigation on the sensitivity of NPS loadings to the rainfall and underlying surface at different scale, and physical basis of the NPS loadings at different scales will provide an insight into the theories of distributed hydrology and NPS simulation model.