Determination of Spatiotemporal Characteristics of Agricultural Non-Point Source Pollution of River Basins Using the Dynamic Time Warping Distance

Kang Wang,Pengxiang Wang,Renduo Zhang,Zhongbing Lin

DOI: https://doi.org/10.1016/j.jhydrol.2019.124303

IF: 6.4

2020-01-01

Journal of Hydrology

Abstract:The aims of this study were to develop a method to determine the spatial and temporal characteristics of agricultural non-point source (ANPS) pollution loads at the basin scale, and to identify the dominant processes and factors responsible. Time series of water flow and NH4+-N, TN, and TP loads at sub-basin outlets in the Qujiang and Fujiang river basins were obtained using a model calibrated with field measured data. The dynamic time warping (DTW) distances of flow rate and ANPS pollution loads between a sub-basin outlet and the basin outlet were calculated using the time series at the sub-basin outlet and the basin outlet. The DTWs were divided into clusters to evaluate the differences in variation, information content, and similarity among ANPS pollution load time series at the various spatial scales. Results showed that the proposed method could successfully identify the critical spatial scale of the basin. In the sub-basins smaller than the critical spatial scale, ANPS pollution loads were primarily affected by the land surface hydrological processes. In the sub-basins larger than the critical spatial scale, ANPS pollution loads were affected mainly by the river hydrological processes. In the Qujiang river basin, the spatial DTW clusters of ANPS pollution loads were similar to those of the water flow. In the sub-basins smaller than the critical scale, distinguishing factors for the ANPS pollution loads clusters were (from most to least important) stream order (i.e., the spatial scale), land use, and soil nutrient content. In the Fujiang river basin, the spatial DTW clusters of the three pollutants were similar. Distinguishing factors for the ANPS pollution loads clusters were (from most to least important) slope, land use, and soil nutrient content. This method developed can be used as a practical technique to characterize spatial and temporal distributions of ANPS pollution in river basins.