Wavelet-ICA Hybrid Spatial Data mining for Multiscale Source Apportionment of Water Pollution in the Latyan Dam Basin

Abstract. This study aimed to elucidate the spatio-temporal, multi-scale latent sources of water pollution and conduct source apportionment in the Latyan Dam basin, Jajrud River, northeast Tehran, Iran, a vital drinking water supply for the Tehran metropolitan area. Over six years, from January 2002 through December 2007, nine monitoring stations recorded monthly time series for nine key water quality parameters: nitrate (NO₃⁻), ammonia (NH₃), sulfate (SO₄²⁻), chemical oxygen demand (COD), biological oxygen demand (BOD), dissolved oxygen (DO), total dissolved solids (TDS), nitrite (NO₂⁻), and fluoride (F⁻)—over six years from January 2002 through December 2007. The record for each parameter was decomposed into low- and high-frequency components using discrete wavelet multiresolution analysis (Daubechies-4). Subsequently, Independent Component Analysis (ICA) was applied to each sub-band to blindly separate independent pollutant source signatures. When applied to the Latyan Dam data, the method discerned multiscale signatures attributable to seasonal agricultural runoff (NO₃⁻, NH₃), episodic industrial effluents (SO₄²⁻, COD, BOD), and urban sewage spikes (TDS, NO₂⁻, F⁻). Clustering analysis divided the stations into three homogeneous regions based on source contributions. Agriculture was identified as the leading contributor across all stations, with secondary industrial influences observed in some regions. Quantitative evaluation demonstrated that Wavelet-ICA at level 5 yielded the largest correlation (0.37) and high mutual information (0.81). However, negative Signal-to-Interference Ratios (e.g., -3.31 at level 5) indicated significant problems with signal overlaps. This source apportionment framework, which is grounded in six years of monthly observations, provides water resource managers with enhanced temporal and spatial resolution of pollution inputs, thereby facilitating targeted mitigation and reservoir management strategies.