Quantitative source apportionment of soil heavy metals on a regional scale: Integrating input inventory, dual-isotope...

Quantifying the sources of soil heavy metal (HM) contaminants in mining-affected agricultural land is vital for effective pollution control and preventing food security issues. However, over large-scale agricultural regions, geogenic inputs, diffuse agricultural pollution (e.g., from fertilisation and irrigation), and discrete industrial sources often overlap spatially, making them difficult to resolve with conventional receptor models such as Positive Matrix Factorisation (PMF). Here, we develop a Spatial Frequency Decomposition (SFD) model that integrates a multi-pathway agricultural flux inventory with HM spatial distribution and validate it using an independent isotope mixing model. Applying the SFD model to a representative agricultural region with a long history of metal mining, we find that Cd inputs are characterised by multiple pathways, whereas Pb loading is predominantly derived from irrigation-borne fluvial sediments. We also note that proximity to river channels amplifies Cd accumulation levels, suggesting large-magnitude inputs associated with historic episodic flood deposition or rainfall-driven surface runoff. The model further identifies industrial point-source clusters for Cd and Pb contamination that spatially coincide with known mining and smelting activities. Arsenic (As) contamination yielded no significant point-source anomalies, consistent with a predominantly diffuse component. Cross-validation confirmed that SFD-derived industrial fractions closely matched isotope benchmarks, whereas the conventional PMF model systematically overestimated the industrial fraction, conflating geogenic enrichment with direct emissions. Our SFD framework provides a spatially explicit method to complement existing source apportionment tools, enabling quantification of HM flux and detection of potential point sources, thus supporting more precise and spatially differentiated sustainable land management strategies.