VNIR estimation of heavy metals concentrations in suburban soil with multi-scale geographically weighted regression

•Spatial relationships between soil chromophores and heavy metals were explored.•Spatial correlations with TFe2O3 were main reason that Cr and Ni can be estimated.•MGWR outperformed traditional regression methods for VNIR estimations of Cr and Ni.•Varied scales and bandwidths contribute to improving model performance.