Road salt causes increasing environmental chloride (Cl-) concentrations which threaten aquatic ecosystems. Environment Canada recommends that road authorities should develop salt management plans including identification of salt vulnerable areas. In this thesis, a Spatial Stream Network (SSN) geostatistical modelling approach was used with seasonal longitudinal field data to develop reach scale models of in-stream Cl- concentrations in three Southern Ontario watersheds. Significance of potential drivers (lane length density (LLD), agricultural & undifferentiated rural land (AURL), and permeability of surficial geology) of stream Cl- were assessed. Results suggest that SSN models are not consistently better than Euclidean models across watersheds. Unexpectedly, LLD was the most important predictor in the rural watershed, and AURL was most important in the urban watershed. Results also show that spatial structure in stream Cl- concentrations was lost under higher flow conditions, which has important implications for when data should be collected to map salt vulnerable areas.