Complex behavior of an urban green infrastructure imaged using time-lapse electrical resistivity tomography and time-series clustering

Urban green infrastructure (GI) is increasingly being implemented to manage stormwater, promote infiltration, and limit loads on drainage networks. However, their design still relies largely on large-scale hydrological performance, with limited attention to site-specific hydrogeological processes. Key mechanisms governing infiltration, solute transport, and seasonal dynamics remain poorly understood, especially in cold-climate cities using de-icing salts, which may threaten buried infrastructure, contribute to groundwater salinization, and degrade receiving ecosystems. Field observations are needed to improve understanding of these processes. This study presents one year of monitoring of water and chloride dynamics around an instrumented GI using high-resolution cross-borehole time-lapse electrical resistivity tomography (TL-ERT) at a roadside site in Laval, Quebec, Canada. Automated geoelectrical monitoring was conducted between October 2024 and November 2025 with four surveys per day, using commercial and open-source hardware. The geophysical approach was complemented by hydrogeological monitoring (water levels and sensors). A time-series clustering method was applied to the inverted dataset to identify distinct temporal behaviors of subsurface units. Results indicate infiltration linked to winter rainfall and spring snowmelt. Seasonal conductivity variations in winter and spring are interpreted as being largely influenced by pore-water conductivity associated with de-icing salt mobilization, while late summer and autumn changes are more consistent with VWC-driven responses. Overall, combining cross-borehole TL-ERT, time-series clustering, and hydrogeological instrumentation provides detailed insight into seasonal water and solute movement in small-scale, heterogeneous urban environments, where point monitoring is often limited. This approach is well suited for studying urban GI, while remaining dependent on auxiliary measurements for process attribution.