Sentinel-1 SAR modeling of a highway in a landslide-susceptible area

Abstract Highways are an essential component of the global transportation network, significantly reducing commute times and facilitating connectivity within each nation. These infrastructures have been constructed over several decades, necessitating continuous monitoring for their preservation and safety. Traditional techniques, such as extensometers, accelerometers and geodetic surveys are widely employed for localized monitoring of critical elements. In contrast, satellite-based methods, which are more recent, offer a broader perspective. Among these, SAR is particularly noteworthy due to its continuous spatial coverage of the entire infrastructure, albeit coarser in time. This coverage eliminates the limitations imposed by the placement of in-situ instruments. This study aims to define a reliable methodology for modeling highways using Sentinel-1 SAR data, which presents challenges and limitations due to its medium spatial resolution. The proposed processing workflow includes several steps: preprocessing of level-1 SAR data using SNAP libraries, application of the Persistent Scatterer Interferometry technique by StaMPS, validation of the resulting displacement time series exploiting in-situ GNSS measurements, and interpolation of the raw displacements in both space and time using cubic splines. This interpolation involves an iterative selection of the number of splines and rejection of outliers. The final deformation model is then estimated along the road centerline, as the radar cannot distinguish between the two lanes. Additionally, the area under study presents challenges due to its complex orography and the presence of multiple active landslides. Satisfactory results are achieved, with displacement models that are coherent across different monitoring techniques and reasonably capture the deformation trends of various road sections.