Figure S3a. Best-fit regression of decay in wave group velocity with frequency (equations [1a] and [1b] in the main manuscript) based on active seismic experiments for each of the 24 seismic sensor locations. Each line is fitted through the average value of 20 repeated sledgehammer impacts (see also Figure 5b in the main manuscript). The thicker the line, the greater the number of seismic sensor locations that share the same \(v_{p0}\) and \(\xi\) values (i.e. Table S3), from a single seismic sensor location sharing the same parameter values (i.e. thinner line), up to three seismic sensor locations sharing the same parameter values (i.e. thicker line). Note that seismic sensor locations presenting the same \(v_{p0}\) and \(\xi\) values were estimated based on a single active seismic experiment encompassing multiple sensors thanks to their geographical proximity. The parameters found by Bakker et al. (2020) for an alluvial channel (in red), and proposed by Tsai et al. (2012) for a generic bedrock site (in blue) are presented for comparative purpose.