Previous studies have shown that sedimentary basins significantly amplify seismic waves and prolong ground motion duration, thereby intensifying earthquake damage. The Kanto basin, as one of Japan’s most densely populated urban regions, exhibits particularly strong ground motion amplification effects due to its thick sedimentary layers and complex geological setting. This study applies the generalized inversion technique to strong motion observation data from the Kanto Basin and its surrounding areas in Japan. By extracting and separating the S-wave Fourier amplitude spectra, we obtain seismic source parameters, quality factors, and site responses, with a focus on elucidating the differences in site response inside and outside the basin as well as their relationships with site parameters such as v_S30 and the sedimentary layer depth Z_3.2. Furthermore, we compare the site amplification results derived from the generalized inversion method and the horizontal-to-vertical spectral ratio （HVSR） method to assess their applicability across different frequency ranges.

A strong-motion dataset covering the period from 2010 to 2019 was constructed, including 1194 earthquake events and 74 strong-motion stations, with moment magnitudes M_JMA ranging from 3.0 to 6.5 and source-to-site distances less than 200 km. Using the two-step non-parametric generalized inversion method, we first separate the path attenuation term, and then apply a reference station constraint to extract source and site terms. The main findings are as follows:

1） The derived source spectra conform to the ω² model in the 0.3−24 Hz frequency range. An empirical relationship between seismic moment M₀ and corner frequency f_c was established: lgM₀＝（22.817±0.797）−3lg f_c, corresponding to an average stress drop of 14.9 MPa. The estimated frequency-dependent S-wave quality factor is Q （ f ）＝67f ^1.19.

2） Correlation analysis of site response and v_S30 at different frequencies shows that, for stations inside the basin, site response and v_S30 are moderately to strongly correlated in the low-frequency range （0.5−2.0 Hz）, while this correlation weakens significantly at higher frequencies. For stations outside the basin, the frequency range of moderate to strong correlation extends up to 8 Hz. These results indicate that v_S30 is limited in its ability to represent basin site effects, whereas it performs better outside the basin. The sensitivity of site response to v_S30 differs between the two regions: the correlation peaks near 1 Hz inside the basin and around 4 Hz outside the basin.

3） Empirical relationships between site response and sediment depth Z_3.2 were established. Stations outside the basin show better correlation in the low-frequency range, whereas stations inside the basin exhibit low correlation across all frequencies, with a sign reversal near 3 Hz. In the low-frequency range （0.3 Hz and 1 Hz）, most basin stations have higher site responses than those outside. In both regions, site response generally increases with increasing depth Z_3.2. However, at high frequencies （5.0 Hz and 10.0 Hz）, site response inside the basin decreases with increasing depth Z_3.2, especially at 10 Hz, reflecting the attenuation effect of thick sediments on high-frequency ground motion.

4） The HVSR method systematically underestimates site amplification, particularly in the low-frequency range （＜1 Hz）, and this underestimation is more pronounced for stations insidethe basin than those outside.

These findings reveal the complexity of site response within the basin, which limits the effectiveness of using single site parameters to construct empirical models. The study identifies distinct site amplification characteristics between intra-basin and extra-basin stations in the Kanto region, and offers insights that may inform future investigations of basin effects on ground motion in other regions such as China.