We present a summary of seismological and geophysical investigations at Amatrice (Central Italy), a village seated on an alluvial terrace and severely stroke by the Mw 6.0 event of August 24th 2016. The high vulnerability alone could not explain the heavy damage (X-XI MCS), whereas the vicinity of the seismic source and the peculiar site effects should be claimed to understand the ground motion variability. After the first mainshock, we investigated the Amatrice terrace for microzonation purposes together with several Italian institutions (Priolo et al., Bull. Earthquake Eng. 2019). In particular: (i) we installed 7 seismic stations as a part of the 3A network (DOI: 10.13127/SD/ku7Xm12Yy9; Cara et al., Sci. Data 2019); we performed (ii) an extensive campaign of 60 single-station ambient noise measurements (downtown stations recorded also few earthquakes), and (iii) several 2D passive seismic arrays aimed at obtaining Vs profiles down to a depth of few tens of meters (Milana et al., Bull. Earthquake Eng. 2019). Earthquake recordings were used to empirically evaluate ground motion amplification effects through spectral ratio approaches, and noise data were collected for defining the spatial distribution of the resonance frequencies. Data analysis reveals a diffuse amplification effect that reaches its maximum values in downtown area with a resonant frequency (f0) of about 2 Hz. Seismic amplification is also characterized by spatial variation and directional amplification, mainly in downtown to the west side of the alluvial terrace, and related to both stratigraphic and topographic effects. This effect tends to decrease and almost vanishes in the central part of the terrace, and it increases again moving towards its eastern edge with a clear shift of f0 towards higher frequencies. Empirical transfer functions were then used to recover the ground motion that could have hit the historical center of Amatrice during the August 24th mainshock, through the convolution with the only record in the vicinity (IT.AMT station experienced a PGA of 0.87 g). The reconstructed peak values are much greater than expected from ground motion models, showing that detailed studies on local site response can largely modify the seismic hazard assessment.