Two Kik-Net seismic strong-motion stations

which are installed on hard ground（FKSH12）and soft site（FKSH14）in Japan

were selected and the corresponding probability site models were established.Fifty site soil profiles were randomly generated for hard ground and soft site via Monte Carlo simulation

and then the site transfer functions（STFs）were calculated as well their standard deviations

respectively.The standard deviations of STFs were discussed in three cases

i.e.

varying subsoil thicknesses（Case I）

varying subsoil shear wave velocity simultaneously（Case II）

and varying subsoil thickness and shear wave velocity simultaneously（Case III）.The results show that the standard deviations of site characteristic frequencies of hard ground are overall larger than those of soft site;the standard deviation of site characteristic frequency of hard ground is the smallest in Case I

followed by Case II

while in Case III the standard deviation of site characteristic frequency is the largest;the standard deviations of soft site characteristic frequencies in case I and Case II are comparative but weakly smaller than that in Case III;the standard deviations of STFs in case III for soft site and hard ground are slightly larger than those in Case I and Case II

but the standard deviations in the three cases are not considerably different;the standard deviations of STF are fairly large in the vicinity of site predominant frequencies with peak values corresponding to peak values of relevant mean STF;the STF standard deviations based on real seismic records are particularly larger than the simulated

while the frequencies at the peak value ranges are consistently matched.