Look at the seismic sampling rate. Seismic spectrum limited by Nyquist frequency, which is equal to 500/sr, where sr - sampling rate in ms. 33 Hz relates to 15 ms, which is not typical for seismic. Probably you have depth seismic data with sampling rate 15 ft.

Above 33 Hz there is no change of acceleration with frequency

The frequency of 33 Hz as a rigid frequency was used in the past because most of the ground motions were available from western USA. Western USA is an active tectonic region, has a fractured rock because of repeated faulting which has occurred due to earthquake occurring at a close interval of 200-300 years and occurring over millions of years. The fractured rocks  give rise to a  very low shear wave velocity and low rock strength which gives rise to  higher value of kappa  (0.03 to 0.06), giving rise to  peak of the response spectrum occur between 5-7 Hz for a Vs 30 of 760 m/sec and the spectrum becoming asymptotic  at around 33 Hz. Whereas,  in stable  continental  regions viz.  North-East-Central  USA, North-West  Europe, Peninsular India, South-East China and  Australia where the interval of earthquake of large magnitude of 6.0 and above is more than 300,000 years i.e. three times in 1 million year. The rocks in this region have not fractured and as such, have shear wave velocity  between 1500 to 3000 m/sec  and kappa value between  0.003 to 0.016 giving rise to peak of the spectrum to occur  between 18 to 35 Hz and the spectrum  becoming asymptotic at around 150 Hz. However, in  a soft soil site, even in a stable continent regions one can get 33 Hz as a rigid frequency.

Thank you Arun sir. Very nice explanation. This is exactly what I was looking for.