88 K. HONDA, T. TERADA, Y. YOSHIDA, AND D. ISITANI. 



latter. These considerations probably account for the smallncss 

 of waves propagated in the northeastern direction. 



The tide -gauge nearest Krakatoa at tlie time of the eruption 

 was that of Batavia. It beautifully recorded two hour waves, 

 but did not trace one hour waves. The absence of this latter 

 period in Batavia raises no serious objection against our sup- 

 position, because the narrow opening to the northeast of the 

 strait, is very unfavourable for the propagation of the energy of 

 the lateral oscillation, as compared with that of the longitudinal. 



Thus the energy of the oscillations was, in its greater part, 

 propagated into the Indian Ocean (PI. LI) and strikingly affected 

 the tide-gauges so far as the ports of Soutliern Africa (PI. LII, 

 Fig. 1-2). Examining the records given in the above cited 

 reports, we may generally distinguish two types of waves, — 

 one includes those types of waves propagated directly from Kra- 

 katoa and the other the stationary oscillation of bays or 

 estuaries excited by the incident waves. Prominent undulations 

 recorded along the coast of India belong to the former type. 

 Comparing the records at Madras and Vizagapatum or Nega- 

 patum and Port Blair, the identity of waves may easily be 

 recognized. We see also the trace of Vizagapatum waves in 

 the Negapatum record and vice versa. Most of these stations 

 are not situated in either bay or estuary possible of oscillation 

 with such a long period. 



For remoter stations, we see generally that the disturbances 

 are chiefly due to the second type, i.e., to the proper oscillation 

 excited by the synchronizing components of the incident waves. 

 Hence for such bays, the periods of oscillation for Krakatoa 

 waves may be calculated from their dimensions, provided a good 

 chart is at hand. Since, at present, we are in want of reliable 



