(iss DR HUGH ROBERT MILL ON THE 



The average results for surface and bottom of the twenty-seven landward and of the 

 seven seaward stations are represented graphically, along with the rainfall, in Plates XL and 

 XII. For this purpose the total rainfall of each month (see Part I. , Tables IV. , V. , VI. , VII. ) 

 is referred to the central day, as it was impossible to make any estimate as to the average 

 distribution throughout the month. The density is, however, entered on the day of the 

 month corresponding to the middle of the period occupied in observing the density of 

 that part of the Area, and is thus not strictly comparable with the rainfall curve. Indeed, 

 the two are in a sense incomparable in nature as well as in period, because the rainfall 

 represents the total cumulative effect of a month, whereas the density is only one phase 

 of a constantly- varying quantity ; but, as the quantities in neither case can be rigorously 

 defined, it is reasonable to assume that the salinity curve corresponds in a general way 

 with the rainfall curve, and to draw such deductions as this assumption may warrant. 



Salinity of Landward Portion. — The mean of all surface density observations in the 

 landward portion of the area gives a curve (Plate XI.) which, although abrupt in its inflec- 

 tions, exhibits remarkable symmetry, and shows a distinct annual periodicity. There is a 

 well-marked maximum in August and September of both years, subsidiary minima in 

 November and June, subsidiary maxima in December and May, and a profound minimum 

 in February. The normal rain curve of this portion of the area presents one very well- 

 marked minimum in May, a double maximum in October and January, with a subsidiary 

 minimum in November. One would expect that in calm weather the salinity curve would 

 be inverse to that of rainfall, but retarded in period. Assuming that such is the normal 

 state of matters, the actual salinity curve — calm weather being necessarily postulated — 

 must be compounded of the normal salinity curve, and of an abnormal curve derivable 

 from the curve of excess or defect of rainfall. It ought therefore to be possible to derive 

 the normal salinity curve, so far as regards its period and amplitude, from consideration 

 of the actual salinity curve in the light of the normal and actual curves of rainfall. The 

 position of the normal salinity curve, i.e., the actual value of salinity at each period, may 

 also be sought after, although with less probability of success. Table XXI. shows the 

 occurrence in time of the minima ( — ) and maxima ( + ) of the actual curves — 



Table XXI. 



Fluctuations of Rainfall and Salinity. 



- + - + - + 



Actual rainfall, . . . ■{ T , October , T , , T T , 



June. September. , ,. ,,-. November. May. July. 



I 



Actual salinity, . . . < August November December February 



+ - + - + + 



~agust November December February „ September 



(early). (early). (late). (early). Jm * 



The correspondence of the later maxima and minima is not very distinct, but in the 

 earlier, where the changes were more rapid and of greater amount, the rainfall minimum 

 or maximum appeared to bring about a salinity maximum or minimum in a little less 

 t hi ii two months. Assuming that the period is six weeks, I subtracted the curve of rain- 



