92 



E. M. HARVEY 



complete external water films ("hard showers") as described 

 above. The rate of absorption of water per hour and the height 

 of the cups above the surface of the water in the burettes was 

 recorded. For the varying water columns used, the rate of absorp- 

 tion was found to be almost directly proportional to the height of 

 the water column supported by the cup. The table below pre- . 

 sents an example of the results obtained. 



HEIGHT* OF CUP 



HOURLY RATE 



RATIO- "^■'^ 

 HEIGHT 



cm. 



CC. 





10 



0.6 



0.06 



30 



4.2 



0.14 



50 



8.8 



0.17 



80 



13.0 



0.16 



100 



15.8 



0.15 



* Height measured from the top of cup to water surface in burette. 



Assuming the average height of the cup in the field to be about 

 30 cm., the rain absorption by the cup may amount to as much 

 as 4.2 cm. per hour, an extent of error which may frequently 

 amount to more than the actual evaporation occurring in a par- 

 tially rainy period. Thus, if during a rainy day the actual evapo- 

 ration from the cup were n cc, and if rain absorption within the , 

 period amounted to m cc. (m being greater than n), it is clear that 

 the reading of the instrument would be not only quantitatively 

 but also qualitatively wrong ; it would appear that during the day 

 there had been an absorption of m— n cc. and the observer could 

 have no way of knowing whether or not any evaporation at all 

 had occurred. With the mercury valve, the instrument would 

 give the integration of the actual loss from the reservoir, minus 

 that of the quite insignificant amounts of water injected into the 

 reservoir by the operation of the valve. With the valve and 

 height of water column assumed in the foregoing paragraphs, it 

 would require about one hundred complete reversals of the valve 

 to produce an error in reading of a single cubic centimeter. No 

 doubt, if it should ever be requisite, the instrument could be 

 arranged so as to record, in some way, the number of reversp-ls 

 of the valve and the error of the reading might thus be still fur- 

 ther reduced. 



