14 The Philippine Journal of Science i9i8 



Table V. — Classification for boiler use. 



Parts per million. 



Less than 90 * Good. 



90 to 200 Fair. 



200 to 430 Poor. 



430 to 680 ' Bad. 



Over 680 Very bad. 



The use of the Berkefeld army filter to clarify turbid waters, 

 as suggested by Leighton, has been discontinued in our field 

 work for several reasons. Comparatively few of the waters 

 examined on the average field trip are turbid. An analy- 

 sis of only the clear portion of a turbid water is ordinarily 

 not of great value, and when it is desired, a clear sample is 

 readily obtained by sedimentation or by filtration through cotton 

 or paper. Turbidity interferes appreciably only with the de- 

 termination of sulphates. Its efiiect can be readily overcome 

 by determining the turbidity of the liquid after adding hydro- 

 chloric acid and before adding barium chloride and subtracting 

 this from the reading obtained after the sulphates have been 

 precipitated. The difference represents the sulphate turbidity, 

 and the amount of sulphates can be determined from the table 

 without appreciable error. In short, the Berkefeld filter has 

 found such limited application in our work that the minor bene- 

 fits derived from its use have not been commensurate with the 

 trouble and inconvenience of carrying it. 



Accuracy of field determinations. — ^While field methods do not 

 claim the exactness and accuracy possible in the laboratory, 

 it is interesting to note that in several cases the values obtained 

 by the two procedures do not differ very widely. As has been 

 previously stated, results obtained in laboratory determina- 

 tions are expressed in terms of two significant figures only. 

 This mode of expression itself involves limits of accuracy that 

 permit a maximum error of about 4 per cent. The average 

 accuracy of field determinations, as stated by Leighton and con- 

 firmed in our own work, is roughly about 5 per cent. Turbidity 

 shows the widest variation, ranging from about 3 per cent with 

 turbidities of 500 to 1,000 parts per million to about 16 per 

 cent with a turbidity of 30 parts per million, the deviation 

 increasing fairly regularly with decreasing turbidities. 



There are several sources of probable errpr of which the 

 following are the most important: 



When using a 100 cubic centimeter graduated cylinder, vol- 

 umes cannot be read more accurately than to the nearest tenth 

 of a cubic centimeter and often not that accurately. Further, 



