6 SMITHSONIAN MISCELLANEOUS COLLECTIONS VOL. 7Q 



of the decade in question or of the group of three decades to 

 temperatures of neighboring fields on both sides has been examined. 

 Also the yearly change in these neighboring fields. The values 

 built up -in this manner are distinguished by brackets in tables iW, 

 3W, 6L and 8L. 



In spite of the unsatisfactoriness of the observations, both as to 

 their number and as to their accuracy, it appears that the values 

 found for the surface temperature fall in good harmony. 



In the isopleth diagrams (on the right in pis. 17-41) which 

 show the distribution of the plus and the minus anomalies both 

 in time and in region from decade to decade and from field to field, 

 we see that in almost all cases a certain connection or system is 

 found in the distribution of the anomalies. It infrequently appears 

 that a minus anomaly is to be found between plus anomalies or 

 vice versa. In general the march of the changes in the signs of 

 the anomalies and in their magnitudes goes gradually along from 

 field to field and from decade to decade. This seems to show that 

 our mean values correspond well with the actual truth even for the 

 single decades. Obviously this is probably, in a yet higher degree, 

 true with the mean of all observations for two groups of three 

 decades each. This inference is easily confirmed by the graphical 

 representation of the values. 



The curves for the single fields in the eastern part of our region 

 up to about 50° west longitude agree in all essential particulars 

 astonishingly well with one another, and change gradually from 

 field to field in a way which shows that they must correspond well 

 with the actual temperature relations, and cannot be changing in 

 a haphazard way (see figs. 16-19). That the agreement is less 

 striking in the western part depends upon conditions of which we 

 shall speak later. 



Our observational material on the air temperatures is less per- 

 fect than that on the surface temperatures, for three reasons : First, 

 the single determinations are ordinarily less satisfactory ; second, 

 the daily amplitude, which we cannot take account of, is much 

 greater than that in the water temperature and accordingly a limited 

 number of air observations must give less satisfactory values than 

 an equal number in the water. Besides, in several cases a some- 

 what smaller number of air observations is available for the compu- 

 tation of a mean temperature value, since the surface temperatures 

 are not always accompanied by a determination of the air tempera- 

 ture. On the other hand the observations were not infrequently 



