III. TEMPERATURE ANOMALIES 



An instantaneous temperature at a given location at any time can be 

 interpreted as a sum of the long-term mean and the difference between the 

 mean and actual temperature^ 



e = + A0 



where A0 i-s the temporal anomaly. The mean and the actual temperatures 

 ■must be known in order to compute the anomaly. Conversely, the instanta- 

 neous temperature can be predicted from the mean temperature if the anom- 

 aly coiild be predicted. Mean surface values can be interpolated from the 

 mean monthly charts, and mean temperatures at 400 feet can be computed 

 with equations (9) through (12) or interpolated from charts based on these 

 equations. If mean values are considered sufficiently accurate for prac- 

 tical application, the rest of the problem is reduced to prediction of 

 anomalies. 



The first logical step would be to study surface anomalies and to 

 formulate analytical, empirical, or statistical ways to predict them. 

 The second part of the problem would be establishment of a relationship 

 between anomalies at the surface and J+00 feet and a method of predicting 

 the anomaly at 400 feet in terms of the surface anomaly. 



Surface anomalies have been studied on several occasions (3^^)^ but 

 results were inconclusive. Their causes and formation processes are 

 complicated. Wo prediction system exists at this time; however, once 

 the surface anomaly is established, it extends over large areas and usu- 

 ally persists for a considerable time (several weeks or months). If val- 

 ues of surface anomalies can be obtained for the entire ocean or a por- 

 tion of it from observed data, it may be assumed without much risk of 

 significant error that the anomalies will be applicable to 5- or 10-day 

 periods. The problem then becomes one of establishing a relationship be- 

 tween anomalies at the surface and ij-00 feet. 



A detailed study of the relationship would require a large amount of 

 data collected over a period of many years at many points. The only loca- 

 tions for which reasonable amounts of data are available are the five 

 North Atlantic ocean weather stations. These data are limited to periods 

 of 5 to l6 years. Consequently, determination of the relationship cannot 

 be exact and complete and will only permit approximations satisfactory 

 for limited applications. 



The range of temperature errors is rather large in individual BT's; 

 therefore, correlation of mean anomaly values at the surface and ifOO feet 

 would produce better results. The monthly mean temperature anomaly is the 

 difference between the mean for a given month and the mean monthly value 

 based on long-term data . Large errors in individual observations are 

 smoothed during computation of monthly means. Since the change in anom- 

 alies is rather slow, monthly mean anomalies can be considered represent- 

 ative of individual anomalies. 



Mean anomalies at the surface and HOO feet can be considered as nor- 

 mally distributed random variables. Table k shows the mean and standard 



17 



