Although pyranometers are calibrated with 

 the exposed internal and external leads pointing 

 in a northly direction, the manufacturer states 

 (personal communication) that failure to keep 

 the instrument oriented in this direction should 

 not introduce serious error or inconsistencies. 



1 have no data which will permit rigorous 

 assessment of errors introduced solely by 

 the failure of the gimbals' support to keep the 

 pyranometer in a horizontal plane. As men- 

 tioned above, the gimbals seem to stabilize 

 the pyranometer rather well except in very 

 rough seas. 



Daily records for clear and overcast con- 

 ditions were selected to determine the vari- 

 ability due to integration; each trace was in- 

 tegrated 10 times. A comparison of results 

 (table 1) shows that the coefficient of variation 

 was 1.24 percent on the day with abundant high 

 clouds, and 0.37 percent on the uniformly 

 overcast day. 



It should be mentioned that for data collected 

 before 1957 and used in the evaluation of in- 

 direct methods of predicting daily totals (see 

 table 2), the daily totals have been reduced by 



2 percent (Drummond, 1961). This change is 

 necessary because pyranometers used before 

 December 1957 were calibrated according to 

 the Smithsonian scale of 1913 rather than the 

 International Pyrhelimetric scale of 1956. 



To recapitulate, the daily radiation totals 

 are believed to be accurate on average within 

 1 7 to 10 percent of the reported value. 



Comparison of Observed and Computed 

 Values 



Although continuous measurements of inci- 

 dent solar radiation falling upon a horizontal 

 plane have been obtained on a majority of 

 cruises and expeditions during this study, the 

 data are not yet sufficient to permit the con- 

 struction of maps giving seasonal annual in- 

 solation values. The data are sufficient, how- 

 ever, to permit a preliminary examination of 

 the relation between observed daily radiation 

 totals and similar totals computed from a 

 selected number of climatological equations 

 which use cloud-cover observations. It will 

 become apparent that these climatological 

 equations do not yield precise estimates of 

 daily insolation and that their use for esti- 

 mating daily totals should be avoided unless 

 the relatively large errors attendant upon their 

 use can be tolerated. 



Four equations were chosen for these com- 

 parisons: that of Kimball (1928), the Savino- 

 Angstrom formulation (Budyko, 1956), that of 

 Black (1956), and that of Laevastu (I960). In 



Table 1. --Variability of 10 successive planlmeter integrations of two daily incident solar radiation records with different cloud 



cover, eastern tropical Pacific, 1958 



Table 2. --Statistical results of 104 comparisons between observed daily totals of incident solar radiation 

 falling upon a horizontal surface and values computed from four climatological equations (Kimball, Sa vino- 

 Angstrom, Black, and Laevastu) or taken from Black's monthly maps, 1956-59 



Source of value 



Observed 



Kimball 

 (1928) 



§a vino- 

 Angstrom 



Black (1956) 



(Budyko 1956) ^fonthly maps Equation (3) 



Laevastu 

 (1960) 



Total radiation for 

 104 days 



Mean daily total 



(g.cal./cm.^) 



Correlation coefficient 



Regression equation. . . . 



Sample standard devia- 

 tion from regression. 



Significance of regres- 

 sion coefficient 



Significant at the 1 -percent level. 



