actually used for computation due to different methods 

 of acquisition. 



No attempt will be made to make a critical evalua- 

 tion of these factors, however, a simple comparison of 

 independently computed net heat flux values may be 

 helpful in the utilization of the charts in this report. 



The region studied by Seckel overlapped that rep- 

 resented by our charts from long. 130° to 170°W, be- 

 tween lat. 20° and 35°N. This comprises 24 5° quad- 

 rangles; consequently, 24 pairs of computed net heat 

 flux values were available for comparison for each of 

 the 24 mo of SeckeFs series, which ran from July 1963 

 through June 1965. 



The mean difference and the root mean square 

 (RMS) difference of net heat flux for each 5° quadran- 

 gle are given in Table 1. 



Positive differences are predominant and indicate 

 that Seckel's values of net heat flux are larger. The 

 RMS differences range from 40 to 106 cal/cm-/day and 

 are generally smallest in areas having the most obser- 

 vations. The RMS difference for the entire sample is 

 72 cal/cm-/day. 



Correlation coefficients were also computed for 

 each 5° quadrangle from paired values of monthly heat 

 flux over the 2-yr period. The results ranged from 0.93 

 to 0.98 for areas lat. 30° to 35°N, 0.88 to 0.95 for areas 

 lat. 25° to 30°N, and 0.62 to 0.91 for areas lat. 20° to 

 25°N. Only three quadrangles had a correlation coeffi- 

 cient smaller than 0.86 which reflects the fact 

 that the seasonal cycle of net heat flux, as represented 

 by both sets of computations, is several times larger 



than the RMS differences. Furthermore, the apparent 

 improvement of correlation with latitude reflects an 

 increase in the seasonal range of heat flux northward 

 from lat. 20° to 35°N. The areas for which the correla- 

 tions were lowest were also the areas with the least 

 observations. 



Averaging over larger areas reduces the effect of 

 random variations which adversely affect the correla- 

 tion by 5° quadrangles between the two sets of val- 

 ues. Figure 2 shows Seckel's monthly total heat flux 

 averaged for the entire region from long. 130° to 170°W 

 and from lat. 20° to 35°N plotted against corresponding 

 averages from our data. The correlation coefficient for 

 the 24 pairs of monthly averages is 0.99 indicating that 

 the effect of random differences has been essentially 

 eliminated. 



Systematic differences, however, are revealed by 

 the regression line which crosses the horizontal axis at 

 12 cal/cm-/day, indicating that Seckel's formulas and 

 processing technique give higher values of the total 

 heat flux. Also, the slope of the regression line differs 

 from 1.0, indicating another systematic discrepancy 

 that stems, at least in part, from Seckel's use of a drag 

 coefficient which varies with wind speed in the equa- 

 tions for evaporation and sensible heat transfer. His 

 formulation gives higher evaporative heat loss with 

 winds stronger than about 7 m/sec (other factors being 

 equal) than those presented in this report, and lower 

 evaporative heat Joss with winds weaker than 7 m/sec. 



Wind speed frequencies tallied from tables in 

 Seckel's report indicate that in the region and period 



Table 1. — Mean difference (upper value) and RMS difference (in parenthesis) 

 of Q, in cal/cm /day for 5° quadrangles 



