Even though we may have correctly predicted the relative effects 

 of condensation, the magnitudes of the computed values may be in 

 error. When latent (and sensible) heat is transferred to the ocean, high 

 stability of the air close to the sea surface greatly inhibits turbulent 

 transfer perhaps reducing the flux to < 10% of the corresponding 

 magnitude for comparable conditions in an unstable boundary layer. To 

 test the effects of such a reduction, we recalculated the values for the 1 ° 

 squares adjacent to the coast between Cape Mendocino and Cape 

 Blanco and reduced the magnitudes by 90% when the computed latent 

 heat fluxes were negative. The resulting long-term monthly means for 

 the 1° squares and months indicated above remained small, but posi- 

 tive, being < 10 W/m ; in the mean. 



Spatial and temporal biases also may have affected the calculations 

 of latent heat flux. The long-term mean for July in the 1 ° square off 

 Cape Blanco Gat. 43°N, long. 124°W) was computed from only 13 

 observations, of which nine values resulted in negative fluxes of latent 

 heat. South of Cape Mendocino, observations from the Blunts Reef 

 Lightship station consistently biased the long-term means by effec- 

 tively weighting the computed fluxes for 1970 and 197 1 by an order of 

 magnitude more than for any other year. By removing the lightship 

 data, the negative values of latent heat flux which appear at this loca- 

 tion during summer (Chart 29 and Charts 31-33) could have been 

 removed, and for July the mean value would have changed from -1 1 .0 

 W/m 2 (condensation) to +5.5 W/m 2 (evaporation). Although we rec- 

 ognized that these biases existed, no attempt was made to systemati- 

 cally remove observations from those areas where densely sampled 

 station data were apparent. 



There are very few direct measurements of negative fluxes of water 

 vapor (latent heat) and none which support a large reduction in the 

 transfer across the air-sea interface under stable conditions. The data 

 reported by Anderson and Smith (1981) suggested only a 12% and at 

 most a 50% reduction in the transfer process (see their equations (7) 

 and (8)) in a stable as opposed to an unstable boundary layer, for com- 

 parable values of moisture flux. Although the magnitudes of latent heat 

 flux remain somewhat in doubt, our computations indicate that heat 

 losses due to evaporation from the sea surface make a relatively unim- 

 portant contribution to the oceanic heat budget during summer in 

 regions adjacent to the coast. 



The net heat exchange across the sea surface, Q x , is the difference of 

 large numbers. Therefore, the uncertainty in Q x may be substantially 

 greater than the errors for the individual exchange processes. For 

 example, during July off Monterey, Calif., the average value of Q s is 

 197 W/m 2 . Conservative estimates of the errore in g s (10%), Q B {10%), 

 Qt( 1 5 % ) . and Q c ( 15%) result in a relative error of 1 7 % in the estimate 

 for (2\. At the same location in December, the error in Q s increases to 

 64% for a long-term mean value of -36.7 W/m 2 . 



The accuracy of each of the long-term monthly mean heat flux 

 components was independently estimated by computing the stand- 

 ard error of the mean defined by: 



SE =SDoA/n 

 Q 



(10) 



where SE^ is the standard error of the mean for each heat exchange 

 component (i.e., Q s , Q s , Q E , Q c . Q N ), SD is the corresponding 

 sample standard deviation, and n is the number of observations. 

 Large values of n and small values of SD C , correspond to mean val- 

 ues Q which closely approximate the population means. Although 

 the computed standard errors reflect observational noise as well as 

 nonseasonal variability, the distributions of SE^ provide an overall 

 view of the reliability of the long-term monthly heat flux estimates. 

 Standard errors of the means and the numbers of observations in 

 each 1 ° square area and long-term month are tabulated in Appendix 

 II . The standard errors are largest for Q £ and Q s and smallest for Q c 



and Q B . The resulting standard errors of the means for Q N ,(SEg ), 

 range from 1.3 W/m 2 at lat. 33°N, long. 118°W in September 

 (h = 2,259) to 57.6 W/m 2 at lat. 22°N. long. 120°W in June 

 (n = 9). Distributions of minimum and maximum values of SE^ 

 generally conform to the pattern of observations shown in Figure 2. 



The net heat exchange standard errors and the numbers of obser- 

 vations for July and December, respectively, are mapped in Figures 

 3 and 4. Minimum values of SE;, coincide with the coastwise ship- 

 ping lanes stretching from southern Baja California to Point Con- 

 ception and with the transoceanic track between San Francisco and 

 Hawaii. In the extreme northwest section of the grid, steady west- 

 erly winds and minimum sea-air vapor pressure differences contrib- 

 ute to low values of SE^ in July (Fig. 3). Within these regions, the 

 95% probability of departure from the population mean is < 10W/ 

 m-. 



High values of SEq occur along and offshore from the coast 

 between Cape Mendocino and Cape Blanco (Fig. 3) and in the 

 southwest section of the grid, and are associated with regions of 

 sparse data, although sporadic sampling of intense winter storms 

 may also lead to large values (Fig. 4). Values between 5 and 10 W/ 

 m 2 occur throughout a large portion of the summary area in July. In 

 the offshore regions from lat. 2 1 °N to 3 1 °N and from lat. 40°N to 

 50°N. 95% probability departures frequently exceed 30 W/m 2 in 

 December. Typical ratios of the standard error to the mean net heat 

 exchange (SE^ IQ X ) range from 0.01 to 0.27 in July, the average 

 ratio being 0.05. In December, large values of SE^ and small val- 

 ues of 2v result in ratios which vary in absolute magnitude from 

 0.04 to 34.8, with an average value of 0.71. 



In areas outside of the primary shipping lanes, monthly mean dis- 

 tributions of Q N are based on approximately equal numbers of 

 observations per 1° square. If there are no systematic sampling 

 errors, then the relative magnitudes of the computed standard errors 

 of the means should be related to the spatial and temporal variabil- 

 ity in2v- Figures 3 and 4 indicate a slight tendency for larger values 

 of SE;, in the northern section of the grid than in the southern area. 

 The seasonal contrast is more apparent. The greater variability 

 associated with winter surface atmospheric properties contributes 

 to values of SE;, which are 50 to 100% larger in December (Fig. 4) 

 than in July (Fig. 3), which indicates that a large part of the 

 monthly variance in Q s may be due to actual intramonth fluctua- 

 tions rather than observational errors. 



Limits on the absolute accuracy of the empirically derived heat trans- 

 fers have not been well established by experimental data. Nevertheless 

 we feel that the general coherence of the values mapped in Appendix I 

 demonstrates that the fields are representative of the expected spatial 

 and seasonal variability over the California Current region. Small-scale 

 features and detail within a single 1 c square area which are not sup- 

 ported by similar values in surrounding squares probably reflect obser- 

 vational "noise" and should be viewed with caution. Smoother 

 distributions could have been produced by taking averages over 2° or 

 5° square areas as Wyrtki (1965) and Roden (1959) have done. How- 

 ever; 1 ° square resolution was retained to preserve detail in the distribu- 

 tions of the individual heat exchange processes, particularly within 300 

 km of the coast. The fine spatial resolution may be less meaningful for 

 the offshore regions where the decrease in the number of observations 

 per averaged value increases the uncertainty in the derived quantities. 



DEPENDENCE OF HEAT EXCHANGE 

 ESTIMATES ON COMPUTATIONAL METHODS 



The procedures followed in this study involved computations of the 

 heat exchange processes for each weather report, which were then 

 averaged to form the long-term monthly mean values. Constant 



10 



