values shown in Appendix I (Charts 1-12) lies within 10% or 

 approximately ± 25 W/m : for typical summer values exceeding 

 250 W/m : in the 1 ° squares adjacent to the coast. Where the means 

 are based on only a few observations per long-term month (i.e., 

 fewer than 10). the uncertainty may be larger. 



Most of the empirical formulae for computations of effective 

 back radiation. Q B , have been derived for overland conditions. Few 

 verifications of the accuracy of the formulae have been made at sea. 

 Comparisons between direct measurements of Q B and predicted 

 values based on Equation (3) in midocean and coastal regions 

 established an upper bound on the accuracy of ±20 W/m : or 

 approximately 20% of the observed daily mean net long-wave flux 

 (Simpson and Paulson 1979; Reed and Halpern 1975). 



The principal error in estimating Q B again arises from uncertainties 

 in the cloud correction term. However, net long-wave flux under clear 

 skies may also depend on the distributions of temperature and humidity 

 above the constant flux layer (Charnell 1967), which are not parame- 

 terized by the empirical formulae. The effect of neglecting humidity 

 variations in the lower atmosphere may be comparable with that of var- 

 iable cloud cover. Measurement errors in sea surface temperature. 7",. 

 and near-surface vapor pressure. e„, of 1°C and 1 mbar, respectively, 

 contribute <3% to uncertainties in estimates of effective back radia- 

 tion. However, an error of 0.125 (1 okta) in estimating total cloudiness 

 introduces a relative error that ranges from <10% for low cloud 

 amounts to more than 50% for high values of cloudiness. Thus, empiri- 

 cal formulae which have been derived for average conditions may not 

 necessarily hold for estimates based on only a few observations. As the 

 numbers of observations increase, the contribution of random errors 

 will generally decrease, which suggests that the accuracy of the long- 

 term mean net long-wave fluxes predicted from Equation (3) lies well 

 within the 20% bound estimated by Simpson and Paulson (1979). In 

 the California Current region, the loss of heat due to net long-wave flux 

 varies from 20 W/m : offshore to 50 W/m : near the coast (Charts 13- 

 24). The associated maximum errors would range from +4 to + 10 W/ 

 m : . 



In western boundary current regions (Husby and Seckel 1975) 

 and in the North Pacific trade wind zone (Seckel 1970), oceanic 

 heat loss due to evaporative processes typically equals and often 

 exceeds the heat gain due to incident solar flux. Along the west 

 coasts of continents, however, latent heat flux is approximately the 

 same order of magnitude as the effective back radiation, which is 

 approximately a factor of 2 smaller than the incoming short-wave 

 radiation. Near the west coast of the United States during summer, 

 latent heat flux decreases by an order of magnitude and therefore 

 may be more comparable in absolute magnitude with sensible heat 

 flux. Although computations of latent heat flux from the bulk aero- 

 dynamic formula may contain large relative errors ranging from 20 

 to 60% (Clark 1967). the effect of data uncertainties on latent heat 

 and net heat exchange computations may not be as serious as in the 

 studies of Seckel (1970) and Husbv and Seckel (1975). 



The accuracy of latent and sensible heat flux computations is 

 affected by uncertainties in the magnitudes of the experimentally 

 determined turbulent transfer coefficients. Qand C„. and measure- 

 ment errors in observations of sea, air, and wet-bulb temperatures 

 and windspeed in the constant flux layer. Errors are also introduced 

 if the assumptions which form the bases for the empirical formulae 

 are not properly satisfied. Values of the neutral exchange coeffi- 

 cients referred to the 10 m level, as summarized by Friehe and Sch- 

 mitt (1976). range from 0.0010 to 0.0016 with typical uncertainties 

 of ±20%. Variations in atmospheric stability and the assumption 

 that the transfer coefficients for heat and water vapor are equal and 

 independent of windspeed. increase the uncertainty in the magni- 

 tudes of the derived turbulent exchange processes. 



Bunker (1976) compared monthly averages of surface meteoro- 

 logical observations obtained from merchant vessels within sub- 

 areas surrounding ocean weather stations (OWS) with 

 corresponding means formed from the assumed higher quality 

 OWS data. Based on comparisons using more than 500 observa- 

 tions, the 95% probability departures for air. sea, and dewpoint 

 temperatures were 0.15°, 0.12°, and 0.27°C, respectively. In our 

 study. 95 % of the 1 ° square means are based on fewer than 500 

 observations/mo, and for 55% of the mean values, the numbers of 

 observations are < 100. The standard errors of temperature corres- 

 ponding to these long-term means are 0.3°, 0.2°, and 0.5°C. For 

 averages formed from < 10 observations, uncertainties in the mean 

 temperatures typically can be > 1 °C and may exceed 2°C. 



The magnitudes of the errors vary as a function of instrument cal- 

 ibration, observing technique, and the reporting schemes used in 

 the historical data files. For example, temperature values in Decks 

 110 and 281, U. S. Navy observations 1920-51, were originally 

 recorded to the nearest whole degree Fahrenheit, and later con- 

 verted to degrees and tenths Celsius; whereas the temperatures 

 archived in Decks 1 18 and 1 19, Japanese ship observations 1937- 

 60, were reported to the nearest whole degree Celsius. Therefore, 

 the temperature values derived from these decks, which constitute 

 18% of the total observations, contain an additional uncertainty of . 

 ±0.5°C (Husby 1980). Because of large uncertainty in random 

 sampling errors, we have not corrected observations of sea surface 

 temperature for a possible systematic bias of 0.7°C in engineroom 

 injection temperatures (Saur 1963) or measurements of air tempera- 

 ture and wet-bulb temperature which may be modified by ship- 

 induced air flow distortion at the 10 m level (Holland 1972). 



Temperature errors contribute directly to uncertainties in estimat- 

 ing sensible heat flux and indirectly affect calculations of latent 

 heat flux by introducing errors in the sea-air vapor pressure differ- 

 ence. Average deviations of 0.2°C in sea-air temperature difference 

 result in relative errors in sensible heat flux which decrease from 20 

 to 4% for temperature differences which increase from 1 ° to 5°C. 

 A more characteristic temperature departure of 0.5°C introduces 

 relative errors which decrease from 50 to 10% for similar sea-air 

 temperature differences. For monthly mean sea-air temperature dif- 

 ferences of 1 °C and windspeeds of 6 to 7 m/s, errors of ± 1 to ±5 

 W/m : may be expected for sensible heat fluxes of 10 W/m 2 . 



An error of 0.5°C in sea surface temperature results in errors in 

 the saturation vapor pressure over water which vary from 0.4 mbar 

 at 10°Cto 1. mbar at 25 °C. Comparable errors occur in estimates 

 of vapor pressure of the air due to erroneous psychrometric mea- 

 surements. Long-term mean sea-air vapor pressure differences over 

 the California Current region generally lie between 1.0 and 5.0 

 mbar. However, in the area off southern Baja California during fall 

 and winter, high sea surface temperatures and relatively colder and 

 drier continental air combine to produce sea-air vapor pressure dif- 

 ferences exceeding 15 mbar for individual observations. Mean dif- 

 ferences are generally < 10 mbar. Therefore, the possible error in 

 computations of latent heat flux may be relatively large. For aver- 

 age deviations of 0.5 mbar. the uncertainty ranges from <4% for 

 large vapor pressure differences to more than 40% at low values. 

 The error associated with monthly mean values of Q £ between 50 

 and 100 W/m : would be approximately ± 10 W/m : . 



In Equations (7) and (8). the vertical fluxes of heat and water 

 vapor have been parameterized as a function of the observed wind- 

 speed. U t0 . Nelson (1977) discussed the sources of error in wind 

 reports in the TDF-1 1 data file and also determined that < 12% of 

 the total wind reports consisted of measured, as opposed to esti- 

 mated, quantities. Approximately 35 % of the reports from the Cali- 

 fornia Current region consisted of Beaufort wind force estimates 



