Table 7.--Attentuation coefficients from photometer lowerings 

 at a station 20 nautical miles west of Oceanside, Calif., 

 1 May 1959 



[H = 0.18; p < 0.975 > 0.99] 



idea of the variability encountered. The values 

 obtained (depth intervals in meters in paren- 

 theses) were: 0.05 (6-11), 0.07 (11-21), 0.15 

 (21-31), and 0.11 (31-41). 



It appears then that under e.xcellent ob. 

 servational conditions the coefficient of vari- 

 ation is of the order of 10 percent. I estimate 

 that measurements of comparable quality to 

 those discussed above are obtained only 30 to 

 50 percent of the time in the northeastern 

 tropical Pacific. The remainder of the time, 

 rapid fluctuations in incident radiation 

 resulting from nonuniform cloud cover and 

 fluctuations in underwater irradiance resulting 

 from vessel surge and from the focusing 

 effects of surface waves tend to give a greater 

 scatter to replicate irradiance measurements 

 at a given "depth." These factors reduce the 

 precision and accuracy of the estimates of 

 attenuation coefficients. 



Variations in irradiance level caused by 

 vessel surge and focusing effects of surface 

 waves cannot be detected with the equipment 

 used below depths where about 25 percent of 

 the incident energy is found. Although changes 

 in incident radiation are still troublesome 

 below this depth, the largest error in esti- 

 mating the attenuation coefficient at depths 

 below 30 m. probably results from errors of 

 depth measurement. 



I must emphasize that the Pend Oreille data 

 provide error estimates which are minimal. 

 Under conditions of varying cloud cover ac- 

 companied by pronounced vessel surge or high 

 wire angles, observations are certainly not as 

 reproducible as those given above. Data from 

 both experiments yield estimates of precision 

 which probably cannot be surpassed with the 

 equipment described. 



The analysis of k data thus far has assumed 

 negligible error in depth determination. Unless 

 one has a depth-sensor attached to the irradi- 

 ance meter, depth is usually computed from 

 wire length and departures in wire angle from 

 the vertical. Computation of instrument depth 

 with angles in excess of 20° and with wire 

 lengths greater than 40 to 50 m. is rather 

 unsatisfactory, although the data available are 



not abundant enough to permit a rigorous 

 assessment of this error. 



To illustrate the problem, data fronn two 

 SCOT Expedition stations are considered. 

 These stations were chosen because wire 

 angles were large and wire lengths were as 

 great as 100 m. In these two examples, the 

 depth indicated by the depth sensor is taken 

 to be the true or "observed" instrument depth. 

 Two different but related questions need to be 

 considered. First, how accurately may one 

 compute instrument depth; and secondly, how 

 well do the computed depth intervals agree 

 with the observed intervals? The answer to 

 the first question tells us if we are assigning 

 our k value to the appropriate depth interval, 

 and the answer to the second tells us how much 

 we may expect the k value to vary from the 

 true value. 



The computed depth is always equal to or 

 less than the observed depth. This relation 

 results from the fact that the computational 

 method assumes the wire to be straight, 

 whereas it is almost certainly curving. 



In Table 8, column 6, the difference be- 

 tween the observed depth and the computed 

 depth is given for the two SCOT stations. The 

 values show that the instrument was often 2 to 

 5 m. and occasionally 9 to 10 m. below the 

 computed depth. This difference means that 

 the k values will frequently be assigned to 

 incorrect depth intervals. It is reasonable to 

 believe that differences between true and 

 computed depth generally increase with wire 

 length, but the few available observations fail 

 to show any such systemic variation. 



Aside from the question of absolute depth, 

 the variation between the observed and com- 

 puted depth interval also needs to be con- 

 sidered, for it is the depth interval that enters 

 directly into the k calculations. Table 8, 

 column 9, gives the difference between the 

 observed and computed intervals. If we as- 

 sume that these data are representative of 

 conditions of high wire angles, then errors as 

 great as 3 m. in 14 will give k values that are 

 22 percent too low. The nnedian value lies 

 between 1 and 2 m. difference, and may be 

 typical. 



It should be pointed out that this assessment 

 of errors must be considered preliminary, 

 although there is no reason to suspect the 

 data. Thus these values are probably typical 

 for the equipment used. 



Optical Depth Determinations 



In addition to the rate of attenuation of 

 diffuse downwelling irradiance (i.e., the diffuse 

 attenuation coefficient) another related 

 quantity, optical depth, is frequently used by 

 investigators concerned with estimates and 

 measurements of primary production. These 

 investigators frequently wish to determine the 



13 



