temperature changes equal to or greater than temperature 

 error of the probes manufactured in 1968 and later. 



As the cost of probes rose in the early 1970's we 

 returned to using T-4 probes. With the merchant ships 

 the amount of wire on the probe was the limiting factor 

 on depth of the XBT observation. We found that the 

 manufacturer's safety margin of excess wire on the probe 

 usually permitted a reliable determination of 

 temperature to 500 m. 



duce bias toward higher temperatures in the remainder 

 of an analog record, and 3) slippage of the friction 

 clutch on the chart drive, which occurred mainly in early 

 years before we became more experienced with the XBT 

 system. Corrections were made at a later time when 

 temperature values were interpolated at 5-m intervals 

 between the surface and 300 m and at 10-m intervals 

 between 300 and 500 m depth, for the computer analysis 

 of temperature fields. 



Initial Processing 



The procedures for initial processing of the observa- 

 tions into digital form on magnetic tape evolved as the 

 project developed. 



As noted earlier, the XBT system used during 1966-69 

 aboard the Californian included an experimental 

 digitizer with a punched paper tape output. Most of the 

 observations through April 1969 were computer 

 translated from this output, which was regularly 

 calibrated with the manufacturer's test canister on the 

 visit to the ships before and after each voyage. In cases of 

 digitizer failure significant points were read by eye from 

 the analog traces, as were all observations for May- 

 December 1969. Some sets of observations from 1967 to 

 1968 were semiautomatically digitized by FNWC in the 

 early stages of development of its XBT digitizing system 

 for computer determination of temperature-depth inflec- 

 tion points. 



When the NMFS Pacific Environmental Group (PEG) 

 was established in Monterey, the 1970 and later observa- 

 tions were digitized on an analog-digital table, under the 

 supervision of McLain and using the facilities and com- 

 puters of FNWC. The digitizing procedures generally fol- 

 lowed those used at FNWC, described by Dale and 

 Stevens (1970), except as modified by McLain at PEG to 

 handle the NMFS data separately from FNWC data, to 

 digitize analogs from T-4 probes to 500 m, and to plot 

 vertical sections. 



Quality Control 



Preliminary vertical sections of the distribution of 

 temperature were constructed, at first by hand and later 

 by computer, for quality control. Saur reviewed each 

 data set for possible errors utilizing the preliminary sec- 

 tions, analog traces, and continuity from section to sec- 

 tion. Locations of observations were plotted to help check 

 positions and an independent check of time and distance 

 between observations was made against the ship's speed. 

 For observations through 1970, copies of marine weather 

 logs on which positions of 6-h weather observations were 

 logged independently of XBT logs were also used to cor- 

 rect time and position errors. 



The data checks were made to eliminate large errors 

 due to instrument failure not detected before digitizing. 

 These were of several types: 1) erroneously high 

 temperatures throughout a trace due to defective ther- 

 mistors or insulation failure from the start, 2) insula- 

 tion failure during the probe descent which would intro- 



Computational Procedures 



The determination of the vertical sections of mean 

 subsurface temperature presented herein involved three 

 steps: 1) Conversion of observed temperatures from 

 each section to temperatures on a standard 

 grid; 2) computation of a seasonally varying mean at 

 each grid point by least squares fit of 12-, 6-, and 4-mo 

 harmonics; and 3) reconstruction of gridded 

 temperature fields from the harmonics, spatial smooth- 

 ing, and contouring of vertical sections. The computer 

 programs used for this were adaptations by Eber of those 

 he prepared at SWFC to map environmental variables in 

 marine weather observations for presentation in Fishing 

 Information. 



1. Conversion to a standard grid. — It was previ- 

 ously mentioned that observations were not taken at the 

 same predetermined location from section to section. 

 The first step was to analyze observed values from each 

 section to a standard rectangular grid, using a procedure 

 from Eber's EDMAP 6 (Environmental Data Manipula- 

 tion, Analysis, and Plotting) program. 



The grid was selected with a distance interval of 92.5 

 km (50 n.mi.) and a depth interval of 10 m. This resulted 

 in a grid of 42 by 51 points representing a vertical section 

 3,800 km (2,050 n.mi.) by 500 m. Distance and depth 

 were converted to grtd coordinate units for the 

 temperature analysis. 



The procedure scanned the data list and fitted 

 temperature values to the grid. Each observation con- 

 tributed to the values at its nearest grid points according 

 to an inverse weighting scheme based on distance from 

 the observation to each of the grid points. The weighting 

 factor decreased to zero at one grid length. If no observa- 

 tion was found within one grid length of a grid point, it 

 was flagged as a "no data" point in that section. 



The procedure can be viewed as a refinement of center- 

 ing the observational data within 185-km (100 n.mi.) by 

 20-m blocks that have a 50% overlap between adjacent 

 (both vertically and horizontally) blocks. However, if 

 there is more than one observation within a block, each is 

 weighted according to the distance to the center of the 

 block and the number of grid points it will affect. (The 



'Unpublished documentation of the EDMAP program is on file at the 

 Southwest Fisheries Center, National Marine Fisheries Service, NOAA, 

 La Jolla. CA 92038. 



