|Fig. 1 Day of Year format, Layer Nine|
I am now using only data with the error flag set to '0', zero, meaning it is an 'acceptable value' with none of the about nine error categories.
In other words the WOD folk have determined that measurement to be free of errors (see WOD User's Manual, Table 12, p. 39, p. 49 PDF).
The exercise decreased the number of rows in my raw SQL tables of data from 974,560,328 rows down to 924,193,524 rows (50,366,804 fewer rows, or about 5%).
|Fig. 2 Month format, Layer Nine|
On the Layer Nine area of the WOD zones, the mean average temperature of the top depth level (0-200m) of the zero-error data decreased by about 2 deg. C in Layer Nine, while the other depth layers were the same (compare Fig. 1 and Fig. 4 with Fig. 3a and Fig. 3b here).
As we go through the layers (0-18) I expect the same results to hold.
|Fig. 3 Year format zero errors, Layer Nine|
|Fig. 4 Year format with errors, Layer Nine|
So I decided to make the change, since WOD recommends it (see "III. QUALITY CONTROL PROCEDURES" at p. 33, p. 43 PDF in the WOD User's Manual).
The upgrade/update applies to the "Year", "Month", and "Day of Year" graph types.
In conclusion, the time and effort was worth it because striving for accuracy is a number one level priority.
The previous post in this series is here.