In early tests, the program occasionally skipped some of the data in the quality control (QC in GHCN2, flag in GHCN1 parlance) files, even in the simplest read-and-write routines. I was unable to discover the cause of the skips.
Part of the problem may line in a change in formatting of GHCN quality control data. In GHCN1, a flag value is given for every monthly value in the corresponding climate file. In GHCN2, NCDC switched a one-entry-per-line format in which only monthly values with questionable data points are listed. Such a format saves a tremendous amount of space, but is more of a challenge (for me) to write appropriate search codes for.
There were no patterns to the data omissions that I could discern, but I continued to revise the search procedures until I at least minimized the problem.
Given the apparent random nature of the skipping, I cannot be completely sure that successful tests on my part represent a successful correction, therefore I recommend you compare the GHCN2 flag output with the original GHCN2 QC files to see if there are any gaps.
Copyright © 1995-2011, David M. Lawrence
Created with the Freeware Edition of HelpNDoc: Free help authoring environment