Geosat GM SDR Recovery & Processing
John Lillibridge - NOAA/NESDIS/ORA/ORAD/LSA: E/RA31
Cynthia Zeigler - NOAA/NESDIS/NODC/ISMD: E/OC3
10/23/2003

Data Sources:

	- NODC 3480 cartridge tapes - except as noted below: 512 of total 549 GM SDRs

	- NASA/GSFC (Anita Brenner) - 17 days, all 1986: 
		86_038 - 043, 051, 070, 071, 074, 075, 198-202, 266

	- NODC 8 mm backup tapes, read at NODC - 15 days, all 1985:
		85_182, 183, 200, 214, 240, 242, 243, 264, 269, 271, 300, 305, 339, 345, 347

	- NODC 8 mm backup tapes, read at LSA - final 5 days, all 1985:
		85_138, 146, 152, 267, 319

Data/Header Processing:

	Headers:
		Correct length restored to 898 bytes
		Add <LF> at end of each header
		Three missing headers recreated from Geosat SDR->GDR logs:
			85_138, 146, 152 (85_146 existed, but was for day 86_096)
		100% recovery of header information

	Data:
		3480 files "reblocked" from 4940 byte to original 260 byte records
		<LF> added to each logical record: new record-length = 261 bytes
		trailing blank lines removed from two files:
			85_103 (9 lines removed); 86_144 (8 lines removed)

	Duplicates:
		3480 files (header &/or data) exact duplicates:
			tape 44, files 1-12 == tape 45, files 1-12
			tape 59, files 1-10 == tape 60, files 2-11
		3480 files NOT exact duplicates:
			85_096: tape 2 files 1,2 and tape 43 files 5,6
			85_105: tape 3 files 5,6 and tape 43 files 7,8
		NODC & GSFC duplicate: 86_198 (NODC data file empty)

Data corruption and errors:

	Missing records (# data records < expected # records from header):
		85_120, 182, 183, 200, 240, 242, 243, 254, 264, 267, 269, 300, 305, 319, 339, 345, 347
		86_006, 051, 266

	10 of these with losses > 1000 data records:
		85_120 (87961), 85_183 (13558), 85_200 (32981), 85_240 (19486), 85_242 (60710)
		85_254 ( 1158),  85_264 (34039), 85_319 (13920), 86_006 (31645), 86_266 (23777)

	Nonetheless: total data return of 99.34% !! (assuming headers are correct)

	Internal corruption (incomplete records NOT 260 bytes long; tape reading errors?):
		85_182, 242, 319 (requires special reading, keyed to "XXX" at end of data record)

	Frame counts "out of bounds" based on start/stop frame counts in header

	Time-tag blunders in correlation between frame count and UTC clock time