The main issues are:
- Completeness: are the same reports available in TAC and BUFR format?
- Timeliness: are the BUFR reports received in good time (as early as the TAC reports)?
- Duplicates - see below.
- Structure: how are the radiosonde reports represented?
For a given station there are different levels of availability: a) no BUFR reports, b) some reports received in BUFR and c) all reports received in BUFR - measured over a period of weeks/months. The timeliness of BUFR data may be poorer, especially before the BUFR reports are fully operational. Exact duplicates (within BUFR) are not too much of a problem but excessive duplication should be avoided (we have seen up to 50 duplicate reports for some radiosonde stations). Near-duplicates (unless one is clearly a correction/replacement for a previous report) can be a problem - which should be used or (ideally) should different elements of the different reports be used?
Structure: the radiosonde BUFR data on the GTS seems to fall into three categories: a) converted from TEMP, usually still as separate parts, b) generated directly in BUFR but low resolution - standard and significant levels combined, c) high-resolution BUFR data. Currently (late 2013) most of the high-resolution reports come from a subset of European stations, in some cases there is a low resolution report (category b) up to 100 hPa and then a high-resolution report for the full ascent. Category a) - data as separate parts - should only be a temporary expedient as it doesn't properly fit into the BUFR transmission rules. It also means that NWP centres have to merge the different parts together. Wind-only "PILOT" reports generally have height as the vertical coordinate although in TAC code the standard levels are reported as pressure (see WMO Publication No. 306 - Manual on Codes, Volume II Regional Codes and National Coding Practices for details of the corresponding heights used in different regions/countries) - it would be much cleaner to report all levels with a height coordinate in BUFR. Some stations report mainly in TEMP code but use PILOT code for significant levels winds - also a nuisance.