Assessing error in global atmospheric temperatures

When dealing with relatively short climate time series (e.g. 21 years), two aspects of dataset error must be addressed. The first deals with measurement error, answering the question, "How precise are the data?". The second question deals with representation, "How representative of all 21-year periods is this particular period?" It is possible to answer the first question using tropospheric temperatures because there have been two fully independent systems in operation since 1979 for comparison. The Microwave Sounding Unit measures emissions from broad vertical layers while radiosondes report data at many specific levels. Generating datasets from each system for true comparison studies is not trivial, however, as unique sampling or response characteristics of each may degrade such comparisons. Several statistical comparisons have been generated to assess the measurement error in each of the system, especially with regard to global variations and trends since 1979. The answer to the second question is perhaps more meaningful, yet is more difficult and suggestions will be offered.