panel cautions that trends in such short periods of record with arbitrary start and end points are not necessarily representative of how the atmosphere is changing in response to long-term human-induced changes in atmospheric composition. Given reliable measurements, as outlined in Recommendation #1, the level of confidence that can be attached to the trends will increase as the period of record of upper air measurements lengthens.
(6) It is not currently possible to determine whether or not there exists a fundamental discrepancy between modeled and observed atmospheric temperature changes since the advent of satellite data in 1979. Measurement uncertainties (Finding #3), modeling uncertainties (Finding #4), and sampling uncertainties (Finding #5) were all considered by the panel as possible causes of the disagreement between models and observations. None of them can be singled out as the dominant factor, nor can any one of them be shown to be unimportant. Surface temperature and lower to mid-tropospheric temperature are different entities, which should not be expected to vary in precisely the same manner in response to human-induced and natural climate forcings during a particular 20-year period of record. Hence, the panel concludes that at least part of the observed disparity between the 20-year changes in surface and mid-tropospheric temperature is probably real, but the measurement, modeling, and sampling uncertainties alluded to above make it difficult to precisely attribute the disparity to any particular sources. A more definitive reconciliation of modeled and observed temperature changes awaits the extension and improvement of the observations and the algorithms used in processing them, better specification of the natural and human-induced climate forcings during this period, and improvement of the models used to simulate the atmospheric response to these forcings.break