On the structure of atmospheric warming in models and observations: Implications for the lapse rate feedback

Abstract

This dissertation investigates the structure of atmospheric warming in observations and general circulation models (GCMs). Theory and GCMs suggest that warming is amplified in the tropical upper troposphere relative to the lower troposphere and the surface -- a phenomenon known as vertical amplification. We assess model and observational agreement using several amplification metrics derived from the satellite-borne microwave sounding unit (MSU) atmospheric temperature trends. An important correction to the satellite microwave record is the removal of temperature drifts caused by changes in diurnal sampling. This correction the principal source of uncertainty in microwave temperature datasets. Furthermore, in three existing datasets, the ratio of tropical warming between the upper troposphere (T24 channel) and the surface (dT24/dTs ~ 0.6 -- 1.3) is lower than that of GCMs (~1.4 -- 1.6). To better understand these issues, we produced an alternate MSU dataset with an improved diurnal correction. We show that existing MSU datasets likely underestimate tropical mid-tropospheric temperature trends. Subsequent improvements to MSU datasets using similar diurnal correction techniques leads to amplification ratios (between T24 and the surface) that are in accord with models. Another measure of tropical tropospheric amplification is the relative warming between the upper troposphere (T24) and the lower-middle troposphere (TLT). We show that most GCMs have excessive T24/TLT amplification compared to satellite microwave observations, even when models are forced with prescribed sea-surface temperatures (SSTs). A number of possible reasons for this discrepancy are assessed. Observational uncertainty in the satellite microwave record is substantial and, when taken into account, many models agree with observations within the observational uncertainty range, though about half of the model ensemble members considered still have significant discrepancies compared to observations. Our findings indicate that the prescribed ozone and stratospheric aerosol forcings do not effect T24/TLT amplification in models. On the other hand, model parameterizations for convection and microphysics and, to a lesser degree, uncertainty in the prescribed SST dataset can influence model amplification behavior and bring models into closer accord with observations. In all, significant T24/TLT discrepancies between models and observations remain, but may be reduced with improved model parameterizations. An underlying motivation for understanding the structure of atmospheric warming is that it is responsible for a large negative lapse rate feedback in future climate simulations. To understand factors that control the global lapse rate feedback across models, we use principal component analysis to find the modes of variability that best explain variance in the local lapse rate feedback. We find that models exhibit marked variability in the lapse rate feedback in the southern hemisphere extratropics. This mode is strongly correlated with the global average lapse rate feedback and is largely a function of the competing influence of tropical and Antarctic surface warming. We show that muted southern ocean sea surface warming and the non-local influence of tropical surface warming contributes to a highly variable lapse rate feedback in the sub-Antarctic across models. This behavior is dissimilar to northern hemisphere high latitudes, which are characterized by strong Arctic amplification and a relatively uniform local lapse rate feedback across GCMs. Climatological Antarctic sea ice extent influences Antarctic warming and, as a result, influences both the meridional profile of warming in the southern hemisphere and the global lapse rate feedback.