 |
| Titel |
The importance of the covariation of the geographical distribution of SSTs and deep convection for tropical tropospheric temperature trends 1980-present |
| VerfasserIn |
Stephan Fueglistaler, Thomas Flannaghan, Stephen Po-Chedley, Isaac Held, Claire Radley, Ming Zhao, Bruce Wyman |
| Konferenz |
EGU General Assembly 2016
|
| Medientyp |
Artikel
|
| Sprache |
en
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 18 (2016) |
| Datensatznummer |
250124197
|
| Publikation (Nr.) |
 EGU/EGU2016-3586.pdf |
|
|
|
|
|
| Zusammenfassung |
| Enhanced upper tropospheric warming relative to the surface in the tropics is a prominent feature
of numerical model simulations, but it has been suggested that models overestimate this warming
compared to observations for the period 1980 to present. Here, we focus on the factors controlling
atmospheric temperature trends in numerical model calculations with prescribed Sea Surface Temperatures
(SSTs). CMIP5 model runs show a remarkably large spread in tropical temperature trends over the period
1980-2008 despite being forced with observed SSTs. Here, we show that the model trends are consistent
with the atmospheric temperature profile being tightly constrained by the surface layer conditions in
regions of deep convection. Large trend differences arise from the use of two different SST data, the
"HURRELL" and the "HadISST1" data. These two SSTs
have very similar tropical average trends, but differ substantially in the warmest percentiles where
most deep convection occurs. The models' temperature trend differences in the tropical troposphere reflect
the trend differences in the regions of highest SSTs.
Further, we show that trend differences in model calculations using
identical SSTs is strongly related to differences in the geographical pattern of strong precipitation
(used as a simple proxy for deep convection) between models, and between ensemble runs of a model.
The time series of precipitation weighted SSTs can explain more than half of the variance in temperature
trends. The variance in trends between ensemble members of the same model, and between ensemble means
of different models, is similar. However, the decrease in variance upon averaging over ensemble members is
modest compared to the expected scaling for independent samples, which provides evidence for
systematic differences between models in their response in the geographical distribution of convection
to changes in SST patterns. |
|
|
|
|
|
|