Tropical Temperature Trends in AMIP Simulations and the Impact of SST Uncertainties

The comparison of trends in various climate indices in observations and models is of fundamental importance for judging the credibility of climate projections. Tropical tropospheric temperature trends have attracted particular attention as this comparison may suggest a model deficiency. One can think of this problem as composed of two parts: one focused on tropical surface temperature trends and the associated issues related to forcing, feedbacks, and ocean heat uptake; and a second part focusing on connections between surface and tropospheric temperatures and the vertical profile of trends in temperature. Here, we focus on the atmospheric component of the problem. We show that two ensembles of GFDL HiRAM model runs (similar results are shown for NCAR's CESM-v1 model) with different commonly used prescribed sea surface temperatures (SSTs), namely the HadISST1 and `Hurrell' data sets, have a difference in upper tropical tropospheric temperature trends (~ 0.1K/decade at 300hPa for the period 1984-2008) that is about a factor 3 larger than expected from moist adiabatic scaling of the tropical average SST trend difference. We show that this surprisingly large discrepancy in temperature trends is a consequence of SST trend differences being largest in regions of deep convection. Further, trends, and the degree of agreement with observations, not only depend on SST data set and the particular `observed' atmospheric temperature data set, but also on the period chosen for comparison. Due to the large impact on atmospheric temperatures, these systematic uncertainties in SSTs need to be resolved before the fidelity of climate models' tropical temperature trend profiles can be assessed.