[Update 31 October 2014]
The summaries of the Climate Dialogue discussion on the (missing) tropical hot spot are now online (see links below). We have made two versions: an extended and a shorter version.
Both versions can be downloaded as pdf documents:
Summary of the climate dialogue on the (missing) tropical hot spot
Extended summary of the climate dialogue on the (missing) tropical hot spot
The (missing) tropical hot spot is one of the long-standing controversies in climate science. Climate models show amplified warming high in the tropical troposphere due to greenhouse forcing. However data from satellites and weather balloons don’t show much amplification. What to make of this? Have the models been ‘falsified’ as critics say or are the errors in the data so large that we cannot conclude much at all? And does it matter if there is no hot spot?
We are really glad that three of the main players in this controversy have accepted our invitation to participate: Steven Sherwood of the University of New South Wales in Sydney, Carl Mears of Remote Sensing Systems and John Christy of the University of Alabama in Huntsville.
Climate Dialogue editorial staff
Rob van Dorland, KNMI
Marcel Crok, science writer
Introduction The (missing) tropical hot spot
The (missing) hot spot in the tropics
Based on theoretical considerations and simulations with General Circulation Models (GCMs), it is expected that any warming at the surface will be amplified in the upper troposphere. The reason for this is quite simple. More warming at the surface means more evaporation and more convection. Higher in the troposphere the (extra) water vapour condenses and heat is released. Calculations with GCMs show that the lower troposphere warms about 1.2 times faster than the surface. For the tropics, where most of the moist is, the amplification is larger, about 1.4.
This change in thermal structure of the troposphere is known as the lapse rate feedback. It is a negative feedback, i.e. attenuating the surface temperature response due to whatever cause, since the additional condensation heat in the upper air results in more radiative heat loss.
IPCC published the following figure in its latest report (AR4) in 2007:
Source: http://www.ipcc.ch/publications_and_data/ar4/wg1/en/figure-9-1.html (based on Santer 2003)
The figure shows the response of the atmosphere to different forcings in a GCM. As one can see, over the past century, the greenhouse forcing was expected to dominate all other forcings. The expected warming is highest in the tropical troposphere, dubbed the tropical hot spot.
The discrepancy between the strength of the hot spot in the models and the observations has been a controversial topic in climate science for almost 25 years. The controversy[i] goes all the way back to the first paper of Roy Spencer and John Christy[ii] about their UAH tropospheric temperature dataset in the early nineties. At the time their data didn’t show warming of the troposphere. Later a second group (Carl Mears and Frank Wentz of RSS) joined in, using the same satellite data to convert them into a time series of the tropospheric temperature. Several corrections, e.g. for the orbital changes of the satellite, were made in the course of years with a warming trend as a result. However the controversy remains because the tropical troposphere is still showing a smaller amplification of the surface warming which is contrary to expectations.
Some researchers claim that observations don’t show the tropical hot spot and that the differences between models and observations are statistically significant[iii]. On top of that they note that the warming trend itself is much larger in the models than in the observations (see figure 2 below and also ref.[iv]). Other researchers conclude that the differences between the trends of tropical tropospheric temperatures in observations and models are statistically not inconsistent with each other[v]. They note that some radiosonde and satellite datasets (RSS) do show warming trends comparable with the models (see figure 3 below).
The debate is complex because there are several observational datasets, based on satellite (UAH and RSS) but also on radiosonde measurements (weather balloons). Which of the dataset is “best” and how does one determine the uncertainty in both datasets and model simulations?
The controversy flared up in 2007/2008 with the publications of two papers[vi][vii] of the opposing groups. Key graphs in both papers are the best way to give an impression of the debate. First Douglass et al. came up with the following graph showing the disagreement between models and observations:
Figure 2. Temperature trends for the satellite era. Plot of temperature trend (°C/decade) against pressure (altitude). The HadCRUT2v surface trend value is a large blue circle. The GHCN and the GISS surface values are the open rectangle and diamond. The four radiosonde results (IGRA, RATPAC, HadAT2, and RAOBCORE) are shown in blue, light blue, green, and purple respectively. The two UAH MSU data points are shown as gold-filled diamonds and the RSS MSU data points as gold-filled squares. The 22-model ensemble average is a solid red line. The 22-model average ±2σSE are shown as lighter red lines. MSU values of T2LT and T2 are shown in the panel to the right. UAH values are yellow-filled diamonds, RSS are yellow-filled squares, and UMD is a yellow-filled circle. Synthetic model values are shown as white-filled circles, with 2σSE uncertainty limits as error bars. Source: Douglass et al. 2008
Santer et al. criticized Douglass et al. for underestimating the uncertainties in both model output and observations and also for not showing all radiosonde datasets. They came up with the following graph:
Figure 3. Vertical profiles of trends in atmospheric temperature (panel A) and in actual and synthetic MSU temperatures (panel B). All trends were calculated using monthly-mean anomaly data, spatially averaged over 20 °N–20 °S. Results in panel A are from seven radiosonde datasets (RATPAC-A, RICH, HadAT2, IUK, and three versions of RAOBCORE; see Section 2.1.2) and 19 different climate models. The grey-shaded envelope is the 2σ standard deviation of the ensemble-mean trends at discrete pressure levels. The yellow envelope represents 2σSE, DCPS07’s estimate of uncertainty in the mean trend. The analysis period is January 1979 through December 1999, the period of maximum overlap between the observations and most of the model 20CEN simulations. Note that DCPS07 used the same analysis period for model data, but calculated all observed trends over 1979–2004. Source: Santer (2008)
The grey-shaded envelope is the 2σ standard deviation of the ensemble-mean trends of Santer et al. while the yellow band is the estimated uncertainty of Douglass et al. Some radiosonde series in the Santer graph (like the Raobcore 1.4 dataset) show even more warming higher up in the troposphere than the model mean.
Not surprisingly the debate didn’t end there. In 2010 McKitrick et al.[viii] updated the results of Santer (2008), who limited the comparison between models and observations to the period 1979-1999, to 2009. They concluded that over the interval 1979–2009, model projected temperature trends are two to four times larger than observed trends in both the lower troposphere and the mid troposphere and the differences are statistically significant at the 99% level.
Christy (2010)[ix] analysed the different datasets used and concluded that some should be discarded in the tropics:
Figure 4. Temperature trends in the lower tropical troposphere for different datasets and for slightly differing periods (79-05 = 1979-2005). UAH and RSS are the estimates based on satellite measurements. HadAt, Ratpac, RC1.4 and Rich are based on radiosonde measurements. C10 and AS08[x] are based on thermal wind data. The other three datasets give trends at the surface (ERSST being for the oceans only while the other two combine land and ocean data). Source: Christy (2010)
Christy (2010) concluded that part of the tropical warming in the RSS series is spurious. They also discarded the indirect estimates that are based on thermal wind. Not surprisingly Mears (2012) disagreed with Christy’s conclusion about the RSS trend being spurious writing that “trying to determine which MSU [satellite] data set is “better” based on short-time period comparisons with radiosonde data sets alone cannot lead to robust conclusions”.[xi]
Christy (2010) also introduced what they called the “scaling ratio”, the ratio of tropospheric to surface trends and concluded that these scaling ratios clearly differ between models and observations. Models show a ratio of 1.4 in the tropics (meaning troposphere warming 1.4 times faster than the surface), while the observations have a ratio of 0.8 (meaning surface warming faster than the troposphere). Christy speculated that an alternate reason for the discrepancy could be that the reported trends in temperatures at the surface are spatially inaccurate and are actually less positive. A similar hypothesis was tested by Klotzbach (2009).[xii]
In an extensive review article about the controversy published in early 2011 Thorne et al. ended with the conclusion that “there is no reasonable evidence of a fundamental disagreement between tropospheric temperature trends from models and observations when uncertainties in both are treated comprehensively”. However in the same year Fu et al.[xiii] concluded that while “satellite MSU/AMSU observations generally support GCM results with tropical deep‐layer tropospheric warming faster than surface, it is evident that the AR4 GCMs exaggerate the increase in static stability between tropical middle and upper troposphere during the last three decades”. More papers then started to acknowledge that the consistency of tropical tropospheric temperature trends with climate model expectations remains contentious.[xiv][xv][xvi][xvii]
We will focus the discussion on the tropics as the hot spot is most pronounced there in the models. Core questions are of course whether we can detect/have detected a hot spot in the observations and if not what are the implications for the reliability of GCMs and our understanding of the climate?
1) Do the discussants agree that amplified warming in the tropical troposphere is expected?
2) Can the hot spot in the tropics be regarded as a fingerprint of greenhouse warming?
3) Is there a significant difference between modelled and observed amplification of surface trends in the tropical troposphere (as diagnosed by e.g. the scaling ratio)?
4) What could explain the relatively large difference in tropical trends between the UAH and the RSS dataset?
5) What explanation(s) do you favour regarding the apparent discrepancy surrounding the tropical hot spot? A few options come to mind: a) satellite data show too little warming b) surface data show too much warming c) within the uncertainties of both there is no significant discrepancy d) the theory (of moist convection leading to more tropospheric than surface warming) overestimates the magnitude of the hotspot
6) What consequences, if any, would your explanation have for our estimate of the lapse rate feedback, water vapour feedback and climate sensitivity?
[i]Thorne, P. W. et al., 2011, Tropospheric temperature trends: History ofan ongoing controversy. WIRES: Climate Change, 2: 66-88
[ii]Spencer RW, Christy JR. Precise monitoring of global temperature trends from satellites. Science 1990, 247:1558–1562.
[iii] Christy, J. R., B. M. Herman, R. Pielke Sr., P. Klotzbach, R. T. McNider, J. J. Hnilo, R. W. Spencer, T. Chase, and D. H. Douglass (2010), What do observational datasets say about modeled tropospheric temperature trends since 1979?, Remote Sens., 2, 2148–2169, doi:10.3390/rs2092148.
[v]Thorne, P.W. Atmospheric science: The answer is blowing in the wind. Nature Geosci. 2008, doi:10.1038/ngeo209
[vi] Douglass DH, Christy JR, Pearson BD, Singer SF. A comparison of tropical temperature trends with model predictions. Int J Climatol 2008, 27:1693–1701
[vii] Santer, B.D.; Thorne, P.W.; Haimberger, L.; Taylor, K.E.; Wigley, T.M.L.; Lanzante, J.R.; Solomon, S.; Free, M.; Gleckler, P.J.; Jones, P.D.; Karl, T.R.; Klein, S.A.; Mears, C.; Nychka, D.; Schmidt, G.A.; Sherwood, S.C.; Wentz, F.J. Consistency of modelled and observed temperature trends in the tropical troposphere. Int. J. Climatol. 2008, doi:1002/joc.1756
[viii] McKitrick, R. R., S. McIntyre and C. Herman (2010) “Panel and Multivariate Methods for Tests of Trend Equivalence in Climate Data Sets.” Atmospheric Science Letters, 11(4) pp. 270-277, October/December 2010 DOI: 10.1002/asl.290
[ix] Christy, J. R., B. M. Herman, R. Pielke Sr., P. Klotzbach, R. T. McNider, J. J. Hnilo, R. W. Spencer, T. Chase, and D. H. Douglass (2010), What do observational datasets say about modeled tropospheric temperature trends since 1979?, Remote Sens., 2, 2148–2169, doi:10.3390/rs2092148
[x] Allen RJ, Sherwood SC. Warming maximum in the tropical upper troposphere deduced from thermal winds. Nat Geosci 008, 1:399–403
[xi] Mears, C. A., F. J. Wentz, and P. W. Thorne (2012), Assessing the value of Microwave Sounding Unit–radiosonde comparisons in ascertaining errors in climate data records of tropospheric temperatures, J. Geophys. Res., 117, D19103, doi:10.1029/2012JD017710
[xii] Klotzbach PJ, Pielke RA Sr., Pielke RA Jr., Christy JR, McNider RT. An alternative explanation for differential temperature trends at the surface and in the lower troposphere. J Geophys Res 2009, 114:D21102. DOI:10.1029/2009JD011841
[xiii] Fu, Q., S. Manabe, and C. M. Johanson (2011), On the warming in the tropical upper troposphere: Models versus observations, Geophys. Res. Lett., 38, L15704, doi:10.1029/2011GL048101
[xiv] Seidel, D. J., M. Free, and J. S. Wang (2012), Reexamining the warming in the tropical upper troposphere: Models versus radiosonde observations, Geophys. Res. Lett., 39, L22701, doi:10.1029/2012GL053850
[xv] Po-Chedley, S., and Q. Fu (2012), Discrepancies in tropical upper tropospheric warming between atmospheric circulation models and satellites, Environ. Res. Lett
[xvi] Benjamin D. Santer, Jeffrey F. Painter, Carl A. Mears, Charles Doutriaux, Peter Caldwell, Julie M. Arblaster, Philip J. Cameron-Smith, Nathan P. Gillett, Peter J. Gleckler, John Lanzante, Judith Perlwitz, Susan Solomon, Peter A. Stott, Karl E. Taylor, Laurent Terray, Peter W. Thorne, Michael F. Wehner, Frank J. Wentz, Tom M. L. Wigley, Laura J. Wilcox, and Cheng-Zhi Zou, Identifying human influences on atmospheric temperature, PNAS 2013 110 (1) 26-33; published ahead of print November 29, 2012, doi:10.1073/pnas.1210514109
[xvii] Thorne, P. W., et al. (2011), A quantification of uncertainties in historical tropical tropospheric temperature trends from radiosondes, J. Geophys. Res., 116, D12116, doi:10.1029/2010JD015487
Guest blog Carl Mears
Thoughts and plots about the tropical tropospheric hot spot.
Carl Mears, Remote Sensing Systems
In the deep tropics, in the troposphere, the lapse rate (the rate of decrease of temperature with increasing height above the surface) is largely controlled by the moist adiabatic lapse rate (MALR). This is true both in complicated simulations performed by General Circulation Models, and in simple, back of the envelope calculations (Santer et al, 2005). The reasoning behind this is simple. If the lapse rate were larger than MALR, then the atmosphere would be unstable to convection. Convection (a thunderstorm) would then occur, and heat the upper troposphere via the release of latent heat as water vapor condenses into clouds, and cool the surface via evaporation and the presence of cold rain/hail. If the lapse rate were smaller than MALR, then convection would be suppressed, allowing the surface to heat up without triggering a convective event. On average, these processes cause the lapse rate to be very close to the MALR. Note that this argument does not apply outside the tropics, because the dynamics become more complex due to the Coriolis force and the presence of large north/south temperature gradients, or in regions with very low relative humidity, such as deserts, where the atmosphere may be far from saturated near the surface and thus the MALR does not apply.
Because the MALR decreases with temperature, this means the any temperature increase at the surface becomes even larger high in the troposphere. This causes the so called hot spot, a region high in the troposphere that shows more warming (or cooling) than the surface. Note that at this point, I haven’t said a thing about greenhouse gases. In fact, this effect has nothing to do with the source of the warming, as long as it arises near the surface. Surface warming due to any cause would show a tropospheric hotspot in the absence of other changes to the heating and cooling of the atmosphere. Never the less, the tropospheric hotspot is often presented as some sort of lynchpin of global warming theory. It is not. It is just a feature of a close-to-unstable moist atmosphere.
Now, I will turn my attention to one of the core questions of this discussion – “can we detect/have detected a hot spot in the observations “. On monthly time scales, there is no question. If we average across the tropics, the temperature of the upper troposphere is strongly correlated with the temperature of the surface, only with larger amplitude (Santer et al., 2005). On decadal time scales, the results obtained depend on the datasets chosen, as Santer 2005 showed for the RSS and UAH satellite datasets and a few homogenized radiosonde datasets. Here we expand this a little further to include more homogenized radiosonde datasets and two of the more recent reanalysis datasets, MERRA and ERA-Interim. Figure 1 shows the ratio of the mid to upper tropospheric temperature trends to surface temperature trends in the deep tropics (20S to 20N). Each point on the graph is the trend starting in January 1979, and ending at the date on the x-axis. The surface temperature is from HADCRUT4. The mid to upper tropospheric data is the “temperature tropical troposphere” product, or TTT, first introduced by Fu and Johanson (2005). For MSU/AMSU, it is equal to 1.1*TMT – 0.1*TLS. This combination has the effect of adjusting for the cooling effect of the stratosphere on TMT by subtracting off part of the stratospheric cooling measured by TLS. The weighting function for this product is centered in the mid to upper tropical troposphere, where we expect the hot spot to be most pronounced.
Fig. 1. Ratio of trends in TTT to trends in TSurf as a function of the ending year of the trend analysis. The starting point is January 1979. The surface dataset used is HADCRUT4. The pink horizontal line is at a value of 1.4, the amplification factor for TTT in reference 1.
Two conclusions can easily be reached from this plot. First, it takes about 25 years (or more) for the measured trend ratios to settle down to reasonably constant values. This is due to the effects of both measurement errors, and “weather noise”. I think that this is part of the cause of the controversy surrounding this topic – we began discussing such trend ratios before we had enough data for the ratios to be stable over time. Second, the values that are ultimately reached depend strongly depend on which upper air dataset is used. For some datasets (HadAT, UAH, IUK, RAOBCORE 1.5, ERA-Interim), the trend ratio is less than 1.0, indicating lack of a tropospheric hotspot. For other datasets (RICH, RAOBCORE 1.4, RSS, MERRA, and STAR), the ratio is greater than one, indicating tropospheric amplification and the presence of a hotspot. CMIP-3 Climate models predicted an amplification value of about 1.4 for the TTT temperature product used here (Santer et al., 2005). Some upper air datasets are in relatively close agreement with these expectations, such as the RSS and STAR satellite data, the older version of RAOBCORE (V1.4), and the MERRA reanalysis (which uses the STAR data as one of its inputs, so it is not completely independent of STAR). Often one or more of these datasets is used to argue that a tropical hotspot exists or does not exist. A more balanced analysis shows that it is difficult to prove or disprove the presence of the tropospheric hotspot given the current state of the data.
In Fig. 2, I have reproduced panel D of Fig. 4 from Santer et al. (2005), except with updated measured data, the addition of the reanalysis data, and the use of CMIP-5 model results. The CMIP-5 model results for 1979-2012 were made by splicing together results from 20th century simulations (before 2005, using measured values of the forcings), and RCP8.5 21st century predictions (after 2005, results using predicted values for the various forcings. For details on this process, see Santer et al., 2012)
Fig. 2. Scatter plot of trend (1979-2012) in TTT as a function of trend in TSurf. The model results cluster around a line with a slope of 1.45, indicating a tropospheric hotspot. For the observed results, HadCRUT4 is used for the surface temperatures, and various sources of tropospheric temperature (Satellites, Radiosondes, and Reanalysis) are used to TTT.
The general story around the hotspot remains unchanged from Santer et al 2005, except that the expected scaling ratio has increased from 1.40 to 1.45 with the use of CMIP-5 data. Two sources of measured data (I realize that a reanalysis is not really a measurement), STAR and MERRA, are reasonably close to the fitted line, while others, such as the HADAT Radiosonde dataset and the UAH satellite dataset, are far from the line. Other datasets are distributed in between. Note that the RSS data point has error bars both in the X and Y direction. These are 90% uncertainty ranges derived from the error ensembles that have been recently produced for the RSS dataset (Mears et al., 2011) and the HadCRUT4 dataset (Morice et al, 2012). (These error ensembles are made up of different realizations of the datasets that are consistent with the estimated errors, including measurement, sampling, and construction errors. The correlations of the errors across both time and location are thus automatically included if the error ensemble members are processed by the user in the same way as the baseline data.) This is the first time that we have been able to put error bars on the observed points on this plot, in both directions, in such a consistent manner.
Looking at Fig. 2., it is obvious that the observed trends in both temperature datasets are at the extreme low end of the model predictions. This problem has grown over time as the length of the measured data grows. (As the comparison time period gets longer, the uncertainty in linear trends both the measured and modeled time series decreases simply because of the longer time period.) For the time being, I am tabling the discussion of this problem and focusing in the discussion of the hot spot. In my mind, the problem of the trend magnitude is more interesting than the argument about the hotspot, and I hope to return to it later in this process. But for now I will stay focused on the hotspot.
Fig. 3. Histograms of the troposphere/surface trend scaling ratio from the RSS/HadCRUT4 error ensembles, and from 33 CMIP-5 model runs.
In Fig. 3, we explore the implications of the RSS and HadCRUT4 error ensembles further. The top histogram shows the range of scaling ratios consistent with the RSS satellite data and the HadCRUT4 surface data when the estimated errors in each are taken into consideration. The bottom histogram shows the range of scaling ratio shown in the 33 CMIP-5 model runs. The two distributions overlap, indicating consistency of this set of observations with the models, though the mean value shown by the observations is clearly lower than that predicted by the models.
It has been suggested that the lack of a tropospheric hotspot (if there is such a lack) is mostly due to errors in the surface temperature datasets, which are (in this story line) suspected of being biased in the direction of too much warming. This seems unlikely. Clearly, the above spread in results for different upper air datasets reveals considerable structural uncertainty (Thorne et al, 2005) for the upper air data, and the error bar on the RSS trend values is much larger than the error bar for the HadCRUT4 value. Also, the various surface datasets are much more similar to each other. To show this, I redo the analysis in Figure 1 using a different surface dataset constructed by NOAA (GHCN-ERSST). The final trend ratios are almost identical to those found using HADCRUT4, and the conclusions reached are unchanged.
Figure 4. Ratio of trends in TTT to trends in TSurf as a function of the ending year of the trend analysis. The starting point is January 1979. The surface dataset used is GHCN-ERSST.
Conclusion: Taken as a whole, the errors in the measured tropospheric data are too great to either prove or disprove the existence of the tropospheric hotspot. Some datasets are consistent (or even in good agreement) with the predicted values for the hotspot, while others are not. Some datasets even show the upper troposphere warming less rapidly than the surface.
Dr. Mears has a B.S. in Physics from the University of Washington (1985), and a PhD. in Physics from University of California, Berkeley (1991), where his thesis research involved the development of quantum-noise-limited superconducting microwave heterodyne receivers. He joined Remote Sensing Systems in 1998. Since then, he has validated SSM/I and TMI winds versus in situ measurements, developed and validated a rain-flagging algorithm for the QuikScat Scatterometer. Over the past several years he has constructed and maintained a climate-quality data record of atmospheric temperatures from MSU and AMSU, and studied human-induced change in atmospheric water vapor and oceanic wind speed using measurements from passive microwave imagers. Dr. Mears was a convening lead author for the U.S. Climate Change Science Program Sythesis and Assessment product 1.1 (the first CCSP report to reach final form), and a contributing author to the IGPP 4th assessment report. He is a member two international working groups, the Global Climate Oberving System Working Group on Atmospheric Reference Observations, and the WCRP Stratospheric Trends Working Group, which is part of the Stratospheric Processes and their Role in Climate (SPARC) project.
B. D. Santer et al., “Amplification of surface temperature trends and variability in the tropical atmosphere,” Science, vol. 309, no. 5740, pp. 1511-1556, 2005.
C. A. Mears and F. J. Wentz, “Construction of the Remote Sensing Systems V3.2 atmospheric temperature records from the MSU and AMSU microwave sounders,” Journal of Atmospheric and Oceanic Technology, vol. 26, pp. 1040-1056, 2009.
J. R. Christy, R. W. Spencer, W. B. Norris, W. D. Braswell, and D. E. Parker, “Error estimates of version 5.0 of MSU-AMSU bulk atmospheric temperatures,” Journal of Atmospheric and Oceanic Technology, vol. 20, no. 5, pp. 613-629, 2003.
P. W. Thorne et al., “Revisiting radiosonde upper-air temperatures from 1958 to 2002,” Journal of Geophysical Research, vol. 110, 2005. (This is the HADAT dataset)
L. Haimberger, “Homogenization of radiosonde temperature time series using innovation statistics,” Journal of Climate, vol. 20, no. 7, pp. 1377-1403, 2007. (This describes the RAOBCORE dataset)
L. Haimberger, C. Tavolato, and S. Sperka, “Towards the elimination of warm bias in historic radiosonde records -- some new results from a comprehensive intercomparison of upper air data,” Journal of Climate, vol. 21, pp. 4587-4606, 2008. (This describes the RICH dataset)
S. C. Sherwood, C. L. Meyer, R. J. Allen, and H. A. Titcher, “Robust tropospheric warming revealed by iteratively homogenized radiosonde data,” Journal of Climate, vol. 21, no. 20, pp. 5336-5352, Oct. 2008. (This describes the IUK dataset)
R. H. Rienecker et al., A. da Silva, “MERRA - NASA’s Modern-Era Retrospective Analysis for Research and Applications,” Journal of Climate, 2011. (This describes the MERRA dataset)
Q. Fu and C. M. Johanson, “Satellite-derived vertical dependence of tropospheric temperature trends,” Geophysical Research Letters, vol. 32, 2005. (This introduces the concept of TTT)
Mears, CA, FJ Wentz, P Thorne and D. Bernie, 2011, “Assessing uncertainty in estimates of atmospheric temperature changes from MSU and AMSU using a Monte-Carlo estimation technique”, Journal of Geophysical Research, 116, D08112, doi:10.1029/2010JD014954. (This discussed RSS error ensembles).
Thorne, P. W., D. E. Parker, J. R. Christy, and C. A. Mears, 2005, “Uncertainties in Climate Trends: Lessons From Upper-Air Temperature Records”, Bulletin of the American Meteorological Society, 86, 1437-1442. (This discusses the idea of structural uncertainty)
Morice, C. P., J. J. Kennedy, N. A. Rayner, and P. D. Jones (2012),Quantifying uncertainties in global and regional temperature change using an ensemble of observational estimates: The HadCRUT4 data set, J. Geophys. Res., 117, D08101, doi:10.1029/2011JD017187. (This discussed HadCRUT4 and the HadCRUT4 error ensembles)
Santer, B. D., J. F. Painter, C. A. Mears, C. Doutriaux, P. Caldwell, J. M. Arblaster, P. J. Cameron-Smith, N. P. Gillett, P. J. Gleckler, J. Lanzante, J. Perlwitz, S. Solomon, P. A. Stott, K. E. Taylor, L. Terray, P. W. Thorne, M. F. Wehner, F. J. Wentz, T. M. L. Wigley, L. J. Wilcox, and C. Z. Zou, 11-29-2012: Identifying Human Influences on Atmospheric Temperature. Proceedings of the National Academy of Sciences, 110, 26-33, 10.1073/pnas.1210514109. (This describes, amoung other things, the construction of the 1979-2012 model datasets by combining 20th century simulations with RCP8.5 21st century predictions)
Guest blog Steven Sherwood
The tropical upper-tropospheric warming “hot spot”: is it missing, and what if it were?
Prof. Steven Sherwood, Director, Climate Change Research Centre, University of New South Wales, Sydney Australia.
In this post I’ll address two issues: first, confidence in tropical lapse-rate* changes and what they would mean for our understanding of atmospheric physics; second, the broader implications for global warming. My main positions on this issue could be summarised as: a) lapse-rate changes differing significantly from those expected from basic thermodynamic arguments would be very interesting, but, b) they would have no clear implications for global warming, and c) evidence that they have occurred is not reliable (which in a way is too bad, because of (a)). A side point is that there are other model-observation discrepancies that I think are more worthy of attention (and are accordingly receiving more attention from the mainstream scientific community).
Confidence and Implications for Atmospheric Physics
I first became interested in the tropical lapse rate (now alternatively known as “hot spot”) issue around 2001, shortly after it was raised in a prominent paper in Science (Gaffen et al. 2000). I had been using radiosonde data to look at wind fields and temperature trends near the tropical tropopause and lower stratosphere. This new problem drew my attention because I was interested in how tropical atmospheric convection (e.g. storms) responds to its environment, one of the grand unsolved problems in atmospheric modelling. Tropical convection was supposed to prevent the kind of lapse-rate changes that were being reported, so what was going on?
I considered various ways that aerosols might alter the convective lapse rate and how to test these hypotheses. Before going too far with this, however, I wanted to assure myself that the reported trends were robust, and began my own analysis of the radiosonde data (or in fact continued it, since I was already using radiosondes to understand change in the lower stratosphere). By 2005, based on my own work and others’ plus a better understanding of the basic challenges, I no longer thought there was credible evidence for any unexpected changes in atmospheric temperature structure. Consequently I dropped this as a research topic (my student who was thinking along these lines, Bob Allen, changed gears to examine the possible impacts of aerosol on the general circulation which did lead to some very interesting results published later; he also showed that wind trends in the tropics were consistent with the hot spot).
Although there has been more to-ing and fro-ing in the literature since then, as described in the opening article for this exchange, I still remain unconvinced that we can observe the small changes in temperature structure that are being discussed. Tests of radiosonde homogenisation methods (e.g., Thorne et al. 2011) show that they are often unreliable. MSU is not well calibrated and its homogenisation issues are also serious, as shown by the range of results previously obtained from this instrument series. To obtain upper-tropospheric trends from Channel 2 of MSU requires subtracting out a large contribution to trends in this channel coming from lower-stratospheric cooling. The latter remains highly uncertain due to a discrepancy between cooling rates in radiosondes and MSU. Tropical ozone trends are sufficiently uncertain so as to render either of these physically plausible (Solomon et al. 2012). I used to think (as do most others) that the radiosondes were wrong, but in Sherwood et al. 2008 we found (to my surprise) that when we homogenised the global radiosonde data they began to show cooling in the lower stratosphere that was very similar to that of MSU Channel 4 at each latitude, except for a large offset that varied smoothly with latitude. Such a smoothly varying and relatively uniform offset is very different from what we’d expect from radiosonde trend biases (which tend to vary at lot from one station to the next) but is consistent with an uncorrected calibration error in MSU Channel 4. If that were indeed responsible, it would imply that there has been more cooling in the stratosphere than anyone has reckoned on, and that the true upper-tropospheric warming is therefore stronger than what any group now infers from MSU data. By the way, our tropospheric data also came out very close to those published at the time by RSS, both in global mean and in the latitudinal variation (Sherwood et al., 2008).
Changes in tropical lapse rate remain an interesting problem in principle, because we know that convective schemes in global atmospheric models need improving, and this could be informative as to what is wrong. Current schemes enforce the theoretical “moist-adiabatic” lapse rate quite strongly. It would not surprise me much if it turned out that they are too heavy-handed in this respect, and that a better model would anchor the upper tropospheric temperature less firmly to the surface temperature. Indeed there is reason to believe that other problems with these models, such as difficulties in generating proper hurricanes or a tropical phenomenon known as the Madden-Julian oscillation, may also derive from the schemes triggering convection too easily and enforcing these lapse rates too vigorously. So I would not at all discount the possibility of these lapse-rate changes occurring, but one needs strong evidence, and we just don’t have that.
Broader implications for global warming
Perhaps the most remarkable and puzzling thing about the “hot spot” question is the tenacity with which climate contrarians have promoted it as evidence against climate models, and against global warming in general.
If I were looking for climate model defects, there are far more interesting and more damning ones around. For example, no climate model run for the IPCC AR4 (c. 2006) was able to reproduce the losses of Arctic sea ice that had been observed in recent decades (and which have continued accelerating since). No model, to my knowledge, produces the large asymmetry in warming between the north and south poles observed since 1980. Models underpredict the observed poleward shifts of the atmospheric circulation and climate zones by about a factor of three over this same period (Allen et al. 2012); cannot explain the warmings at high latitudes indicated by paleaoclimate data in past warm climates such as the Pliocene (Fedorov et al. 2013); appear to underpredict observed trends in the hydrological cycle (Wentz et al. 2007, Min et al. 2011) and in their simulated climatologies tend to produce rain that is too frequent, too light, and on land falls at the wrong time of day (Stephens et al. 2010). Finally, the tropical oceans are not warming as much as the land areas, or as much as predicted by most models, and this may be the root cause of why the recent warming of the tropical atmosphere is slower than predicted by most models (there is a nice series of posts about this on Isaac Held’s blog). What makes the “hot spot” more important than these other discrepancies which, in many cases, are supported by more convincing evidence? Is it because the “missing hot spot” can be spun into a tale of model exaggeration, whereas all the other problems suggest the opposite problem?
Let us suppose for the moment that the “hot spot” really has been missing while the surface has warmed. What would the implications be?
The implications for attribution of observed global warming are nil, as far as I can see. The regulation of lapse rate changes by atmospheric convection is expected to work exactly the same way whether global temperature changes are natural or forced (say, by greenhouse gases from fossil fuel burning).
The implications for climate sensitivity are also roughly nil. The total feedback from water vapour and lapse-rate changes depends only on the changes in relative humidity in the upper troposphere, not on the lapse rate itself (see Ingram, 2013). In fact, in climate models where the lapse rate becomes relatively steeper as climate warms (as would be the case with a missing hot spot), the total warming feedback is very slightly stronger because the increased lapse rate increases the greenhouse effect of carbon dioxide and other well-mixed greenhouse gases. So a missing hot spot would not mean less surface warming, at least according to our current understanding.
Moreover, the discrepancy with models was opposite from 1958-1979 (Gaffen et al. 2000)—that is to say, the observed tropical upper-tropospheric warming was evidently stronger than expected. But the world was warming then too. So if this interesting phenomenon is real, it probably is not connected to global warming.
Fig. 1. Weaker upper-tropospheric warming and hence weaker water-vapour feedback actually implies, on average, slightly stronger overall positive feedback due to lapse rate and water vapour combined (from Ingram 2013).
Anyone who wants to argue that the “missing hot spot” implies something as to the future (say, that global warming will be less than current models predict) needs to come up with an alternative model of climate that agrees just as well with observations, obeys physical laws, predicts the absence of a “hot spot,” and predicts less future global warming (or whatever other novel outcome). This is how science advances---through the consideration of multiple hypotheses. If a new one comes along that fits the observations, I’ll gladly consider it.
Currently none of the explanations I can see for the “missing hot spot” would change our estimate of future warming from human activities, except one: that the overall warming of the tropics is simply slower than expected. It does seem that global-mean surface warming is starting to fall behind predictions, and this is particularly so in the tropical oceans (though not, curiously, on land). Possible causes are (a) aerosols, solar or other forcings have recently exerted a stronger (temporary) cooling influence than we think; (b) negative feedbacks from clouds have kicked in; or (c) the oceans are burying the heat faster than we expected. If (b) were true, we would revise our estimates of climate sensitivity downward. There are observations supporting options (c) and to a small extent (a), but there is plenty of room for new surprises. If it is (c) (which appears most likely), we then have to decide whether this is a natural variation or if it is a feature of global warming. In the former case the heat will soon come back; in the latter, the oceans will delay climate change more effectively than we thought. Another decade or so of observations should reveal the answer.
*for readers unfamiliar with the term “lapse rate,” it is the rate at which air temperature decreases with altitude.
Dr. Steven Sherwood is professor at the Climate Change Research Centre of the The University of New South Wales in Sydney. He received his M.S. degree (1989) in Eng. Physics/Fluid Mechanics at the University of California San Diego, USA and his Ph.D. degree (1995) in Oceanography at the Scripps Institution of Oceanography.
Sherwood studies how the various processes in the atmosphere conspire to establish climate, how these processes might be expected to control the way climate changes, and how the atmosphere will ultimately interact with the oceans and other components of Earth. Clouds and water vapour in particular remain poorly understood in many respects, but are very important not only in bringing rain locally, but also to global climate through their effect on the net energy absorbed and emitted by the planet. Tropospheric convection (disturbed weather) is a key process by which the atmosphere transports water and energy and in the process creates clouds, but it is also a turbulent phenomenon for which we have no basic theory and which observations cannot yet fully characterise.
Sherwood leads a research group that applies basic physics to complex problems by a combination of simple theoretical ideas and hypotheses and directed analyses of observations.
Allen, R. J., S. C. Sherwood, J. R. Norris and C. Zender, Recent Northern Hemisphere tropical expansion primarily driven by black carbon and tropospheric ozone, Nature, Vol. 485, 2012, 350-355.
Fedorov, A. V., C. M. Brierley, K. T. Lawrence, Z. Liu, P. S. Dekens and A. C. Ravelo (2013). "Patterns and mechanisms of early Pliocene warmth." Nature 496(7443): 43-+.
Gaffen, D. J., B. D. Santer, J. S. Boyle, J. R. Christy, N. E. Graham and R. J. Ross, Multidecadal changes in the vertical temperature structure of the tropical troposphere, Science, 2000, V. 287, 1242-1245.
Ingram, W. (2013). "Some implications of a new approach to the water vapour feedback." Climate Dynamics 40: 925-933.
Min, S. K., X. B. Zhang, F. W. Zwiers and G. C. Hegerl (2011). "Human contribution to more-intense precipitation extremes." Nature 470(7334): 378-381.
Sherwood, S. C., C. L. Meyer, R. J. Allen, and H. A. Titchner, Robust tropospheric warming revealed by iteratively homogenized radiosonde data. Journal of Climate, Vol. 21, 2008, 5336-5352.
Solomon, S., P. J. Young and B. Hassler (2012). "Uncertainties in the evolution of stratospheric ozone and implications for recent temperature changes in the tropical lower stratosphere." Geophysical Research Letters 39.
Stephens, G. L., T. L'Ecuyer, R. Forbes, A. Gettlemen, J.-C. Golaz, A. Bodas-Salcedo, K. Suzuki, P. Gabriel and J. Haynes (2010). "Dreary state of precipitation in global models." Journal of Geophysical Research 115: D24211.
Thorne, P. W. et al., A quantification of uncertainties in historical tropical tropospheric temperature trends from radiosondes, J. Geophys. Res., Vol. 116, 2011, D12116.
Wentz, F. J., L. Ricciardulli, K. Hilburn and C. Mears, 2007, How much more rain will global warming bring?, Science, Vol. 317, 233-235.
Guest blog John Christy
Why should we care about the tropical temperature?
John R. Christy, Distinguished Professor, Department of Atmospheric Science, Director Earth System Science Center, The University of Alabama in Huntsville
One important part of climate change research is to document the amount of change that can already be attributed to human activity. In other words we want to know the answer to the question, “How has the climate changed specifically because of the enhancement of the natural greenhouse effect caused by extra emissions due to human progress?” These rising emissions come primarily from energy production using carbon-based fuels which emit, as a by-product, the ubiquitous and life-sustaining greenhouse gas, carbon dioxide (CO2). From about 280 ppm in the 19th century, the current concentration of CO2 has risen to about 400 ppm.
So, what has the extra CO2 and other greenhouse gases done to the climate as of today? Climate model simulations indicate that a prominent and robust response to extra greenhouse gases is the warming of the tropical troposphere, a layer of air from the surface to about 16 km altitude in the region of the globe from 20°S to 20°N. A particularly obvious feature of this expected warming, and is a key focus of this blog post, is that this warming increases with altitude where the rate of warming at 10 km altitude is over twice that of the rate at the surface. This clear model response should be detectible by now (i.e. 2012) which gives us an opportunity to check whether the real world is responding as the models’ simulate for a large-scale, easy-to-compare quantity. This is why we care about the tropical atmospheric temperature.
There are two aspects to this tropical warming that are sometimes confused. One aspect is the simple magnitude of the warming rate, or temperature trend, of the entire troposphere. This metric quantifies the amount of heat that is accumulating in the bulk atmosphere. A well-established result of adding greenhouse gases to the atmosphere is that heat energy (in units of joules) will accumulate in the troposphere which can be detected as a rise in temperature. [The fundamental issue of the effects of greenhouse warming is: how many joules of heat are accumulating in the climate system per year?]
We don’t know at what rate that accumulation might occur as other processes may come into play which reduce or magnify it. For example, with extra greenhouse gases, the rate at which the joules are allowed to escape to space may be reduced by additional responses, causing even more heating. On the other hand, there could be an increase in cloudiness which may limit the number of joules (from the sun) which enter the climate system, thus causing a cooling influence. A reaction of the climate system to extra CO2 that promotes even more accumulation over what would have happened due to CO2 alone is a positive feedback, while one that limits the accumulation of joules is a negative feedback. In the climate system, there are numerous feedbacks of both signs, all interdependent and intertwined.
The second aspect of enhanced temperature change is the amount of amplification the higher altitude layers will experience relative to the surface warming as noted earlier – which is linked to the first aspect and is a feature discussed as a complement to the first aspect. In simple thinking, if enough joules are added to the troposphere to increase its temperature by 1 °C throughout, one would expect a uniform 1 °C warming from the surface to the top of the troposphere. However, as seen in the way the real atmosphere behaves on monthly and yearly time scales, the surface temperature change tends to be less than 1 °C while the upper troposphere warms to more than 1 °C. Since there is a reduction in the expected increase of the surface temperature given the number of joules added, this phenomenon is called a negative lapse-rate feedback to surface temperature (even though the upper air heats up more.) So the models anticipate that there will be a strong amplification of the surface temperature change as one ascends through the troposphere. [So, if someone claims that surface and upper air trends agree in magnitude, then they are also claiming that this is not consistent with the enhanced greenhouse effect since, according to models, the temperature trend of those two levels should not agree.]
Thus, there are two ideas to test in the tropics, (1) the overall magnitude of the layer-average temperature rise and (2) the magnification or amplification of the surface temperature change with height.
Balloons and satellites
Measurements of tropical tropospheric temperature have been performed by balloons that ascend through the air and radio back the atmosphere’s vital statistics, like temperature, humidity, etc. Due to a number of changes in these instruments through the years research organizations have spent a lot of effort to remove such problems and create homogenous or consistent databases of these readings. For this study we shall assume that the average of four major and well-published datasets (known as RATPAC, RAOBCORE, RICH and HadAT2) will serve as the “best guess” of the tropical temperatures at the various elevations (see Christy and Hnilo, 2007, Christy et al. 2010, 2011 for descriptions and earlier results.)
For a layer-average of the tropospheric temperature there are two satellite-based tropospheric datasets (known as UAH and RSS) which have by independent methods combined the readings from several spacecraft carrying microwave instruments into a time series beginning in late 1978. There are dozens of publications which detail the methods used by the various groups to generate both balloon and satellite products. Through the years each group has updated their products as new information has come to light, and we use the latest versions as of June 2013.
The time frame we shall consider here will begin in Jan 1979 and end in Dec 2012 as this is the time we have output from models and from observations, both balloons and satellites. It is also the period for which the greatest amount of accumulation of heat energy (joules) should be evident due to the increasing impact of the rising greenhouse gas concentrations.
To examine the simple magnitude of full-tropospheric trends we look at two layers as represented by what satellites measure which are roughly the average temperature of the surface and to about 10 km (lower troposphere or TLT) and surface to about 17 km (mid-troposphere or TMT). TMT gives more weight to the region between 500 (5.5 km) and 200 hPa (12 km) where the warming is expected to be most pronounced according to models, so the figures will focus on TMT. We can simulate the satellite layer using both balloon data and model output for direct, apples-to-apples comparisons (Fig. 1.)
Figure 1. Time series of the mid-tropospheric temperature (TMT) of 73 CMIP-5 climate models (rcp8.5) compared with observations (circles are averages of the four balloon datasets and squares are averages of the two satellite datasets.) Values are running 5-year averages for all quantities. [There are four basic rcp emission scenarios applied to CMIP-5 models, but their divergence occurs after 2030. Thus, for our comparison which ends in 2012, there are essentially no differences among the rcp scenarios.] The model output for all figures was made available by the KNMI Climate Explorer.
We see that all 73 models anticipated greater warming than actually occurred for the period 1979-2012. Of importance here too is that the balloons and satellites represent two independent observing systems but they display extremely consistent results. This provides a relatively high level of confidence that the observations as depicted here have small errors. The observational trends from both systems are slightly less than +0.06 °C/decade which is a value insignificantly different from zero. The mean TMT model trend is +0.26 °C/decade which is significantly positive in a statistical sense. The observed satellite and balloon TLT trends (not shown) are +0.10 and +0.09 °C/decade respectively, and the mean model TLT trend is +0.28 °C/decade. In a strict hypothesis test, the mean model trend can be shown to be statistically different from that of the observations, so that one can say the model-mean has been falsified (a result stated in a number of publications already for earlier sets of model output.) In other words, the model mean tropical tropospheric temperature trend is warming significantly faster than observations (See Douglass and Christy 2013 for further information.)
Regarding the second aspect of temperature change, we show the vertical structure of those changes in Fig. 2 where we display the temperature trend by vertical height (pressure) as indicated by the four balloon datasets (circles), their average (large circle) and 73 model simulations (lines of various types).
Figure 2 Temperature trends in °C/decade by pressure level with 1000 hPa being the surface and 100 hPa being around 16 km. Circles represent the four observational balloon datasets, the largest circle being their mean. The lines represent 73 CMIP-5 model simulations (identities in Fig. 3) with the non-continuous lines representing models sponsored by the U.S. The large black dashed line is the 73-model mean. The pressure values are very close to linear with respect to mass but logarithmic with respect to altitude, so that 500 hPa is near 5.5 km altitude, 300 hPa near 9 km altitude and 200 hPa about 12 km altitude.
Figure 3 Caption for Fig. 2, identifying model runs and observational datasets.
The models (especially) show increasing trends as altitude increases to 250 hPa (about 10 km) before decreasing toward the stratosphere (~90 hPa). In comparing model simulations with the observations it is clear that between 850 and 200 hPa, all model results are warmer than the average of the balloon observations, a result not unexpected given the information in Fig. 1.
The amount of the amplification of the value of the surface trend with elevation in Fig. 2 is somewhat difficult to discern as each model has its own surface trend magnitude. To better compare the amplification effect, we normalize the pressure-level trend values by the trend of the surface value for each dataset and model simulation.
Figure 4. Value of the 1979-2012 temperature trend at various upper levels divided by the magnitude of the respective surface trend, i.e. the ratio of upper air trends to surface trends. Model simulations are lines with the average of the models as the dotted line. Squares are individual balloon observations (green – RATPAC, gray RAOBCORE, purple – RICH and orange – HadAT2) with the averages of observations the gray circles.
Figure 4 displays the ratio, or amplification factor, that observations and models depict for 1979-2012 in the tropics (see Christy et al. 2010 for further information). The mean observational result indicates the values are between +0.5 and +1.5 through the lower and middle troposphere (850 to 250 hPa). [The observational results tend to have greater variability due to the denominator (surface trend) being relatively small. Viewing Fig. 2 shows that the observations are rather tightly bunched for absolute trends in comparison to the model spread.] The models indicate a systematic increase in the ratio from 1.0 at the surface with amplification factors well above +1.5 from 500 to 200 hPa. What this figure clearly indicates is that the second aspect of this discussion, i.e. namely the rising temperatures with increasing altitude, is also over-done in the climate models. The differences of the means between observations and models are significant.
While there is much that can be discussed from these results, we wonder simply why the models overwarm the troposphere compared with observations by such large amounts (on average) during a period when we have the best understanding of the processes that cause the temperature to change. During a period when the mid-troposphere warmed by +0.06 °C/decade, why does the model average simulate a warming of +0.26 °C/decade?
Unfortunately, a complete or even satisfactory answer cannot be provided. Each model is constrained by its own sets of equations and assumptions that prevent simple answers, especially when all of the individual processes are tangled together through their unique complex of interactions. The real world also presents some baffling characteristics since it is constrained by the laws of physics which are not fully and accurately known for this wickedly complex system.
An interesting feature of the models is that almost all show greater year-to-year variability than observations (Fig. 1.) The average model annual variance (detrended) of anomalies is 60 percent greater than that of the observational datasets. This is a clue that suggests the models’ atmospheres are more sensitive to forcing than is the real climate system, so that an increase in greenhouse forcing in models will lead to a greater temperature response than experienced by the actual climate system. But saying the climate models are too sensitive only identifies another symptom of the issue, not the cause.
We want to know why the extra joules of energy that increasing CO2 concentrations should be trapping in the climate system are not found in Nature’s atmosphere compared with what the models simulate.
Could the extra joules be absorbed by the deep ocean and prevented from warming the atmosphere (Guemas et al. 2013)? This requires extremely accurate measurements of the deep ocean (better than 0.01 °C precision) which are not now available comprehensively in space and time. Current studies based only on observations suggest this enhanced sequestration of heat is not happening.
Could there be a separate process like enhanced solar reflection by aerosols that is keeping the number of joules available for absorption at a smaller level relative to the past? The interaction of aerosols with the entire array of climate processes is another fundamental area of research that has more questions than answers. How do aerosols affect cloudiness (more?, less?, brighter?, darker?). What is the precise, time-varying distribution of all types of aerosols and what exactly does each type do in terms of affecting the absorption and reflection of the joules in all frequencies? The IPCC typically shows very large error ranges for our knowledge of the aerosol effects, so there is a possibility that models have significant and consistent errors in dealing with them (IPCC 2007 AR4 Fig SPM.2).
Clouds and water vapor
Could there be a complex feedback response in the way the real atmosphere handles water vapor and clouds that acts to enhance the expulsion of joules to space under extra greenhouse forcing so they don’t accumulate very rapidly? Of the many processes that models struggle to represent, none are more difficult than clouds and water vapor. As recently shown by Stevens and Bony (2013) different models driven by an identical, simplified forcing produced very different results for cloudiness. This is my favorite option in terms of explaining the lack of joule-accumulation. As my colleague Roy Spencer reminds us, if you think about it, the atmosphere should have 100 percent humidity because it has an essentially infinite source of water in the oceans. However, precipitation prevents that from happening, so precipitation processes are apparently in control of water vapor concentrations – the greenhouse gas with the largest impact on temperature. This means the way precipitation and clouds behave (both in causing changes or responding to them) when slight changes occur in the environment is key in my view. We have actually measured large temperature swings that were preceded by changes in cloudiness in our global temperature measurements. So a response to the extra CO2 forcing by clouds and water vapor, which have a massive impact on temperature, could be the reason for the rather modest temperature rise we’ve experienced (Spencer and Braswell, 2010).
Or, could there be natural variations that completely overcome small enhancements in greenhouse-joule-trapping? These variations have demonstrated the ability to drive large temperature swings in the past, but we cannot simulate or predict them well at all. For that we need extremely accurate ocean simulations along with accurate representations of clouds, precipitation and water vapor (among other things).
The bottom line is that, while I have some ideas based on some evidence, I don’t know why models are so aggressive at warming the atmosphere over the last 34 years relative to the real world. The complete answer is probably different for each model. To answer that question would take a tremendous model evaluation program run by independent organizations that has yet to be formulated and funded.
What I can say from the standpoint of applying the scientific method to a robust response-feature of models, is that the average model result is inconsistent with the observed rate of change of tropical tropospheric temperature - inconsistent both in absolute magnitude and in vertical structure (Douglass and Christy 2013.) This indicates our ignorance of the climate system is still enormous and, as suggested by Stevens and Bony, this performance by the models indicates we need to go back to the basics. From this statement there is only a short distance to the next - the use of climate models in policy decisions is, in my view, not to be recommended at this time.
J.R. Christy is Distinguished Professor of Atmospheric Science at the University of Alabama in Huntsville and Director of the Earth System Science Center. He is Alabama’s State Climatologist. In 1989 he and Dr. Roy Spencer, then of NASA, published the first global, bulk-atmospheric temperatures from microwave satellite sensors. For this achievement they were recognized with NASA’s Medal for Exceptional Scientific Achievement and the American Meteorology Society’s Special Award for developing climate datasets from satellites. Christy has served on the IPCC panels as Contributor, Key Contributor and Lead Author and has testified before the U.S. Congress, federal court, many state legislatures and regulatory boards on climate issues.
Christy, J. R., W. B. Norris, R. W. Spencer, and J. J. Hnilo. Tropospheric temperature change since 1979 from tropical radiosonde and satellite measurements, J. Geophys. Res., 2007. 112, D06102, doi:10.1029/2005JD006881.
Christy, J.R., B. Herman, R. Pielke, Sr., P. Klotzbach, R.T. McNider, J.J. Hnilo, R.W. Spencer, T. Chase and D. Douglass, (2010): What do observational datasets say about modeled tropospheric temperature trends since 1979? Remote Sens. 2, 2138-2169.
Christy, J.R., R.W. Spencer and W.B. Norris, 2011: The role of remote sensing in monitoring global bulk atmospheric temperatures. Int. J. Remote Sens., 32, 671-685, DOI:10.1080/01431161.2010.517803.
Douglass, D. and J.R. Christy, 2013: Reconciling observations of global temperature change: 2013. Energy and Env., 24 No. 3-4, 414-419.
Guemas, V., F.J. Doblas-Reyes, I. Andrea-Burillo and M. Asif, 2012: Retrospective prediction of global warming slowdown in the past decade. Nature Clim. Ch., 3, 649-653, DOI:10.1038/nclimate1863.
Spencer, R.W. and W.D. Braswell, 2010: On the diagnosis of radiative feedback in the presence of unknown radiative forcing. J. Geophys. Res., 115, DOI:10.1019/2009JD013371.
Stevens, B. and S. Bony, 2013. What Are Climate Models Missing? Science. 31 May 2013. Doi:10.1126/science/1237554.