Recently, a number of articles in the mainstream press have pointed out that there appears to have been little or no change in globally averaged temperature over the last two decades. Because of this, we are getting a lot of questions along the lines of “I saw this plot on a denialist web site. Is this really your data?” While some of these reports have “cherry-picked” their end points to make their evidence seem even stronger, there is not much doubt that the rate of warming since the late 1990’s is less than that predicted by most of the IPCC AR5 simulations of historical climate. This can be seen in the RSS data, as well as most other temperature datasets. For example, the figure below is a plot of the temperature anomaly (departure from normal) of the lower troposphere over the past 35 years from the RSS “Temperature Lower Troposphere” (TLT) dataset. For this plot we have averaged over almost the entire globe, from 80S to 80N, and used the entire TLT dataset, starting from 1979. (The denialists really like to fit trends starting in 1997, so that the huge 1997-98 ENSO event is at the start of their time series, resulting in a linear fit with the smallest possible slope.)
In this figure, the thick black line is from a climate data record derived from microwave sounding satellite (MSU and AMSU ) measurements. Each of the thin light blue lines represents the temperature anomaly time series for the same atmospheric layer from one of 33 IPCC climate model simulations that I have analyzed. I have adjusted each individual time series so that its average is 0.0 for the 1979-1988 period. This has no effect on the trend of each line, but it does make it easier to see long term changes in the plot.
The dips in the simulated model temperatures in 1983 and late 1991 are due to the eruptions of El Chichón and Mt. Pinatubo. These eruptions spewed enough volcanic ash into the stratosphere to block part of the incoming sunlight and cool Earth’s surface and troposphere. The cooling can easily be seen in the measured satellite data in 1992-1993. The cooling event in 1983 happened by chance at more or less the same time as an El Niño event in 1983-84, making it harder to see. (Note that the same events warmed the stratosphere. See our TLS dataset.) The year-to-year variability of the measured data is dominated by El Niño/La Niña events, with an overall warming trend (0.123K/decade) on longer timescales.
The plot shows that the measured temperature rise is within the envelope of model predictions up until the late 2000’s. After that time, observed temperatures are sometimes less than any model prediction, and are clearly different than the mainstream model behavior. This slow-down in the warming, often called the “warming hiatus”, has become a major research topic over the last several years, and a source of much controversy across the blogosphere. In this post, I offer my view on the cause of the hiatus. Some of the following discussion is distilled from a moderated debate I took part in under the auspices of the Climate Dialogue website
Does this slow-down in the warming mean that the idea of anthropogenic global warming is no longer valid? The short answer is ‘no’. The denialists like to assume that the cause for the model/observation discrepancy is some kind of problem with the fundamental model physics, and they pooh-pooh any other sort of explanation. This leads them to conclude, very likely erroneously, that the long-term sensitivity of the climate is much less than is currently thought.
The truth is that there are lots of causes besides errors in the fundamental model physics that could lead to the model/observation discrepancy. I summarize a number of these possible causes below. Without convincing evidence of model physics flaws (and I haven’t seen any), I would say that the possible causes described below need to be investigated and ruled out before we can pin the blame on fundamental modelling errors.
Also, a philosophical comment -- often, we are predisposed to the position that a given effect is due to a single cause. Part of the reason for this is probably human nature. We like to distill complex things into simple stories or parables. The other part is that for most of the science courses we take in school, simple experiments are presented that demonstrate the fundamental ideas in the topic under study. Single causes are often the case in laboratory experiments -- these experiments are usually designed to isolate a single causative effect. In “real-world” science, such as the study of Earth’s climate, things are very unlikely to be as clear cut. Instead, each observed “effect” will be due to the combination of numerous causes. My point is that I do not expect the disagreement between models and observations over the past 15 years to be due to a single cause. It is much more likely to be due to some combination of the possible causes listed below.
The possible causes for the model/observation discrepancies can be grouped into several categories:
Errors in Model “Forcing”
Internal Variability (Random Fluctuations) in the Climate System
Errors in Fundamental Model Physics
The first 3 causes have no effect on the long-term sensitivity of the climate to increased CO2 and only some of the fundamental model physics errors (4th cause) would change the long-term sensitivity. Some of the causes described below might delay the warming, but we would end up at the same point in the future, or we might see changes in the geographical patterns of the warming signal, but the overall change would be the same. In this post, I’ll address the first three possibilities in more detail. I am not an expert in modeling, so I will leave the discussion of possible model errors to others with more expertise.
As a data scientist, I am among the first to acknowledge that all climate datasets likely contain some errors. However, I have a hard time believing that both the satellite and the surface temperature datasets have errors large enough to account for the model/observation differences. For example, the global trend uncertainty (2-sigma) for the global TLT trend is around 0.03 K/decade (Mears et al. 2011). Even if 0.03 K/decade were added to the best-estimate trend value of 0.123 K/decade, it would still be at the extreme low end of the model trends. A similar, but stronger case can be made using surface temperature datasets, which I consider to be more reliable than satellite datasets (they certainly agree with each other better than the various satellite datasets do!). So I don’t think the problem can be explained fully by measurement errors.
Error in Model Forcings:
Model forcings are changes external to the climate system that can change the state of the climate. These include things like the output of the sun (clearly external), the rise in the concentration of CO2 (CO2 is located within the climate system, but is not directly affected by changes in climate), changes in aerosols caused by volcanoes and/or industrial pollution, and changes in other trace gases such as methane or ozone. For climate simulations, these forcing variables serve as inputs to the program. Any errors in the forcings data input to the model can lead to errors in the output. A simple case of garbage in -- garbage out. For the plots in the figure above, the forcings come from (perhaps imperfect) measurements for the period up to 2005. After 2005, the models used predicted forcing values derived from estimated future emissions (Representative Concentration Pathways, or RCPs, in IPCC jargon). Many of these forcings may indeed contain errors, not only in the predicted values, but in some cases, even the pre-2005 measured values. The one forcing that is not in doubt is the concentration of CO2. Carbon Dioxide has continued to rise as predicted.
Ozone: Temperature changes in the upper troposphere and lower stratosphere have been shown to be very sensitive to the stratospheric ozone concentrations used in the models (Solomon et al. 2012). These effects appear to extend below the tropical tropopause, low enough to affect tropospheric temperature trends. The ozone dataset used in the IPCC AR5 model simulations is that with the most conservative ozone trend. If one of the other datasets had been used, the models would have shown less upper tropospheric warming.
Volcanic Aerosols: Volcanic events, such as the El Chichón and Pinatubo eruptions discussed above (and other smaller events before 2000), are well represented in the stratospheric aerosols datasets used to force the 20th century simulations for IPCC AR5. After 2000, the level of stratospheric aerosols in the input datasets was allowed to decay to zero, increasing the modeled surface temperature. In real life, however, observations indicate that the background level of stratospheric aerosols has increased over the 2000-2010 period (Solomon et al. 2011), probably due to a large number of small volcanic eruptions (Neely et al. 2013). Santer et al.(2014) showed that these aerosols can lead to reduced warming. This is almost certainly a real effect, though the exact amount of temperature reduction caused by the volcanoes remains uncertain without further modeling and observational studies. Volcanic aerosols, like the other possible causes listed here, are unlikely to explain the discrepancy by themselves.
Anthropogenic Aerosols: Like sulfate aerosols in the stratosphere, tropospheric sulfate aerosols, which are mostly caused by human sources such as unfiltered fossil fuel pollution, cool the surface. Over the last two centuries, tropospheric sulfate aerosols have increased and decreased in a complicated way, as economic booms came and went, preferred fuel sources changed, and as environmental regulations such as the Clean Air Act went into effect. Sulfate aerosols caused by human activities increased substantially in the post 1998 period. Much of this increase is due to increased coal burning in China, which more than doubled during the 2003-2008 period. Kaufmann et al (2011), used a statistical model to show that this increase cancels much of the effect of increasing CO2 during the period from 2000-2008. Some of the IPCC AR5 climate models internally generate anthropogenic sulfate aerosol loading, while others use the output of separate emission models as input. In both cases, the huge economic growth and subsequent coal burning in China were not part of sulfate aerosol predictions and thus the cooling effect of these aerosols would be underestimated in the final climate model projections. Schmidt et al. (2014) estimated that the difference accounts for slightly less than ¼ of the model/observation discrepancy.
Solar Irradiation: The total output of the sun varies on an approximately 11 year cycle. During the latest cycle, the solar output stayed low longer than usual (2007-2010), and then now appears to be peaking at a lower level than that of previous several cycles. The IPCC AR5 models assumed that this current solar cycle would be more similar to previous ones, and thus they overestimated the solar irradiance input from about 2008 to 2013. This is certainly a factor, but by itself probably explains only about 10% of the model/observation discrepancy (Schmidt et al. 2014).
Stratospheric Water Vapor: The amount of water vapor in the stratosphere has also decreased since 2000, which cancels the effects of increased CO2 (Solomon et al 2010), and reduces the warming signal. It is unclear whether this reduction in vapor constitutes a forcing term, or whether the reduced stratospheric water vapor is a poorly understood negative feedback term. Injection of water vapor into the stratosphere occurs only in the most intense deep convection events, which typically occur only in regions with the highest sea surface temperatures, the tropics. The distribution of the regions of high temperature can be changed by other climate factors, and thus by anthropogenic forcing.
Part of the cause of the hiatus could simply be due to bad luck, that is, the last 15 years could have been cooler than normal simply because of random fluctuations in the climate system. There are many modes of variability in the climate system on time scales of a year, decades, and even longer. Even when such modes are well represented in a climate model, they most likely occur at different times than the real-world events, leading to differences between the modeled and observed time series.
The longer, multi-year time scales tend to be associated with modes of variability that include changes in ocean currents and temperature. Excess heat can go into the ocean system, sink below the surface and not reappear for years or decades. The huge mass of the ocean gives it a huge capacity to store heat, and also means that currents respond slowly to changes in wind.
Many recently published papers describe an increase in heat subduction into the deep ocean. The idea is this -- increased trade winds cause increased turbulence in the upper ocean. In the western Pacific, this has the effect of mixing warm surface waters into the deep ocean at rates greater than normal, removing heat from the climate system. At the same time, the increased trade winds lead to more large-scale upwelling in the eastern tropical Pacific. This results in sort of a permanent La Niña state, with more cold water on the surface near the equator. This cold water sucks heat out of the atmosphere, leading to cooler temperatures worldwide than would be observed otherwise.
Since about 2000, the Pacific Ocean has been in the negative phase of the Interdecadal Pacific Oscillation (IPO). The negative phase is characterized by enhanced trade winds and cooler temperatures in both the eastern tropical Pacific and along the eastern continental margins and warmer temperatures in the western extratropical Pacific. The Pacific Ocean was in this negative IPO phase from about 1945 to about 1975, a period that also exhibited little warming on the global scale. This idea is nicely presented in the paper by England et al. (2014), and the accompanying news article by Kosaka (2014). Other papers that address this issue include (Balmaseda et al. 2013),(Kosaka and Xie 2013),(Meehl et al. 2013). The change in wind speed in the tropical Pacific can easily be seen in the RSS wind speed data measured using satellite microwave imagers. In the figure below, I show a map of wind speed trends from 1998-2012. A large trend in the central tropical Pacific is apparent. A similar trend is also present in reanalysis output, and is manifested by an increase in sea surface height in the western tropical Pacific.
My view is that the subduction of heat into the ocean is very likely a significant part of the explanation for the model/observation discrepancies. What is less clear is whether or not this subduction is due to random fluctuations in the climate, or some sort of response to anthropogenic forcing. An important question is now ‘how long will the enhanced trade winds continue?’. The trade wind anomaly lessened during 2013, but we do not know whether this change will persist over the next few years and lead a positive phase of the IPO, or if the IPO will take longer to flip to its other phase.
I’ll conclude by reiterating that I do not expect that the hiatus and model/observation discrepancies are due to a single cause. It is far more likely that they are caused by a combination of factors. Publications, blog posts and media stories that try to pin all the blame on one factor should be viewed with some level of suspicion, whether they are written by climate scientists, journalists, or climate change denialists.
Balmaseda, M. A., K. E. Trenberth, and E. Källén, 2013: Distinctive climate signals in reanalysis of global ocean heat content. Geophysical Research Letters, 40, 1754–1759.
England, M. H., and coauthors, 2014: Recent intensification of wind-driven circulation in the Pacific and the ongoing warming hiatus. Nature Climate Change, 4, 222–227.
Kaufmann, R. K., H. Kauppi, M. L. Mann, and J. H. Stock, 2011: Reconciling anthropogenic climate change with observed temperature 1998–2008. Proceedings of the National Academy of Sciences, 108, 11790–11793.
Kosaka, Y., 2014: Increasing wind sinks heat. Nature Climate Change, 4, 172–171.
Kosaka Y. and S.-P. Xie, 2013: Recent global-warming hiatus tied to equatorial Pacific surface cooling. Nature, 501, 403–407.
Mears, C. A., F. J. 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, doi:10.1029/2010JD014954.
Meehl, G. A., A. Hu, J. M. Arblaster, J. Fasullo, and K. E. Trenberth, 2013: Externally Forced and Internally Generated Decadal Climate Variability Associated with the Interdecadal Pacific Oscillation. Journal of Climate, 26, 7298–7310.
Neely, R. R., and Coauthors, 2013: Recent anthropogenic increases in SO2 from Asia have minimal impact on stratospheric aerosol. Geophysical Research Letters, 40, 999–1004.
Santer, B. D., and Coauthors, 2014: Volcanic contribution to decadal changes in tropospheric temperature. Nature Geoscience, 7, 185–189.
Schmidt, G. A., D. T. Shindell, and K. Tsigaridis, 2014: Reconciling warming trends. Nature Geoscience, 7, 158–160.
Solomon, S., J. S. Daniel, R. R. Neely, J.-P. Vernier, E. G. Dutton, and L. W. Thomason, 2011: The Persistently Variable “Background” Stratospheric Aerosol Layer and Global Climate Change. Science, 333, 866–870.
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, doi:10.1029/2012GL052723.