A new work by Santer et al. in the Journal of Climate shows that trends observed in sea surface temperature between 1988 and 2019 [SST], Tropospheric temperature [TLT and TMT], and all tropospheric water vapor [TWV] generally agree to varying degrees with the climate model trends over the same period. The study uses ratios between observed trends in these variables to examine how well the ratios meet model expectations, assuming that the models provide “truth” in such comparisons. Particular attention is paid to the inconsistency between TWV humidification rates and tropospheric satellite temperature warming rates: total water vapor has increased faster than one would expect for the weak rate of satellite-observed tropospheric warming (but both are still below average climate) model trends in CMIP5 or CMIP6).
While the paper itself does not pinpoint tropospheric temperatures as flawed, the paper’s widespread coverage uses the same biased headline, such as that from DailyMail.com: “Satellites May Have Underestimated the Warming of the Planet for Decades”. The coverage largely ignored what was in the paper, which was far less critical of satellite temperature trends and should have been more newsworthy. For example: (1) The paper shows that SST warming is well below the climate model’s expectations of both CMIP5 and CMIP6, which could have been an important conclusion; (2) the possibility that the satellite-based TWV will rise too quickly (permitted in the paper and addressed below), and in particular (3) the possibility that the TWV is not a good proxy for mean and upper tropospheric warming anyway (discussed below).
As others have shown, free tropospheric vapor (not well captured by TWV) would be the correct proxy for free tropospheric warming, and the fact that climate models maintain constant relative humidity with altitude as it warms is not based on fundamental physical processes ( as the authors imply), but on arbitrary humidification assumptions that are implicit in convective model parameterizations. Observations show that the humidity in the free troposphere does not increase as strongly with the tropospheric temperature as in the GFDL climate model. Thus, a weak tropospheric warming measured by satellites could be an indication of weak water vapor feedback in the free troposphere, which in turn could explain the weaker than (in the model) expected surface warming. A possible reason for a high distortion of the TWV trends is also addressed, which is consistent with the trend behavior of the other variables.
Evidence presented in Santer et al. (2021)
I have been asked by several people to write a new article in the Journal of Climate by Santer et al. (Using Climate Model Simulations to Constrain Observations), which has as one of its conclusions the possibility that satellite-based warming estimates of tropospheric temperature may be too low. Based on my initial examination of the paper, I conclude that there is nothing new in the paper that would cast doubt on the humble nature of satellites’ tropospheric warming trends – unless one believes in climate models as evidence in this one Case we don’t need observations anyway.
The new study focuses on the period 1988-2019 so that full integrated water vapor recovery over the ocean can be used by the satellite-based instruments SSM / I and SSMIS. The recent warming of the surface and troposphere has actually been accompanied by an increase in water vapor in the troposphere, and the quantitative relationship between temperature and vapor is used by the authors as a guide to determine whether the tropospheric warming rates from satellites have been unrealistically low.
Most of the relevant conclusions in the new work come from its Fig. 9, which I have annotated in Fig. 1 below for the sake of clarity.
The problem with using integrated water vapor as a proxy for tropospheric warming is increasing
A key conclusion of the paper is that total integrated water vapor has increased faster than the SST trends suggest, while tropospheric temperature has increased less rapidly (assuming the models are correct that SST warming in the troposphere is significantly increased should). This shifts the observations from the climate model-based regression lines in Fig. 1a, 1b and 1b.
The problem with using TWV humidification as a proxy for tropospheric warming is that while TWV is strongly coupled to SST warming, how well it is naturally coupled to free tropospheric warming (above the boundary layer), is very uncertain. TWV is dominated by boundary layer water vapor, while it is a medium to upper tropospheric warming (and thus in the TMT satellite measurements), which is strongly related to how much the humidity increases at these high altitudes (Po-Chedley et al., 2018 ).
This high-altitude region is not well represented in the TWV retrievals. Satellite-based TWV polls use the relatively weak water vapor line near 22 GHz and are therefore primarily sensitive to water vapor in the lowest layer of the atmosphere.
In addition, these queries are dependent on assumptions about the profile shape of water vapor in the atmosphere. If global warming is accompanied by preferential humidification of the lower troposphere (through increased surface evaporation) and a thickening of the humid boundary layer, the all-important increase in free tropospheric humidity might not be as strong as assumed in these withdrawals, which is based on regional profile differences across different sea surface temperature regimes.
As Spencer & Braswell (1997) and others have shown, the ability of the climate system to cool into space depends heavily on changes in humidity in the upper troposphere during warming (see Fig. 2). The upper troposphere has very low levels of water vapor in both relative and absolute terms, but these low levels of steam in the upper 75% of the troposphere have dominant control over cooling to space.
As indicated in Fig. 2, water vapor in the lowest levels of the troposphere is largely controlled by surface evaporation. If the surface warms up, increasing evaporation moistens the boundary layer, and a constant relative humidity is a pretty good rule of thumb there. However, in the middle and upper troposphere, the air extracted from the precipitation systems has a decisive influence on the humidity. The proportion of condensed water vapor that is removed by precipitation determines how much is left to humidify the environment. The humidity of the free troposphere, which in clear air drops thousands of kilometers away from any precipitation system, was determined days to weeks beforehand when this air rises in these precipitation systems. As Renno, Emanuel and Stone (1994) demonstrated using a model with an explicit atmospheric water cycle, the precipitation efficiency determines whether the climate is cool or warm by controlling the main greenhouse gas water vapor.
Important, We don’t know how precipitation efficiency changes with warming, so we don’t know how strong the water vapor feedback is in the real climate system. We know that tropical rain systems are more efficient than higher latitude systems (as many of us anecdotally know from visits to the tropics, where even flat clouds can produce torrential rainfall). Global warming is expected to be accompanied by an increase in precipitation efficiency, and recent research is beginning to support this view (e.g. Lutsko and Cronin, 2018). This would mean that the absolute (specific) humidity of the free troposphere may not increase as much as climate models assume, resulting in less surface warming (as observed) and less tropospheric amplification of surface warming (as observed).
Since climate models do not yet take into account the precipitation microphysics, which determine the changes in precipitation efficiency with warming, The behavior of the models in relation to temperature and humidity in the free troposphere should not be used as “truth” when evaluating observations.
While climate models tend to maintain a constant relative humidity throughout the troposphere during warming, which causes strong positive water vapor feedback (e.g. Soden and Held, 2006) and thus leads to strong surface warming and even greater tropospheric warming, there are differences here between models respect. In the Po-Chedley et al. (2018, her Fig. 1a) there is a factor of 3 variation in the feedback of the lapse rate between the models, which is a direct measure of how much tropospheric amplification there is of surface warming (the so-called “hotspot”). This gain is in turn directly related (you get r = -0.85) how much additional water vapor is discharged into the free troposphere (also in its Fig. 1a).
What happens to free tropospheric moisture in the real world?
In the real world, it is not clear that free tropospheric water vapor maintains constant relative humidity when warmed (which would result in severe surface warming and even greater tropospheric warming). We do not have good long-term measurements of water vapor changes in the free troposphere on a global basis.
Some researchers have argued that seasonal and regional relationships can be used to infer feedback from water vapor, but this seems unlikely. How the entire system changes over time with warming is not so certain.
For example, if we use satellite measurements near 183 GHz (e.g. available from the NOAA AMSU-B instruments since late 1998), which are very sensitive to the vapor from the upper troposphere, we find in the tropics that the tropospheric temperature and humidity changes appearing over time in satellite observations are very different from the GFDL climate model (Fig. 3).
Further details on the results in Fig. 3 can be found here.
Possible distortions in satellite based water vapor trends
While satellite polls from TWV are known to be fairly accurate compared to radiosondes, minor changes in the vertical profile of water vapor during global warming can potentially cause bias in TWV trends. The Santer et al. (2021) mentions the possibility that the total vertically integrated atmospheric water vapor trends reported by satellites since mid-1987 may be too high, but does not address the reasons for this.
This topic has preoccupied me for many years because the TWV trend since 1988 (only available over the ocean) has been increasing faster than we would expect based on the warming trends of sea surface temperature (SST) combined with the assumption of constant relative humidity at depth the troposphere (see Fig. 1a, 1b, 1c above).
How can such a retrieval bias occur? The TWV retrieved is proportional to the heating of a passive microwave TB near the weak 22.235 GHz water vapor absorption line above the radiometrically cold (reflective) sea surface. As such, it depends on the temperature at which the water vapor emits microwave radiation.
The TWV restoration depends on the assumed shapes of the vertical profile of water vapor in the troposphere, ie on what heights and thus what temperatures the water vapor emits. These assumed vertical profile shapes are based on radiosonde data (weather balloon) from different regions and different seasons with different underlying sea surface temperatures. However, these regional and seasonal variations in shape may not reflect the changes in shape during warming. If the predominant part of the humidification takes place in the boundary layer during long-term warming (see Fig. 2 above, below 800 hPa pressure height), with possibly slight thickening of the boundary layer, but little humidification in the upper troposphere, then the retrieved TWV could be strongly distorted because the additional water vapor emits microwave radiation from a lower (and therefore warmer) altitude than is assumed during the search. This leads to a high distortion of the recovered water vapor over time as the climate system warms and humidifies. As the head of the NASA AMSR-E Science Team, I asked the developer of the TWV retrieval algorithm about this possibility a few years ago, but never received an answer.
The new Santer at al. Study ignores radiosonde evidence backing our UAH satellite temperatures
That being said, it’s also worth noting that the new study doesn’t even refer to our 2018 results (Christy et al., 2018) showing that the most stable radiosonde data sets support the temperature trends of the UAH satellites.
The new study by Santer et al. does not provide convincing evidence that satellite readings of tropospheric temperature trends are unrealistically low, and media coverage of their study has been skewed in this regard. Their conclusion (which they admit is ambiguous) depends on belief in climate models of how warming of the upper troposphere is related to increases in total tropospheric water vapor (TWV). Because TWV is not very sensitive to changes in water vapor in the upper troposphere, and these changes largely determine how much tropospheric amplification of surface temperature trends will occur (e.g. the “tropical hotspot”), TWV cannot determine whether tropospheric temperature trends are realistic or not.
In addition, there is some evidence that the TWV trends themselves are heavily skewed, which the study authors admit as a possible explanation for the trend relationships they calculated.
The present observations, as they are in Santer et al. Studies broadly agree with the view that global warming is advancing at a significantly slower rate than that predicted by the latest climate models, and that much of the disagreement between models and observations is due to incorrect assumptions in those models.
1) The warming of the SST was significantly less than the models predict, especially in the tropics
2) The tropospheric enhancement of surface warming was weak or absent, suggesting weaker positive water vapor feedback in nature than in models
3) In turn, weak water vapor feedback helps explain weak SST warming (see ).
4) Recent published research (and preliminary evidence in Figure 3 above) supports the view that water vapor feedback from climate models is too strong and therefore current models should not be used to validate observations in this regard.
5) Satellite-based total water vapor trends cannot be used to infer water vapor feedbacks because they are likely to be highly skewed due to assumptions about the vertical profile and because they are unlikely to reflect how water vapor in the free troposphere has changed with warming, which a great impact on the feedback of water vapor.
Christy, J. R., R. W. Spencer, W. D. Braswell and R. Junod, 2018: Investigation of space-based atmospheric mass temperatures in climate research.
Intl. J. Rem. Sens., DOI: https://doi.org/10.1080/01431161.2018.1444293
Lutsko, N. J. and T. W. Cronin, 2018: Increasing the precipitation efficiency with surface heating in the radiation-convection equilibrium. J. von Adv. Modell. Earth Sys., DOI: https://doi.org/10.1029/2018MS001482.
Po-Chedley, S., K. C. Armor, C. M. Bitz, M. D. Zelinka, B. D. Santer and Q. Fu, 2018: Sources of Intermodel Spread in the lapse rate and water vapor feedbacks. J. Klima, DOI: https://doi.org/10.1175/JCLI-D-17-0674.1.
Renno, N.O., K. A. Emanuel and P. H. Stone, 1994: Radiation convection model with explicit water cycle: 1. Formulation and sensitivity to model parameters, J. Geophys. Res. – Atmos., DOI: https://doi.org/10.1029/94JD00020.
Santer, B. D., S. Po-Chedley, C. Mears, J. C. Fyfe, N. Gillett, Q. Fu, J. F. Painter, S. Solomon, A. K. Steiner, F. J. Wentz, M.D. Zelinka and C.-Z. Zou, 2021: Using climate model simulations to restrict observations. J. Klima, DOI: https://doi.org/10.1175/JCLI-D-20-0768.1
Soden, B. J. and I. M. Held, 2006: An assessment of climate feedback in coupled ocean-atmosphere models. J. Klima, DOI: https://doi.org/10.1175/JCLI3799.1.
Spencer, R.W. and W.D. Braswell, 1997: How dry is the tropical free troposphere? Implications for Global Warming Theory. Bull. Amer. Meteor. Soc., 78, 1097-1106.