Greenhouse Warming: Fact, Hypothesis, or Myth?

Just recently I (Douglas V. Hoyt, author of the book "The Role of the Sun in Climate Change", and more than 80 scientific publications) have started looking at the Greenhouse Warming Hypothesis. On this and linked web pages, we are "kicking the tires", so to speak. Every new hypothesis needs its critics to keep things honest and that's what we are doing here. So spend a little time with me "kicking the tires" before buying or not buying the hypothesis.

We start by looking at changes in the global mean temperature or global temperature anomalies as mea-sured by surface thermometers and by satellites. As input data I used the approximately 3000 stations from the Climate Research Unit (CRU) in the UK. The Goddard Institute for Space Physics (GISS)uses a similar data set. GISS uses 1951-1980 as a base period to calculate temperature ano-malies. CRU uses 1950-1979. We use the year 1965. GISS divides the globe into about 8000 regions. CRU divides the globe into 2592 regions. We divide the globe into 720 equal areas. GISS and I use air temperature measurements only. CRU also use sea surface temperatures. Despite these differences in methods, GISS, CRU, and our calculations give nearly the same results. This exercise confirms that our computer program is running properly. It is important to note that we duplicate the GISS and CRU trends only if no corrections for urban warming are made. GISS, CRU and our trends are virtually identical, but are all these results correct?

In the plot of the time series above, they are smoothed with an 11 year running means so the two derivations of the time series can be compared easily. The red line is our reconstruction and the black line is the GISS reconstruction.

What Happens When the Base Climate Interval is Changed?

Next we changed the base year in our calculations from 1965 to 1980 (blue line in figure above). Other-wise our calculations were identical. Note that using the 1980 base year gives lower temperatures in recent years and higher temperatures in the earlier years. Why are the results such sensitive functions to the choice of base year? The so-called "global" measurements are not really global at all. At best, they sample 40% of the globe. The coverage by land surface thermometers slowly increased from less than 10% of the globe to about 40% in the 1960's, but has decreased rapidly to less than 20% in recent years. Depending upon the base year or interval chosen, a different time history of coverage will result. This different coverage history will lead to different climate reconstructions as shown in the figure above. These differences in trends arise from differences in areal coverage which are a function of baseline interval chosen.

The Temporal Variations in Areal Coverage

Changes in areal coverage of the land based thermometers are shown in the attached figure. Using either 1965 or 1980 as base years gives increasing coverage to the 1960's with sharply decreasing coverage since. The slight temporal differences in coverage for these two years give rise to different temperature trends. It is important to realize that only 8% of the climate variability can be categorized as global with the remaining 92% being regional variations. Thus, even if the global variations were eliminated, climate fluctuations at the local and regional level will still occur. If the thermometer net-work is sampling a portion of the globe, as is now the case, the combined regional variations could be interpreted as a global change even if no change is occurring. We will revisit this point in the next section of this essay.

Returning to our climate reconstruction, it is not clear that 1965 is a better base year than 1980. Al-though it provides more coverage, it has a greater temporal variation in its coverage. However, based upon a climate reconstruction using hundreds of different base periods, we conclude that using 1980 instead of 1965 provides a better reconstruction of climate. Or, in other words, today's climate is much like it was in the 1940's and 1860's. A study by Robeson in Climatic Change in 1995 showed that the CRU temperatures for 1880 were too low by 0.25 C, which is a completely independent confirmation of our results. Also J. Murray Mitchell in 1970 showed that there was a cooling between 1860 and 1880 as we have also found. Finally Wu et al. (1990) found the period around 1850-1880 is warmer than CRU and GISS have. Thus, four independent studies have found that the CRU and GISS temperature recon-structions for 1850 to 1880 are incorrect. This era was slightly warmer than they claim.

Despite our reservations about the temperature reconstructions which have been made so far, it is instructive to look at the recent reconstruction by Mann, Bradley and Hughes (Nature, vol. 392, pp. 779-787), which uses various proxies such as tree rings.

Mann's reconstruction is shown in the figure above as the blue line and bold blue line with a 11 year smoothing. In this version of Mann's temperature reconstruction, the temperatures have been adjusted so that the long-term trends agree with the 6000 borehole measurements of Huang et al. (1997). Note that the modeled solar irradiance (black line), given in our 1993 JGR article (vol. 98A, pp. 18895-18906), has many similarities to the temperature variations with the puzzling exception around 1800 about which Lamb (1977) says "the extent of snow and ice... attained a maximum as great as...any since the last major Ice Age." It is likely both the temperature reconstruction and solar irradiance reconstruction have some uncertainties associated with them. Despite this, the two series correlate well, particularly in recent years where the correlation is as high as 0.7.

This reconstruction uses tree rings to get the short term variations. However, tree rings respond not only to temperature, but also to precipitation, snow cover, snow melt, the concentration of carbon dioxide (i.e., fertilization), nitrogen fertilization, and so forth. The location of treelines can also be expected to respond to these same factors. These spurious effects can effect the tree ring reconstruc-tions, particularly on the longest time scales, and are not easily removed. Briffa et al. (Phil. Trans. B, 1999) comment on these problems saying "this partial non-climatic enhancement of twentieth century tree growth, particularly if acts in tandem with temperature forcing, will bias the coefficients in any regression-based equation estimating tree growth as a function of recent measured temperatures. Hence, the magnitude of the modern warming might be overestimated in the context of earlier recon-structed variability." The various forcing factors might interact in a non-linear fashion which could lead to greater uncertainties in reconstructing climate variations, particularly on the longest time scales, than is commonly assumed.

In contrast, borehole reconstructions respond primarily to temperatures. They are good at getting long-term trends, but have poor temporal resolution. Combining the two reconstructions takes advantage of the strengths of the two techniques and supports the existence of a Medieval Warm Period which is warmer than the present. To see this reconstruction using this approach, click here. Oxygen-18 isotope records from the Carribean also show similar climate variations.

Using Mann's reconstructions, the correlation of temperature and model solar irradiance around 1800 (covering the years 1701-1800) is about 0.4. In recent years it is about 0.6. Yet centered around 1880, the correlation is actually negative and reaches a value of about -0.4. The reason for the negative value is that a large number of volcanic eruptions occurred between 1811 and 1912. These eruptions cooled the climate and masked the solar signal. The volcanic eruptions occurred near solar maxima in nearly every case, so while the solar irradiance model would say the Earth would warm, the volcanic forcing would make it cool.

Another important point to make is that the temperature anomaly for the last 30 years of Mann's record equals 0.10 C. For 1400 to 1800, before any anthropogenic influence, the mean anomaly was -0.16 C. Therefore, the net positive temperature anomaly is 0.26 C. Since 1800 the anthropogenic forcing due to greenhouse gases should have reached half the value it will reach for a doubling of the greenhouse gases. Thus, if the recent 0.26 C warming is attributed solely to greenhouse gases, a doubling of greenhouse gases will give a net warming of 0.52 C. This number must be considered an upper limit. To put this another way, solar forcing can explain most of the twentieth century warming, but there is an unexplained residual warming of about 0.16 C that can be attributed to anthropogenic greenhouse gases. Looking to the future, we can anticipate roughly a 0.34 C additional warming in the summer months when greenhouse gases double circa the year 2070. The corresponding figure for annual means is about 2.25 times larger giving rise to an estimated warming of 0.77 C by 2070. For other discussions of empirical methods of the effect of doubling greenhouse gases, here are our calculations.

One final added comment (March, 2001): the Mann reconstruction assumes constant EOF's over 1000 years (i.e., constant spatial patterns of temperature variations). This dubious assumption, which does not appear to be true for the last 100 years, will tend to suppress climate variations and could explain why Mann is unable to detect the Little Ice Age and the Medieval Warm Period. Outputs of GCMs could be used to check the validity of the constant EOF assumption.

Is There a Warming of the Earth since 1979?

In the plots above and from the GISS reconstruction, there is a definitive warming of the Earth since 1979. Yet satellite observations using the Microwave Sounding Unit(MSU) show no warming over the same interval as shown in the figure below from Pielke et al. (JGR, 1998). Note that in this figure, three methods, two using balloons, show no trend for 1979-1996. So are these three measurements correct or are the surface observations correct?

We sub-sampled the MSU observations so their spatial and temporal sampling was identical to the surface observations. Then we ran our program to calculate temperature anomalies. We found that the anomalies and trends in the MSU data were not quite the same for both full and partial sampling. Care-ful examination of the trends shows that the spatial-temporal sampling for the land surface stations has substantial uncertainties in the yearly means (of the order 0.070 C). How these yearly uncertainties may effect the long-term trend determinations is not clear, but, from the above attached figure, about 0.12 C of the 0.23 C trend from 1979 to 1994 may be arising from spatial-temporal under-sampling by the surface network. This means that only about 0.11 C warming at the surface may be real.

A few other points are worth adding. Of the 2907 stations in the database, only 161 (or 5.5%) have complete temporal coverage from 1900 to 1990. All but 19 of these stations are in the United States. The US, with the most complete record anywhere, has no trend in temperatures during this century. In 1989 and 1990 about 30% of the stations ceased reporting. This may account for the difference in glo-bal temperature trends derived from surface observations when compared to balloon and satellite ob-servations. Support for this idea comes from the fact that 135 stations in the USSR ceased observing at the end of 1989. Subsequently there appeared to be a warming in the USSR but this warming is not supported by pressure observations. Thus, it appears half or more of the reported global warming from ground observations is arising from this change in station coverage. It is possible that as much as 0.2 C of the 0.25 C warming for 1979-1999 can be explained by this change in stations, although more study is required to refine this number. Other locations where the surface network has notable problems in-clude South Africa, Nigeria, Timbuktu, Algeria, Peru, central and coastal Brazil, the Seychelles, Diego Garcia, New Guinea, and several Polynesian islands.

In the two tables below we summarize the trends for the troposphere and surface respectively, as de-termined by a variety of techniques. The IPCC model predictions are also given. Most methods show little or no warming.

The MSU, balloon, and sea level derived temperature trends differ markedly from the surface obser-vations, as analyzed by the UK Met Office, and differ even more from the climate model predictions. Our own analysis, using the surface network, does not give as much warming in the last two decades. We now turn our attention to other possible problems with these time series.

Drifts in the MSU Time Series?

Roy Spencer and John Christy address this topic very adequately. We quote the summary of their conclusions: "There isn't a problem with the [MSU] measurements that we can find," Spencer explai-ned. "In fact, balloon measurements of the temperature in the same regions of the atmosphere we measure from space are in excellent agreement with the satellite results." Dr. Christy explained fur-ther, "In particular, we've examined these two `breaks' claimed by Hurrell and Trenberth. Even in these disputed intervals, we find excellent agreement between the two independent, direct atmospheric temperature measurements from balloons and satellites." The confirmation of the satellite observa-tions by the radiosondes virtually eliminates any possibility there could be a problem with the satellites such as esoteric claims that orbital decay is masking the appearance of a trend. The culprit in the disagreement between the two temperature almost certainly lies in problems with the surface network, to which we now turn.

Drifts in the Surface Temperature Network?

When the surface and MSU temperature trends are compared region by region, we note two things:

These two facts lead us to conclude that the surface network trends are varying in physically incongru-ous ways. Therefore, it is highly probable that the quality control on the surface network is inadequate to detect long-term trends (at least since 1979). This a problem above and beyond the lack of full global sampling already discussed. Many of the greatest discrepancies between surface and MSU observations occur in urban regions.

Is Urban Warming Causing the Apparent Upward Global Temperature Trend?

Urban areas have many buildings with vertical walls. The walls act like light traps or cavities and they capture and store heat during the day and then re-emit it at night. Cities and towns are enormous col-lections of Trombe walls. Thus, urban areas are warmer than the surrounding country and, in particu-lar, are warmer at night. (This has been known since at least 1850 when people were commenting on the urban heat island around the pioneer town of St. Louis). Thus, the urban heat islands are warmer with smaller diurnal variations. With growing population and growing urban construction, all urban regions can be expected to be getting progressively warmer. These upward temperature trends are probably accelerating because of the exponential growth in population.

Airports have similar heat islands. Here, huge paved runways soak up sunshine and act like huge Trombe walls laying on the ground, keeping the locality warmer. The enormous burning of gasoline also helps keep it warmer than the surrounding country. Today a large number of surface temperature observations are located at airports. Increased air traffic and larger airports cause these sites to warm up as well.

Removing the faulty city and airport observations always has the effect of bringing the surface and MSU temperature trends into better agreement.

Changing Skyline Hypothesis

Every thermometer station is located such that the thermometer screen and adjacent region view both the earth's surface and the sky. The sky is cold compared to the surface so there is net flux of radiant energy away from the station to the sky. Over the years, the portion of the sky seen by the average weather station may decrease because of growing trees or new nearby buildings. If these trees grow such that an additional 1% of the sky is blocked, then the temperature immediately around the radiation screen will rise by about 0.2 C. This temperature increase occurs because the thermometer is now embedded in a deeper and warmer cavity then before. The growing trees will also decrease wind velocities at the site, leading to further warming.

For example, if the station is located in a field with a line of trees about the station that are 20 feet high and 150 feet away in 1979, and are 30 feet high and 150 feet away in 1996, the temperature of that site will warm by 0.2 C. This warming is caused by increased trapping of thermal radiation which can no lon-ger radiate as easily to the cool sky. The warming will be greatest at night when the contrast between surface temperatures and sky temperatures is greatest and will tend to be greater in northern regions then in tropical regions. This effect will show up in rural stations.

A nice aspect of this hypothesis is that it explains why the surface stations are warming, but the sate-llite observations of the free atmosphere above the stations are cooling. The explanation, in brief, is that climate is slightly cooling and the surface network is spuriously warming. It also explains why the diurnal cycle of temperatures is decreasing even at rural locations and why these changes are greater in the arctic regions then elsewhere. No other climate hypothesis predicts this behavior. We will name this hypothesis the "Changing Skyline Hypothesis". To our knowledge, there has been no discussion or consideration of this possibility anywhere in the scientific literature.

Thomas Karl states: "Since 1950 all of the increase of temperature across the U.S.A. is due to an in-crease in the minimum temperature (about 0.75 degrees C/ Century or 1.5 degrees F/Century ) with no change in the daily maximum temperature. This caused a decrease in the diurnal temperature range." Subsequently, this type of behavior has been observed at other locations and is stronger as one goes towards the polar regions. It now appears most of the observed global surface warming of recent decades is occurring at night.

In 1995, James Hansen noted: "Models show that daytime warming will be almost as great as night-time warming" [for a greenhouse gas forcing]. In a 1997 paper in Climate Dynamics, climate modeler Watterton points out that the observed decrease in the range of diurnal temperatures "are not consis-tent with their being produced by the observed increase in greenhouse gases."No climate model yet devised can fully account for the observed diurnal variations and hence the recent observed warming. Watterton suggests that there are major errors in the way climate models treat clouds. A climate model experiment that comes close to explaining the results is given by Hansen et al. (1995) in Atmospheric Research. Their model experiment implies that a greenhouse gas doubling, accompanied by a 1.2% increase in low level clouds, will reduce the diurnal cycle by 0.21 C compared to an already observed decrease of about 0.5 C. This experiment implies clouds are acting as a strong negative feedback, so that when greenhouse gases double, the global warming will equal 0.67 C. Even this number may be high since they are unable to explain the full 0.5 C change, meaning even more clouds causing more cooling may be required. However, a recent study by Kaiser (Kaiser, D. P. (1998) Analysis of cloud amount over China, 1951-1994, Geophysical Research Letters, 25, 3599-3602) shows that cloud cover in China is decreasing along with the decrease in diurnal cycle amplitude. Thus, Hansen's model is not consistent with observations. Neither greenhouse gases nor changing cloud cover can account for the decreasing diurnal temperature cycle in the surface observations, which is the primary evidence that is used to claim the warming is caused by increased greenhouse gases.

Alternatively, the surface observations could be responding to a changing skyline, which would make the surface and satellite observations consistent with each other and yet leave the climate models mostly intact, but this idea also implies the greenhouse warming is very small. Perhaps the Changing Skyline Hypothesis is valid. It certainly fits all the observations whereas all alternative explanations have failed.

To validate or refute the Changing Skyline Hypothesis will require detailed site-by-site studies of temporal variations of the skylines at many stations. The postulated changes in the skylines are small and would not be readably noticeable. Measurements of present skylines and comparison to old photo-graphs of the sites is one approach. Another approach is a present day measurement of the skyline and a modeling of the tree shapes going back in time to deduce previous skylines. There is every reason to believe that the skylines about nearly all stations are varying. Should the Changing Skyline Hypothesis be confirmed, then all our knowledge about climate changes in the last century (and the theories to explain them) will be in doubt.

In brief, the stations were designed to determine the climatology. They were not designed to detect climate change and their quality control is insufficient for this task.

Finally, some correspondents have expressed doubt that the changing skyline can have any effect on temperature measurements and there is no experimental determination of the effect. To counter this, we quote Greely (1886): "It was once the experience of the writer in investigating the cause of unduly high temperatures, recorded by a thermometer well placed in a standard lourve shelter, that the distur-bing element was a painted tin roof, some fifty feet distant. This roof, reflecting the sun's heat at an obtuse angle, raised the temperature at times from two to three degrees."

Trombe Wall Effect, Changing Skylines, or Greenhouse Effect?

Everywhere today we hear that the Earth is warming up due to increased carbon dioxide in the atmos-phere. Yet the MSU satellite observations fail to see any warming. The surface observations, which are the main support for the greenhouse effect hypothesis, are poorly made and contaminated with urban effects and changing skyline effects. The surface temperature observations are made mostly in the 1% of the global areas that are heavily populated and built up. The spatial-temporal sampling of the surface network is not global and this under-sampling alone appears to be causing a spurious war-ming trend of 0.12 C (out of the observed 0.23 C warming trend between 1979 and 1994).

The old fashioned technique of using 30 year normals also forces climatologists to preferentially cho-ose faulty or inappropriate locations. A station must make measurements within a specified 30 year window or it will not be included in the GISS or CRU temperature reconstructions. This technique alone causes them to throw out about 40% of the observations made before 1870. There are analysis meth-ods which allow the use of all the data. Climatologists have not used them. There are analysis techni-ques that allow one to separate the urban heat island effect from other causes of climate change. Cli-matologists have not employed these techniques either. They are not published on this web page. When (and if) we get funding, we will show how to make the temperature trend analysis correctly. For those interested, here are the monthly temperature anomalies for GISS and MSU for 1997, each sca-led to give a yearly mean anomaly of 0.00 C. The two time series disagree quite a bit from month to month. For another argument that the surface observations are incorrect and the satellite observations are correct, see this discussion.

You Be the Judge

A physical theory is considered to be successful if its predictions and reality agree. Here is a GREEN-HOUSE WARMING SCORECARD of more than 30 predictions made by the greenhouse warming theory compared to actual climate changes or model predictions. This is your chance to judge the success of the greenhouse effect.

The Curious Case of the Missing Climate Feedbacks

Ramanthan (JGR, vol. 84, pp. 4949-4958) says "the direct radiative effects of doubled CO2 can cause a maximum surface warming [at the equator] of about 0.2 K, and hence roughly 90% of the 2.0-2.5 K surface warming obtained by the GCM is caused by atmospheric feedback processes described abo-ve." Or to put it another way, the initial global warming of 0.2-0.3 C is amplified through positive feed-backs by about a factor of 10 to give a 2.5 C warming for a doubling of carbon dioxide. These feed-backs are predominantly increased moisture and decreased cloud cover, which can respond very fast to any initial perturbation to the climate.

It is now known that the sun's brightness varies by about 0.1% over an 11 year cycle. Using this for-cing and the above postulated feedbacks, we would expect to see a 0.5 C oscillation in global tempera-tures with a period of 11 years. Observations indicate that the temperature oscillation is about 0.05 C and it is difficult to detect. This implies 1) the feedbacks don't really exist as the models have them or 2) there is some long time constant that damps out the solar variations. Supporting the first contention is that the model solar irradiances on very long time scales (shown in the figure above) can explain the observed climate variations only if the feedbacks are weak or non-existent. Thus, we have the curious case of the missing climate feedbacks when tested against solar forcing. Why then should we think that these positive feedbacks will exist for greenhouse gas forcing?

The simple carbon dioxide greenhouse effect causes a warming of about 0.25 C for a doubling. The "enhanced" greenhouse effect causes roughly a 2.5 C warming. The "enhanced" portion comes from positive or amplifying feedbacks, of which the strongest is claimed to be increased water vapor con-centrations. This hypothesis remains unproven.

Three Final Points

There are three important points to make about the reported warming of the last 20 years:

1. The warming has occurred mostly at night and not during the day. This result is inconsistent with a warming caused by greenhouse gases, but is consistent with urban heat island and other surface effects.

2. The reported warming has occurred only at the surface and not in the upper atmosphere. This type of warming is completely opposite to what is predicted if greenhouse gases are the cause. Again these observations are consistent with problems in the surface measurements.

3. The warming has occurred primarily in the Northern Hemisphere mid-latitudes with little in the polar and tropical regions. This result is consistent with urban influences, but is incompatible with the climate warming predicted from greenhouse gases which predict it to be largest in the polar regions.

In short, the reported warming is inconsistent with warming due to greenhouse gases in its tem-poral, vertical, and geographical distribution. The reported warming is consistent with problems in the surface network.

More Discussions

1. TEMPERATURE EFFECTS OF A CARBON DIOXIDE DOUBLING

For those interested, here are our calculations of how a doubling of carbon dioxide will change climate.

2. CAUSES OF CLIMATE CHANGE. Here is the better explanation for climatic changes in the 20th century and earlier.

3. A LESSON IN MODELING For information on modeling process, read this.

4. THE CLIMATE AND OTHERS SPEAK

Here is what the climate modelers and others have to say about the models and related topics.

5. FUTURE CLIMATE CHANGES

For information on the future of climate, read this section.

As final note, Sir Napier Shaw in his Manual of Meteorology said: "Every theory of the course of events in nature is necessarily based upon some process of simplification of the phenomena and is to some extent therefore a fairy tale."

Questions, comments, and suggestions should be sent to:

Douglas V. Hoyt,

The views presented here are those of the author alone, based upon 30 years of research in climate change and related topics. Copyright - 1997 - All Rights Reserved. Use of this article is for personal use only. Any other use is strictly prohibited. For any other use, permission must be in writing from Douglas Hoyt at

LINKS TO OTHER PAGES ON THIS SITE

TEMPERATURE RECONSTRUCTIONS | GREENHOUSE SCORECARD | SELECTED FEEDBACK

RECENT WARMING | DOUBLING CO2 | FUTURE CLIMATE

A MODELING LESSON | THE MODELERS SPEAK | PUBLICATIONS

TOP TEN SITES | BIASED THINKING

Back to Climate Change Page Back to English Version

You are visitor No.:

See here many interesting
statistics about this site

Which countries see us?
Who are our visitors?

TEMPERATURE MEASUREMENTS METHOD (troposphere)	Trend (degrees C/decade) for 1979-1997
Weather balloon thermistors (Angell/NOAA)	-0.070
Thickness derived temperature (1000-300 mb) (Pielke et al, JGR, 1998)	-0.029
Weather balloon thermistors (Parker/UK)	-0.020
MSU observations (Spencer and Christy, 1998), corrected for orbital decay and for east-west orbital drift	-0.010 +0.035 (1979-2001)
MSU observations (Wentz and Schabel, 1998), corrected for orbital decay only	+0.070
IPCC model predictions (2001)	+0.40

Temperature Measurement Method (surface except TOPEX observations)	Trend (degrees C/decade) for 1979-1997
Thickness derived temperature (1000-850 mb) (Pielke et al, JGR, 1998)	-0.056
Northern Hemisphere summer temperatures from tree rings through 1994 (Briffia et al., Nature, 1998)	-0.038
Mean & standard deviation of 12 lower tropospheric data sets (Santer et al., JGR, 1999) [for 1979-1993]	+0.008 +/- 0.066
Our analysis without urban heat island corrections	+0.055
TOPEX/Poseiden observations showing 0.9 mm/year global mean sea level rise for 1992-2001 which gives upper limit to ocean temperature rise	+0.045 (maximum, 1992-2001) +0.025 (probable, 1992-2001)
UK Met Office analysis	+0.150
IPCC model predictions (2001)	+0.360