Global Warming

* Everything should be made as simple as possible, but not simpler – A. Einstein

Introduction

In this article I will develop a simple* model for understanding the complex phenomenon of global warming.  By global warming I mean the rising trend in average surface temperature of the entire Earth that has occurred over the past century.  I start by describing the fundamental principles driving global temperature, which are quite straightforward.  Basically, the Earth receives energy from the sun, which warms the Earth, and radiates energy out into space, which cools the Earth.  Radiation increases with higher temperature and so the temperature of the Earth is that temperature at which energy outgo equals energy income.  How the Earth radiates energy depends on properties of its surface and atmosphere.  Here is where the complexity comes in because many things affect the ability of the Earth to radiate energy and so affect the temperature.   This article deals with the three major factors, greenhouse gases, solar effects and the effect of aerosols.

The model developed here is greatly simplified yet it contains enough elements to allow a nonexpert to gain an understanding of how greenhouse gases work to warm the climate and how clouds affect it.  The model is used to describe an important recent development in climate science, the cosmic ray climate driver hypothesis.  This theoretical mechanism has been cited by many global warming skeptics as an alternate explanation for warming besides greenhouse gases. This article shows that although the theory is useful for explaining early 20th century warming, it does not explain recent warming. 

Elementary climate dynamics

The ultimate driver of climate is the sun, or more specifically, the energy bestowed upon the Earth as sunshine.  The sun produces a total power output of some 4 x 1026 watts.  At Earth’s orbital distance this power output works out to about 1370 watts per square meter of surface perpendicular to the sun’s rays.  This energy flux is called the solar constant or S.  The Earth presents a circular barrier to this power output of area πR2, where R is the radius of the Earth.  It follows that the amount of energy intercepted by the Earth is then πR2S.  The Earth reflects about 30% of incident energy, the rest is absorbed.  The fraction reflected is called the albedo or A.  With this I can write the following expression for the rate of energy absorption by the Earth:

1.                  Energy absorption rate = πR2 S (1-A)

The Earth also radiates energy into space.  This radiation is in the form of infrared (IR) radiation, not visible light as is much of the sun’s energy.  All bodies radiate energy as a function of temperature.  The amount of radiation that is emitted by one square meter of surface at a given temperature is given by the Stefan-Boltzmann equation:

2.                  Radiated energy flux   =  ε σ T4

Here T is the absolute temperature in degrees Kelvin (K).  The Kelvin temperature scale is simply the familiar Celsius temperature with 273.15 added to it.  It measures absolute temperature; 0 K is absolute zero, the coldest temperature possible.  The parameter σ is the Stefan-Boltzmann constant and is equal to 5.67x10−8 watt m-2·K-4.  Finally, ε is the average emissivity of the earth.  Emissivity is a property of the material emitting radiation that relates a material’s actual radiative properties to that of an ideal black body. A black body has an emissivity of one for all wavelengths and represents the maximum radiative potential.  Real materials have values of ε between 0 and 1 as a function of wavelength.  The ε in equation 2 refers to an average value over the relevant wavelengths and various surface types of the Earth.   

Equation 2 refers to an energy flux, that is, the energy transfer rate per unit area (square meter).  To determine the total radiation emitted by the Earth I multiply this flux by the surface area of the Earth, which can be approximated as the area of a sphere: 4 πR2.  Multiplying equation 2 by this area gives the rate that energy is radiated by the Earth:

3.                  Energy radiation rate  =  4 πR2 ε σ T4

Over time the Earth’s average temperature will adjust so that energy absorbed equals energy radiated.  Setting equation 1 and 2 equal to each other and rearranging gives the following equation related Earth’s average temperature, ε, the solar constant, and albedo:

4.                  ε σ T4 =  ¼ S (1-A)  =  I  = 240 watts/m2

For a solar constant of 1370 watts/m2and albedo of 0.3, the right hand side of equation 4 works out to 240 watts/m2.  I will call this the quantity the incoming solar radiation or insolation (I)  It refers to the average amount of solar energy absorbed per square meter of surface.  The actual average surface temperature of the Earth is about 15º C (288 K).  If this value is inserted into equation 4 and the result solved for ε, one obtains:

5.                  ε  =  I / (σ T4) = 240 / (5.67 x 10-8 · 2884)  =  0.615

Table 1.  Emissivity of various terrains (typical values observed over the entire range of IR wavelengths)

Material

Emissivity

Reference

Ice and snow

0.94-0.99

1

Vegetation

0.94-0.99

1

Soil

0.75-0.99

1

Seawater

0.98-0.99

2

Grassland

0.95

3

Barren Land

0.93

3

Forest

0.96

3

Urban

0.95

3

This value is the apparent emissivity of the Earth as viewed from space.  Table 1 lists emissivity for a variety of terrain types.  Emissivities for many materials do not vary much with wavelength, making the use of an average value reasonable.  As Table 1 shows, the emissivities from various types of terrain are not much different from each other either, making the use of a single average value to describe the Earth’s surface emissivity a good approximation.  The emissivity of the Earth’s surfaceS) appears to be greater than 0.95.  A value of 0.96 is used on some models3 and will be used here.  With a 0.96 emissivity, equation 4 yields a temperature of 258 K or −15º C.  An Earth surrounded by an IR-transparent atmosphere would be a frozen ice ball.  The Earth is not frozen because the clouds and atmosphere are not transparent to IR radiation.  Some of the radiation emitted from the Earth is absorbed by the clouds and atmosphere and re-radiated.  This reduces the Earth’s apparent emissivity from the 0.96 value for the surface to an apparent value of 0.615.

Factors affecting the earth’s albedo and emissivity: the effect of clouds

Table 2 shows presents radiative properties of clouds.  Clouds cover about 63% of the Earth’s surface on average.5 The albedo of the Earth is given by the weighted sum of surface average albedo (ASURF) and the cloud average albedo (ACLOUD):  A = 0.3 = 0.37∙ASURF + 0.63∙ACLOUD.  Overall radiation measurements indicate that in the absence of cloud, solar insolation would be 53.5 watt m-2 greater.5  Thus, a cloudless Earth would see a solar insolation of 240 + 53.5 = 293.5 watt m-2.  From equation 4, this value for I is consistent with an albedo of 0.14.  Thus, ASURF = 0.14.  It follows that ACLOUD = 0.39 = (0.3 − 0.37∙0.14)/0.63.  The 0.25 difference in albedo between clouds and the surface also affects climate.  In the section above I showed that an Earth with an IR-transparent atmosphere would have a temperature of −15º C.  This calculation ignored the effect of clouds on albedo.  An Earth with an IR transparent atmosphere would not have any clouds; its albedo would be 0.14 rather than 0.3. With A = 0.14 and ε = 0.96, equation 4 gives T = 271 K or −2º C, not −15º C.

Table 2.  Radiative properties of clouds

Type

Fraction coverage5

Effective ε4

Albedo5

Low

26%

0.68

0.39

Medium

18%

0.55

0.39

High

19%

0.33

0.39

Surface

37%

0.96

0.14

Radiation absorbed by clouds is re-radiated out into space from a surface having a lower effective emissivity than the surface below it.  This lower effective emissivity mostly reflects the lower temperature of clouds due to their elevation.  Clouds thus act to reduce the overall emissivity of the Earth, warming it.  They serve as an IR radiation shield.  Those readers who have ever used a carport in cold climes know that despite the lack of enclosure, no frost collects on the car’s windows overnight.  Frost is caused by radiation from the car cooling it below the air temperature, allowing frost to form.  With the carport top blocking this radiation, frost is avoided.  Clouds serve a similar function for the surface beneath them.

Table 2 shows the effective emissivities of clouds as a function of their elevation. These values are averages over plots of cloud emissivities as functions of wavelength obtained by Shippert et al.4  Low cloud is defined as clouds with a base height below 3 km and high clouds have base heights above 6 km.  Medium-high cloud is in between.  Higher clouds are more effective shields because they re-radiate at lower temperatures.  Using data from Table 2 the average emissivity of Earth with clouds would be 0.37·0.96 (cloudless surface) + 0.26·0.68 (low cloud) + 0.18·0.55 (medium-high cloud) + 0.19·0.33 (high cloud) = 0.694 (apparent emissivity of Earth as viewed from space). With a 0.694 emissivity, equation 4 yields of temperature of 279 K or +6º C.  With clouds in an IR-transparent atmosphere, the Earth would be 8 degrees warmer than without clouds, but it still would be colder than its actual temperature of +15º C. 

Note that in this calculation cloud cover is treated as an opaque screen; no radiation from the surface is assumed to penetrate through the cloud layer, it is all either absorbed or reflected back to the surface.  These opaque screens themselves radiate outward into space with a lower effective emissivity, reducing the Earth’s emissivity from the 0.96 value of the surface to the value given in Table 2.  In actuality, thin cloud is not strictly opaque and the interaction of clouds with radiation is more complex than this shield model indicates.  Nevertheless this approach gives a reasonably good semi-quantitative picture of the effect of clouds.  For example, an Earth completely covered by high cloud would be warmer: with A = 0.39 and ε = 0.33, equation 4 gives a temperature of 325 K (52º C).  In contrast, an Earth covered by low cloud would be cooler: with A = 0.39 and ε = 0.68, equation 4 gives a temperature of 271 K (−2º C).  These results are consistent with more sophisticated models that also show warming from high cloud and cooling from low cloud. The exact amounts predicted are different, but one should not expect detailed accuracy from such a simple model.

Factors affecting the earth’s emissivity: the greenhouse effect

Certain gases in the atmosphere called greenhouse gases absorb and re-radiate IR radiation emitted from the surface, just like clouds.  Unlike clouds, the greenhouse-gas containing atmosphere cannot be modeled as an opaque screen, but rather, as a translucent screen that partially shields the ground emissions.   This effect serves to reduce the effective emissivity of the cloudless sky from the 0.96 of the surface to a new value (εT) that takes into account this partial shielding from greenhouse gases.  The shielding provided by greenhouse gases is dependent of the wavelength of the IR radiation.  Figure 1 shows a plot of the absorption of IR radiation by the cloudless atmosphere as a function of wave number.  Wave number is the number of wavelengths per centimeter, so larger values means shorter wavelength.  From this figure one can see effects of individual greenhouse gases: strong absorption by water vapor at 100-600 cm-1 and 1400-1600 cm-1, by ozone at 1060 cm-1, and by carbon dioxide (CO2 ) at 600-700 wavenumber.6 

Figure 1.  Difference between surface radiation and energy radiated into space, indicative of IR radiation absorbed by the atmosphere6

Table 3 shows a summary of four kinds of IR shields lying over the Earth’s surface: the three types of cloud and the translucent atmosphere.  The simple model I am developing here holds that the total emissivity of the Earth (which we know to be 0.615) is simply the weighted sum of the effects of all these shields.  The weights are the percent coverage of each type of screen.  The effective emissivities for three of the four screens are known (Table 2) and are reproduced in Table 3.  The contributions from these three sum to 0.3385.  The rest of the emissivity comes from surface radiation through the cloudless, translucent atmosphere, which has an effective emissivity of εT.  If I assume that εT is higher than that for all clouds, then it will only affect the radiation from the 37% of the surface not covered by clouds.  A value for εT can be obtained by subtracting the cloud-provided shielding from the overall emissivity of 0.615 and dividing the difference by the cloudless fraction of the Earth’s surface: (0.615−0.3385)/0.37 = 0.747 for εT (see Table 3).  Since 0.747 is greater than 0.68 the assumption that εT was higher than the emissivity for any type of cloud was valid.

Once again this translucent screen analogy to the IR-absorbing atmosphere is a simplification of a more complex situation.  Absorption and re-radiation of IR radiation occurs throughout the atmosphere and interacts with clouds.  A completely accurate assessment of this phenomenon requires very complex radiation balance models.  Nevertheless, the results obtained with this simple model (see below) are quite sufficient to gain a reasonably good understanding of how CO2 affects the climate.

Table 3.  Effect of various IR shields on overall emissivity of the Earth

Type of Cover

%Coverage5

Emissivity4

Contribution

High Cloud

19%

0.33

0.0627

Med Cloud

18 %

0.55

0.0990

Low Cloud

26%

0.68

0.1768

Surface

37%

εT = 0.747

0.2764

Overall

100%

0.615

0.3385+0.37εT

The value of εT depends on the concentration of greenhouse gases in the atmosphere such as CO2.  Thus, adding CO2 to the atmosphere affects εT, which affects the earth’s overall emissivity as shown by equation 6:

6.                  ε   =  0.3385 + 0.37 εT

Changes in overall ε produced by rising CO2 affect temperature according to equation 4. 

The analysis of the translucent atmosphere is more complex than the effect of clouds, which were treated as opaque barriers with a fixed emissivity.  The translucent atmosphere both absorbs and transmits (allows to pass through) radiation from the surface.  This situation can be described by a radiation balance on the translucent atmosphere:

            RADIATION INTO ATMOSPHERE   =        RADIATION OUT OF ATMOSPHERE

7.                              σ εS TS4                                      =        2 σ εA TA4                 +                         (1− εA) σ εS TS4 

      surface radiation into atm       =       atm. radiation into space & back to surface       surface radiation transmitted through atm into space

Here εS is the emissivity of the surface (0.96) and εA is the intrinsic emissivity of the atmosphere (a function of greenhouse gas levels).  TS and TA are the temperature (in Kelvin) of the surface and the atmosphere, respectively.  The factor (1− εA) is called the transmission of radiation through an absorbing medium.  The factor 2 occurs because the atmosphere radiates both outward into space and inward towards the surface.  Equation 7 can be solved for TA as follows:

8.                  TA4  =   ½ εS TS4        TA = (½ εS)¼ ∙ TS  =  0.83∙288  = 240 K

The total emissions from the translucent atmosphere into space are given by:

9.                  total emissions = σ εT TS4  =    σ εA TA4  + (1− εA) σ εS TS4 

Note than the 2 is absent because only the radiation going into space from the atmosphere is considered here.  Equation 8 can be substituted into equation 9 to obtain the result:

10.              εT   = εS (1 − ½ εA)

For the actual value of εT of 0.747 from Table 3 and εS = 0.96, equation 10 gives a value of 0.4438 for εA.  The effect of IR absorbers on emissivity is described by Beer’s Law, which states that for a given wavelength, the logarithm of the transmission through a mixture is equal to the negative of the sum of the absorbance (Ai) of each IR-absorbing species present:

11.              ln(1─ εA) =  − ∑ Ai   for each wavelength

The symbol ∑ means “the sum of, (in this case absorbances) for all i”.  The absorbance of species i is given by Ai = ai∙b∙Ci, where Ci is the concentration of species i, b is the length of light path through the i-containing medium and ai is the molar absorptivity.   The molar absorptivity is a function of wave number (η); that is, ai should be written as a function, ai(η).  For the application here, b is dependent on the thickness of the atmosphere, which is a constant independent of Ci. Hence ai(η) and b can be combined into a single parameter ki(η) and equation 11 becomes:

12.              ln(1─ εA) =  − ∑ ki(η)∙Ci     εA =  1 − exp[−∑ ki(η)∙Ci]  for each wavelength

In fact, an IR spectra like Figure 1 gives the value of ki(η) implicitly as a function of wave number since the spectra is a measure of atmospheric absorbance for a fixed Ci.  I will assume ki(η) is a constant equal to its average value over all η.  I will consider only two species, water vapor (H2O) and CO2.  For these assumptions, equations 11 and 12 can be written for the entire spectrum of emitted radiation as follows:

13.              εA  =  1 − exp[−(AH2O + ACO2)]   =  1 − exp(−kH2OCH2O − kCO2∙CCO2)

For εA = 0.4438, the sum of AH2O and ACO2 is equal to 0.5865.  Based on the bands in Figure 1 it is clear that water vapor is responsible for most of the IR absorption by the atmosphere.  Let us assume 79% of absorbance is due to water vapor, that is, AH2O = 0.79·0.5477 = 0.4634.  This leaves ACO2 = 0.1232. The “base level” for CO2 is set at 350 ppm (0.035%), that is, the assumed Earth temperature of 288 K occurs at CCO2 = 0.035.  From this I obtain kCO2 = 0.123/0.035 = 3.52 and equation 13 becomes:

14.              εA =  1− exp(−AH2O − 3.52∙CCO2) 

To obtain a value for AH2O for situations other than the base case I assume that changing global temperature does not affect average relative humidity, only absolute humidity.  With this assumption the effect of temperature on the average concentration of water (CH2O) of the atmosphere is governed by the Clausius-Clapeyron equation:

15.              ln[CH2O/288CH2O]  = ln[AH2O/288AH2O]  =  −ΔHVAP/R∙[1/T − 1/288] →  AH2O = 288AH2O∙exp(5294∙[1/288 − 1/T])  =  0.4634∙exp(18.382 − 5294/T) 

Here ΔHVAP is the heat of vaporization of water and R is the gas constant.  The term ΔHVAP/R has the value 5294 K.   288CH2O is the concentration of water vapor at the standard temperature of 288 K.  The ratio CH2O/288CH2O is equal to the ratio AH2O/288AH2O where 288AH2O is the value of AH2O at the standard temperature of 288K, which has already been determined as 0.4634.  With these values AH2O becomes a simple function of T as shown in equation 15.  Substituting equation 15 into equation 14 gives:

16.              εA =  1− exp(−0.4634∙exp(18.382 − 5294/T)  − 3.52∙CCO2) 

The greenhouse model is now complete.  For a given value of CCO2, εA as a function of T is obtained from equation 16.  The value of εA is used in equation 10 to obtain εT as a function of T.  This value is used in equation 6 to obtain ε for the Earth as a function T.  This value of ε then used to calculate temperature using equation 4.  The value of temperature obtained is a function of T and must equal T.  This will be true for a particular value of T which can be obtained using trial and error (the goal seek function in Microsoft Excel™ makes this easy to do).   This procedure was used to calculate the change in temperature from 288K (ΔT) produced by changes in CO2 level from 0.035% (Δ%CO2). The results are plotted in Figure 2.

Figure 2.  Temperature change and forcing as a function of change in CO2 level relative to 350 ppm

This simple model has a number of limitations.  It ignores the effect of changing levels of water vapor on clouds.  If changes in water vapor affect lower clouds preferentially, then increases in water vapor will exert a cooling effect that will offset at least some of the greenhouse warming effect of water vapor itself.  On the other hand, if increases in water vapor affects high cloud more than lower cloud, then the cloud-mediated effect of water vapor could be greater than the effect calculated here.

The model does a reasonably good job of predicting the temperature increase for a doubling of CO2 level from 280 ppm (the preindustrial value) to 560 ppm.  The value obtained from this model is +1.7º C, which is pretty close to the +1.85º C corresponding value obtained from more sophisticated models. The model underestimates the effects of low levels of CO2.  For example, it predicts temperature would be 2.3º C cooler without any CO2 in the atmosphere.  More sophisticated models predict 3-4º C (= 9-12% of total 33º C greenhouse effect).7  On the other hand, the model overestimates the effects of higher CO2 levels; quadrupling CO2 levels from pre-industrial levels is predicted to increase temperature by 4.4º C as compared to 3.7º C from more sophisticated models. 

It is important to note than this model made use of an adjustable parameter, sometimes called a fudge factor.  This adjustable parameter was the fraction of the atmospheric absorbance under base conditions that was assumed due to water vapor.  The value of 79% was selected in order to calibrate the model with respect to more sophisticated models.  What this means is the greenhouse model presented here is representational, not predictive.  One cannot use it to accurately predict the direct effect of rising CO2 on temperature; the results obtained with it will not exactly match the results of the complex models.  The purpose of this model is to provide an easy to understand (and manipulate) model that can represent the results of more complex models.  The results obtained with it will not exactly equal the results of the complex models, but it will get the approximate magnitude and direction correct. 

An example of the use of the simple model: converting Earth into Venus

As an example of how this model can be used, let us consider what would happen if greenhouse gases rose incessantly.  Equation 10 shows εT varying between 0.96 for εA = 0 (transparent) to 0.48 for εA = 1.0 (opaque).   Note that 0.48 is the effective emissivity of an opaque emitter like a cloud or carport at the atmospheric temperature (TA) of 240 K.  According to the simple model, with very high levels of greenhouse cases εA would become one and the translucent atmosphere would become an opaque screen like a cloud with emissivity of 0.48.  The emissivities of low (ε = 0.68) and medium (ε = 0.55) cloud would be replaced by the more effective screen (ε = 0.48) provided by the IR-opaque atmosphere.  This would reduce overall ε from 0.615 to 0.452, resulting in a warming to 38º C.  Associated with higher temperatures would be higher levels of water vapor, which would mean more clouds.  If this cloud showed up as low or medium cloud, the cooling effect of increased albedo would reduce the warming effect of the greenhouse gases.  If increased cloud showed up as high cloud too, then both emissivity and albedo would be reduced, resulting in increasingly warmer temperatures all the way to the 52º C obtained earlier for 100% coverage by high cloud.  This example shows that even in an extreme case, this simple model correctly predicts that a Venus-like “runaway greenhouse effect” in which the oceans boil away simply can not happen on Earth no matter how high greenhouse gas levels get.  The temperatures predicted here are not predicted exactly, of course, but the qualitative result (they are below the boiling point) is valid.

As an exercise, let’s see what would happen to an Earth in Venus’s orbit.  The solar constant for Venus is 2614 watts m-2, much higher than Earth’s 1370 watts m−2, reflecting its closer proximity to the sun.  Let’s cover the Earth with maximum-cooling low cloud.  What would be the temperature be according to the simple model?  With A = 0.39, S = 2614, and ε = 0.68, equation 4 gives T = 319 K, or 46º C (115º F).  Since Venus actually is completely covered by clouds, this calculation seems reasonable enough and suggests that a Venus with liquid water and life could exist, although would be a pretty hot place by Earth standards.  This was the rationale for pre-1950’s science fiction stories that depicted a steamy hot swamp-like Venus. 

But how realistic is it for all the cloud to be low-lying cloud?  If here on Earth there is high cloud cover equal to 75% of low cloud cover, shouldn’t this be the case for a warmer Earth?  A 100% low-cloud covered Earth would most likely have at least 75% high cloud coverage as well, which would lower ε from 0.68 to around 0.4.  This would raise the temperature to 364 K (91º C), at which point water vapor would make up a substantial portion of the atmosphere.  With such a wet atmosphere, cloud cover would likely be complete at low, medium and high elevations.  With 100% high cloud cover, ε would be 0.33, and equation 4 gives the temperature as 382K (109º C).  The oceans can now boil, filling the atmosphere with water vapor. 

The mass of the oceans is more than 300 times that of the atmosphere.  Boiling away the oceans would increase the mass of the atmosphere enormously.  Let’s imagine a 100-fold increase, that is, surface atmospheric pressure rises to 100 times that on Earth (the actual surface atmospheric pressure on Venus is about 96 atmospheres so such large atmospheres can certainly exist on terrestrial-type planets).  Such a thick atmosphere would expand the elevation at which clouds can form.  On Earth, the 0.33 emissivity for high cloud provides a measure of the temperature of the atmosphere at which high cloud forms.  This temperature is 218 K because (218/288)4 = 0.33.  On Earth, atmospheric pressure in the troposphere (the thick portion of the atmosphere under 11 km elevation) roughly follows this relation:

17.              P  = PS(T/TS)5.26

Here the subscript s refers to surface.  I note that at the level of high cloud T/TS = ε¼.  With this I can write:

18.              PHC = PS∙ε1.32  =  1.0∙0.331.32  = 0.23 atmospheres

Thus, high cloud is associated with a pressure of about 0.23 atmospheres.  At pressures much below this, thick clouds apparently do not form, perhaps the air is too thin to keep ice crystals in suspension.  If I assume that high cloud occurs where PHC = 0.23, then I can use equation 18 to predict ε for high cloud for atmospheres thicker than the Earth’s.  The relation is:

19.              ε  =  (0.23/ PS)0.76  =  0.33 PS−0.76 

For PS = 100 atmospheres, ε becomes 0.01.  For extremely thick clouds, it is probably more appropriate to use the albedo for Venus itself, which is covered by thick cloud.  The albedo of Venus is 0.75.8  Using A = 0.75, ε = 0.01 and S = 2614 in equation 4 gives T = 733K (460º C)—hot enough to melt lead.  The actual average temperature of Venus8 is 464º C, remarkably close to this calculation.  A hot Earth in Venus’s orbit would gradually lose its water to photolysis by solar ultraviolet rays, producing hydrogen and oxygen.  The former would escape and the latter would oxidize sulfur compounds released by volcanism to sulfuric acid.  The heat would bake out CO2 from surface carbonates, maintaining a high atmospheric pressure despite loss of water by photolysis.  The end result would be a bone dry, extremely thick atmosphere composed mostly of CO2 with the high temperatures maintained by thick high clouds composed of sulfuric acid, which is exactly what Venus is today.  Thus, the simple model presented here can explain how an Earth in Venus’s orbit would become like Venus, while an Earth in Earth’s orbit will stay Earth-like.

More sophisticated models in which a more exact radiation balance is calculated for each wavelength of the spectrum can give more detailed predictions.  I will use these results, which are expressed in terms of forcings (see next section) in my subsequent discussion of the greenhouse effect. 

The concept of forcings

More physically realistic models do not provide a simple relation between greenhouse gases or cloudiness and effective emissivity like the simplistic model just presented.  Rather than expressing the effects of greenhouse gases on ε in equation 4, the practice is to treat the effect of greenhouse gases as forcings (F), effective changes in insolation (I), the average amount of energy reaching the top of the atmosphere.  These forcings produce a temperature effect ΔT.  The effect of a forcing on temperature can be determined from equation 4 by placing I + F and T + ΔT in for S and T and then subtracting the unmodified equation 4 as follows:

20.              ε σ (T + ΔT)4  =  I  + F   =  ε σ (T4 + 4T3ΔT + 6 T2ΔT2 + 4T ΔT3+ ΔT4) 

21.              ε σ (T)4            =  I

Subtracting equation 21 from 20 and dividing by T4 gives

22.              ε σ 4 [(ΔT/T) + 6 (ΔT/T)2 + 4T (ΔT/T)3+ (ΔT/T)4]  =  F / T4

If ΔT is small compared to T, the higher-power terms of (ΔT/T) on the left hand side of equation 19 will be so small that I can neglect them.  In this case equation 22 becomes:

23.              4 ε σ ΔT/T  = F / T4

Multiplying both sides by T4 and solving for ΔT gives

24.              ΔT   =  F / (4εσT3)  =  λ F  =  1/(4∙0.615∙5.67x10-8∙2883) ∙ F  = 0.3 F

The parameter λ depends on ε and T.  For T = 288 K and ε = 0.615, its value is 0.3.  A value for forcing can be obtained from the simple greenhouse model constructed earlier.  The radiation balance in equation 7 had a term, 2σεA∙TA4, reflecting radiation from the atmosphere into space and back to the surface.  The change in the half of this term going back to the surface is the greenhouse forcing.  Thus I can write:

25.              F   =  Δ(0.37∙σεA∙TA4) =  0.37∙5.67 x 10-8∙2404∙ΔεA = 69.3∙ ΔεA =  69.3∙(εABεA)  =  69.3 εA  30.8

Here the Δ symbol refers to the change in back radiation (the forcing) produced by the factor (greenhouse gases) under consideration.  The factor 0.37 is included because this forcing only applies to the 37% of the surface not covered by clouds.  The parameter BεA refers to the base value of εA, for T = 288 and CCO2 = 0.035%, which is equal to 0.4438, see the section above.  A forcing is the “pure” effect of a variable, without including any feedback effects from increased water vapor.  Thus, equation 14 with AH2O held constant at 0.4634 should be used to calculate the value of εA in equation 25: 

26.              F =  69.3∙[1− exp(−0.4634 − 3.52∙CCO2)] – 30.8  =  38.5  − 69.3∙exp(−0.4634 − 3.52∙CCO2)]  =  38.5 − 43.6∙exp(−3.52∙CCO2)

The CO2 forcing calculated by equation 26 is plotted in Figure 2.  Values for λ can obtained by dividing the ΔT value in Figure 2 by F.  Values around 0.45 are obtained.

Greenhouse gas forcings calculated from complex models are approximated by equations having a logarithmic form:

27.              F  =  5.35∙ln[CCO2/BCCO2]

Here BCCO2 is the base CCO2 level of 0.035%.  The values from equation 26 obtained for CO2 levels between 280 and 560 ppm, which includes the entire range of interest, can be fit to an equation of the form given in equation 27 with K = 5.4.  This close match is deliberate, as the water-CO2 split of 79%:21% was chosen to make the forcing behavior accurately reflect the forcing results obtained from more sophisticated models.9  Unlike equation 26, forcings obtained from equation 27 with K = 5.359 can be considered as accurate assessments of the direct effect of CO2 on climate forcing.  Of course, climate forcing isn’t the result I desire, temperature is.  And to obtain temperature from a forcing, a value of λ is needed.  As shown above, the simple model provides a value of 0.45 for λ when effects of water vapor were included.  This value is different from the 0.3 value determined above. The 0.3 value for λ applies to effects of forcings without any considering any other effects.  For example, a doubling of CO2 levels will produce a forcing of 3.7 = 5.35∙ln(2) watts/m2 according to equation 27.  The effect of CO2 all by itself would be 0.3∙3.7 = 1.1º C.  The simple model gives the same result if water vapor content is held constant.  But with rising temperature caused by CO2, there will be rising H2O, which will exert its own effect on temperature.  This secondary effect from H2O is called a feedback effect.  By ignoring cloud formation and assuming constant relative humidity to account for water-vapor greenhouse effects, the simple model gives a value of 0.45 for λ.  More sophisticated models give different values depending on how they handle the greenhouse effects of water vapor.  The consensus value for λ from these models is 0.5.  For λ = 0.5 I can write

28.              ΔT  = 2.68∙ln[CCO2/BCCO2]

Figure 3.  Temperature relative to 1961-90 compared CO2 forcing (data from refs 10-15)

Figure 3 shows a plot of  equation 28 as “CO2 forcing”.  Also shown is a plot of global temperature.  Both plots are shown in terms of deviations from a 1961-1990 basis period.  That is, the temperatures are temperatures relative to the average over these years and BCCO2 is the average CO2 level during these years.  The CO2 forcing shows a gradual rise that accelerates after 1970, which is roughly consistent with rising temperature.  The extent of the rise is about the same, but the monotonic nature of the CO2 forcing model fails to account for periods of flat and declining temperature in the temperature trend.  Clearly, more is needed to explain the trends in global temperature since 1850 than just the greenhouse effect.

Sun-related factors

Another obvious effecter of climate is the sun. The sun exerts two types of effects.  The first effect is through the solar constant S as described in equation 4.  Solar insolation (I) is directly related to S:

29.              I =  ¼ S (1-A)  = 0.175 S  à F = ΔI  =  0.175 ΔS

Figure 4 shows a plot of the solar constant over the past three decades.  The solar constant is not actually constant.  Rather, it shows a 1 watt/m2 cyclical variation that is closely correlated with the 11-year sunspot cycle.  Equation 29 shows that this cyclical fluctuation produces a cyclical forcing of 0.18 watt / m2.  Multiplication by λ = 0.5 gives a temperature effect of 0.09º C.

Figure 4.  Total solar irradiance (solar constant) since 1976 (data from refs 16-18)

Figure 5 shows a plot of the global temperature since 1850.  The grey circles and line shows the annual values for global temperature relative to the 1961-1990 average.  The heavy black line shows the trend obtained from a moving 20-year linear regression analysis.  The actual plot fluctuates around the trend value.  The parallel black lines show the “track” within which the temperature remains approximately 82% of the time. That is, it is the value met or exceeded roughly twice every 11 years (top and bottom of each 11-year solar cycle).  The distance between these lines is 0.24º C.  It reflects the size of the short-term fluctuations in global temperature.  Note that the monotonic nature of the CO2 forcing means that CO2 cannot explain these temperature fluctuations.  The existence of them is proof of the operation of important factors other than CO2 in global temperature change.  An obvious question is whether some of these fluctuations could be explained by solar effects.  At most, 0.09º C of this fluctuation can be caused by the cyclical fluctuations in solar constant shown in Figure 4.  Apparently there is more involved in these fluctuations than can be explained by fluctuations in the solar constant.

Figure 5.  Global temperature relative to 1961-1990 average (data from refs 11-15)

Another solar effect on climate: the cosmic ray hypothesis

Danish physicist Henrik Svensmark has proposed that the sun exerts an indirect effect on climate that amplifies the direct effect of solar irradiance (see Figure 6).  Solar wind modulation by the sun’s magnetic field varies with the solar cycle (as measured by sunspot number).  When solar activity is high (lots of sunspots) the solar magnetic field strength is high and vice versa.  The solar magnetosphere shields the Earth from galactic cosmic rays that enter the solar system from distant sources.  As a result, cosmic ray flux on Earth varies inversely with solar magnetosphere strength and with sunspots.  Cosmic rays produce ionization in the lower atmosphere, which leads to the formulation of tiny particulates that serve as cloud condensation nuclei resulting in more cloudiness.  Cloudiness affects climate as we have seen earlier.

Figure 6.  Cosmic ray climate mechanism (figure from ref 19)

Figure 7 shows a plot of the solar cycle defined by sunspots and total solar irradiance (TSI).  Also shown is the negative of cosmic ray flux (−CRF).  The plots are shown as differences from their average/trend values divided by their standard deviations.  This puts the disparate data on the same scale and allows a good comparison of the cycles shown by all three parameters.  All three variables show closely aligned cycles.  Temperature deviations from the trend shown in Figure 5 are also plotted.  A three-year moving average is used to smooth the temperature data.  An apparent correspondence can be seen between temperature and the solar cycle as defined by sunspots, TSI or –CSF.  The temperature oscillation associated with the solar cycle shown in Figure 7 is 10 years long on average with 0.15º C average magnitude.  The size of these oscillations is too large for TSI to be the sole cause of them.  The cosmic ray mechanism provides another way for the sun to exert a cyclical effect on temperature.

Figure 7.  Cycles in Temperature, TSI, Sunspots and −CRF (sunspot data from ref 20, CRF data from refs 21,22)

Figure 8 shows a plot of low cloud cover versus cosmic ray flux (CRF).  A good correspondence is evident.  The correlations in Figures 7 and 8 show that low cloud cover should be inversely proportional to TSI and sunspot number.  Because low cloud cover exerts a cooling effect, CRF also exerts a cooling effect.  This means than −CRF should exert a warming effect, which is what was observed in Figure 7.  That is, the CRF-mediated cloud effect should add to the TSI cycle, effectively amplifying it.  This second effect, in tandem with TSI, is a likely explanation for the cyclical temperature fluctuation aligned with the solar cycle.

Figure 8.  Cosmic rays versus low cloud cover (from reference 23)

Measuring the size of the cosmic ray effect: estimation using Figure 8 and the simple model

Figure 8 indicates that the solar cycle shows an average change in low cloud cover over a cycle of about 1.6%, say from 25.2% to 26.8%, which implies a change of fraction cloudless from 37.8% to 36.2%.  Equation 6 becomes:

30.              ε   =  0.3494 + 0.354 εT

Albedo changes by +0.004 = 1.6%∙(0.39-0.14), from which solar insolation decreases from 240 to 240∙(1-0.304)/(1-0.30) =  238.6 watts m-2.  This value is substituted for 240 in equation 4 to give

31.              T = (238.6/εσ)¼

The procedure is as before.  Here CCO2 = 0.035 and equation 16 gives εA as a function of temperature.  Equation 10 gives εT from εA, from which T is calculated by equations 30 and 31.  The trial and error solution is a 0.44º C decrease in temperature due to the increased cloud cover.  Added to this is the effect of TSI, which I have previously calculated as 0.09º C to yield a total solar cycle effect of 0.53º C.  We have seen that the actual size of cyclical fluctuations is 0.15º C, less than a third this value.

Measuring the size of the cosmic ray effect: empirical approach from past excursions in CRF

Another approach to measuring the size of the cosmic ray effect is to use historical “natural experiments” in which cosmic rays were known to have varied and temperatures with them.  Once such natural experiment is shown in Figure 9 when a very long-term variation in cosmic rays was impressed upon the Earth’s climate by the solar system’s periodic passage through the galaxy’s spiral arms, which are rich in cosmic ray sources.  The result of this movement is a 185 million year cycle in cosmic rays during which cosmic rays fluctuate by about 80% of their current magnitude.  Associated with this cycle is a temperature cycle of average magnitude 2.8º C.  This suggests a 0.035º C temperature effect per 1% change in cosmic ray flux.

Figure 9.  Cosmic ray flux versus temperature over a very long time scale (fig from ref 24)

Using this sensitivity, the effect of CRF on temperature could in principle be calculated in terms of the deviation from its modern value.  CRF data is only available beginning in 1951 (Figure 10).  These data cannot be used directly with the 0.035º C/percent temperature effect of CRF obtained from Figure 9 because different measurements of CRF show different magnitude changes (see Figure 10).  Data from the Haleakala-Huancayo detectors show about a 5% variation per solar cycle suggesting a 0.17º C temperature effect due to the cosmic ray variations of  the solar cycle.  On the other hand, the Climax detectors show a 16% variation suggesting a 0.56º C temperature effect.  To these values the TSI-mediated effect of 0.09º C must be added to give estimates of 0.26º and 0.65º C.  The theoretical value of 0.53º C falls in between these two values.

Figure 10.  Cosmic Ray Flux from various detectors (data from refs. 20-21).

Using proxies to predict historical solar effect

When the data in Figure 10 are normalized by plotting them in terms of standard deviations, the CRF data from the different locations correspond very well.  Indeed, Figure 7 showed CRF data from all three detectors expressed in a normalized fashion and averaged into a single composite profile.   Figure 7 also showed the good correlation between CRF and sunspot number.  It should be possible to extend CRF back before 1950 using this correlation.  Figure 11 shows an extension of Figure 7 backwards in time.  Taking the place of TSI is the total open solar magnetic flux (OSF) obtained from a paper by Lockwood.25  This paper gives OSF in terms of a running 11-year average.  To be consistent, sunspot number and CRF are presented as 11-year moving averages also.  The data are scaled in the same way as in Figure 7, the values are the differences from the average value over 1951-2000 and divided by the standard deviation over the same period.

Figure 11. Trends in SSN20, OSF25, and –CRF21,22 expressed in terms of standard deviations from 1950-2000 mean

Figure 11 shows that solar activity as measured by sunspots or OSF rose significantly during the first half of the 20th century.  Global warming skeptics often point out that solar activity is at record levels so temperatures should be.  They are right, solar activity is very high today, but it has been very high for 50 years.  The major rise in solar activity that resulted in the current high levels of solar activity occurred in the first half of the 20th century, during which temperature rose as well (see Figure 5).  The simultaneous rise in both temperature and solar activity strongly suggests that solar activity is responsible for the early 20th century temperature rise (but not the more recent post-1970 temperature rise).  I desire to calculate the magnitude of this solar effect to verify the hypothesis that solar factors produced the early 20th century rise in global temperature.

Figure 11 shows that the size of the solar activity rise during the first half of the 20th century is measured at 4.5-6 standard deviations, depending on the measure (sunspots vs. OSF) used.  These standard deviations are for 11-year averages.  The standard deviation of an average is the square root of N times smaller than that of the raw data, where N is the number of things being averaged.  The standard deviations in Figure 11 are smaller than those in Figure 7 by a favor of 3.32, the square root of 11.  The amplitude of an 11-year cycle of annual data points is 2.6 standard deviations.  This follows because two points out of every 11 (the cycle maximum and minimum) define the cycle amplitude, while the other 9 (82%) fall closer to the trend than the cycle amplitude.  On a normal bell curve, 82% of the area under the curve lies within ±1.3 standard deviations of the average (trend) value.  Thus, a standard deviation in Figure 7 is 2.6 times smaller than the cycle amplitude.  Since the standard deviations in Figure 11 are 3.32 times smaller than the standard deviations in Figure 7, they are also 8.62 (=3.32x2.6) times smaller than the cycle amplitude.  Given a cycle size, one divides by 8.62 to obtain the temperature to standard deviation calibration needed to convert Figure 11 into a temperature effect.

So far I have four estimates for the size of the solar-driven temperature cycle, which range from 0.15º to 0.65º C.  It is tempting to go with the 0.15º C value from Figure 7, since that is a “real observed effect”.  Unfortunately if one looks as a continuation of Figure 7 back to 1900, one sees that over this time the neat correspondence between temperature and sunspots is no longer evident.

Figure 12.  Temperature fluctuations (3 YMA) compared to the sunspot cycle 1900-1960

Other factors such as the el Niño Southern Oscillation (ENSO) give rise to significant departures from the temperature trend and can obscure the effect of cyclical forcings produced by the solar cycle.  A more sophisticated analysis is needed to elucidate the solar-related temperature cycles from ENSO-type effects.  Values of 0.09 to 0.11º C for the size of the solar cycle temperature effect have been reported in the literature.26-8  A consensus value of 0.10º C will be used here.  This value is quite close to the 0.09º C cycle size consistent with the TSI-driven cycle.  If the solar cycle is as small as 0.1º C, there is no room for a CRF-mediated solar effect on temperature.  This finding contradicts the evidence in Figure 8, from which a solar cycle effect of around 0.5º C was estimated using the simple model.  It ignores the direct evidence shown in Figure 9 from which a solar cycle effect of 0.2-0.65º C was estimated.  Finally a 0.1º C solar cycle effect means a temperature calibration of 0.012º C (=0.1/8.62) per standard deviation in Figure 11.  With this calibration, the 4.5-6 standard deviation rise in the early 20th century would amount to 0.05-0.07º C, too small to explain the 0.3-0.35º C temperature rise then, even with the ~0.1º C contribution from the CO2 greenhouse effect added on.  The solar effect needs to be 3-4 times larger in order to explain the early 20th century temperature rise.

There is a discrepancy between the empirically-observed temperature effect of the solar cycle of 0.1º C and the 0.2-0.65º C size consistent from the CRF mechanism and the early 20th century temperature rise.  This issue can be resolved by noting that the apparent size of the temperature effect from an 11-year cyclical forcing will be depressed if the response between the forcing and the temperature effect is delayed.  Such a delayed response serves to reduce the size of the temperature response.  It acts like a filter or moving average to smooth out the cyclical fluctuations.  The more smoothing, the smaller the output relative to the input forcing.  I define the damping factor as the ratio between what the output would be in the absence of damping and the actual result.  It varies from 1 (no damping) to infinity (completely damped).  Appendix A discusses the approach I took to estimating this damping factor.  Based on my result and a number of literature results a best guess value of 3.3 was obtained.  Thus, the temperature cycle caused by the solar cycle would be about 0.33º C without any damping.  This empirical value was averaged with the other three cycle size estimates (0.22, 0.53 and 0.65) to obtain a value of 0.43±0.10º C for the size of the solar cycle for use in the next section.

Use of cyclical amplitude and Figure 11 to calculate past solar effect on temperature

I now have means to convert the profiles shown in Figure 11 in terms of standard deviations (of 11-year averages) into temperatures.  With a cycle size of 0.43º C, dividing by 8.62 yields 0.05º C per standard deviation as given in Figure 11.  The sunspot and OSF data in Figure 11 were put in terms of temperature by multiplying them by 0.05.  The two trends were averaged together to obtain a composite solar forcing estimate and a trend line obtained from this composite.  This trend is plotted in Figure 13 as the red line.  Equation 28 was used to calculate the CO2 effect on temperature with a 1950 basis year.  A value of 0.5 was assumed for λ.  This value was added to the red line to obtain the effects of solar and CO2 combined, which is also plotted in Figure 13 as the blue line.

Figure 13.  Temperature trend compared to solar forcing with added CO2 forcing

Shown for comparison is the actual temperature trend line (also adjusted to a 1950 basis year) in black.  Solar activity can account for nearly all of the early 20th century warming, but essentially none of the recent warming.  When the effect of CO2 is added (with λ = 0.5) the recent warming can be explained, but the early 20th century warming predicted by the two forcings is larger than the actual amount observed.  I have not yet considered the third major effecter of climate, aerosols, which tend to exert a cooling effect.  Natural aerosols, which mostly come from volcanic eruptions, do not show a trend over time, but aerosols produced by human industrial activity do show a rising trend until the late 20th century when pollution control efforts have reduced their growth.  A model taking into account the rising trend in aerosols would show a smaller overall temperature rise over the 20th century, which should improve the fit of the solar+CO2 model shown in Figure 13.  In the next section the effect of aerosols is discussed.

Effect of aerosols

Aerosols, tiny solid particles suspended in air act that act like they are dissolved (hence aero = air + sol = solution), affect climate in a number of ways.  They can exert direct effects by reflecting sunlight or re-radiating surface radiation like the water droplets in clouds do.  They can also exert indirect effects by serving as cloud condensation nuclei, much like cosmic-ray produced ions do in the cosmic ray-mediated solar effect.  In both their direct and indirect effects, aerosols can exert warming or cooling effects. 

Figure 14.  Emissions of aerosols: anthropogenic carbon29 and sulfur dioxide30 emissions plus volcanic sulfates31-32

Figure 14 shows a plot of aerosol releases into the atmosphere.  Three human-generated sources and one natural source are shown.  The human sources are black carbon (BC) or soot, organic carbon (OC) and sulfur dioxide (SO2).  The natural source is volcanic sulfate deposits from the North Pacific region.31  I could not find global data, and since the net effect of most of this data was a low-level “noise factor” the exact values from year to year are not really very important for my purposes.  What is important is the effect of “big” volcanic eruptions like that of Mt. Pinatubo in 1992.  Using a list of major eruptions which described their relative size,30 I was able to use the 1912 eruption of Katmai, which appears as a huge spike in the North Pacific data, to estimate values for major eruptions in 1883 (Krakatoa) and 1992 (Pinatubo).

Sulfur dioxide (SO2) itself is a gas, not an aerosol.  It dissolves in water to form sulfurous acid (H2SO3) which reacts with oxygen in the atmosphere to produce sulfuric acid (H2SO4).  Much of this sulfuric acid falls as acid rain but small droplets of acid solution can dry to small acid particles that function as an aerosol.  Because not all SO2 goes to form aerosols, the level of SO2 emissions cannot be compared directly to the carbon-type aerosols also shown in Figure 14.  That is, human-generated sulfate-type aerosols are not necessarily several times more important than carbon-type aerosols as the relative quantities emitted suggest.  Also, although the volcano plot in Figure 17 is based on measurements of sulfates, these values are not directly comparable to human SO2 emissions.  If they were, even major volcanoes would exert a negligible effect on climate compared to human activities, which is known not to be true.  For this reason the volcanic sulfates are plotted on the same scale as the carbon particulates. 

The effects of aerosols are so complex it is not possible to give even a grossly simple model to account for their effects.  All that can be said about them is the direction of their effects.  BC produces a warming effect; OC and sulfates produce cooling effects; and the cooling effects dominate.  Since the human-generated carbon and sulfur dioxide trends are roughly similar it is unnecessary to include all three factors into a model.  If I simply assume that the cooling due to human SO2 additions offsets warming produced by BC this leaves the OC trend to explain the net cooling produced by human aerosol additions.  The volcano aerosol contribution can be added to the human contribution to give total effective aerosols (TA):

32.              TA = OC  + α∙V

Here V refers to the volcanic sulfate emissions.  The factor α reflects the importance of volcanic aerosols relative to human aerosols; it is an adjustable parameter.  The effect of TA on temperature is negative and proportional, with another adjustable parameter (γ) describing the relative importance of aerosols on temperature as compared to solar and greenhouse gases.  From this I can write the following expression for temperature:

33.              T = solar effect  +  λ∙5.35∙ln(CO2/CO20)  − γ∙(TA-TA0)

In this case the subscript 0 stands for the values of CO2 and TA in the 1950 base year used in Figure 13.  The introduction of human-generated aerosols will produce a net cooling effect reducing the overall rise in temperature predicted by the model.  How important these human aerosols are depends on the values of both α and γ.  If α is large then natural aerosols will dominate over human aerosols and the term (TA-TA0) will average close to zero.  Since I am interested in trends, this average value is what matters.  Thus if volcanic emissions dominate, the situation will be like that in Figure 13, except random factors from volcanism could obscure the trends.  In any case adding aerosols to the model would accomplish nothing.  The more interesting case is when volcanic aerosols are about the same size as human emissions.  In this case the high level of volcanism early in the twentieth century (see volcanic trend in Figure 14) helps explain the low temperatures of that time, while the rising level of human-produced aerosols serves to explain declining temperatures in the 1950’s and 1960’s, neither of which was well-explained by the solar+greenhouse model shown in Figure 13.  I tried varying α between 0.5 and 2 and saw little qualitative effect on how well the model represents the facts, so I set α at 1.0.

Figure 15.  Effect of λ on best fit obtained when aerosol effects are included

With this value for α, the model contains a single adjustable parameter γ for a given value of λ.  Figure 15 shows the model represented by equation 33 for three values of λ.  In each, γ was adjusted to obtain a best fit of the temperature data.  The fit obtained with these values varied between 0.044º C and 0.053º C average error.  The best fit was obtained with λ = 0.87.  Table 4 summarizes the results obtained when various combinations of parameters were used to fit the temperature data.  A number of interesting observations can be made.  Note that aerosols by themselves had no ability to fit the temperature data.  The best fit was obtained with zero impact of aerosols (i.e. a constant relative temperature of 0º C).  This makes sense because the effect of aerosols over time is negative (cooling) while the temperature trend is positive (warming).  Adding aerosols to the solar effect gave no improvement to fit at all.  Once again this makes sense since the total temperature rise predicted by just the solar effect is less than what has occurred.  Obviously adding a negative trend will not improve the fit.

Table 4 Model fits obtained with various conditions

Parameters used

Best fit parameter values

Average error (º C)

Just solar

λ = 0, γ = 0

0.083

Just greenhouse effect

λ = 0.5, γ = 0

0.068

Just aerosols

λ = 0, γ = 0

0.177

Solar + greenhouse

λ = 13, γ = 0

0.072

Solar + aerosols

λ = 0, γ = 0

0.083

Greenhouse + aerosols

λ = 0.54, γ = 0.01

0.067

Solar + greenhouse + aerosols

λ = 0.87, γ = 0.087

0.044

Of considerable interest are the two factor models involving the greenhouse gases.  Combining aerosols with greenhouse effect did not improve the fit significantly and combination with solar gave a worse fit.  That is, adding the effect of either solar or aerosol forcings did not significantly improve the fit obtained with greenhouse gases alone (that is Figure 5).  Yet it was apparent that the greenhouse-alone fit is inadequate (see Figure 5).  It is only when all three parameters are combined that results obtained that are significantly better than those obtained with a one-forcing model.  The improvement is substantial—a 35% reduction in the average error.  Noteworthy is the fact that both CRF-mediated solar effects and greenhouse effects in combination with aerosols must be considered.  Either one by itself doesn’t do the job.

The reason for this is the “two-step” shape of the temperature rise.  Greenhouse effects are too weak to explain the early 20th century temperature rise unless large values of λ are used.  But then the late 20th century rise is too large.  Adding in aerosols cannot help the fit since they show a cooling trend.  Only when solar effects are added, which add an extra temperature gain in the early 20th century without one in the late 20th century can the fit be improved.  What this says is that the CRF-mediated solar effect is necessary to adequately explain the 20th century temperature rise (as skeptics maintain) but inclusion of this mechanism makes it even more clear that recent warming reflects the greenhouse effect.  That is, if the CRF-mediated solar effect is considered, it makes the case for the greenhouse effect as the dominant factor in recent warming even stronger.

What do higher values of λ mean?

In the section on forcings, I showed that for pure forcings the value for λ is 0.3.  When the feedback effect of water vapor was added to the CO2 effect in the simple model the results obtained were consistent with a λ value of 0.45.  The consensus value amongst sophisticated models is 0.5.  Now Figure 15 suggests that to explain recent warming still higher values for λ (~0.87) are needed to fit observed temperature data well.  What sort of mechanisms are consistent with such high values of λ?

As noted earlier, increased water vapor does not just exert a greenhouse effect.  It will likely increase cloud cover.  The reason is that cloud formation depends on two factors, the availability of water vapor whose condensation forms the droplets making up clouds and the availability of cloud condensation nuclei that serve as catalysts for condensation.  When water vapor is available in excess, then it is the amount of nuclei that limits cloud formation.  When nuclei are in excess supply, then the amount of water vapor limits cloud formation. 

The cosmic ray effect discussed earlier operates through control of the amount of nuclei.  When CRF is high there is increased supply of nuclei and so more cloud.  But this effect only happens for low cloud. High cloud is unaffected by CRF.6  Apparently nuclei are abundant for high cloud, cloud formation is limited by water availability, and changes in levels of nuclei don’t affect cloud formation.  This makes sense because even small amounts of cosmic rays will produce large amounts of ions high in the atmosphere.  It is only near the surface, where few cosmic rays can penetrate, that ions can be in short supply and so cloud formation can be responsive to variations in CRF.  What this means is high cloud should be responsive to increases in water vapor levels from temperature rises produced by other forcings.

Similarly, the fact that the CRF effect happens with low cloud suggests that water vapor isn’t the limiting factor in low cloud formation.  Increasing water vapor should not affect low cloud much.  What this means is the most likely cloud-related feedback effect from rising water vapor would be increases in high cloud cover—which is warming.  That is, higher values for λ make sense.  To explore this idea, I employed the simple model to calculate the effects of a direct proportional relation between increased water vapor and increased high cloud.  That is, a 1% increase in water vapor was assumed to produce a 1% increase in high cloud.  This situation is expressed by the following equation:

34.              Percent high cloud cover = %HC = 19%∙exp(18.382-5294/T)

Equation 16 is used to calculate εA as a function of T and %CO2.  Equation 10 is used to calculate εT.  Overall ε is calculated as was done in Table 3:

35.              ε  =  26%∙εLC + 18%∙εMC + %HC∙εHC + (56% − %HC)∙εT 

Here εLC, εMC, and εHC are the emissivities for low, medium and high cloud, respectively, that is 0.68, 0.55 and 0.33.  Changes in %HC also affect albedo:

36.              A = 0.39∙(0.44 + %HC) + 0.14∙(56% − %HC)

With values for ε and A obtained from equations 34-36, T is calculated from equation 4 and compared to the value of T used in equations 16 and 34.  Agreement is reached by trial and error as previously done. This procedure was done for CO2 levels of 280 and 560 ppm.  The temperature increase obtained was 2.8º C, which is remarkably close to the consensus value of 2.9º C for the effect of doubling CO2, considering all feedbacks.33   This compares to the value of 1.7º C obtained earlier by considering just the greenhouse effect of water vapor or a 1.1º C obtained from just considering the effects of CO2 by itself.  All three of these values reflect the same CO2 forcing of 3.7 watts m-2, but with different feedback effects assumed.  Each assumption gives rise to a different value for λ: 0.3 for no feedback, 0.45 for just the water vapor greenhouse effect and 0.76 for the greenhouse effect of water vapor plus the effect on high cloud.  For comparison, the consensus values for the latter two values are 0.5 and 0.78.   This last value is reasonably close to the 0.87 value I obtained from the model fit in Figure 15.

Note that this entire discussion of λ is being applied to just the CO2 forcing.  Lambda actually applies to all forcings.  For solar forcings, however, the temperature effect was mostly determined empirically by considering the size of the temperature cycle associated with the solar cycle and from ancient temperature/CRF response data.  The impact of λ was built into the estimate.  For aerosols no effort at calculating a forcing was made.  The size of the aerosol effect was simply fit to the temperature record and so any feedback effects of water vapor on λ were included.  Only for CO2 is λ relevant since an accurate forcing can be calculated for CO2, from which a temperature effect can be obtained by multiplication by λ.  The real effect of CO2 will include all feedbacks and so unless the feedbacks all cancel each other out, λ will almost certainly not be equal to 0.3 (the value with no feedbacks).  Table 4 shows that when just greenhouse factors are considered, a value for λ of about 0.5 is required to explain recent warming.  Adding in a crude representation of the aerosol effect suggests higher levels of λ may be needed to explain recent warming.  This means that the feedback effects of higher CO2 on cloudiness are likely to be warming and the effects of a doubling of CO2 are likely to be closer to 3º C than 1º C as argued by skeptics.

Conclusion

CO2-mediated greenhouse warming is the dominant mechanism responsible for global warming over the past three decades.  If present CO2 output trends continue atmospheric CO2 could more than double from its current level by the end of this century.  A doubling in CO2 level would produce a forcing of about 3.7 watts m-2 according to equation 27.  With λ = 0.76, the value obtained with the simple model considering all feedbacks, the predicted temperature increase would be 2.8º C.  This value falls comfortably within the 2.2-3.9º C value predicted from more sophisticated models.33

References

1.      Emissivity of natural materials (www.icess.ucsb.edu/modis/EMIS/html/em.html)

2.      Emissivity of seawater (www.icess.ucsb.edu/modis/EMIS/html/seawater.html)

3.      Jin, Menglin and Shunlin Liang, (2006) “An Improved Land Surface Emissivity Parameter for Land Surface Models Using Global Remote Sensing Observations” Journal of Climate, 19 2867-81. (www.glue.umd.edu/~sliang/papers/Jin2006.emissivity.pdf)

4.      Shippert, T. R, S. A. Clough, P. D. Brown, W. L. Smith, R. O. Knuteson and S. A. Ackerman (1998) “Spectral Cloud Emissivities from LBLRTM/AERI QME” Proceedings of the Eighth Atmospheric Radiation Measurement (ARM) Science Team Meeting, 688-92. (www.arm.gov/publications/proceedings/conf08/extended_abs/shippert_tr.pdf)

5.      Kirkby, J, (2002) “Cloud: A particle beam facility to investigate the effect of cosmic rays on clouds”, Proc. IACI Workshop (http://cloud.web.cern.ch/cloud/documents_cloud/kirkby_iaci.pdf)

6.      Mlynczak, M.G. D.G. Johnson, and J. I. Applin (2002) “The Far-Infrared Spectroscopy of the Troposphere (FIRST) Project” NASA Earth Science Technology Conference (http://esto.nasa.gov/conferences/estc-2002/Papers/B4P2(Mlynczak).pdf)

7.      Schmidt, G. A., “Water vapour: feedback or forcing?”, RealClimate (www.realclimate.org/index.php/archives/2005/04/water-vapour-feedback-or-forcing/)

8.      http://geosci.uchicago.edu/~rtp1/ClimateBook/UtilityScripts/planets.py

9.      Intergovernmental Panel on Climate Change, Climate Change 2001, Cambridge Univ. Press, New York , 2001. (www.grida.no/climate/ipcc_tar/wg1/222.htm#635)

10.  A. Robertson1, J. Overpeck, D. Rind, E. Mosley-Thompson, G. Zielinski, J. Lean, D. Koch, J. Penner, I. Tegen, and R. Healy (2001) "Hypothesized Climate Forcing Time Series for the Last 500 Years", Journal of Geophysical Research, 106,(D14) 14783-14803 (www.ncdc.noaa.gov/paleo/pubs/robertson2001/robertson2001.html)

11.  Climatic Research Unit  School of Environmental Sciences University of East Anglia.  HADCRUT3v global temperature series (www.cru.uea.ac.uk/cru/data/temperature/#sciref).  Based on data from references 8-11.

12.  Brohan, P., J.J. Kennedy, I. Haris, S.F.B. Tett and P.D. Jones, 2006: Uncertainty estimates in regional and global observed temperature changes: a new dataset from 1850. J. Geophysical Research 111, D12106.

13.  Jones, P.D., New, M., Parker, D.E., Martin, S. and Rigor, I.G., 1999: Surface air temperature and its variations over the last 150 years. Reviews of Geophysics 37, 173-199.

14.  Rayner, N.A., P. Brohan, D.E. Parker, C.K. Folland, J.J. Kennedy, M. Vanicek, T. Ansell and S.F.B. Tett, 2006: Improved analyses of changes and uncertainties in marine temperature measured in situ since the mid-nineteenth century: the HadSST2 dataset. J. Climate, 19, 446-469.

15.  Rayner, N.A., Parker, D.E., Horton, E.B., Folland, C.K., Alexander, L.V, Rowell, D.P., Kent, E.C. and Kaplan, A., 2003: Globally complete analyses of sea surface temperature, sea ice and night marine air temperature, 1871-2000. J. Geophysical Research 108, 4407

16.  C.Frohlich, 2000, "Observations of Irradiance Variations”, Space Science Rev. 94, 15-24

17.  C. Frohlich, 2006, "Solar Irradiance Variability Since 1978: Revision of the {PMOD} Composite During Solar Cycle 21", Space Science Rev. in press

18.  Dataset (version d41_61_0702) obtained from PMOD/WRC, Davos, Switzerland.  It includes unpublished data from the VIRGO Experiment on the cooperative  ESA/NASA Mission SoHO

19.  Shaviv, Nir J, Science Bits blog http://www.sciencebits.com/ice-ages

20.  National Geophysical Data Center, Group Sunspot Numbers (Doug Hoyt re-evaluation) (www.ngdc.noaa.gov/stp/SOLAR/ftpsunspotnumber.html)

21.  Space Physics Data System, University of New Hampshire, Neutron Monitor Datasets, (http://ulysses.sr.unh.edu/NeutronMonitor)

22.  Moscow Neutron Monitor (http://helios.izmiran.rssi.ru/cosray/main.htm)

23.  Marsh, N., H. Svensmark, (2003) “Galactic cosmic ray and El Niño-Southern Oscillation trends in International Satellite Cloud Climatology Project D2 low-cloud properties”, Journal of Geophysical Research, 108(D6), pp. AAC 6-1.

24.  N. Shaviv & J. Veizer, "A Celestial driver of Phanerozoic Climate?", GSA Today 13, No. 7, 4, 2003.

25.  M. Lockwood, (2002) “Long-Term Variation in the Open Solar Flux and Possible Links to Earth’s Climate”, Proc. SoHO 11 Symposium ESA-SP-508, pp 507- 522.

26.  Douglass, D.H., and B.D. Clader (2002) “Climate sensitivity of the earth to solar irradiance”, Geophysical Research Letters, 29(16), 1786.

27.  White, W.B., J. Lean, D.R. Cayan, and M.D. Dettinger (1997) “Response of global upper ocean temperature to changing solar irradiance”, Journal of Geophysical Research 102(C2), 3255–3266.

28.  Shaviv, N. J. (2005) “On climate response to changes in the cosmic ray flux and radiative budget”, Journal of Geophysical Research, 110, A08105.

29.  Junker, C. and C. Liousse (2006) “A global emission inventory of carbonaceous aerosol from historic records of fossil fuel and biofuel consumption for the period 1860–1997”, Atmos. Chem. Phys. Discuss., 6, 4897–4927 (www.atmos-chem-phys-discuss.net/6/4897/2006/acpd-6-4897-2006.pdf)

30.  Netherlands Environmental Assessment Agency, Sulfur Dioxide Emissions (www.mnp.nl/hyde/emission/SO2/)

31.  Briffa K.R., P.D. Jones, F.H. Schweingruber and T.J. Osborn (1998) “Influence of volcanic eruptions on Northern Hemisphere summer temperature over the past 600 years”, Nature 393, 450-455.

32.  Yalcin, K., C.P. Wake, K.J. Kreutz, M.S. Germani, and S.I. Whitlow. (2007)  "Ice core paleovolcanic records from the St. Elias Mountains, Yukon, Canada", Journal of  Geophysical. Research, 112, in press.
Data from: NOAA/NCDC Paleoclimatology Program, Boulder CO, Eclipse Icefield Volcanic Aerosol Reconstruction. IGBP PAGES/World Data Center for Paleoclimatology, Data Contribution Series # 2007-040.
(ftp://ftp.ncdc.noaa.gov/pub/data/paleo/icecore/trop/eclipse/eclipse-volcanic.xls)

33.  Annan, J.D. and J.C. Hargreaves, “Using multiple observationally-based constraints to estimate climate sensitivity”, Draft paper, 2006.
(www.jamstec.go.jp/frcgc/research/d5/jdannan/GRL_sensitivity.pdf)