Do clouds control climate?

Clouds have a net average cooling effect on the earth’s climate. Climate models assume that changes in cloud cover are a feedback response to CO2 warming. Is this assumption valid? Following a study with Euan Mearns showing a strong correlation in UK temperatures with clouds, we  looked at the global effects of clouds by developing a combined cloud and CO2 forcing model to sudy how variations in both cloud cover [8] and CO2 [14] data affect global temperature anomalies between 1983 and 2008. The model as described below gives a good fit to HADCRUT4 data with a Transient Climate Response (TCR )= 1.6±0.3°C. The 17-year hiatus in warming can then be explained as resulting from a stabilization in global cloud cover since 1998.  An excel spreadsheet implementing the model as described below can be downloaded from http://clivebest.com/GCC

Best fit  to data with TCR=1.4C

Best fit(Acalc)  to data(H4)  using  TCR=1.4C

A basic uncertainty for climate science is in understanding the net effect of clouds on the radiative balance of the earth [3].  Clouds regulate solar heating by increasing the planet’s albedo while simultaneously absorbing infrared (IR) from the surface. This interplay between albedo and greenhouse effect of clouds is complex and varies with latitude and with season. The net radiative forcing from cloudy regions is So(1-α) –F, where F is the outgoing IR and α is cloud albedo. On a global scale the Earth Radiation Budget Experiment (ERBE) measurements have shown a net cooling of around -13 watts/m2, which is four times that expected from a doubling of CO2 alone [3].  However, more recent measurements from the Clouds and the Earth’s Radiant Energy System CERES [5] show that the net average cooling effect of clouds is larger (-21 W/m2) (Figure 1b). It is often assumed that changes to cloud cover are a feedback to CO2 forcing rather than an independent phenomenon. A change in climate can induce cloud changes which then feedback into the initial climate change. This effect is built into most Climate models, which then result in a mean cloud feedback of between 0-2 W/m2/°C [6]. Radiative forcing from increasing CO2 levels is rather well understood [7], but its direct impact on cloud cover is unclear. Feedbacks cannot be too large compared to the Planck response as otherwise they soon become unstable \Delta{T} = \frac{\Delta{T_0}}{(1-f)} as f approaches 1.

Global cloud cover variations measured by a number of satellites under the guidance of the International Satellite Cloud Climatology Project (ISCCP) are subject to uncertainty linked to data acquisition methods [10], and viewing biases [11]. However, we have found previously [12] that using sunshine hours at surface as an inverse-proxy for cloud cover confirms the ISCCP results over the UK. In private correspondence, NASA have also provided assurance that data acquisition and corrections are now reliable and that the ISCCP data are therefore robust.

Figure showing the ISCCP global averaged monthly cloud cover from July 1983 to Dec 2008 over-laid in blue with Hadcrut4 monthly anomaly data. The fall in cloud cover coincides with a rapid rise in temperatures from 1983-1999. Thereafter the temperature and cloud trends have both flattened. The CO2 forcing from 1998 to 2008 increases by a further ~0.3 W/m2 which is evidence that changes in clouds are not a direct feedback to CO2 forcing.

Figure 1a showing the ISCCP global averaged monthly cloud cover from July 1983 to Dec 2008 over-laid in blue with Hadcrut4 monthly anomaly data. The fall in cloud cover coincides with a rapid rise in temperatures from 1983-1999. Thereafter the temperature and cloud trends have both flattened. The CO2 forcing from 1998 to 2008 increases by a further ~0.3 W/m2 which is evidence that changes in clouds are not a direct feedback to CO2 forcing.

CERES measured data on global cloud forcing. Reflected short wave radiation reduces surface heating by ~44 watts/m2 which is offset by cloud absorption of outgoing IR therefby increasing the greenhouse effect from clouds by ~26watts/m2. This results in a net cooling effect from cloudy skies globally of -22 watts/m2. This figure is used to define the NCF = 0.91. [5]

Fig 1b: CERES measured data on global cloud forcing. Reflected short wave radiation reduces surface heating by ~44 watts/m2 which is offset by cloud absorption of outgoing IR therefby increasing the greenhouse effect from clouds by ~26watts/m2. This results in a net cooling effect from cloudy skies globally of -22 watts/m2. This figure is used to define the NCF = 0.91. [5]


Figure 1a shows the ISCCP global averaged cloud cover [8] compared to Hadcrut 4 global temperature anomalies [4]. Until 1998 cloud cover decreased in line with increasing CO2 levels, which may support the existence of a CO2 feedback. However, since 1998 both cloud cover and temperatures have remained flat while CO2 forcing has continued to rise. This is evidence that cloud cover does not depend simply on CO2 forcing alone and may itself be a major natural driver for climate change. There is no direct evidence that cloud cover varies in response to CO2 and the ISCCP data discussed here is cyclic in nature which cannot be explained by unidirectional CO2 forcing. We have therefore developed a model that treats clouds and CO2 forcing independently and separately. Mearns and Best [10] have reported evidence that changes in cloud cover can explain approximately 40% of the UK surface temperature changes since 1956 especially during summer months (June, July, August)[10]. We now apply essentially the same model on a global scale using data from the ISCCP [9] that we downloaded from the US National Oceanic and Atmospheric Administration (NOAA) web site [ref] since the NASA web site has been disabled [ref] referenced to measured surface temperature data from CRU-Hadley (HADCRUT4)[8]. ISCCP cloud data is available beyond 2008 but is not yet in the public domain.

 Cloud forcing model. We define the net cloud-forcing factor (NCF) as the resultant balance between albedo and Green House (GH) effects for clouds. In effect NCF is used to describe the ratio of the combined forcing (cloud transmissibility and GH effect) of clouds relative to that for clear skies. Effectively (1-NCF) is the net cooling factor of clouds with respect to clear skies. Radiative energy balance is then given by

(1-CC)\times{S_0} + CC\times{NCF}\times{S_0} = \epsilon \sigma\ T^4

where, S0 for clear skies is taken as a global average 240 W/m2 [13]. Therefore for each month, m, the incoming net insolation is

S(m) = (1-CC(m))S_0 + CC(m) \times NCF \times S_0

The calculated temperature change for Tcalc (m) is then given by

\Delta{T(m)} = \frac{(S(m)-S(m-1))}{3.5}

where 3.5 Wm-1°C-1 is the Planck response  DS/DT for 288K and is the increase in IR radiation for a 1oC rise in surface temperature. We initialize the model by normalising the first data point Tcalc(July 1983) = Thcrut(July 1983) and then calculate all subsequent  monthly temperatures based only on the measured changes in ISCCP Cloud Cover (CC).

T_{calc}(m) = T_0(m-1) + \Delta{T(m)}

We fix NCF = 0.91 as measured by the CERES for global net cloud forcing.

The CO2 radiative forcing model: The change in CO2 forcing for month (m) is calculated using the formula [7]

S(m) = CS \times 5.3\ln(\frac{CO_2(m)}{CO_2(m-1)})

where ΔS  is the monthly change in radiative forcing,CO2(y)/ is the concentration of CO2 in the atmosphere for month y, and CS is a factor representing climate sensitivity. CO2 values are the measured monthly Mauna-Loa data [14]. The model with CS=1.0 then corresponds to an equilibrium climate sensitivity (ECS) of 1.1°C. However, when the model is applied to contemporaneous temperature data, CS corresponds instead to the transient climate response (TCR). Climate models with net positive feedbacks yield larger values of ECS of between 2 to 5°C [15].  The model value of CS is to be determined empirically from the data.

We apply the model to calculate global temperature anomalies from variance in global cloud cover after normalising the start point (July 1983) to the measured global average temperature and then compare model output to measurements of actual temperature variance as recorded by HadCRUT4. Our criterion for goodness of fit between the model and HadCRUT4 is based on the use of χ2 per degree of freedom (χ2/df). For Χ2 we take a measurement error of 0.1°C for the monthly anomalies.  The χ2 results found by varying CS values with NCF fixed at 0.91 are shown in Figure 2b. A minimum in χ2 is found for CS = 1.45 corresponding to TCR = 1.6 °C.  The error on CS is determined by how much variation is needed to shift  χ2/df by 1.

Fig2

Fig 2 a) Results of the model calculations (Tcalc) for the best fit value of CS =1.45 (TCR=1.6 °C) compared to monthly Hadcrut4 anomaly data.
b) Variation of χ2 per degree of freedom calculated between the predicted and the measured anomalies calculated for different CS values taking NCF=0.91 as measured by CERES.

Clouds and CO2 alone cannot explain all the variations in monthly global temperatures. It is known that explosive volcanic eruptions and ENSO[E1]  also have transient effects on global temperatures, and for this reason it is no surprise that the minimum Χ2/df > 1.0. However the main trend is well reproduced by the model as shown in Figure 2a which compares the best-fit value of CS to the real measured data. The general warming trend until 1998 can mostly be explained by the fall in cloud cover during that period. The flattening off in temperature since 1998 coincides with a leveling off in global cloud cover. To explain observed warming over the full period by a CO2 dependent term alone with clear skies (NCF=1) would require TCR = 2.2 °C resulting in an approximate linear increase of 0.7 °C over this time period. Examining the yearly change in cloud forcing shows that it increased from 225.5 W/m2 in 1984 to 262.2 W/m2 in 1999, or an increase in forcing of ~0.7 watts/m2. CO2 forcing with TCR = 1.6 °C increased by 0.54 W/m2 over the same time period.  This result demonstrates that more than half of the rapid warming observed in the 1980s and 1990s can be explained by a decrease in cloud cover. Since 1999 net cloud forcing has remained approximately constant (-0.2 W/m2), while CO2 forcing has increased by a further 0.58 W/m2.

Results for the summer cloud analysis for a) Northern Hemisphere with model TCR=1.0C and NCF=0.9 and b) Southern Hemisphere with model TCR = 1.65C and NCF=0.91. The Hadcrut4 data and the model data for each year are the averaged results for June, July and August in case a) and for December, January and February for case b). In the latter case the year is assigned to that of December. There is a clear difference in dependence with CS for the Northern and Southern Hemispheres. Changes in cloud cover have a greater impact in the northern hemisphere than in the southern hemisphere. This affects the best fit values for CS for each hemisphere.

Results for the summer cloud analysis for a) Northern Hemisphere with model TCR=1.0C and NCF=0.9 and b) Southern Hemisphere with model TCR = 1.65C and NCF=0.91. The Hadcrut4 data and the model data for each year are the averaged results for June, July and August in case a) and for December, January and February for case b). In the latter case the year is assigned to that of December. There is a clear difference in dependence with CS for the Northern and Southern Hemispheres. Changes in cloud cover have a greater impact in the northern hemisphere than in the southern hemisphere. This affects the best fit values for CS for each hemisphere.

There are marked seasonal variations in cloud cover for each hemisphere – particularly in the southern hemisphere. In order to isolate differences between the long-term effects of clouds in each hemisphere we have made summer averages of temperature and cloud cover (June, July, August (JJA) for Northern Hemisphere (NH) and December, January, February (DJF) for Southern Hemisphere (SH)) and then compared the model with the hemispheric HadCRUT4 anomaly data. For the average summer hemispheric insolation we take a value of S0=312 W/m2 which is 240 W/ m2 corrected for the angle of the sun for summer months. The results of this analysis are shown in Figure 3.  There is a clear difference between the northern hemisphere and southern hemisphere. The response to cloud forcing in the northern hemisphere is stronger with fixed NCF=0.91, and leads to a lower c2 fitted value for CS (TCR = 1.0 ± 0.3°C). The southern hemisphere shows a smaller cloud forcing response with correspondingly larger values for CS (TCR = 1.65 ± 0.3°C). This difference is most likely due to dominance of oceans in the southern hemisphere. By studying each hemisphere separately and by eliminating as far as possible seasonal effects, the global result is confirmed. These results demonstrate that over half the warming observed between 1983 and 1999 is due to a reduction in cloud cover mainly effecting the northern hemisphere. The apparent slowdown in warming observed since 1999 coincides with a stabilization of global cloud cover. In an analysis of cloud and temperature variance in the UK, Mearns and Best [12] reach a similar conclusion which is that approximately 50% of net warming since 1956 is due to a net reduction in cloud cover. However, in that study NCF was estimated empirically to be 0.54, significantly lower than the CERES value of 0.91 used here. A lower NCF factor means that clouds are having a larger effect and the difference between the global and UK results may reflect latitude and the fact that UK data are land based only.

In conclusion, natural cyclic change in global cloud cover has a greater impact on global average temperatures than CO2. There is little evidence of a direct feedback relationship between clouds and CO2. Based on satellite measurements of cloud cover (ISCCP), net cloud forcing (CERES) and CO2 levels (KEELING) we developed a model for predicting global temperatures. This results in a best-fit value for TCR = 1.4 ± 0.3°C. Summer cloud forcing has a larger effect in the northern hemisphere resulting in a lower TCR = 1.0 ± 0.3°C. Natural phenomena must influence clouds although the details remain unclear, although the CLOUD experiment has given hints that increased fluxes of cosmic rays may increase cloud seeding [19].  In conclusion, the gradual reduction in net cloud cover explains over 50% of global warming observed during the 80s and 90s, and the hiatus in warming since 1998 coincides with a stabilization of cloud forcing.

References

1. Randall, D. A. Cloud Feedbacks. Frontiers in the Science of Climate Modeling (2006).

2. Randall, D. A. & Wood, R. A. Climate Models and Their Evaluation. (Cambridge Univ. Press: Cambridge [u.a.], 2007).

3. V. Ramanathan, R.DCess, E.F. Harrison, P.Minnis, B.R. Barkstrom, E. Ahmad, D. Hartmann, Cloud-Radiative Forcing and Climate: Results from the Earth Radiation Budget Experiment, Science, Vol 243, P 57, 1989

4. Jones, P. D., D. H. Lister, T. J. Osborn, C. Harpham, M. Salmon, and C. P. Morice (2012), Hemispheric and large-scale land surface air temperature variations: An extensive revision and an update to 2010, J. Geophys. Res., 117, D05127,

5. Richard P. Allan, Combining satellite data and models to estimate cloud radiative effects at the surface and in the atmosphere, RMetS Meteorol. Appl. 18: 324–333, 2011

6. Bony, S. et al. How Well Do We Understand and Evaluate Climate Change Feedback Processes? Journal of Climate 19, 3445–3482 (2006).

7. Myhre, G., E.J. Highwood, K.P.Shine,, F.Stordal, New Estimate of Radiative ForcingDur to Well Mixed Greenhouse Gases, Geophys. Ress. Lett. 25, 2715-2718, 1998

8. Monthly averaged ISCCP cloud data Data derived from file MnCldAmt.nc, Catalog. (2009) http://www.ncdc.noaa.gov/thredds/catalog/isccp/catalog.html

9. Morice, C. P., J. J. Kennedy, N. A. Rayner, and P. D. Jones (2012), Quantifying uncertainties in global and regional temperature change using an ensemble of observational estimates: The HadCRUT4 dataset, J. Geophys. Res., 117,

10. Rossow, W. B. & Schiffer, R. A. Advances in Understanding Clouds from ISCCP. Bulletin of the American Meteorological Society 80, 2261–2287 (1999).

11. Evan, A. T.; A. K. Heidinger, and D. J. Vimont (2007). Arguments against a physical long-term trend in global ISCCP cloud amounts. Geophy. Ress. Lett 34 (L04701)

12. E.W. Mearns, C.H. Best, Strong coherence between cloud cover and surface temperature variance in the UK

13. K.E. Trenberth, J.T. Fasullo & J. Kiehl, Earth’s Global Energy Budget, Bulletin of the American Meteorological Society (2009)

14. Keeling, C. D. et al. Atmospheric carbon dioxide variations at Mauna Loa Observatory, Hawaii. Tellus 28, 538–551 (1976).

15. Solomon, S. et al. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. (Cambridge Univ. Press: Cambridge [u.a.], 2007).

16. J. Kirby et al. Role of sulphuric acid, ammonia and galactic cosmic rays in atmospheric aerosol nucleation, Nature 476, 429–433 (2011).

Note: I have posting this only now because after a long review process the paper was finally rejected. I am beginning to despair of any outsider ever getting anything published in a climate science journal!

This entry was posted in AGW, Climate Change, climate science, Oceans, Science and tagged , , . Bookmark the permalink.

24 Responses to Do clouds control climate?

  1. Pingback: NASA Satellite Climatology Data | Energy Matters

  2. Pingback: Global Temperatures Rose As Cloud Cover Fell In the 1980s and 90s | NOT A LOT OF PEOPLE KNOW THAT

Leave a Reply to pdtillman Cancel reply