Estimating Cloud Feedback From Observations

Guest Post by Willis Eschenbach

I had an idea a couple days ago about how to estimate cloud feedback from observations, and it appears to have panned out well. You tell me.

Figure 1. Month-to-month change in 5° gridcell actual temperature ∆T, versus gridcell change in net cloud forcing ∆F. Curved green lines are for illustration only, to highlight how many of the datapoints fall outside those lines in each of the four quadrants. Results have been area-weighted, giving a slightly smaller slope (-1.7 W/m2 per degree) than initally reported (- 1.9 W/m2 per degree). Data colors indicate the location of the gridcell, with the Northern hemisphere starting with blue at the far north, slowly changing to yellow and to red at the equator. From there, purple is southern tropic, through pink to green for the farthest south latitudes. Updated.

Cloud feedback is what effect the changing clouds have if the earth warms. Will the clouds act to increase a warming, or to diminish it? The actual value of the cloud feedback is one of the big unknowns in our current understanding of the climate.

The climate models used by the IPCC all say that as the earth warms, the clouds will act to increase that warming. They all have a strong positive cloud feedback. My thunderstorm and cloud thermostat hypothesis, on the other hand, requires that the cloud feedback be strongly negative, that clouds act to decrease the warming.

My idea involved the use of what are called “gridded monthly climatologies”. A monthly climatology is a long-term month-by-month average of some climate variable of interest. “Gridded” means that the values are given for each, say, 5° latitude by 5° longitude gridbox on the surface of the planet.

My thought was to obtain the monthly actual temperature gridded climatology. This is the real temperature “T” as measured, not the anomaly. In addition, I would need the gridded net cloud forcing “F” from the ERBE (Earth Radiation Budget Experiment) data. Net cloud forcing is the balance of how much solar energy the clouds reflect away from the earth on the one hand, and on the other, how much the same clouds increase the “greenhouse” downwelling longwave radiation (DLR). Net cloud forcing varies depending on the type, thickness, altitude, droplet size, and color of a given cloud. Both positive and negative cloud forcing are common. By convention, positive net cloud forcing (e.g. winter night-time cloud) is warming, while a negative net cloud forcing (e.g. thick afternoon cumulus) is cooling.

Remembering that a cloud feedback is a change in net forcing in reference to a change in temperature, I took the month to month differences of each of the two climatologies . I did this in a circular fashion, each month minus the previous month, starting from February minus January, around to January minus December. That gave me the change in temperature (∆T) and the change in forcing (∆F) for each of the twelve months.

The ERBE satellite only covers between the Arctic and Antarctic circles, the poles aren’t covered. So I trimmed the polar regions from the HadCRUT absolute temperature to match. Then, the HadCRUT3 absolute temperature data are on a 5° grid size, while the ERBE satellite data is on a 2.5° grid. Since the grid midpoints coincided, I was able to use simple averaging to “downsample” the satellite cloud forcing data to correspond with the larger temperature gridcell size.

The results of the investigation are shown in Figure 1. The globally averaged cloud feedback is on the order of -1.9 watts per square metre for every one degree of monthly warming.

This result, if confirmed, strongly supports my hypothesis that the clouds act as a very powerful brake on any warming. At typical Earth surface temperatures, the Stefan-Boltzmann equation gives about five watts per square metre (W/m2) of additional radiation  per degree. That is to say, to warm the surface by 1°C, the amount of incoming energy has to increase by about 5 W/m2. This, of course, means that if there were no feedbacks, a doubling of CO2 (+3.7 watts per square metre per the IPCC) would only cause about 3.7/5 or about three-quarters of a degree of warming. The models jack this three-quarters of a degree up to three degrees of warming by, among things, their large positive cloud feedback.

But this analysis says that the cloud feedback is strongly negative, not positive at all. As a result, a doubling of CO2 could easily cause less than eight-tenths of a degree of warming. If the cloud negative feedback is actually -1.9 W/m2 per degree as shown above, and it were the only feedback, a doubling of CO2 would only cause half a degree of warming …

If confirmed, I think that this is a significant result, so I put it up here for people to check my math and my logic. I’ve fooled myself with simple mistakes before …

Code for the procedures and data is appended below.

All the best,

w.

PS – please, no claims that the “greenhouse effect” is a myth or that DLR doesn’t exist or that DLR can’t transfer energy to the ocean. I’m beyond that, whether you are or not, and more to the point, there are plenty of other places to have that debate. This is a scientific thread with a specific subject, and if necessary I may snip such claims (and responses) to avoid thread drift. If so, I will indicate such excisions.

NOTE: The slope of the trend line in Figure 1 is now properly area-adjusted, making the following section andFigure 2 superfluous. .[UPDATE] I’ve gone back and forth about whether to area-average. The problem is that the gridcells are not the same size everywhere. The usual way to area-average is to multiply the data by the cosine of the mid latitude, so I have done that.

Figure 2 shows the area-adjusted version. Still a significant negative feedback from clouds, but smaller than in the non-adjusted version.

FIGURE 2 REMOVED

Figure 2. Area adjusted cloud feedback. Note the lower estimate of the cloud feedback, a bit smaller than my initial estimate. Color of the dots indicates latitude, ranging from blue at the furthest north through cyan to the equator, then in the southern hemisphere through yellow to red at the furthest south.

Note that we still see the same form in the four quadrants. It is still rare for a large temperature drop to be associated with anything but a rise in the cloud forcing.

I’m still not completely happy with this method of area-adjusting, because it adjusts the data itself. But I think it’s better than no area-adjusting at all. The best way would be to convert both of the datasets to equal-area cells … but that’s a large undertaking and I think the final result won’t be much different from this one.

[UPDATE] Here’s the two hemispheres:

Figure 3. Northern Hemisphere Cloud Feedback. Color of the dots indicates latitude, ranging from blue at the furthest north through yellow in the subtropics, to red at the equator.

Figure 4. Southern Hemisphere Cloud Feedback. Color of the dots indicates latitude, ranging from green at the furthest south through pink in the subtropics, to purple at the equator.

[UPDATE] To better inform the discussion, I have made up the following maps of the variables of interest, month by month. These are the monthly absolute temperatures T, the monthly net cloud forcings F, and the month by month changes (deltas) of those variables, ∆T and ∆F.

Figure 5. Absolute temperature (T)

Figure 6. Net Cloud Forcing (F)

Figure 7. Change in absolute temperature (∆T)

Figure 8. Change in net cloud forcing (∆F)

APPENDIX: R code to read and process the data (not including the updated charts). I've tried to keep wordpress from munging the code, but it likes to either put in or not put in carriage returns.

===================================

# data is read into a three dimentional array [longitude, latitude, month]

diffannual=function(x){# returns month(t+1) minus month(t)

x[,,c(2:12,1)]-x

}

# rotates the circle of months by n

rotannual=function(x,n){

if (n!=0) {

if (n>=0){

x[,,c((n+1):12,1:n)]

} else {

x[,,c((13+n):12,1:(12+n))]

}

} else

x

}

#_averages_2.5°_gridcells_into_5°_gridcells,_for_[long,lat,mon]_array

downsample=function(x){

dx=dim(x)

if (length(dx)==3){

reply=array(NA,c(dx[1]/2,dx[2]/2,dx[3]))

for (i in 1:dx[3]){

reply[,,i]=downsample2d(x[,,i])

}

} else {

reply=downsample2d(x)

}

reply

}

# averages 2.5° gridcells into 5° gridcells for [long, lat] 2D array

downsample2d=function(x){

width=ncol(x)

height=nrow(x)

smallforcing=matrix(NA,height/2,width/2)

for (i in seq(1,height-1,2)){

for (j in seq(1,width-1,2)){

smallforcing[(i+1)/2,(j+1)/2]=mean(c(x[i,j],x[i+1,j],x[i,j+1],x[i+1,j+1]),na.rm=T)

}

}

as.matrix(smallforcing)

}

# EEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEEE End Functions

# LLLLLLLLLLLLLLLLLLLLLLLLLL LOAD DATA ----- gets the files from the web

# HadCRUT absolute temperature data

absurl="http://www.cru.uea.ac.uk/cru/data/temperature/absolute.nc"

download.file(absurl,"HadCRUT absolute.nc")

absnc=open.ncdf("HadCRUT absolute.nc")

download.file("http://badc.nerc.ac.uk/browse/badc/CDs/erbe/erbedata/erbs/mean5jan/data.txt","albedojan.txt")

download.file("http://badc.nerc.ac.uk/browse/badc/CDs/erbe/erbedata/erbs/mean5feb/data.txt","albedofeb.txt")

download.file("http://badc.nerc.ac.uk/browse/badc/CDs/erbe/erbedata/erbs/mean5mar/data.txt","albedomar.txt")

download.file("http://badc.nerc.ac.uk/browse/badc/CDs/erbe/erbedata/erbs/mean5apr/data.txt","albedoapr.txt")

download.file("http://badc.nerc.ac.uk/browse/badc/CDs/erbe/erbedata/erbs/mean5may/data.txt","albedomay.txt")

download.file("http://badc.nerc.ac.uk/browse/badc/CDs/erbe/erbedata/erbs/mean5jun/data.txt","albedojun.txt")

download.file("http://badc.nerc.ac.uk/browse/badc/CDs/erbe/erbedata/erbs/mean5jul/data.txt","albedojul.txt")

download.file("http://badc.nerc.ac.uk/browse/badc/CDs/erbe/erbedata/erbs/mean5aug/data.txt","albedoaug.txt")

download.file("http://badc.nerc.ac.uk/browse/badc/CDs/erbe/erbedata/erbs/mean5sep/data.txt","albedosep.txt")

download.file("http://badc.nerc.ac.uk/browse/badc/CDs/erbe/erbedata/erbs/mean5oct/data.txt","albedooct.txt")

download.file("http://badc.nerc.ac.uk/browse/badc/CDs/erbe/erbedata/erbs/mean5nov/data.txt","albedonov.txt")

download.file("http://badc.nerc.ac.uk/browse/badc/CDs/erbe/erbedata/erbs/mean5dec/data.txt","albedodec.txt")

albnames=c("albedojan.txt","albedofeb.txt","albedomar.txt","albedoapr.txt","albedomay.txt","albedojun.txt","albedojul.txt","albedoaug.txt","albedosep.txt","albedooct.txt","albedonov.txt","albedodec.txt")

# read data into array

forcingblock=array(NA,c(52,144,12))

for (i in 1:12){

erbelist=read.fwf(albnames[i],skip=19,widths=rep(7,13))

erbelist[erbelist==999.99]=NA

erbeout=erbelist[,13][which((erbelist[,1]>-65) & (erbelist[,1]

length(erbeout)

forcingblock[,,i]=matrix(erbeout,52,144,byrow=T)

}

# DOWNSAMPLE FORCING DATA TO MATCH TEMPERATURE DATA,

# and swap lat and long to match HadCRUT data

smallforcing=aperm(downsample(forcingblock),c(2,1,3))

smallforcing[1:72,,]=smallforcing[c(37:72,1:36),,]# adjust start point

# GET ABSOLUTE DATA, TRIM POLAR REGIONS

absblock= get.var.ncdf(absnc,"tem")

smallabs=absblock[,6:31,]

#dim(absblock)

# GET MONTH-TO-MONTH DIFFERENCES

dabs=diffannual(smallabs)

dforcing=diffannual(smallforcing)

dim(dforcing)

#SAVE DATA

save(forcingblock,smallforcing,smallabs,dabs,dforcing,file="erbe_cloud_forcing.tab")

# make cosine weight array

cosarray=array(NA,c(72,26,12))

cosmatrix=matrix(rep(cos(seq(-62.5,62.5,by=5)*2*3.14159/360),72),72,26,byrow=T)

cosmatrix=cosmatrix/mean(cosmatrix[1,])

cosarray[,,1:12]=cosmatrix

cosarray[,,2]

# GET CORRELATION, SLOPE, AND INTERCEPT

#cor(dabs,dforcing,use="pairwise.complete.obs")

module=lm(dforcing~dabs)

m=module$coefficients[2]

b=module$coefficients[1]

#Plot Results

par(mgp=c(2,1,0))

plot(dforcing~dabs,pch=".",main="Cloud Feedback, 65°N to 65°S", col="deepskyblue3",xlab="∆ Temperature (°C)",ylab="∆ Cloud Forcing (W/m2)")

lines(c(m*(-20:15)+b)~c(-20:15),col="blue",lwd=2)

textcolor="lightgoldenrod4"

text(-20,-60,"N = 18,444",adj=c(0,0),col= textcolor)

text(-20,-70,paste("Slope =",round(m,1),"W/m2 per degree C of warming"),adj=c(0,0),col= textcolor)

text(-20,-80,paste("p = ","2E-16"),adj=c(0,0),col= textcolor)

Advertisements

178 thoughts on “Estimating Cloud Feedback From Observations

  1. The only thing that comes to mind is this from William Briggs:
    http://wmbriggs.com/blog/?p=195
    Now I’m going to tell you the great truth of time series analysis. Ready? Unless the data is measured with error, you never, ever, for no reason, under no threat, SMOOTH the series! And if for some bizarre reason you do smooth it, you absolutely on pain of death do NOT use the smoothed series as input for other analyses!
    Gridded data might qualify for pre-smoothed, though I don’t know what processes were involved in making the gridded data. Best to rule out that this isn’t some artifact of data processing.

  2. Anthony Watts says:
    October 8, 2011 at 4:22 pm

    The only thing that comes to mind is this from William Briggs:
    http://wmbriggs.com/blog/?p=195

    Now I’m going to tell you the great truth of time series analysis. Ready? Unless the data is measured with error, you never, ever, for no reason, under no threat, SMOOTH the series! And if for some bizarre reason you do smooth it, you absolutely on pain of death do NOT use the smoothed series as input for other analyses!

    Gridded data might qualify for pre-smoothed, though I don’t know what processes were involved in making the gridded data. Best to rule out that this isn’t some artifact of data processing.

    Thanks, Anthony. As far as I know, almost all of the gridded data used all across the board is done by a weighted average, with the weighting done by distance from the centre of the gridcell.
    I disagree with William Briggs somewhat regarding smoothing. Used cautiously, and with regard to the niceties, it can be valuable for revealing underlying patterns that may not be visible using just the raw data. His second point, however, is well taken.
    I’m not sure how I could “rule out that this isn’t some artifact of data processing”, which was kinda why I’m putting it out here, for people to poke holes in it … science at work, many thanks for the blog.
    w.

  3. Looks like a random line drawn through an amorphous cloud of points.
    This is the same problem I have Dressler, Lindzen & Choi, and Spenser & Braswell. There might be something there, but it sure is well hidden, if it is there. At the end of the day I am less certain of any relationship after all the smoothing and statistical gymnastics to tease something of value from the cloud, than I am with just the scatter plot cloud.

  4. Doug in Seattle says:
    October 8, 2011 at 4:41 pm

    Looks like a random line drawn through an amorphous cloud of points.
    This is the same problem I have [with] Dressler, Lindzen & Choi, and Spenser & Braswell. There might be something there, but it sure is well hidden, if it is there.

    That’s why we invented mathematics, all right. Part of the problem that there are over eighteen thousand data points, so the underlying pattern is obscured. That doesn’t mean it’s not there, just that this particular visualization breaks down under the weight.
    w.

  5. Willis,
    An interesting analysis that makes sense. One nit. OK two nits.
    “That is to say, to warm the surface by 1°C, the amount of incoming energy has to increase by about 5 W/m2. This, of course, means that if there were no feedbacks, a doubling of CO2 (+3.7 watts per square metre per the IPCC) would only cause about 3.7/5 or about three-quarters of a degree of warming.”
    These numbers confused the day lights out of me because SB Law shows a bit over 5 watts/m2 to raise the average surface temperature (15 degC) by one degree, but the IPCC meme is that CO2 doubling = 3.71 watts/m2 = + 1 degree. I had to go back to AR3 to figure out how they came up with it. The answer is that they calculate +1 degree against the “effective black body temperature” of the earth, which is about -20 degC. Given SB Law is P=5.67*10^-8*T^4 with P in watts/m2 and T in degrees K, the math comes out correctly. 3.71 watts/m2 to go from -20 degC to -19 degC and a bit over 5 watts/m2 to go from plus 15 to plus 16. Or, as I prefer to put it, CO2 doubling = 3.71 watts/m2 = +1 degree as seen from space = ~ 0.6 degC “on average”. There being no such thing as “average” when it comes to surface temperature, the change at -40 in winter at high latitudes is larger than 1 degree, and the change in temperature at +40 in the tropics is not only less than 1 degree its far less than the “average” of 0.6 too, more like 0.2. Of course that’s assuming that the 3.71 watt/m2 is uniform. I’ve never checked for myself, but logic suggests not.
    The other nit is time constant. I haven’t really thought this through all the way. If the 5 time constants exceeds one month, then some of the cloud feedback shows up in the next month or perhaps the month after that…given that you’ve gone all the way around the calendar however doing a month over month analysis, I’d think the average for the year would be representative.

  6. That’s why we invented mathematics, all right. Part of the problem that there are over eighteen thousand data points, so the underlying pattern is obscured. That doesn’t mean it’s not there, just that this particular visualization breaks down under the weight.
    Perhaps a contour map of point density would help.

  7. The ‘Ninja Claw Star’ plot you show is rather interesting, it looks like a bi-functional ‘cloud’.
    I wonder if you could crop the central data points and then split the data points into two pairs; the very shallow (which your line describes) and an almost vertical one. Do the two sets of points suggest two types of clouds. .

  8. Try coloring datapoints belonging to gridcells >50% land a different color to those with >50% water. I suspect there may be at least two quite distinct patterns hiding in this mess. Clouds over land behave in a qualitatively different fashion from clouds over ocean.

  9. I’ve updated Figure 1 in the head post to highlight the differences in the four quadrants.
    w.

  10. Roh234 says:
    October 8, 2011 at 5:45 pm

    Which program is the code?

    It’s written in the computer language “R”, if that’s the question.
    w.

  11. Steve at Climate Audit did some work on this recently, showing that monthly centering hides the true trend.
    http://climateaudit.org/2011/09/28/monthly-centering-and-climate-sensitivity/
    “In our recent discussion of Dessler v Spencer, UC raised monthly centering as an issue in respect to the regressions of TOA flux against temperature. Monthly centering is standard practice in this branch of climate science (e.g. Forster and Gregory 2006, Dessler 2010), where it is done without any commentary or justification. But such centering is not something that is lightly done in time series statistics. (Statisticians try to delay or avoid this sort of operation as much as possible.)
    When you think about it, it’s not at all obvious that the data should be centered on each month. I agree with the direction that UC is pointing to – a proper statistical analysis should show the data and results without monthly centering to either verify that the operation of monthly centering doesn’t affect results or that its impact on results has a physical explanation (as opposed to being an artifact of the monthly centering operation.)”

  12. Ian H says:
    October 8, 2011 at 5:32 pm (Edit)

    Try coloring datapoints belonging to gridcells >50% land a different color to those with >50% water. I suspect there may be at least two quite distinct patterns hiding in this mess. Clouds over land behave in a qualitatively different fashion from clouds over ocean.

    There’s a whole host of other ways to slice and dice this data. I just want to make sure that I’m looking at something real first.
    I think it is real, for reasons like the very small number of data points in the third quadrant above. What that means is that when temperatures drop a ways, net cloud forcing rarely drops. Instead, when temperatures drop, a majority of the time the clouds act to warm the surface.
    w.

  13. I’ll echo Ian H above. Split the data set by land/water.
    I think you’re on the right track, particularly regarding tropical thunderstorms.
    I’ve got another piece to drop in later. That should happen in a couple months.

  14. Dishman says:
    October 8, 2011 at 6:09 pm

    I’ll echo Ian H above. Split the data set by land/water.

    Easy for you to say … seriously, I need two things for that, a land/water mask, and 28 hours in a day …
    One step at a time, dear friends …
    w.

  15. Willis Eschenbach says:
    October 8, 2011 at 6:00 pm
    “I’ve updated Figure 1 in the head post to highlight the differences in the four quadrants.”
    That really helped a lot. Now you can see in the second and fourth quadrants how the data produces the slope. I agree with others above, it was hard to see in the first version you put up.
    Thanks; good tweak.

  16. Kevan Hashemi says:
    October 8, 2011 at 5:02 pm

    Clouds reflect sunlight back into space. Whatever effect they have upon long-wave radiation is entirely secondary to the fact that they block the sun’s heat from reaching the surface of the planet. …

    While I admire your certitude, whether DLR is “entirely secondary” is the question, not something you can assert.
    Half of the time (nights) the DLR effect is the only effect … so it can hardly be “secondary” then. And in many other circumstances, the gain in DLR outweighs the loss of solar. You have to average it out 24/7/365 for the entire globe, you can’t just do a simple simulation as you propose. Even complex simulations (climate models) don’t get it right.
    w.

  17. This is an extremely important finding.
    Nobody can really figure this out based on the data we have available. But I think Willis has done this in the appropriate way matching spatial changes.
    The result is fully -2.9 watts/m2/K lower than the climate models have and would rewrite the theory if confirmed.
    The total Feedbacks would then net out to -0.9 watts/m2/k versus the theory at +2.0 watts/m2/K.
    Interesting that that this value is very close to what the missing energy is (given the calculations are that we are at +0.7K to date).
    Trenberth even wrote a paper showing a mysterious negative “Radiative Feedback” for which there was no source. Maybe Willis found it. Note the -2.9 watts/m2/K differential is very close to Trenberth’s estimate.
    http://img638.imageshack.us/img638/8098/trenberthnetradiation.jpg

  18. Easy for you to say … seriously, I need two things for that, a land/water mask, and 28 hours in a day …
    KNMI lets you download monthly temps by land only and sea only for hadcrut. they also show a cloud cover for land only from CRU and a sea only from COADS. Would that help instead of building a land/sea mask?
    On the 28 hours in a day, I can help there too, as I came across this problem some time ago. My solution was to begin also working nights.

  19. With 18,000 data points, it looks as if any dependent variable is perpendicular to this data set.
    I see absolutely no correlation. If there is a relationship in there, it is hiding pretty good.
    What the heck is cloud forcing? I understand the units W/m^2… but not the variable.

  20. Look at the “outliers”.
    100W/m^2 with NO TEMP change… I’d like to explain that!… How can you have -100 W/m^2 and no temp change?

  21. R^2 = 0.11 means there’s no significant pattern.
    In other words nothing to see here. Move along now.

  22. Maybe I am missing something here, but I have difficulty seeing how a month on month temperature change can cause a month on month change in cloud forcing.
    I’d expect the cloud forcing to be a direct function of temperature. Warmer = more clouds = increase in the cloud forcing. What matters is the temperature, not whether the temperature has increased/decreased.
    So I’d expect to see a similar (or larger) effect if cloud forcing were plotted against monthly temperature.

  23. davidmhoffer says:
    October 8, 2011 at 6:34 pm (Edit)
    Easy for you to say … seriously, I need two things for that, a land/water mask, and 28 hours in a day …

    KNMI lets you download monthly temps by land only and sea only for hadcrut. they also show a cloud cover for land only from CRU and a sea only from COADS. Would that help instead of building a land/sea mask?

    Anomalies only, not absolute temperatures, AFAIK. Also no data on the net cloud forcing, just albedo.
    w.

  24. I’m so old I remember being mocked at Climate Audit when I suggested that we knew neither the magnitude nor the sign of the water vapor feedback.
    ===============

  25. Paul Westhaver says:
    October 8, 2011 at 6:40 pm

    Look at the “outliers”.
    100W/m^2 with NO TEMP change… I’d like to explain that!… How can you have -100 W/m^2 and no temp change?

    Take a breath, Paul. We haven’t allowed for such things as the change in snow albedo. So your idea, that the only ∆F is clouds, is not correct. Here’s another possibility, cold wind sweeps away the clouds for a month, solar radiation goes way up but is counteracted by the wind.
    w.

  26. Dave Springer says:
    October 8, 2011 at 6:41 pm

    R^2 = 0.11 means there’s no significant pattern.
    In other words nothing to see here. Move along now.

    Dave, that’s not true. The question is not the size of the r^2. The question is, is the relationship significant? Clearly the answer is yes, highly so.
    The small R^2 is related to all of the other confounding variables – winds, surface albedo, land versus sea, and the like. All of those increase the “noise”, and thus decrease the R^2 … but what they don’t do is, they don’t affect the relationship between ∆T and ∆F.
    If you look at the data, the variation is quite large … which is why if we can explain even 10% of that variation by means of cloud forcing, it is a significant result. Yes, it doesn’t explain it all … but it does explain -1.9 W/m2 per degree of the variation.
    w.

  27. Willisl
    Anomalies only, not absolute temperatures, AFAIK. Also no data on the net cloud forcing, just albedo>>>
    Select monthly observations, there are several sections under “Land”. The last one (just before the Tmin and Tmax data sets) is CRU Land TS 3.1 1901 to 2009 in absolute temps. I thought there was an equivelant for SST, but I was wrong. There is however MOHMAT which has marine air temps from 1856 to 2007 in absolute temps. The need to correlate to net forcing slipped my mind, sorry. They do have Tmin and Tmax however…would it be of value to correlate cloud cover to those? I’d think that Tmax in particular would lead increase in cloud cover (tropics in particular) but one the cloud cover reached a certain point, Tmax increases would start dropping off? Would be better still to do that same cut at the data but break it up by night versus day.

  28. Not that I believe much in models, but couldn’t you model an earth with clouds of different kinds, in 100% fashion with some of the IPCC assumptions? That is, a model of the earth with 100% “A” – “D” type clouds, to see what the effect is?
    Yes, a lot of work, but in the end you would start to get a handle on how clouds effect earth. These could give you the areas to look at: which clouds are most important to temperature change. You could then run different cloud assumptions through the IPCC models of the earth, and see how these different cloud assumptions change earth’s temperature. If there are reasonable assumptions that lead to negative feedback, then it’s up to the modelers to prove their assumptions.

  29. @Willis
    [SNIP – off topic. I don’t censor criticism. I censor off-topic posts, as I said I would. – w.]
    Censoring criticism is not very scientific so if you don’t want an angry Marine dogging your ass about it for the foreseeable future I suggest you not resort to censoring your critics.

  30. Willis:
    “I’m not sure how I could “rule out that this isn’t some artifact of data processing”, which was kinda why I’m putting it out here, for people to poke holes in it … ”
    Seed your grids with noise and see what pops out the other end.
    On Briggs: If you’re going to go about averaging things all about the place then you’ve two options. Use the currently accepted methodologies or justify the worth of your own bespoke methodology. If your take on being a data masseuse is the accepted practice of the Gridder Gang then roll with it. Your results are valid for the methodology, within the methodology, and that’s all that needs be said about it.
    If you’re going bespoke then it’s your onus to tell us why your custom treatment is better than the one accepted by the field. Note well that it’s a mere trifle if the currently accepted methodology is known to be so gobsmackingly wrong as to claim that the sun rises daily by erupting from the Earth’s crust somewhere just south of Billings, Montana. It is largely political, but that’s science when you get statistics deeply involved and measurements are difficult to make. If you don’t know which your method is, or why and under what conditions it’s valid, then you need take a step back.
    For example:
    1. Why use a simple mean() rather than the average of the variance weighted by Euclidian distance? (I don’t know R, it may be doing that for all I can tell. In which case, why that?)
    2. Why use month-to-month differences rather than a single static baseline. (Including seasonal differences in insolation as necessary.)
    “I don’t know” is not a valid answer. “Trenberth/Mann does it,” is.
    Apologies in advance if I come over harsh in this.

  31. Very cool if it holds up Willis. Not only does there appear to be negative feedback when it is warm, but positive feedback when it is cold.

  32. “The climate models used by the IPCC all say that as the earth warms, the clouds will act to increase that warming.”
    Is this correct ? I was under the impression that the claim was that increasing CO2, increases humidity rather than cloud cover, which increases warming.

  33. If I understand what you are doing Willis. if you have a 2.5 grid of delta cloud forcing and a ( after downsampling) a 2.5 grid of delta T, you really might consider the fact tat your different data points might benefit from scaling due to spherical geom. or use an equal area dataset for both.
    Not sure, but that might effect the spread of your data.. the information from a2.5 grid at the equator should have more weight than the same information from high or low lat? head scratching

  34. Just to be certain, this result is a comparison of the HadCrut 1961-90 average monthly temperatures against the 1985-89 monthly average cloud forcing?

  35. Looks like about 1% of the points are contributing to this correlation, so this may be a small area of the surface, and therefore the forcing should be divided by a corresponding amount. Can the correlation be applied as if it is global?

  36. What? You mean this regression, or something like it, has NEVER been done before with the raw data? What the #$% have the tens of thousands of so-called “climate researchers” been doing with the tens of billions of dollars pouring into that industry? I have a hard time believing that such an obvious preliminary back-of-the-envelope calculation has not been done and does not form the default assumption about cloud feedback in the same way as radiative physics provides us with a default value for the expected warming effect of any change in CO2 concentration.
    I’m glad you’re out there, Willis. But don’t you feel a bit like the child in The Emperor’s New Clothes — this really is a pretty simpleminded experiment, which one would think anyone would carry out before making assertions about cloud feedback.

  37. Elegantly done, with no obvious holes. I would like to suggest a different dot coloring scheme from the land/water one others above have suggested, and much easier to code. Can you color the dots by latitude, say from red at the equator to blue at the Arctic circle and green at the Antarctic circle? I expect this scheme to illuminate your tropical cloud thermostat theory, and could also contrast the mostly water southern temperate latitudes with the northern temperate latitudes that have a much larger proportion of land.

  38. Very cool if it holds up Willis. Not only does there appear to be negative feedback when it is warm, but positive feedback when it is cold.
    Agreed. That would be a very significant discovery.
    The Spencer post linked above, seems to show there are significant forcings internal to the climate system. And Spencer seems to think they operate via clouds. Something I’ve thought for a while because there is too much variability in the climate to be caused by the accepted forcings. This variability gets hand waved away as noise.
    Which means clouds aren’t exclusively a feedback.

  39. I have the notion that the feedback varies with latitude, as the cloud types vary somewhat with latitude. I agree with Doug in Seattle, where he says (October 8, 2011 at 4:41 pm):

    Looks like a random line drawn through an amorphous cloud of points.
    This is the same problem I have Dressler, Lindzen & Choi, and Spenser & Braswell. There might be something there, but it sure is well hidden, if it is there.

    That is the first thing I saw, how much it looked like Dressler. And the r^2 value is awfully low.
    But I am not so sure that the amorphous cloud will exist if is broken down by latitude ranges. I am thinking sorting it out will show more relationship with latitude as a factor.

  40. mt says:
    October 8, 2011 at 8:22 pm
    Just to be certain, this result is a comparison of the HadCrut 1961-90 average monthly temperatures against the 1985-89 monthly average cloud forcing?
    ##
    no he pulled hadcru absolute
    absurl=”http://www.cru.uea.ac.uk/cru/data/temperature/absolute.nc”
    download.file(absurl,”HadCRUT absolute.nc”)
    absnc=open.ncdf(“HadCRUT absolute.nc”)
    But its not clear where he aligns the time.

  41. steven mosher says:
    October 8, 2011 at 8:11 pm

    If I understand what you are doing Willis. if you have a 2.5 grid of delta cloud forcing and a ( after downsampling) a 2.5 grid of delta T, you really might consider the fact tat your different data points might benefit from scaling due to spherical geom. or use an equal area dataset for both.
    Not sure, but that might effect the spread of your data.. the information from a2.5 grid at the equator should have more weight than the same information from high or low lat? head scratching

    Thanks, Mosh. Before reading your comment I updated the head post with a stab at area-averaging the data, and mentioned my own head-scratching. I’m not real happy with it, an equal-area weighting system might be better.
    I took the data the other way, to a 5° x 5° grid. I thought about breaking the temperature down to 2.5 x 2.5, probably could do it without much error since average monthly surface temperature is a fairly smooth 2d function. I just would rather average data than induce data …
    Give me your thoughts …
    w.

  42. mt says:
    October 8, 2011 at 8:22 pm

    Just to be certain, this result is a comparison of the HadCrut 1961-90 average monthly temperatures against the 1985-89 monthly average cloud forcing?

    Close. Actually it’s a comparison of the month to month changes in those two datasets, and the HadCRUT is the gridded absolute monthly averages.
    w.

  43. Nice work Willis,
    I hope this isnt a silly question, however given the work on cosmic ray influence on cloud formation is there any relationship that can be overlaid here?

  44. Willis, you say “Easy for you to say … seriously, I need two things for that, a land/water mask, and 28 hours in a day …” Agreed.
    FWIW, my first reaction was to try to break down the main figure to simpler parts. We surmise that clouds in daytime work differently to clouds at night. So separate them out. Then, the similar question of over land/over water. Then, the special case of ice/snow and albedo competitions*.
    Then, it starts to get murky as you think further. Daytime where, night time where? Clear enough at the Equator, rather different at a Pole. Different night lengths allow different times for temperatures to change. If they can’t equilibrate overnight, is there good math in taking a grid cell average over a month? For this example, is a weighted distance infill of grid cells appropriate, especially id a discontinuity affects cloud behaviour?
    You mind is sharper than mine, so at about this point of thought I’d give up and go watch South Park, realising that the huge effort is premised on whether feedback is + or -, which is important for some, but which becomes less sexy with each new approach. This is not a denigration, it is a praise for your heroism, for your patience exceeding mine.
    *“I wonder if the snow loves the trees and fields, that it kisses them so gently? And then it covers them up snug, you know, with a white quilt; and perhaps it says “Go to sleep, darlings, till the summer comes again.”
    ― Lewis Carroll, Alice’s Adventures in Wonderland & Through the Looking-Glass

  45. UnfrozenCavemanMD says:
    October 8, 2011 at 8:40 pm

    Elegantly done, with no obvious holes. I would like to suggest a different dot coloring scheme from the land/water one others above have suggested, and much easier to code. Can you color the dots by latitude, say from red at the equator to blue at the Arctic circle and green at the Antarctic circle? …

    See the update to the head post. Not quite your color scheme, but using your idea.
    w.

  46. Scott says:
    October 8, 2011 at 9:36 pm

    Nice work Willis,
    I hope this isnt a silly question, however given the work on cosmic ray influence on cloud formation is there any relationship that can be overlaid here?

    No can do. Remember that these are the average monthly values, and as far as I know there’s little annual swing in cosmic ray levels. Nor do cosmic rays vary by gridcell in a a significant manner, as far as I know.
    w.

  47. Willis…. In my experience, outliers precisely measured, contain a great deal of information.
    Normalize for wind speed of pressure?

  48. R. Craigen says:
    October 8, 2011 at 8:36 pm

    What? You mean this regression, or something like it, has NEVER been done before with the raw data? …
    I’m glad you’re out there, Willis. But don’t you feel a bit like the child in The Emperor’s New Clothes — this really is a pretty simpleminded experiment, which one would think anyone would carry out before making assertions about cloud feedback.

    I haven’t seen it done before, doesn’t mean it hasn’t been done.
    And as for it having taken a while to do … well, I’ve been studying the climate extensively for some years now, and it took me this long to think of it … I wouldn’t call it an obvious plan of attack, except perhaps in hindsight …
    w.

  49. Hmm, Willis, You might want to consider other observation datasets. Hadcru has a 1 degree, so you could get a 3 degree from that and resample the 2.5 up to three. use the raster package to resample..
    I could do a 2.5 grid for you using GhcnV3, but I’m in the middle of a long data crunching run, so it would be a couple 3 or 4 days.. I’m verified against hadcru… Giss has an equal area…..

  50. Willis if you need a land water mask just ask.
    In my package RghcnV3
    readMaskDeg1()
    I’ve included a 1 degree land ocean mask in the external directory of the package sources.
    you can resize using the aggregrate() function.

  51. steven mosher says:
    October 8, 2011 at 10:26 pm

    Hmm, Willis, You might want to consider other observation datasets. Hadcru has a 1 degree, so you could get a 3 degree from that and resample the 2.5 up to three. use the raster package to resample..
    I could do a 2.5 grid for you using GhcnV3, but I’m in the middle of a long data crunching run, so it would be a couple 3 or 4 days.. I’m verified against hadcru… Giss has an equal area…..

    Thanks, Mosh. I’d use GISS equal area if I had it as actual temperatures, not anomalies. Is that available?
    w.

  52. Willis,
    I tried running your R code, so first some R stuff:
    1. The package “ncdf” is used, so somewhere it needs to say
    require(“ncdf”)
    and of course you have to have that installed.
    2. In the download from absurl, I found it necessary to add mode=”wb”
    3. The line starting erbeout=erbelist has been mangled, and I haven’t worked out what it should say. It looks like a problem with a < sign.
    However, I’ve looked at the data, and it seems, re mt, that you are comparing just the monthly averages of the 1961-1990 temp with the 1985-9 ave of albedo. So it’s just the seasonal change averaged over those years, over the many grid cells.
    I can’t see how you can interpret that averaged seasonal change as a feedback, or indeed attach any useful causality. The temp rise is of course mainly from the sun angle, and while that rise could then have an effect on clouds, there is obviously much else that changes seasonally – for example, monsoons, movement of the westerly airstreams. And your analysis heavily emphasises the localisation – temp in a gridcell vs could forcing in a gridcell, whereas cloud systems are likely responding to temperatures elsewhere as well – they do blow around.

  53. Willis,
    I watched about the first third of a talk by Samuel Stechmann this morning, “Multiscale Models for Atmospheric Convection and Waves”. I’ll get to the rest tomorrow morning (much better than watching the news). Anyway, the lecture covered the MJO and gravity waves (I did not know such things existed). One of the motivations of the work was that the GCMs do a poor job at emulating these oscillations. Your post stirred a thought related to these oscillations, specifically the MJO. By their very nature, such oscillations have to be self propagating or there would be no oscillation.
    I have not dug deeply into either Spencer & Braswell or Lindzen & Choi, but a quick reading gives me the impression that their models, as well as your analysis here deal with feedback over time but the positions are stationary. Perhaps, the correlations would be tighter if the moving waves set the reference frame, rather than using geographically fixed positions. It would make the math a little hairy. Just a thought.
    One of the problems with the GCMs is that many of the processes that influence the heat flux is that they occur on scales much smaller than the grid size in the simulation. The only way to deal with this is to parameteizer the process so that the output is as close a possible to observations averaged over the grid cell. Since thunderstorms and their influence on atmospheric processes necessary for wave propagation are clearly sub-grid phenomena, I am not surprised that GCMs perform poorly in capturing the MJO and other wave phenomena.
    http://online.kitp.ucsb.edu/online/turbulence11/stechmann/
    The following talk by Joao Teixeira from CalTech discusses low level clouds in the northeastern Pacific and how parameters might be set. It’s yet another significant problem for the GCMs, because if the proper amount of clouds is improperly emulated, then the improper amount of heat enters or leaves the ocean with significant affects on ocean and air temperatures. These errors are then carried forward throughout the simulation. Mmmm.
    Turbulence, Clouds and Climate Models http://online.kitp.ucsb.edu/online/turbulence11/teixeira/
    Home page of KITP: http://www.kitp.ucsb.edu/

  54. Well I’ll leave this one to all the PhD Physicists out there. I think this is just about the first “Paper” posted here at WUWT, for which I can say enequivocally, that I have no understanding of anything that is presented.
    There once was a time, when I could pick up an issue of “Reviews of Modern Physics”, and find papers in which I could not understand any word longer than four letters.
    Well here I can understand most; but not all of the words; but that is the sum total of my comprehension. It’s totally bamboozling to me.

  55. Now what could be causing that vertical spread of southern hemisphere data? That is really interesting. Thanks for the color coding. Now what happens when each hemisphere is looked at separately?

  56. Separate by latitude to see differences, then by longitude to see if these differences are ‘robust’..

  57. Willis,
    I’m puzzled by your implementation of area weighting. You actually modified the values by the weighting. That is unusual, and if you want to do it that way, you should use sqrt(cosarray). You can do weighting in lm(),
    just add ,weights=cosarray (and don’t change dabs etc).

  58. The ground-based temperature measurements are, at best, a proxy measurement of the gridcell’s temperatures. They’re a point measurement of a distributed variable. The error bars are similarly the instrumental variance – not the actual estimate of deviation from the gridcell’s “true” temperature, which is necessarily larger.
    Thus, this exact same experiment would be far more interesting if applied to the satellite data instead.

  59. Nick Stokes says:
    October 8, 2011 at 10:51 pm

    Willis,
    I tried running your R code, so first some R stuff:

    3. The line starting erbeout=erbelist has been mangled, and I haven’t worked out what it should say. It looks like a problem with a < sign.

    Thanks for the comments, Nick. Let me try it again:
    erbeout=erbelist[,13][which((erbelist[,1]>-65) & (erbelist[,1] <65 ) )]
    If there are more problems, let me know.

    However, I’ve looked at the data, and it seems, re mt, that you are comparing just the monthly averages of the 1961-1990 temp with the 1985-9 ave of albedo. So it’s just the seasonal change averaged over those years, over the many grid cells.
    I can’t see how you can interpret that averaged seasonal change as a feedback, or indeed attach any useful causality. The temp rise is of course mainly from the sun angle, and while that rise could then have an effect on clouds, there is obviously much else that changes seasonally – for example, monsoons, movement of the westerly airstreams. And your analysis heavily emphasises the localisation – temp in a gridcell vs could forcing in a gridcell, whereas cloud systems are likely responding to temperatures elsewhere as well – they do blow around.

    If you were correct, Nick, then the results would be random, with no relationship between ∆T and ∆F. But in fact there is such a relationship. For example, big drops in temperature are usually associated with a rise in cloud warming, and hardly ever associated with more cloud cooling. In other words, when it gets colder, the response of the clouds is a warming.
    If you have an alternate explanation for that statistically significant relationship, I’m all ears … you are right that the clouds are responding to a host of things, seasons and monsoons and the like. But with all of that … large temperature drops almost never are associated with clouds cooling, only with warming.
    I will indeed see about getting the 1985-89 HadCRUT anomaly data, so I can add it back in to the 61-90 absolute data and then re-average to get close to the 1985-89 values. I would be surprised if it makes a difference, however, because the anomalies are much smaller than the monthly variations.
    I also have another absolute dataset that I want to test it against. Unfortunately, it’s land-only, but it should be interesting.
    w.

  60. Willis,
    As I am not mathematician, or a scientist at that, I have only my instinct and common sense to go with.
    I have this nagging question concerning the underlying of the whole idea of cloud feedback:
    You state that clouds have a positive feedback at nights. My common sense tells me clouds only mitigate the cooling should there be no clouds. To say clouds have positive feedback would say the atmosphere actually gets warmer at night. I was taught it retains the warming by “blanketing” the surface.
    What am I seeing wrong here?

  61. If cloud feedback was positive – as stated by the IPCC – it is difficult to see how exponential heating of the Earth’s surface would not occur. The argument is more CO2, therefore more heat, therefore more evaporation and therefore more clouds and therefore more heat and so on.
    The geological record shows this has never happened, so there has to be a balancing mechanism in the Earth’s climate – negative cloud feedback makes perfect sense.

  62. From the standpoint of feedback considerations, it might be important to distinguish between two cloud types, which are fundamentally different. The first type I will call an atmospheric advection cloud. These are created by the interaction of two different air masses moving against each other. Thus they have little thermal contact with the surface except via LWIR. Surface conditions have little to do with the formation of these clouds, so I would expect minimal direct ground feedback in this case.
    The second type I will call a surface convection cloud. As the formation of these clouds do depend on surface temperature, I would expect a strong, probably negative, feedback associated with them. So, in studies of this kind, I think the word ‘cloud’ by itself may be incomplete if we do not make a distinction between surface convection clouds and atmospheric advection clouds.

  63. Willis Eschenbach says:
    October 9, 2011 at 12:12 am
    For example, big drops in temperature are usually associated with a rise in cloud warming, and hardly ever associated with more cloud cooling. In other words, when it gets colder, the response of the clouds is a warming.
    This looks to me like it could be a visual of the known water cycle, (which is excluded from the AGWhype because it doesn’t suit to have negative feedback in CO2 is driving temps). With the water cycle taken out of the atmosphere temp would be 67°C. That’s a lot of cooling by the water cycle taking up heat from the surface in water vapour and forming clouds, (clouds form when water vapour gets cold enough, at dew point it condenses out into water to form clouds so the heat is already released and temps beneath clouds will drop). Cloud cover impeding convected rising heat may well temporarily keep temps higher beneath, but prolonged such will impede the Sun’s thermal infrared from reaching the surface to heat it to continue the cycle, a ‘year without a summer’ type situation could result. Regardless rising CO2 ‘backradiating’ like mad..

  64. I added Northern and Southern Hemisphere graphs to the head post, colored by latitude.
    w.

  65. Willis Eschenbach says:
    October 9, 2011 at 12:12 am
    “But in fact there is such a relationship. For example, big drops in temperature are usually associated with a rise in cloud warming, and hardly ever associated with more cloud cooling. In other words, when it gets colder, the response of the clouds is a warming.”
    “cloud warming” and “cloud cooling”.
    Are you sure you don’t mean more cloud formation and less cloud formation?
    And if so, how is that any kind of revelation?
    Water vapour transports energy, it doesn’t add energy. Clouds are a channel or a conduit, they are not a “forcing”. The term “forcing”, in respect to the atmosphere, is a logical fallacy. To claim that clouds, and even water vapour can cause warming is a circular argument.
    Temperature is simply a measure of perceived or perceivable, thermalised energy. The water vapour content of the atmosphere, is not determined by the water vapour content of the atmosphere. It is determined by almost everything but.
    I live in the UK where we are immersed in the Gulf Stream. The Gulf Stream keeps the UK unseasonably mild. This extra energy is not provided by the clouds, it is merely transported by them from the warmer region of the Gulf of Mexico near to the Equator.
    Clouds may effect localised temperature but they cannot and do not effect the global “energy budget” as it is referred to. Clouds effect diffusion of incoming and out going energy, not the amount of energy. The confusion originates in the fallacy that all incoming solar EMR is SW. It is more than 50% LW
    Clouds intercept incoming solar EMR. Some of the SW frequencies are reflected, while more of the LW frequencies are diffused through the system. This diffused energy has been rebranded DLR. Yet it is simply diffused solar EMR. Near the Equator some of the energy emitted by the ground causes some atmospheric warming but, over most of the globe the opposite is the case. As can be seen in the global radiosonde data.
    Like it or not Willis, these facts are not conducive to the current “greenhouse effect” hypothesis.

  66. Willis, I find the difference between North & South extremely interesting, it shows your “Thermostat” very well.
    Over the sea in the South and towards the equator in the North the values are much less scattered.
    The effects of the land masses in the North seem to be causing a lot of scattering.

  67. “In other words, when it gets colder, the response of the clouds is a warming.”
    Well, you don’t know that – correlation is not causation. It could equally be that the clouds cause the cooling (despite the sign of the forcing). But more likely the common root cause – the seasonal rise and fall of the sun. When the sun declines, it cools and you get more clouds, which do give a positive forcing, but which can’t outweigh the loss of sunlight.
    Personally, I’m not bothered by the different time periods – as you say, you’re just looking at climatologies.

  68. Someone needs to volunteer to adjust the GCMs to predict the observations that you are using.

  69. @Willis Eschenbach
    You make a point about cloud and DLR:
    “Net cloud forcing is the balance of how much solar energy the clouds reflect away from the earth on the one hand, and on the other, how much the same clouds increase the “greenhouse” downwelling longwave radiation (DLR).”
    Then a big generalized leap:
    “That is to say, to warm the surface by 1°C, the amount of incoming energy has to increase by about 5 W/m2. ”
    This generalization does not hunt. The type of energy and type of surface makes a difference. For instance if the surface is ice and the energy is shortwave it will raise the temperature very little. If the surface is water and the energy is shortwave it will raise the temperature very much.
    More to the point, if the energy is longwave and the surface is water it will not raise the temperature very much. This is predicted by the physics and experimentally confirmed.
    http://tallbloke.wordpress.com/2011/08/25/konrad-empirical-test-of-ocean-cooling-and-back-radiation-theory/
    This is extremely relevant to your hypothesis, Willis. Snipping challenges to underlying assumptions in your thesis is not very scientific.
    I’m putting this comment in a macro and will repeat it ad infinitum in every thread of yours until you acknowledge it. If that doesn’t work I’ll start placing it in threads you don’t control.
    Are we clear on what I think about censoring critics and how I intend to respond to you now?
    [UPDATE: Dave, you apparently have mistaken me for someone who gives a flying flock about your ideas. If you wish to make a fool of yourself by screaming, thinking that repeating your claims more loudly or more frequently will somehow make them true or will convince people to pay attention to you, I fear I can’t help you.
    So, this time I’ll leave your comment here, in the hopes that it will convince you to be polite and go away. You remember “polite”? That’s where, when someone says “Please discuss that claim on another thread, we’re discussing something else on this thread”, you say “OK”, and you discuss it elsewhere.
    If you want to prove that you are a jerk, Dave, go ahead and keep repeating your claims, and persevere in trying to butt into a conversation where those kinds of claims are not being discussed. However, people here won’t get to read them, I’ll snip them all day long. WE ARE DISCUSSING ANOTHER MATTER HERE, how hard is that to understand?
    So if you wish to pound yet again on your one-note drum, have the common courtesy and simple decency to do it somewhere else. I will snip it here, but not because it is scientific nonsense. It is nonsense, but that’s not the reason. I will snip it because we are talking about something else on this thread, and your contribution is neither appropriate nor wanted.
    PS: Threats don’t work with me. You say that you are “putting this comment in a macro and will repeat it ad infinitum in every thread of yours until you acknowledge it.” Yeah, I’m shaking in my boots with fear now, that’s a brilliant plan, everyone appreciates being spammed, the lurkers will love you, and you’ll never get banned for that kind of childishness … you don’t seem to get it. REPEATING YOUR CLAIM DOESN’T MAKE IT TRUE. So go ahead, spam your message on every site you can find. A site discussing cross-stitching methods for needlepoint? Perfect, you’ve proven you don’t care in the slightest what others want to talk about, you are an equal-opportunity bore and you insist on your right to talk “radiation can’t heat the ocean” no matter what the thread is about. So go ahead post your claims on the cross-stitching site and every other one you can find, that’ll show the world you’re not a man to be messed with … – w.]

  70. @Willis Eschenbach
    [SNIP – please discuss your claims that radiation can’t heat the ocean elsewhere. -w.]

  71. Hi Willis. I’m not getting fig 3 showing up. Also legend for fig 4 Southern hemisphere gives colours for regions from the North to the equator. Last time I looked, the equator is in the North in the Southern hemisphere.
    [Thanks, fixed the note. Figure 3 shows up for me, is anyone else having problems with it? – w.]

  72. I can only agree that an analysis of this nature is long overdue. Of course we are all happily wagging our tails because it seems to confirm what many (most?) of us here believe. However, I would feel a lot more confident of the validity of the outcome if the suggested sub-sets of land/water, day/night, etc could somehow be done. Should such individual results hold form in the sense of intuitive logic, the value of the overall outcome here would be vastly strengthened. On the other hand, should individual analyses give counter-intuitive results, it would deminish the overall result as perhaps being an artifact of some sort. Whatever such outcome, this certainly is an area where some serious grant money would be well spent!

  73. @Willis Eschenbach
    You missed the memo from NASA.
    [SNIP – you missed the memo about me snipping off-topic comments. w.]

  74. Are we seeing a ‘corkscrew’ plot as we march from the poles to the equator?
    Does the slope revolve like a propeller as we move from cold-pole to hot-equator?

  75. hi willis,
    it might be useful to run a series of fits sliced by latitude. this would also remove the need for area weighting. it would also give u some sense of the stability of the effect at least on a spatial basis.
    d

  76. Philip Bradley says:
    October 8, 2011 at 6:43 pm
    Maybe I am missing something here, but I have difficulty seeing how a month on month temperature change can cause a month on month change in cloud forcing.
    I’d expect the cloud forcing to be a direct function of temperature. Warmer = more clouds = increase in the cloud forcing. What matters is the temperature, not whether the temperature has increased/decreased.
    So I’d expect to see a similar (or larger) effect if cloud forcing were plotted against monthly temperature.
    >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
    That is what I saw looking at afternoon thunderstorm formation in the summer along the east coast of the USA. The number of afternoon thunderstorms per week decreased going from Florida to Fayetteville NC/Rocky Mt. This seemed to be the Northern most point with numerous summer thunderstorms. It is also the “Snow Line” that is the point where you can expect at least some snow each winter.
    Other independent analysis:
    “…Pall ́ et al (2004a) correlated the earthshine data with International Satellite Cloud Climatology Project (ISCCP) cloud data to construct from the latter a proxy record of the Earth’s reflectance. They showed from that proxy that the Earth’s albedo decreased by about 6 W/m2 from 1985 to 2000, while direct earthshine observations from 1999-2003 revealed that the decline had stopped and even reversed to an increasing trend in reflectance. The ISCCP project however, recently released the FD product, which contain estimates of the top of the atmosphere (TOA) albedo. These data are based on ISCCP cloud properties, plus some modeling estimates to derive shortwave and longwave fluxes, and they indicate a more muted long-term albedo decrease from 1983 to 2000…. Meanwhile, data collection has continued, and in the past couple of years, all datasets in the aforementioned literature have been re-calibrated and/or re-analyzed, so it is now proper to make a new comparison….
    Over the 1999-2007 period, there is an increasing reflectance trend from 1999 to 2003, but after that year the reflectance does not seem to vary. In fact, except for 2003, the Earth’s reflectance as measured by earthshine has not changed since 2001.”
    http://bbso.njit.edu/Research/EarthShine/literature/Palle_etal_2008_JGR.pdf
    This indicates a decrease in cloud cover/albedo during a time period when the overall temp was increasing and an increase/plateau when the overall temp “stagnated” see Dr. Roy Spencer’s Data: http://www.drroyspencer.com/wp-content/uploads/UAH_LT_1979_thru_September_2011.png.
    I think the correlation is badly confounded because of the effects of other factors and different cloud types. It is interesting that the southern hemisphere data is much more tightly clustered compared to that of the northern hemisphere perhaps indicating the ocean effect.
    I agree that separating land vs ocean data should be the next step. We already know that the oceans effect temperature from the other work that has been done.
    AMO+PDO= temperature variation – one graph says it all http://wattsupwiththat.com/2010/09/30/amopdo-temperature-variation-one-graph-says-it-all/

  77. In my view snipping comments because the thread originator decides they are off topic is not a good direction for this site . . . . . on the grounds of abuse, stalking etc it definitely makes sense. It leaves wattsupwiththat open to charges of censorship. . . probably the last thing it needs in the climate debate. . . And yeah this comment is off topic . . but it would be pretty pointless to snip it though . . . if there’s widespread disagreement with it then fair enough you’ll have snipping policy support for the future . . . . .

  78. I think that in figure 4 south should replace north.
    Figure 4. Southern Hemisphere Cloud Feedback. Color of the dots indicates latitude, ranging from blue at the furthest south through yellow in the subtropics, to red at the equator. [Thanks, fixed. w.]

  79. How does this work match up to Roy Spencer’s calculations.
    I do find it frustrating different scientists can use the same data and equations and yet seem to come up with diametrically opposed answers, such as Spencer and Dessler.

  80. Paul Homewood says:
    October 9, 2011 at 6:06 am
    How does this work match up to Roy Spencer’s calculations.
    I do find it frustrating different scientists can use the same data and equations and yet seem to come up with diametrically opposed answers, such as Spencer and Dessler.
    >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
    Welcome to the workings of true science.
    It is perfectly acceptable for different scientists to have different interpretations of the same data and form different hypotheses, as long as NO ONE is censored and it leads to experiments to prove or disprove those hypotheses
    The problem comes in when a clique of scientists decide to censor anything that does not agree with their thoughts.
    Science advances one funeral at a time. — Max Planck

  81. Willis, both graphs 3 & 4 have the same note about coloring of dots, description for #4 cannot be right. Interesting how tight the SH is compared to NH. One might assume the difference is land mass ocean ratio, and/or a difference in cloud cover related to that ratio. I would think on land there are a lot of differences in localized thermal gradients, whereas the ocean would be more homogenous in albedo and topography. The topography of the land causes changes in water vapor, for example, air moving over mountains loses water content. From the first graph, it is interesting that solid water 0°C cools as clouds.

  82. Steven Mosher,
    Have you ever thought of doing aerosol masks? I’ve been thinking it’d be good to mask areas where there are large concentrations of man made and land based aerosols good for cloud nucleation where levels are consistently high). Then look at clouds, crf, and albedo and ocean heat.

  83. Willis,
    I’m out of my league here – poor in math and knowing nothing of climate science. So let me just jump in with both feet firmly in my mouth:
    From Wikipedia, approximately 53% of incoming radiation from the sun is infrared.
    From NSIDC(.org) approximately 30% of incoming solar shortwave radiation is reflected back into space. The remaining 70% strikes the earth and is re-emitted as longwave (infrared) radiation.
    ———————————–
    The above stuff I got from the Internet, so if I’m wrong already, forgive me.
    ———————————–
    Assume 1,000 Watts of radiation strike the earth from the sun.
    530 Watts is longwave and will have a positive forcing (warms the earth).
    470 Watts is shortwave of which 30% is reflected back, leaving 70% or 329 Watts to strike the earth and be re-emitted as longwave (infrared).
    Regarding the clouds, do they prevent the incoming shortwave (the 530 W) from reaching the earth’s surface to the same extent that they will prevent the outgoing shortwave (the 329 W) from leaving the earth?
    Seems to me that clouds would always have a negative forcing (cooling effect) when incoming infrared is greater than outgoing, at least to a first order approximation.
    So what is wrong with my logic here?

  84. jim hogg says:
    October 9, 2011 at 5:46 am
    I“n my view snipping comments because the thread originator decides they are off topic is not a good direction for this site . . . . . on the grounds of abuse, stalking etc it definitely makes sense. It leaves wattsupwiththat open to charges of censorship. . . probably the last thing it needs in the climate debate…. “
    ___________________________________________________________________
    Actually I do not see it that way at all.
    Willis stated very clearly that he wished the thread to stay on topic and focused because he is looking on feed back about his work. As he said there are plenty of other posts on WUWT where those who wish to can discuss the topics he wishes to avoid.

  85. Does the difference in slopes between the last two graphs show a real world effect? Could this be part of a real description of the planetary climate engine?

  86. *****
    Paul Westhaver says:
    October 8, 2011 at 6:40 pm
    100W/m^2 with NO TEMP change… I’d like to explain that!… How can you have -100 W/m^2 and no temp change?
    ****
    Not unprecedented. In winter here, there can be a day or two of very low clouds, fog & rain. Little solar input to the surface. Afterwards an Arctic high moves in w/crystal-clear skies (lots of solar input), but the temp drops to well-below freezing.
    Large or fast airmass movement can overcome the general tendency of more watts-input = higher temps, at least on a local scale & for limited time.

  87. Willis, thanks for adding the latitude and hemisphere data. As expected, it clearly shows the temperature moderating effect of the Southern oceans, relative to the large continental land masses in the North. I find the robustness of the R=-0.27 R2=0.08, regardless of how the data is sliced to be remarkable, and would want to double and triple check it. We know from Dr. Spencer’s post above that cloud feedback, regardless of parameter, can only explain a small fraction of the net radiative forcing, and 8 percent seems within the realm of possibility. We know from Svensmark that cosmic ray modulation of clouds can supply a radiative forcing. If clouds are the thermostat, then cosmic rays, volcanoes, ocean currents, industrial aerosols, etc. all have a finger on the dial.

  88. The paucity of points in the third quadrant, especially near the extremes of latitude, suggests that the straight line is an approximation of a true curve that is concave-down. Put into English, that suggests that negative cloud feedback becomes stronger as the temperature anomaly gets hotter, especially at temperate latitudes.

  89. Willis I like your approach – let the data speak for itself and see how it matches the science.
    A few points:
    1. My concern is not that largish grids introduce smoothing but that using monthly data does. The effects of cloud cover are, more-or-less, instantaneous and on different days of the month feedback could go either way.
    2. There is also implicit smoothing – you assumed the effect is the same for all latitudes for all times of the year which could mask seasonal trends. You have enough data to look at 5 sub-sets: northern hemisphere summer, northern hemisphere winter, tropics, southern hemisphere summer, and southern hemisphere winter. For summer you could take April to September (as close to the period between the equinoxes as is possible with monthly data).
    My hunch is that your conclusion is right but the closer you can get to showing the ‘fuzzy’ plot of negative feedback is the balance of more clearly defined positives and negatives the stronger your argument will be.

  90. Gail . . I appreciate your point but the danger is that some readers will wonder if comments are being deleted from ulterior motives . . . it also provides scope for petty smearing – the kind that travels well on the net . . .In an ideal world I’d agree with you . . . this one is far from ideal though as the climate wars have shown all too often. Honesty and fairness have got to go out of their way to be seen or heard, lest they be swamped.

  91. Dave Springer says:
    October 9, 2011 at 5:01 am
    “the downwelling shortwave radiation that is the ocean’s only source of heating. Downwelling longwave radiation has no effect on ocean temperature.”
    ———————————————–
    Dave, how is it possible that shortwave radiation is the only source of ocean heating?
    Dave, how is it possible that longwave radiation from the sun has no effect on ocean temperature?
    Thanks.
    [UPDATE – See, this is the problem. Dave starts off with his claims, and right away someone (reasonably) says “Hey, that makes no sense at all”, and then we’re off the rails. I’ll leave this here to illustrate the point, but I will snip further responses along the same line. If you want to dispute the DLR/ocean issue with Dave, please do so elsewhere. w.]

  92. Willis,
    I made an error in my earlier post. Of the radiation reaching the earth, only about 1% is longwave (> 4 um). The other 99% is shortwave. Sorry to display my ignorance but that’s how I learn.
    ———————-
    So virtually all incoming solar radiation is shortwave. That explains Dave’s comments.
    Which leads me to a new question: Are clouds better at blocking shortwave radiation (negative) or trapping outgoing longwave radiation (positive)?

  93. Nick Stokes says:
    October 9, 2011 at 3:31 am

    “Giss in absolute?”
    Since only the monthly differences are used, I don’t think it matters.

    Thanks, Nick. I’m using the monthly differences in the absolute (actual measured) temperatures. GISS, unfortunately, deals only in the data with the monthly average absolute temperatures removed from the data. As a result, even using monthly differences doesn’t get back to what I need, which is the actual difference in temperatures.
    w.

  94. it also seems particularly strained to insist on explaining conduction, phase change and convective transport in terms of radiation.
    ghg-man got bitten by a radiative spider in this comic series so long ago that now it only appears quaint and superstitious.

  95. jim hogg says:
    October 9, 2011 at 5:46 am

    In my view snipping comments because the thread originator decides they are off topic is not a good direction for this site . . . . . on the grounds of abuse, stalking etc it definitely makes sense. It leaves wattsupwiththat open to charges of censorship. . . probably the last thing it needs in the climate debate. . . And yeah this comment is off topic . . but it would be pretty pointless to snip it though . . . if there’s widespread disagreement with it then fair enough you’ll have snipping policy support for the future . . . . .

    Jim, I’ve had my threads hijacked before by endless specious claims that the DLR can’t heat the ocean. I have discussed these claims AT LENGTH, and even written a whole article about them. So there’s lots of places that are appropriate for discussing those claims. And the SIF folks that believe those ideas are more than happy to monopolize the conversation.
    If you think that refusing to discuss those ideas here is “censorship” when I have welcomed them elsewhere and even devoted an entire thread to them, then I fear I can’t help you. I will not debate or discuss whether DLR can heat the ocean or not on this thread, been there, done that elsewhere, go there if you want to discuss it.
    w.

  96. Ron Manley says:
    October 9, 2011 at 9:07 am

    Willis I like your approach – let the data speak for itself and see how it matches the science.
    A few points:
    1. My concern is not that largish grids introduce smoothing but that using monthly data does. The effects of cloud cover are, more-or-less, instantaneous and on different days of the month feedback could go either way.

    Thanks, Ron. While you are right that on different days the clouds will go different ways, the large dataset allows us to see things like that there is little data in the lower left quadrant, etc.

    2. There is also implicit smoothing – you assumed the effect is the same for all latitudes for all times of the year which could mask seasonal trends. You have enough data to look at 5 sub-sets: northern hemisphere summer, northern hemisphere winter, tropics, southern hemisphere summer, and southern hemisphere winter. For summer you could take April to September (as close to the period between the equinoxes as is possible with monthly data).
    My hunch is that your conclusion is right but the closer you can get to showing the ‘fuzzy’ plot of negative feedback is the balance of more clearly defined positives and negatives the stronger your argument will be.

    Hey, it’s early days, there’s a whole host of other slicing and dicing of the data that can be done. I wanted to get the basic idea out in the open early, so if there were issues with the analysis I wouldn’t spend time on a blind alley.
    w.

  97. Willis this looks very interesting , it’s going to take time to digest but one big problem jumps right out at me. OLS regression on this kind of data will give BADLY wrong estimates of any linear relationship that is there.
    Linear regression ASSUMES there is negligible error in the independent variable. If this is not the case the mathematical derivation of OLS simple does not apply. The net result is a REDUCTION in the estimated slope. One may imagine there is a grey area centred around the fitted line but in fact it can be shown to always reduce the estimated slope. This is called regression attenuation.
    In short , it works well for nice clean lab experiments where you can control the quantity plotted on x axis. If you have noise or other effects, like correlated or uncorrelated cyclic phenomena it will reduce the ols estimator. Neither is this just a minor error that can be ignored in a rough estimate, it may be a factor or 3 ro 4 with the sort of cloud you are looking at.
    This can be corrected for to some extent , but this requires knowledge of that nature and relative magnitude of the noise/error quantities that is often difficult or impossible. It is often better to try to avoid this happening that try to correct it.
    Ignoring this problem is ubiquitous in climate science. This is principally why Dessler, Forster & Gregory etc get a positive feedback , their maths is bad and they under estimate the slope, leading to a false conclusion of +ve feedback.
    Fitting straight lines to scatter plots basically and fundamentally wrong.
    I hope this does not undo all your hard work , but it does need to be accounted for. Now to digest the rest.
    best regards.

  98. Steve from Rockwood says:
    October 9, 2011 at 10:19 am (Edit)

    Willis,
    I made an error in my earlier post. Of the radiation reaching the earth, only about 1% is longwave (> 4 um). The other 99% is shortwave. Sorry to display my ignorance but that’s how I learn.
    ———————-
    So virtually all incoming solar radiation is shortwave. That explains Dave’s comments.
    Which leads me to a new question: Are clouds better at blocking shortwave radiation (negative) or trapping outgoing longwave radiation (positive)?

    Your question is the great unknown in climate science, and is what I’m attempting to answer in this thread.
    w.

  99. One other big difficulty with this approach is that both the data sets are time series. The time element and relationship of relative changes in the two varialbes is at the heart of causality as is rate of change. (Radiation causes a rate of change of temperature, not a simple temperature change).
    In doing this sort of scatter plot you are throwing away most of the information that relates the two quantities.
    For example , if you do a similar phase space plot (which is what this is) but joint the dots you can will see loops and lines that tell you more about how the two quantities relate. Here you lose that.
    I don’t mean to discourage you but I think these points are useful and important to realise. I went though exactly this same process some months back , so I’ve made the same mistakes and found the short-comings.
    You may find it instructive to look at dT(t)/dt since that is where the response to a radiative forcing is , not in T(t). It may also be the key to separating forcing and feedback, the latter is a T(t) dependency.
    I don’t see your objection to the cosine weighting . ERBS data is W/m2 so you must relate it to an area related term.
    As a means of reducing the 2.5 gridded data I’d suggest a 1:2:1 binomially weighted average in lat , then the same in longitude. Alternatively interpolate the 5 degree data with a Catmull-romm spline (quite easy actually)
    HTH 😉

  100. PS, one way to test what I say about regression attenuation is to invert the axes and redo the ols fit. If the result is good one slope will be the inverse of the other. I think you will get a shock how different they are. Clearly the result should not depend on which way you plot the data … unless I’m right about OLS.

  101. Here’s an example of what the two ols estimators on a scatter plot like yours:
    http://tinypic.com/view.php?pic=2cqcv28&s=7
    It’s not a total loss since the correct slope should be somewhere between the two. Though probably not at ‘in the middle’ , the two wrong results could be used to delimit a range where the right answer lies. This may be enough if you just want to show a negative slope.
    After that you have to argue about what the “slope” represents physically.

  102. Willis you are a liar and coward and you make me sick to stomach, GO FUCK YOURSELF!
    REPLY: I’m allowing this post, which will be your last, to show what kind of person you are in your own words. And this is why you are banned, permanently – Anthony

  103. Looking at the third plot there maybe some mileage in breaking down latitude bands. There you may be able to get a less hazardous data set where OLS may be more applicable.
    If you want a sea mask , you could use hadSST which has some kind of NaN value for land or partially land grid boxes.

  104. Willis, could you redo the analysis by latitude band? Splitting by hemisphere showed a fairly large difference. Perhaps the feedback strength varies by the type of clouds most prevalent in tropics v. temperate latitudes.

  105. P. Solar says:
    October 9, 2011 at 12:03 pm

    One other big difficulty with this approach is that both the data sets are time series. The time element and relationship of relative changes in the two varialbes is at the heart of causality as is rate of change. (Radiation causes a rate of change of temperature, not a simple temperature change).
    In doing this sort of scatter plot you are throwing away most of the information that relates the two quantities.
    For example , if you do a similar phase space plot (which is what this is) but joint the dots you can will see loops and lines that tell you more about how the two quantities relate. Here you lose that.

    These are gridcell monthly averages, about twenty thousand of them. Which among them are you proposing to connect with dots, and on what basis? Please think this stuff through before tossing out un-finished ideas.

    … You may find it instructive to look at dT(t)/dt since that is where the response to a radiative forcing is , not in T(t). It may also be the key to separating forcing and feedback, the latter is a T(t) dependency.

    Please re-read what I have done. ∆T is the change in temperature with time, which is the discrete form of the derivative you are using, dT(t)/dt. No way to put this politely. You’re suggesting that I do what I just did, you’re not paying attention.

    I don’t see your objection to the cosine weighting . ERBS data is W/m2 so you must relate it to an area related term.

    Nick Stokes above already pointed out the proper way to deal with the weighting. Read the thread.

    As a means of reducing the 2.5 gridded data I’d suggest a 1:2:1 binomially weighted average in lat , then the same in longitude. Alternatively interpolate the 5 degree data with a Catmull-romm spline (quite easy actually)

    Lots of possible ways to do it, but since Nick solved the problem already, not necessary … plus the fact that there’s no way I’d do a smooth as you suggest, that’s just losing information.
    w.

  106. P. Solar says:
    October 9, 2011 at 12:11 pm

    PS, one way to test what I say about regression attenuation is to invert the axes and redo the ols fit. If the result is good one slope will be the inverse of the other. I think you will get a shock how different they are. Clearly the result should not depend on which way you plot the data … unless I’m right about OLS.

    Report back with your results …
    w.

  107. P. Solar says:
    October 9, 2011 at 1:03 pm

    If you want a sea mask , you could use hadSST which has some kind of NaN value for land or partially land grid boxes.

    Brilliant, I love unconventional solutions like that. I’ll take a look.
    w.

  108. Willis Eschenbach says:
    October 9, 2011 at 11:31 am
    Steve from Rockwood says:
    October 9, 2011 at 10:19 am (Edit)
    Willis,
    I made an error in my earlier post. Of the radiation reaching the earth, only about 1% is longwave (> 4 um). The other 99% is shortwave. Sorry to display my ignorance but that’s how I learn.
    ———————-
    So virtually all incoming solar radiation is shortwave. That explains Dave’s comments.
    Which leads me to a new question: Are clouds better at blocking shortwave radiation (negative) or trapping outgoing longwave radiation (positive)?
    Your question is the great unknown in climate science, and is what I’m attempting to answer in this thread.
    w.

    How much would this change what happens to the clouds?

    http://journals.ametsoc.org/doi/abs/10.1175/2010JAS3282.1
    Abstract
    This paper focuses on two shortcomings of radiative transfer codes commonly used in climate models. The first aspect concerns the partitioning of solar versus infrared spectral energy. In most climate models, the solar spectrum comprises wavelengths less than 4 μm with all incoming solar energy deposited in that range. In reality, however, the solar spectrum extends into the infrared, with about 12 W m−2 in the 4–1000-μm range. In this paper a simple method is proposed wherein the longwave radiative transfer equation with solar energy input is solved. In comparison with the traditional method, the new solution results in more solar energy absorbed in the atmosphere and less at the surface. Li, J., C. L. Curry, Z. Sun, F. Zhang, 2010: Overlap of Solar and Infrared Spectra and the Shortwave Radiative Effect of Methane.

  109. aaron
    “Have you ever thought of doing aerosol masks? ”
    OT. willis wants a tight thread, lets give him that seeing as he has devoted a bunch of time to the work

  110. FYI
    Longwave infrared (LWIR) begins at 5 micrometers. For all practical purposes sunlight contains no LWIR energy. What little it does have never makes it to the surface as greenhouse gases block it. Also for all practical purposes the earth emits nothing but longwave infrared. There is a wide stark divide in the frequency domains of sunlight and thermal emission from the earth.
    Moreover, different materials, not just just greenhouse gases but surface materials such as ice, water, rocks, soil, and vegetation have radicially different absorption and emission characteristics across all frequency ranges. It borders on ludicrous to lump all electromagnetic energy into “Watts” and all matter into “surface” or “atmosphere”. When you make simplifications that border on ludicrous you get answers that border on ludicrous. Garbage in, garbage out. All the fancy statistical math in the world won’t restore highly critical details of how EMR interacts with different forms of matter once they are left out.

  111. Peter Miller says:

    If cloud feedback was positive – as stated by the IPCC – it is difficult to see how exponential heating of the Earth’s surface would not occur. The argument is more CO2, therefore more heat, therefore more evaporation and therefore more clouds and therefore more heat and so on.

    The piece you are missing is that infinite series sometimes converge to finite values. For example, the geometric series 1 + (1/2) + (1/4) + (1/8) + (1/16) + … converges to the value 2. Hence, if a positive feedback responded to a warming by CO2 of 1 C by producing an additional half C of warming, and then the feedback responded to that half C of warming by producing an additional quarter C of warming, and so on…then what you would end up with is not exponential heating but simply the multiplication of the original heating by a factor of 2.

  112. Steve from Rockwood says:
    October 9, 2011 at 9:37 am

    Dave Springer says:
    October 9, 2011 at 5:01 am
    “the downwelling shortwave radiation that is the ocean’s only source of heating. Downwelling longwave radiation has no effect on ocean temperature.”
    ———————————————–
    Dave, how is it possible that shortwave radiation is the only source of ocean heating?

    Because the sun is its only significant source of energy and for all practical purposes there is zero longwave infrared energy in sunlight. But I see you already discovered for yourself that I was correct.
    Now you need to discriminate between the ways by which the ocean absorbs and gives up energy. Radiation from the sun is the only thing that penetrates it. That is it’s only source of energy. Longwave infrared is totally blocked by a surface layer that is only a few microns thick. This skin doesn’t mix downwards. Viscosity overwhelms all other forces in skin layer. The skin layer evaporates. Evaporation is the primary way that the ocean gives up absorbed solar energy – 70% of it. 10% is given up by conduction and the remaining 20% is via radiation. This is RADICALLY different from the way land and ice gain and lose energy. There is no comparison between the two. Greenhouse gases are quite effective over land because their mode of operation is via radiation and radiation IS the primary way that land surfaces give up solar heat. Radiative emission from the ocean on the other hand is not the primary means of giving up energy so greenhouse gases play commensurately smaller role there.
    Eschenbach refuses to acknowledge the difference between land and water and how they are quite different in the way they absorb and release solar energy. He’s decided that bringing up actual, well established physics that are counter to his anedotally acquired beliefs is “thread hijacking” and has promised to deprive you of this information so that he may pursue his ridiculous wild geese flying through the clouds.
    It’s really kind of amusing. An ostrich with his head stuck in the sand chasing after wild geese…
    I’ll clip this comment and post it to tallbloke’s website. Tallbloke has forgotten more science than Willis ever knew.
    http://tallbloke.wordpress.com/2011/08/25/konrad-empirical-test-of-ocean-cooling-and-back-radiation-theory/

  113. @Myrrh
    > How much would this change what happens to the clouds?
    > http://journals.ametsoc.org/doi/abs/10.1175/2010JAS3282.1
    It doesn’t change anything. Total solar insolation is 1200 Watts at top of atmosphere. Energy in longwave infrared range (>5um) is 1% of that or 12 Watts. Virtually none of that 12 Watts makes it to the cloud deck as it is absorbed by O3, CO2, CH4, H2O, sulfate aerosols, nitrogen compounds, and a host of other molecules between the cloud deck and TOA.

  114. Dave, how is it possible that shortwave radiation is the only source of ocean heating?
    Dave, how is it possible that longwave radiation from the sun has no effect on ocean temperature?
    Thanks.
    [UPDATE – See, this is the problem. Dave starts off with his claims, and right away someone (reasonably) says “Hey, that makes no sense at all”, and then we’re off the rails. I’ll leave this here to illustrate the point, but I will snip further responses along the same line. If you want to dispute the DLR/ocean issue with Dave, please do so elsewhere. w.]
    UPDATE TO UPDATE No Willis, if you had the first clue about the physics you’d know that the response was not reasonable. There is practically no longwave infrared energy in sunlight. The responder apologized and corrected himself later in this thread for his error. ow if only Willis had the means, motive, and self-esteem to correct himself we could make some progress and people could actually learn some science from one of Willis’ misadventures into statistical analysis which, by all indications, is another anecdotal skill he picked up via osmosis while either surfing, sailing, fishing, trying to get a dishonarble mental discharge from the Army, or whatever his source of knowledge which decidedly does not include going to school or studying any science textbooks.
    [Reply: You are invited to submit an article of your own if you wish. You might be surprised at the comments. ~dbs, mod.]

  115. Willis,
    Good work. Applause, applause.
    you wrote: That’s why we invented mathematics, all right. Part of the problem that there are over eighteen thousand data points, so the underlying pattern is obscured. That doesn’t mean it’s not there, just that this particular visualization breaks down under the weight.
    Usually, the more information you have on a graph, the better. This might (repeat might) be one of the exceptions: I recommend a separate panel for each region. Also, I recommend a separate analysis for each region, with a test of whether the regression coefficient is the same in all regions. Lastly, I recommend a region BY month breakdown, and a test of whether the slope varies across the year, between regions, of if there is a region BY month interaction. These are not criticisms, these are requests for follow-up analyses.
    Mike McMillan wrote this: Perhaps a contour map of point density would help.
    Think of our recommendations as complimentary, not conflicting.
    It’s true that the R^2 value is small, but small effects over large regions and long times make non-trivial differences. Besides, the sign is opposite to the “conventional wisdom”.
    Kevan Hashemi wrote: Clouds reflect sunlight back into space. Whatever effect they have upon long-wave radiation is entirely secondary to the fact that they block the sun’s heat from reaching the surface of the planet. Here’s a simulation that shows the effect:
    It is important to have actual data backing up important claims. Willis has actual data at the “explantory level” that counts: actual clouds and actual temperatures.
    Dave Springer wrote: R^2 = 0.11 means there’s no significant pattern.
    And yet, the AGW promoter Andy Dessler got an even smaller R^2, probably with the wrong sign, published in Science.
    Dave Springer wrote: This generalization does not hunt. The type of energy and type of surface makes a difference. For instance if the surface is ice and the energy is shortwave it will raise the temperature very little. If the surface is water and the energy is shortwave it will raise the temperature very much.

    More to the point, if the energy is longwave and the surface is water it will not raise the temperature very much. This is predicted by the physics and experimentally confirmed.

    That’s pretty incidental to the empirical relationship that Willis has reported. Exactly how much of this is longwave, and exactly how much is shortwave, and exactly how much is on land and exactly how much is at sea, and exactly how much is am and how much pm (see Willis’ work on the TAO/TRITON data) can be subjects for subsequent research and analysis.
    Willis, I see that you got requests for more work, not just mine. Have fun.
    And keep up the good work.

  116. Willis says: “These are gridcell monthly averages, about twenty thousand of them. Which among them are you proposing to connect with dots, and on what basis? Please think this stuff through before tossing out un-finished ideas.”
    First of all Willis, don’t get pissy. I am trying to make useful and constructive criticism. I’ve already done this kind of analysis and more and can probably point out some errors and make useful suggestions. If anyone is ” tossing out un-finished ideas” it’s you , as you readily admit in your post.
    You specifically asked for people to check over what you’d done so don’t get offended when they do.

  117. Just looking at the pretty colors 🙂 it would appear that some of the commenters are correct: you need to divide it up between landmass and ocean.
    As a preliminary proxy for that, do latitude bands. 0 – 65 at 5 degree cells gives 13 cell bands. Do it 5/4/4, that is, 0-25, 25-45, and 45-65. I predict that the 0-25 band will look very much the same whether Northern or Southern hemisphere, but the others will be quite different because the land distribution is quite different. You really want to separate out land and ocean, but bands might be simpler as a first cut.
    Regards,
    Ric

  118. Willis says: These are gridcell monthly averages, about twenty thousand of them. Which among them are you proposing to connect with dots, and on what basis? Please think this stuff through before tossing out un-finished ideas.
    Why don’t you think it through before dismissing it? Part of your problem is dumping that much incompatible data it one bucket. Read my posts , you may learn something. Don’t let it go to your head that Anthony let you post on his blog.
    Please re-read what I have done. ∆T is the change in temperature with time, which is the discrete form of the derivative you are using, dT(t)/dt. No way to put this politely. You’re suggesting that I do what I just did, you’re not paying attention.
    Yeah, but you are plotting against dR/dt , so once again you have not even though about what I said before rejecting it.
    I don’t see your objection to the cosine weighting . ERBS data is W/m2 so you must relate it to an area related term.
    Nick Stokes above already pointed out the proper way to deal with the weighting. Read the thread.
    This thread was already >110 posts by the time I got to read it. Most of that is the usual garbage , bitching and inane comments. Life’s too short. I did scan the whole thread for anything you’d posted and picked up on help Nick had provided you with.
    Your post said you were still unhappy about the idea of weighting, I was saying it was correct to do it and why. Again I see no reason for your off-hand response.
    As a means of reducing the 2.5 gridded data I’d suggest a 1:2:1 binomially weighted average in lat , then the same in longitude. Alternatively interpolate the 5 degree data with a Catmull-romm spline (quite easy actually)
    Lots of possible ways to do it, but since Nick solved the problem already, not necessary … plus the fact that there’s no way I’d do a smooth as you suggest, that’s just losing information.
    If you want to move from a 2.5 grid to a 5 deg grid, how do you imagine you can do that without “just losing information”? The fact that you later throw away the whole time dependency of BOTH datasets makes you comment about losing information rather comical.
    Don’t get your back and get narky, just read the posts, you may learn something. If you are unable to learn anything I’m sorry, I tried let you benefit from some of the things I’d learnt in doing a similar analysis, I don’t intend to spend my days arguing about it.
    best.

  119. PS, one way to test what I say about regression attenuation is to invert the axes and redo the ols fit. If the result is good one slope will be the inverse of the other. I think you will get a shock how different they are. Clearly the result should not depend on which way you plot the data … unless I’m right about OLS.
    Willis says: “Report back with your results …”
    My result IS just this, that doing OLS on this kind of data is fundamentally wrong and I’ve said why. I don’t need to reproduce your errors to prove it to you. Do it yourself. [hint reverse the arguments you pass to lm() ]
    I did post a link to a graph where I have done this on similar data and and link to a thread where I covered this with TroyCA. He did listen and agreed.
    If you want to dismiss the “positive feedback” results of Dessler, F&G et al this is the easiest and most incontrovertible route. Since the results of this flawed method are the only empirical justification for positive feedback in models and that is the only reason the models produce alarming warming this seems to me to be the crux of the whole AWG battle.

  120. To spell this out: Dessler (and others) do a similar scatter plot method and get an incorrect slope for similar reasons. It is a mathematical result that the ols estimator WILL be less than the true “slope”. How much less depends on the amount of noise. The data has more noise than signal. The regression attenuation will be very significant.
    Since the ols value is less than the neutral 3.3 W/m2/K of the plank response he comes to the erroneous conclusion that he has demonstrated positive feedback.

  121. “”””” Steve from Rockwood says:
    October 9, 2011 at 10:19 am (Edit)
    Willis,
    I made an error in my earlier post. Of the radiation reaching the earth, only about 1% is longwave (> 4 um). The other 99% is shortwave. Sorry to display my ignorance but that’s how I learn.
    ———————-
    So virtually all incoming solar radiation is shortwave. That explains Dave’s comments. “””””
    Well take heart Steve; ignorance is NOT a disease; we are all born with it. The Library of Congress is one of the world’s largest repositories of information about most of which, most of us are totally ignorant.
    But I’m happy to see that you found some correction for your earlier statement that 53% of the solar spectrum was “longwave”.
    It would be useful, if persons who are interested in this whole climate energy business acquainted themselves with certain rather basic pieces of information. The first I would suggest is the specific shape of the Planck Black Body radiation curve.
    The version I would most strongly recommend is the universal normalized BB curve as published in the standard Optics Text; Modern Optical Engineering, by the late Warren J Smith, who worked for many years at Infra_Red Industries in Santa Barbara California.
    He plots the Spectral Radiant Emittance (W.cm^-2.micron^-1) as a fraction of the peak value at the spectrum peak, as a function of the wavelength also scaled to the spectral peak wavelength.
    So the relative wavelength goes from 0.15 to 50, and the vertical axis has two scales ; a linear one from 0.0 to 1.0, and a logarithmic one from 1.0 down to 10^-5. From this one can deduce the behavior of almost any thermal radiator, at least qualitatively and often quite quantitatively.
    Some important general properties (of BB radiation) are.
    1/ (almost) exactly 25% of the total energy is radiated at wavelengths shorter than the peak wavelength, and 75% at longer wavelengths.
    2/ Only 1% of the total energy survives at wavelengths shorter than 0.50 times the peak wavelength.
    3/ 50% of the total energy is emitted below about 1.5 times the peak wavelength (closer to 1.45).
    4/ Only 1% of the total energy is emitted at wavelengths longer than 8 times the peak wavelength.
    5/ The Spectral Radiant Emittance is down to 10^-5 at 0.20 times the peak wavelength and 40 times the peak wavelength.
    The other most important relationship is the Wien Displacement Law, which says that the Temperature (Kelvins) times the peak wavelength, is a constant (2897.8 Kelvin.microns.
    Apologies to the chemists and molecular spectroscopists who think in terms of Wave numbers (frequency) rather than wavelength. Despite the rationale for using frequency; given that
    E = h.(nu) , I have yet to see my first plot of the solar spectrum plotted on a frequency scale.
    Translated to the solar spectrum TSI, in its near BB guise at roughly 6,000 K. Other sources give 5900 K as the best fit BB curve fit, or 5770 as the value best meeting the TSI value which was 1353 Wm^-2, when that was published, so likely 5800 K at today’s 1366 Wm^-2.
    So Wien’s law gives about 0.5 microns for the TSI spectral Peak, so 98% of the energy lies between 0.25 microns, and 4.0 microns. and the 50% division is at about 740 nm.
    As for “visibility”, The commission on Colorimetry of the Optical Society of America cites Standard Luminosity values for the Photopic Luminosity curve of the eye from about 385 nm to 760 nm. at which points the value is 10^-4. This is not the extreme of the eye sensitive range. The dark adapted eye is capable of seeing single photons, at least in the blue-green range, so 10^-4 is not invisible. Warren Smith gives eye sensitivity values in the near infra-red out to 1.1 microns wavelength, where the sensitivity is 10^-11..
    I tend to place the solar energy “short wave” radiation as being the 200 nm to about 2.0 microns range even though about 7% of solar energy survives past that. But the “long wave” part of the infra-red spectrum has some effect from about 2.0 and certainly by 4.0 microns.
    A final key plot to have is the air mass one ground level solar spectrum. This plot corresponds roughly to the clear air surface maximum irradiance; before any albedo effects are considered.
    Well not all, because the first major return of solar spectrum energy to space, in elastic scattering processes, is the Raleigh scattering of the short wavelengths which result in the blue sky. The blue sky looks the same looking down on the blue planet, as it does looking up, so the blue sky you see looking ap remote from the solar disk, represents exactly the amount of solar energy in the shorter wavelengths that is immediately lost to space, and is a large part of the loss from the 1366 Wm^-2.
    The big mystery to me, is this almost childish fascination with the major climate calamity that is purportedly created from a source that is modelld quite closely by the LWIR emission from an ordinary bottle of drinking water at about room Temperature. It emits a near BB spectrum at about 400 Wm^-2 with a spectral peak of about 10.1 microns; corresponding to the average surface Temperature of the planet, as immortalized By Dr Trenberth’s amazing cartoon drawing.
    Meanwhile, there is virtually total neglect of the effect of water both as a vapor or as clouds, on the real energy source driving our climate; the energy supplied by the sun.
    FEEDBACK networks normally operate, by sensing the system OUTPUT, and modifying the system INPUT. One does not normally connect a feedback network to some largely irrelevant noisy output, which is what the surface LWIR actually is.
    On any average mid-summer day in the north, the earth can enjoy a total surface Temperature range from as low as -90 C at places around Vostock in Antarctica, up to about +60 deg C in North African and middle eastern deserts. The surface LWIR emittance covers at least an 11 to 1 range, from nearly double the sacred average of 390 W/m^2, to about 1/6th of that value.
    And that entire range can exist simultaneously; so we worry about whether or if, the average may have changed by 0.6 deg C in the last 150 years, and that from a quite corrupted and improper global sampling methodology.
    When you connect the major climate feedback (water) to the INPUT (the sun) you find, that the feedback can never ever be positive. (more) Water can only lower the solar energy that reaches the earth surface; it can never increase it. CO2 also acts to reduce incoming solar energy, but not to the extent that water does.

  122. Interesting and informative post George. However you seem to have misunderstood what ‘positive’ is all about.
    “Water can only lower the solar energy that reaches the earth surface; it can never increase it.”
    I’m not aware of anyone saying that is not the case. That part is negative. However, when you look at the similar effect of outgoing IR a similar argument would lead to saying this can only be positive. The big question is the overall, net effect on ERB.
    Unfortunately, for the reasons I’ve stated above, this cannot be resolved by doing invalid OLS regression on scatter plots.
    That Willis makes this error in a blog post is excusable. That the likes of Dr. Dessler do this in the published literature is rather less admirable.

  123. Troy’s post I linked, looking at water vapour vs CERES radiation budget had an r2 of 0.03
    He initially posted a “slope” of -0.19 , when he did it the other way around he got -6.4
    We’re not talking +/-10% here.
    Least squares does not work unless your x variable is a CONTROLLED variable.
    In Willis’ case here I expect it will give a much more negative slope the other way around. So it may establish the sign of whatever it is measuring.
    In the Dessler / F&G analysis the fundamental question is which side of 3.3 is it. This means the attenuation can change the “net feedback” from negative to positive.

  124. Figure 1 states: “Data colors indicate the location of the gridcell, with the Northern hemisphere starting with red at the far north, slowly changing to yellow through the tropics, and to red at the far southern latitudes.”
    Did you mean to mention blue dots in there somewhere?
    Jim

  125. Willis Eschenbach wrote: Report back with your results …
    Too often you get angry instead of reading. You have a nice analysis. Why ruin it by ignoring its shortcomings? P. Solar is correct: you can not estimate the slope that you want to estimate. It’s well known.

  126. Willis wrote: These are gridcell monthly averages, about twenty thousand of them. Which among them are you proposing to connect with dots, and on what basis? Please think this stuff through before tossing out un-finished ideas.
    OK, so you don’t have time. Let it go at that instead of the pseudo-intellectual dismissal.

  127. I may be completely misunderstanding what you’re doing here, but if you are subtracting the temp of the month ahead from the current month I can see why Nick Stokes implies there could be a seasonal bias – and one which would presumably apply in one direction for one half of the year in each hemisphere (days lengthening or shortening) and then in the reverse direction for the rest of the year (for that hemisphere). The hemispheres would be out of phase and you’d probably expect a 4 quadrant plot which is what you show. If you treated the hemispheres separately AND split the year into 2 parts for each hemisphere you could run the subtraction both forward and backward for each of the 4 hemisphere/season combinations. IF each forward/backward run showed the same trends you could then rule out the seasonal signal (which may be much of your ‘noise’). If there’s still a negative trend in temp with increasing cloud, you’ve got a very interesting result.

  128. “”””” P. Solar says:
    October 9, 2011 at 10:54 pm
    Interesting and informative post George. However you seem to have misunderstood what ‘positive’ is all about.
    “Water can only lower the solar energy that reaches the earth surface; it can never increase it.”
    I’m not aware of anyone saying that is not the case. That part is negative. However, when you look at the similar effect of outgoing IR a similar argument would lead to saying this can only be positive. The big question is the overall, net effect on ERB. “””””
    “”””” I’m not aware of anyone saying that is not the case. “””””
    Well P, here at WUWT, we are simply inundated with people constantly posting that that IS the case. I can count on the fingers of one hand, the number of times anyone ELSE has ever posted that.
    Essentially 100% of the input energy driving the earth climate system comes from one source; the sun. And over 75% of that input solar energy that reaches the surface, ends up getting stored in the deep oceans; as deep as 700 metres. That heat store ultimately drives the weather and the climate. To the extent that the captured solar energy decreases (due to more water in the atmosphere), the earth must cool down.
    It may be months for variations in the stored ocean energy to show up in the climate. On the other hand, variations in the escape of LWIR from earth, as a result of GHGs or clouds; or anything else, takes place in time scales many orders of magnitude faster than the transport of stored ocean energy. Those effects cannot stop the escape of that energy; only delay it. The loss of that input solar energy is however permanent; it is NOT recoverable.
    The point is that very short term changes in the atmosphere condition as it relates to GHGs or clouds, is completely swamped by the permanent change in the energy captured by the earth from the sun, as a result of those transient events.
    The LWIR emission from clouds, is isotropic, and it is trivially obvious that at least for a small isothermal cloud, the result is that 50% of that emission is lost to space, and only 50% of it returns to earth., and most of that (3/4) simply results in prompt evaporation from the ocean surface, which transports huge amounts of latent “heat” of evaporation back into the atmosphere where convection processes quickly move iot to the upper atmospher for loss to space. Yes for larger clouds the tops could be cooler than the bottoms; but even that is not assured, since the cloud tops are also directly heated by the sun. Unless the escape delay of LWIR surface emissions RAISES the amount of incoming solar energy that the earth captures, that CANNOT be considered to be a positive feedback. The input driving signal to this sytem is ONLY the sun.

  129. P. Solar had said:
    P. Solar:
    October 9, 2011 at 12:03 pm

    One other big difficulty with this approach is that both the data sets are time series. The time element and relationship of relative changes in the two varialbes is at the heart of causality as is rate of change. (Radiation causes a rate of change of temperature, not a simple temperature change).
    In doing this sort of scatter plot you are throwing away most of the information that relates the two quantities.
    For example , if you do a similar phase space plot (which is what this is) but joint the dots you can will see loops and lines that tell you more about how the two quantities relate. Here you lose that.

    Well … no, in a word. So I replied:

    Willis says: “These are gridcell monthly averages, about twenty thousand of them. Which among them are you proposing to connect with dots, and on what basis? Please think this stuff through before tossing out un-finished ideas.”

    Now Mr. Solar is back again, vis:

    First of all Willis, don’t get pissy. I am trying to make useful and constructive criticism. I’ve already done this kind of analysis and more and can probably point out some errors and make useful suggestions. If anyone is ” tossing out un-finished ideas” it’s you , as you readily admit in your post.

    You likely can “point out some errors”, P. Solar, I can learn from anyone. But what happened was you got so bothered you forgot to answer my question, despite quoting it in your post. Perhaps when try to answer it, you’ll know why I asked you to think this one through before uncapping your electronic pen.
    So buck up here, P., the opportunity is not lost yet, give us the benefit of your knowledge. Which of the 18,000+ dots are you proposing that I connect up to make a phase space plot? I was kinda thinking of connecting the dot that is my home gridcell in December with the dot that is the Rio De Janeiro gridcell during Mardi Gras, but I’m sure you can join my wife in telling me why that choice is not possible …
    w.

  130. P. Solar also complained about my reply, where I had said:

    Nick Stokes above already pointed out the proper way to deal with the weighting. Read the thread.

    P. Solar, who is upset because I though his comment was inane, says:

    This thread was already >110 posts by the time I got to read it. Most of that is the usual garbage , bitching and inane comments. Life’s too short. I did scan the whole thread for anything you’d posted and picked up on help Nick had provided you with.

    In other words, he’s too busy to read the “inane comments” before posting one of his own.
    w.

  131. P. Solar says:
    October 9, 2011 at 9:08 pm

    PS, one way to test what I say about regression attenuation is to invert the axes and redo the ols fit. If the result is good one slope will be the inverse of the other. I think you will get a shock how different they are. Clearly the result should not depend on which way you plot the data … unless I’m right about OLS.
    Willis says: “Report back with your results …”
    My result IS just this, that doing OLS on this kind of data is fundamentally wrong and I’ve said why. I don’t need to reproduce your errors to prove it to you. Do it yourself. [hint reverse the arguments you pass to lm() ]

    No, that’s not a result. A result is where you ACTUALLY DO THE HARD WORK with the data I’ve provided, and you tell and show us what ACTUAL RESULTS of your inquiry are. I’m not going to do it, I don’t have time or interest. You claim to be interested, but you’re all lip. So go and actually do the test you are proposing.
    Then, as I requested … report back with your results. If you can establish what you say, the same way I do, by doing the math and reporting back that I’m wrong, I’ll be the first to admit it.
    Until then, it’s just more smoke like your claim about connecting up the dots … but then refusing to tell us which dots.
    w.

  132. Septic Matthew says:
    October 10, 2011 at 6:37 am

    Willis Eschenbach wrote: Report back with your results …
    Too often you get angry instead of reading. You have a nice analysis. Why ruin it by ignoring its shortcomings? P. Solar is correct: you can not estimate the slope that you want to estimate. It’s well known.

    I am estimating the slope of ∆ Forcing with respect to ∆ Temperature.
    You blithely say “you cannot estimate the slope you want to estimate”.
    While it is possible that you are correct, your claim would be more credible if you explained:
    a) what you think I’m trying to estimate.
    b) exactly why I cannot, under any circumstances or with any possible mathematical models, estimate that slope.
    c) just how, if you are correct that I cannot estimate what I claim to estimate, it is still a “nice analysis”?
    d) how this relates to my asking P. Solar to provide backup for his claims.
    w.
    PS – Generally, when someone says to me “It’s well known” that just means that they can’t find any facts to substantiate their claim. I’d avoid that circumlocution if I were you …
    PPS – I would also invite you to tell me about P. Solar’s claim that I should “connect up the dots” in the analysis … which dot’s should I connect? Because he seems at a loss to answer that one.

  133. Willis: b) exactly why I cannot, under any circumstances or with any possible mathematical models, estimate that slope.
    You can not estimate that slope using linear regression, as you are attempting. You might try with some vector autoregressive methods, but they have problems as well.
    “It’s well known” in this case refers to statistical texts on the non-uniqueness of regression estimates with bivariate (and multivariate) normal distributions.

  134. Septic Matthew says:
    October 10, 2011 at 6:44 am

    Willis wrote:

    These are gridcell monthly averages, about twenty thousand of them. Which among them are you proposing to connect with dots, and on what basis? Please think this stuff through before tossing out un-finished ideas.

    OK, so you don’t have time. Let it go at that instead of the pseudo-intellectual dismissal.

    Egads, what are you, the new bitchy Miss Manners of the blogs? If you think his plans are so great, you can explain them to us. He hasn’t done so, he hasn’t said which of the dots to connect to make his magic phase diagram, he didn’t provide citations or explanations for his claims. I pointed it out to him. And? … sure, I might have been nicer, I could have blown in his ear and rubbed his tummy while I told him very nicely what I told him crudely, that he hadn’t thought his claim all the way through. And that would be better how?
    In any case, one man’s ceiling is another man’s floor, so here’s your chance to shine, Septic. You can step into the breach and do P.’s work for him, and come back and tell us which dots to connect to make the phase diagram.
    Or you could take your act down a notch. Your choice, but bitching about what P. Solar and I have to say doesn’t help. Stand back and let us figure it out, why do you want to fight another man’s battles, particularly when you don’t know the answer about the dots?
    w.

  135. Dixon says:
    October 10, 2011 at 6:46 am

    I may be completely misunderstanding what you’re doing here, but if you are subtracting the temp of the month ahead from the current month I can see why Nick Stokes implies there could be a seasonal bias – and one which would presumably apply in one direction for one half of the year in each hemisphere (days lengthening or shortening) and then in the reverse direction for the rest of the year (for that hemisphere). The hemispheres would be out of phase and you’d probably expect a 4 quadrant plot which is what you show. If you treated the hemispheres separately AND split the year into 2 parts for each hemisphere you could run the subtraction both forward and backward for each of the 4 hemisphere/season combinations. IF each forward/backward run showed the same trends you could then rule out the seasonal signal (which may be much of your ‘noise’). If there’s still a negative trend in temp with increasing cloud, you’ve got a very interesting result.

    That’s an interesting way to go about it, Dixon. I’ll have to think about that some, and how to best do it.
    I’m also a bit concerned about the idea of “seasonal effects”. If the nature of the seasons is that the clouds tend to warm the earth as the temperature drops with the seasons, and they tend to cool the earth as the temperature warms with the seasons … isn’t that an actual, verifiable effect and as such one that we don’t want to remove? This whole question is a bit of a head-scratcher for me.
    w.

  136. Willis,
    there will be some climate-related presentations at the Joint Statistical Meetings in San Diego, July 28 – Aug2, 2010. I am working to extend your analyses of the TAO data. There will be about 5,000 statisticians present. TAO/TRITON is in the tltle. The program will be posted online next May or thereabouts. Meet me and we can discuss our work.
    Maybe by then you’ll have figured out how to connect the dots.

  137. [SNIP – off topic. If you want to claim that visible sunlight cannot warm things, you’ll have to do it elsewhere. – w.]

  138. [Reply: You are invited to submit an article of your own if you wish. You might be surprised at the comments. ~dbs, mod.]
    I refuse to join any club that would have me for a member. :-p

  139. Septic Matthew says:
    October 10, 2011 at 12:05 pm

    Willis,
    there will be some climate-related presentations at the Joint Statistical Meetings in San Diego, July 28 – Aug2, 2010. I am working to extend your analyses of the TAO data. There will be about 5,000 statisticians present. TAO/TRITON is in the tltle. The program will be posted online next May or thereabouts. Meet me and we can discuss our work.
    Maybe by then you’ll have figured out how to connect the dots.

    Thanks, Matthew, best of luck with your analysis. I’d love to discuss all of these questions over a beer if you are ever around the San Francisco Bay area.
    Regarding connecting the dots, I already know how I connect them. What I haven’t be able to do is to get P. Solar to tell me how he thinks they should be connected, despite his obvious willingness to advise me that I’m doing it wrong …
    w.
    PS – I think you must mean “2012” above.

  140. Septic Matthew says:
    October 10, 2011 at 11:57 am

    Willis: b) exactly why I cannot, under any circumstances or with any possible mathematical models, estimate that slope.

    You can not estimate that slope using linear regression, as you are attempting. You might try with some vector autoregressive methods, but they have problems as well.
    “It’s well known” in this case refers to statistical texts on the non-uniqueness of regression estimates with bivariate (and multivariate) normal distributions.

    Last time you said I could not estimate that slope at all. Now you say I can’t do it with linear regression. While you may be right, so far all you have done is change your claim. It seems you are saying that some slopes can be estimated with linear regression, but not this particular one. However, that is just another assertion. WHY is this particular dataset not viable for linear regression, and what mathematical methods did you use to establish that fact?
    It’s a serious question, Matthew, what math did you use to come up with your conclusion? I’ve shown my work … now it’s your turn. Bring out the math to substantiate your point.
    Finally, just replacing “it’s well known” with “it’s in the textbooks” is not moving forward as far as increasing our sum of knowledge …
    w.

  141. 161.Myrrh says:
    October 10, 2011 at 12:14 pm
    [SNIP – off topic. If you want to claim that visible sunlight cannot warm things, you’ll have to do it elsewhere. – w.]
    I specifically said I didn’t want to argue about it here, I merely gave it as background to where I was coming from, traditional physics, what I was posting was on the figures Steve gave for infrared, 53%, and I was adding to that. Stupidly, I thought you’d be interested.
    So, now I know where you’re coming from.
    [RESPONSE: I’m coming from my request not to discuss extraneous issues, which you have arrogantly abrogated. Calling your idee du jour “background” does not make it suddenly relevant or on-topic. Am I interested? Perhaps, everyone has something to teach me. BUT NOT ON THIS THREAD. I don’t know if you are really that dumb or just trolling, but what part “not interested in that on this thread, please take it elsewhere” is unclear to you? -w.

  142. Willis: Now you say I can’t do it with linear regression.
    Sorry, the claim that you couldn’t estimate the slope with linear regression was stated first. I carried it forward implicitly, without restating it.
    WHY is this particular dataset not viable for linear regression, and what mathematical methods did you use to establish that fact?
    The X and Y variables are both random variables. Your computation of the statistical significance assumes that they are Normally distributed random variables. That gives you a bivariate normal time series. Regression methods assume that the X variables are fixed; or, that you are conditioning on the observed values of the X variables. Either way, the estimate of the regression coefficient of Y on X is biased. I missed the part where you derived the correctness of the formula for computing the observed significance level. Paraphrasing you, you did not put in the mathematics to establish that fact. How to estimate the slope of the regression of a bivariate normal time series is taken up in, for example, in “New Introduction to Multiple Time Series Analysis”, by Helmut Lutkepohl, Springer, 2006, corrected second printing 2007.
    PS – I think you must mean “2012″ above. good catch, thanks

  143. I’ve added monthly maps of the variables of interest (T, F, ∆T, ∆F) to the head post above.
    w.

  144. Septic Matthew says:
    October 10, 2011 at 3:19 pm

    The X and Y variables are both random variables. Your computation of the statistical significance assumes that they are Normally distributed random variables. That gives you a bivariate normal time series.

    Matthew, thanks, now we’re getting down to the good part, where I learn something. However, I don’t understand your statements above. Seems to me that if X and Y were normally distributed random variables as you say, that there would be absolutely no statistical relationship between them.
    But it is obvious from looking at Figure 1 that the arrangement of the dots is far, far from random.
    So what am I missing here?
    w.

  145. I have never understood how causally a net warming effect from clouds could be proposed. It is becoming well accepted that an unrestrained dynamic system will tend to maximize the rate of entropy creation. Radiation from the sun has relatively low entropy. Its the heat divided by the temperature after all and 5000°K is a much bigger divisor than the 288°K of the Earth’s surface. The heat flow in and out has to be the same or the temperature would be changing quickly.
    Consider 5000 units of heat coming in from the sun at 5000°K, being converted to 288°K at the Earth’s surface, and radiating out to deep space at 4°K. The entropy change, sun to earth is 5000/28 – 5000/5000 = 16.4 The entropy change, earth to deep space is 5000/4 – 5000/288 = 1233. Clouds that cause net warming have to interfere more with the high entropy generation of earth to deep space than the low entropy generation from sun to earth. That would be the direct opposite of maximum entropy generation.
    Furthermore, it is well known that shorter light wavelengths are more easily scattered than long wavelengths. The sky is blue and the sun at sunset is red after all. The long wave radiation from the earth will more easily penetrate clouds than the short wave radiation from the sun. The clouds have to be cooling. The only way clouds could have a net warming effect if for their to be more clouds at night than during the day. That should be easily measurable. In fact as Willis has described in the Thermostat Hypothesis, the opposite is observed. Cumulus clouds and thunderstorms tend to start to build up at mid day and dissipate in the evenings. I believe stratus clouds have more to do with large weather systems where the time scale is longer than a day and do not seem to respond to the daily cycle as much.

  146. There’s an inconsistency between the order of the longitudes in the ERBE file and the HadCRUT file. The longitudes in the ERBE file go from 1.25 degrees to 358.5 degrees. The longitudes in the HadCRUT run from -177.5 degrees to 177.5 degrees. Your program incorrectly assumes that the order for the two files is the same. The simplest fix is the interchange the first 36 longitudes in the HadCRUT file with the last 36. By my calculation doing so increases the slope for the unweighted calculation from -1.9 to -1.4, and for the weighted calculation from -1.7 to -1.2.

  147. Joseph Dunn says:
    October 10, 2011 at 6:08 pm

    Whoa, I think you’re right, very good catch. Means I have to start over … such is life. That’s why I wanted to put this all up first …
    All the best,
    w.

    [UPDATE: I’ve updated all of the graphics in the head post. Thanks, Joseph. -w.]

  148. Willis,
    I know that “anomalies” are not always popular, but it seems that they might be helpful in addressing the issues Dixon brought up about seasonal affects. (Others may have brought up seasonal affects too — I just haven’t had time to look thru the whole thread).
    It is clear that in the fall (for either hemisphere), the ΔT is negative. Over land in the N hemisphere, this is often -4C to -8 C. Why not subtract out the average change to find out how each particular year performs? I think this is one layer deeper yet for your analysis.
    I am thinking something like “At 45 N, 95 W, the temperature drops an average of 6C from Oct to Nov. and the average forcing is 35. In 1986 the drop was only 5 C and the forcing was 40. Thus a high forcing led to a small drop in temperature.” (the numbers are simply for purpose of illustration — they do not represent real data) By repeating this for every year, you could start to get a trend thru time. But doing this at various locations, you could start to get a trend around the globe.
    And as I think about it, I would suggest comparisons with the month before or after. Something like “At 45 N, 95 W, the temperature was above average in Oct, 1986. The forcing was above average in Sept, suggesting that the high forcing caused the temperature to be high in the next month. Also, the forcing was high in Nov. This suggests that the high temperature caused the forcing to be high in the next month.” This would help establish whether clouds are a forcing or a feedback.
    Of course, now you have thousands (millions?) of potential trends to consider

  149. [SNIP – I’m not discussing it, and I’m not listening to you argue about it. I told you I’d snip off topic posts. Stop trying to sneak them in. – w.]

  150. Shrug, this is to you, not trying to “sneak” anything in… I suppose it’s time I accepted that most climate ‘scientists’, I use the word loosely because the general interest of this attracts scientists and others from different disciplines, see a completely different world to that which traditional physics teaches. How then do you account for the phenomenon of ‘the sun burns up the mist and clouds’ as we wait for it do so? Rhetorical. I’ve given up taking these alternate universe view seriously.

  151. Willis, first let me repeat my first comment, that this is good work. A few day ago I referred to you as a gifted amateur, in the tradition of Florence Nightingale. It’s early days to say you’ll have the same impact (hardly anybody has, but she was an ideal), but your fresh looks at extant data sets should be followed up by others, and I always look forward to them.
    However, I don’t understand your statements above. Seems to me that if X and Y were normally distributed random variables as you say, that there would be absolutely no statistical relationship between them.
    That’s the special case of independence. You can have normally distributed random variables that have a correlation coefficient that is non-zero. That indeed, is what you found for this case. Estimating the regression of Y on X, or of X on Y is problematical. Mathematically each regression is well-defined (that was the case that Sir Francis Galton was working on with heights of British men and their sons when he consulted a mathematician.) However, to estimate the regression of Y on X or of X on Y from data leads to a problem. Look up “Deming Regression” — used in industry to evaluate the comparability of two methods for assaying the same quantity ( say, testosterone in blood), where neither method is a “gold standard”.
    Enough for now. Keep up the good work.
    The Joint Statistical Meetings are recurrently held in San Francisco, but won’t be for the next 5 years. You’re bracketed between San Diego and Seattle.

  152. “Data colors indicate the location of the gridcell, with the Northern hemisphere starting with red at the far north, slowly changing to yellow and to red at the equator. From there, purple is southern tropic, through pink to green for the farthest south latitudes.”
    Red in the north AND red at the equator? And what about blue? I guess it is blue at the equator, and not red?
    Besides this: Really cool work! 😉
    [REPLY: Thanks, fixed. -w.]

  153. To use your language from a previous post:
    * the cause of ΔT is the tilt of the earth and changes in TOA insolation.
    * a 1st order affect is land vs ocean (clearly visible on your maps of ΔT)
    * clouds would be a 1st or 2nd order effect.
    Until you have separated out the effect of the seasons (and the effect of land vs ocean), I think finding a signal in the clouds will be tough. For example, whether a region gets more cloudy or less cloudy in the fall, it will cool.

  154. I’ve had a chance to review the data and I’ve come to the realization that your analysis is missing a major confounding factor, namely the incoming solar radiation. The quantity being analyzed is the net cloud forcing, namely the total outgoing radiation attributable to clouds. This does not measure the efficiency of clouds in dealing with a fixed level of incoming radiation, the relevant quantity if we’re interested in feedback from CO2 related changes. Stated more simply, the more sunlight coming in, the more sunlight is reflected off of clouds. Furthermore, increasing sunlight causes the air to heat up giving rise to the potential for a spurious correlation.
    With this in mind I redid the regression analysis adding incoming solar radiation as one of the explanatory variables. In the following printout, dncf is the change in the net cloud forcing, dtem is the change in the temperature and dtsi is the change in the solar forcing. The tsi is determined from the ERBE file as the sum of the outgoing long wave radiation, the outgoing shortwave radiation and the net radiation. After a review of the scatter plot of dncf vs dtsi, I concluded that there is an obvious non-linear relationship between the two variables, which is reasonably approximated by a cubic spline of three degrees of freedom. So with that introduction, here’s the output of the analysis.
    [sourcecode]
    Call:
    lm(formula = dncf ~ ns(dtsi, 3) + dtem, data = df2, weights = df2$area)
    Residuals:
    Min 1Q Median 3Q Max
    -75.538 -4.112 0.089 4.138 53.575
    Coefficients:
    Estimate Std. Error t value Pr(>|t|)
    (Intercept) 42.46878 0.45574 93.19 <2e-16 ***
    ns(dtsi, 3)1 -23.68547 0.31780 -74.53 <2e-16 ***
    ns(dtsi, 3)2 -104.67009 1.03840 -100.80 <2e-16 ***
    ns(dtsi, 3)3 -61.00918 0.46796 -130.37 <2e-16 ***
    dtem 1.38452 0.03608 38.37 <2e-16 ***

    Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
    Residual standard error: 7.575 on 18439 degrees of freedom
    (4020 observations deleted due to missingness)
    Multiple R-squared: 0.554, Adjusted R-squared: 0.5539
    F-statistic: 5726 on 4 and 18439 DF, p-value: < 2.2e-16
    [/sourcecode]
    The R squared is now 55%, with about 51% attributable to the incoming radiation. The sensitivity to the temperature has now flipped signs. Given this model the uncertainty in the coefficients is nil, but that does not preclude the possibility of other confounding factors. Furthermore, as Roy Spencer so frequently notes, it’s not clear whether changes in clouds drive changes in temperature or vice versa.

  155. Joseph Dunn says:
    October 11, 2011 at 12:03 pm (Edit)

    I’ve had a chance to review the data and I’ve come to the realization that your analysis is missing a major confounding factor, namely the incoming solar radiation. The quantity being analyzed is the net cloud forcing, namely the total outgoing radiation attributable to clouds. This does not measure the efficiency of clouds in dealing with a fixed level of incoming radiation, the relevant quantity if we’re interested in feedback from CO2 related changes. Stated more simply, the more sunlight coming in, the more sunlight is reflected off of clouds.

    Joseph, you’ve entirely misunderstood what I wrote. I have completely included the incoming solar radiation. It is called “Insolation”. It is reflected in the map as well as in the phase diagrams.
    w.

Comments are closed.