FFTs Cloud feedback and Stuff

Many "sums" have been done using FFTs and convolution.
It started out with Spencer:

Changed to this with "Bart"

However there seems to me to be problems with all of this.

1.  the data being used is from 10 years only
2.  the data from clear to cloudy sky is not simultaneous
3.  the  data is average over 1 month so can never be safely used to subtract clear from cloudy - the data is smeared over 1 month and can never be data from the same region.
3a. Albedo of soil and water are very different- cloud over water will show a large TOA flux difference wheras the cloud over land will show less outward going flux.
water albedo= 0.02 approx (at some angles)

ground albedo = 0.1 to 0.5
Clouds albedo = 0 to 0.8
4.  What about the "insulating" effect of clouds at night. Shouldn't this be included in any flux calculations?
5.  There is no way the data available from Spencer/Dessler/Bart can show the accurate change in flux due to clouds

the Plots below are simple spectrums using the FFT function in excel. Note that any thing over 60 years period has very little resolutuion. and these are from 200 year records not 10 year!!.

Also when doing FFT on a non infinite series the trend should be removed before the fft transform is applied as the termination of data at either end causes problems. Using a FFT windowing function will also improve performance.

Comparison between data with a trend and detrended data:

Ocean albedo calc

Measurement of ground albedo

Variation/validation  of Albedo


  1. I go over several of these problems thoroughly at CA. Yes, it is difficult to tease out correlations which are longer term than the data. But, indications are that the estimate is reasonably close to reality. The relevant estimated transfer function is very long term (low frequency bandwidth), so a lot of your effects manifest themselves in the higher frequency region and do not affect the response. And, the phase response of -180 degrees at low frequency indicating negative feedback is well nigh incontrovertible.

    In any case, it's a LOT more believable and rigorous than linear regressions on a phase plane, which is a perfectly potty way to try to estimate feedback. - Bart

  2. Bart
    You may go over a FEW of the prolems but from what I have read, you have not even touched on the non-simultaneous nature of the signals. This in my opinion is a killer

    The data corresponds to the last 10 years. The global temperature in the previous 10 years no way reflects the last.
    To extend your method to timescales greater than the data is available is just plain wrong. You may be able to derive a correlation over these 10 years but next year the temp anomaly could be -1K or +1k - who knows? Your "simulation" will not predict anything beyond the current 10 years.
    Your simulation comes up with a negative forcing more cloud=lower temperature. plotting temp against offset cloud should therefore give an R2 better than the non offset - it doesn't