Linkage of LoD to annual earth mass redistribution by global sea ice extent as a proxy.
Author: Tim Channon
LoD (Length of Day) contains a strong annual signal which is good shape match (r2=0.91 at a time lag of just over 5 months) with global sea ice extent as a proxy for annual mass redistribution, ultimately driven by variation in insolation and earth movement.
This suggests no lunar involvement, a disconnected between this part of the dataset and the rest.
The author noted a major mismatch between the spectra at earth distance for solar radiation and the same periods in LoD data. Recalling global sea ice looked similar this was compared and found to be a good match.
Northern snow cover and Arctic sea ice extent are closely allied. Unfortunately there is no global snow cover dataset (confirmed in a private conversion with Rutgers Snow Lab).
The sea ice dataset (NOAA/NSIDC d02135) is low resolution but is the only long term global dataset. This is sufficient for the present analysis.
Software was used to synchronise to and clone with accurate amplitude and phase the annual cycle of LoD and sea ice extent data to the 5th harmonic producing a pair of data models. Daily LoD was re-referenced to the sample rate of the monthly sea ice data.
This was normalised and plotted with a data time offset at a suitable lag, 5 months, then empirically fine tuned using a Farrow interpolar. Best r2 is at 5.16 months lag, value 0.91, a few cycles are shown as figure 1.
A direct visual confirmation of the above is important. This required processing of the dataset to remove major confounding features which are unrelated.
LoD data needed removal of the majority of the lunar orbit signal and long term change.
This was accomplished with suitable low and high pass filters.
Period in years, left to right, high pass, low pass LoD, low pass sea ice, all end corrected, of no importance here. The high pass is a compromise attempting to avoid the 0.1666/2 sample rate but allow the 0.2y 5th harmonic. the data not meeting Nyquist anyway. Low pass is arbitrary and keeping the known 2 to 5 year variations. A lot of weather noise is allowed.
Filtered datasets over a common time span but without relative time lagging.
LoD data contain a lot of weather noise where as usual the huge 1983 En Nino is clearly present.
Time lagged processed data plotted for visual comparison, r2=0.67
None of this provides insight into what exactly is moving mass and might or might not involve snow/ice, which is frozen water originating as liquid elsewhere.
I’m expecting dispute from certain readers who have other ideas on what is going on, so be it, is how progress is made.
Spectra, the annual components and the effect of the bandpass filter.
The NOAA/NSIDC data source is about as cussed as it could be, bizarre file and directory structure, pairs of files by month in different directories. I don’t think they want anyone to actually see the data, only their political spin presentation.
How accurate is it? Not particularly, needs better than monthly but is from a mishmash of satellites anyway.
Also note missing data is not marked.
And read the notes in the files because the data is broken as is, needs an offset added for some of the time.
Alternatively trust me. (scripted download and mangle into a time series)
There is more data if you need it, intermediates, and so on, ask.