Jump to content

Archived

This topic is now archived and is closed to further replies.

VillagePlank

Evidence Of A Link Between The Moon And Weather ?

Recommended Posts

Well, would you believe it ...

I've think I might have found evidence of a link between the moon and the weather (particularly temperature) - I know, I know, I nearly fell of my chair, too.

Here's what I did,

I got hold of the daily CET temperature records from the HadObs site, and then I transformed the data from the irritating tabular format the MetO use into a format that is easy for databases to process easily (a pivot query) and removed all the -999 records. I calculated the moon-phase for each day using a system of 0..7 where 0 is a new moon, and 4 is a full moon, and calculated for each date in the series.

I created a summary set of data for each day of the year computing the average, standard deviation, minimum temp reported for that day, maximum temp reported for that day. I also computed a low threshold (Av-Std*n), and a high threshold (Av+Std*n) where n is 1,2 or 3 (ie the number of standard deviations away from the mean we are prepared to say that a CET day is exceptional - ie that it is either below the low threshold or above the high threshold.

There was no point in computing 4stds because hardly any days could be considered exceptional.

I then calculated the quantity of exceptional days in the dataset and grouped the numbers by moon-phase. Here's the table:

post-5986-0-05908100-1295097673_thumb.pn

Now it is pretty difficult to discern any immediate pattern from that so I drew the charts:

post-5986-0-88036400-1295098031_thumb.pn

post-5986-0-48116100-1295098036_thumb.pn

post-5986-0-58187500-1295098042_thumb.pn

Preliminary conclusions from this - derived from nearly 90,000 days of data - must be that exceptional days ie where they fall out of some standard deviation multiple are higher the closer they are after the new moon. In English, during or straight after a new moon you'd expect out of the ordinary weather that affects temperature more often, and that weather is more likely to average before the new moon. The calm before the storm one might say ...

So, what about MB's predictions? Well, the 1st Feb's moonphase index is 7, and rest are all .... 0's - so the conclusion from empirical data is that he has a reasonable chance of getting at least out of the ordinary temperatures. I think it's reasonable to assume, then, that the moon has something to do with the "method" given references to the "ancients"

I'll do some randomised testing (etc etc) so should be able to report the significance of this within the next week or so - and I'll report the full algorithms (etc) on my blog as well.

EDIT: Here's the chart of extraordinary days shown by month:

post-5986-0-78068500-1295100367_thumb.pn

Share this post


Link to post
Share on other sites

Well, would you believe it ...

I've think I might have found evidence of a link between the moon and the weather (particularly temperature) - I know, I know, I nearly fell of my chair, too.

Here's what I did,

I got hold of the daily CET temperature records from the HadObs site, and then I transformed the data from the irritating tabular format the MetO use into a format that is easy for databases to process easily (a pivot query) and removed all the -999 records. I calculated the moon-phase for each day using a system of 0..7 where 0 is a new moon, and 4 is a full moon, and calculated for each date in the series.

etc.

Would be interesting to see how those graphs look for each year of the record. i.e. how consistent is the correlation across the years.

Compiling a series for some 300 years is fine for averages and thus making general probabilistic forecasts, but doesn't tell you much about whether it would be useful as a forecasting tool on any given day, since there may be some years where the correlation is atrocious.

Share this post


Link to post
Share on other sites

Would be interesting to see how those graphs look for each year of the record. i.e. how consistent is the correlation across the years.

Compiling a series for some 300 years is fine for averages and thus making general probabilistic forecasts, but doesn't tell you much about whether it would be useful as a forecasting tool on any given day, since there may be some years where the correlation is atrocious.

Yes, and also the direction of extremity - was it colder, or warmer ...

Share this post


Link to post
Share on other sites

Would be interesting to see how those graphs look for each year of the record. i.e. how consistent is the correlation across the years.

Compiling a series for some 300 years is fine for averages and thus making general probabilistic forecasts, but doesn't tell you much about whether it would be useful as a forecasting tool on any given day, since there may be some years where the correlation is atrocious.

Here you go ...

post-5986-0-94149500-1295102142_thumb.pn

post-5986-0-30828100-1295102148_thumb.pn

Was the moon closer to us in 1950? ;)

Share this post


Link to post
Share on other sites

Yes, and also the direction of extremity - was it colder, or warmer ...

Indeed.

Have you thought about analysing the Hadobs precipitation series? If the moon were to play a part, I wonder whether the extremities would extend to other weather observation factors.

It is certainly interesting to analyse data and try and find something, but I am doubtful of how meaningful it is.

Here you go ...

post-5986-0-94149500-1295102142_thumb.pn

post-5986-0-30828100-1295102148_thumb.pn

Was the moon closer to us in 1950? ;)

That's one hell of a large variance

Share this post


Link to post
Share on other sites

Indeed.

Have you thought about analysing the Hadobs precipitation series? If the moon were to play a part, I wonder whether the extremities would extend to other weather observation factors.

It is certainly interesting to analyse data and try and find something, but I am doubtful of how meaningful it is.

Yup - I agree, I was surprised to see anything at all.

Share this post


Link to post
Share on other sites

Here's the quantities for whether the temperature goes up or down split out ...

post-5986-0-33593100-1295103277_thumb.pn

post-5986-0-54417000-1295103284_thumb.pn

One thing to remember about these annual charts is this is just a count of exceptional days per year, and have no moonphase calculation in them; so, one needs to look at the mean, mode, and median - the closer they are the more standard the distribution is - ie the less skewed the data is.

Share this post


Link to post
Share on other sites

Here you go chaps - a complete prediction of the CET monthly average for 2011 - using only the CET temp series, Moon phases, and a lot of averaging:

post-5986-0-00460300-1295108261_thumb.pn

Giving an annual CET of 9.69C. Notably cold in May, and notably warm in April.

I am looking for a correlation in ups and downs not in magnitude - so will use the Pearson Product Moment correlation test to see if the ups and downs of the anomaly match.

And whilst I am at it -

...Coldest day will be 19th Dec 2011 (just after new moon)

...Warmest day will be 1st August 2011 (the day new moon starts)

Let's see what moon-phases can really do!

(I am as sceptical as the next man, and don't really expect this to work very well - but we shall see)

Share this post


Link to post
Share on other sites

Incidentally, if you are following MB's forecast, an 'event' appears in this "method", too

post-5986-0-96561800-1295109438_thumb.pn

... the classic two-day British cold spell. This method thinks it's two days earlier, though - although, given the size and rate of the drop, I'd say that the winds will be quite brisk on the day in question.

Share this post


Link to post
Share on other sites

Right, here's how I did it:

I put the dates for 2011 in a spreadsheet, I calculated the moon-phase for each day, and then inserted the average temperature for that day over the whole series.

I then computed what the deviation from that average is for each of the eight moon phases for each day. I also computed the relative change for each moon-cycle for each month, figured out it's % change over the whole calendar year, made it symmetrical about zero (by deducting 0.125 off the value) and then multiplied it by 750.

For each day in the spreadsheet, I took the average for that day and added the difference according to the moon-cycle for that day multiplied by the monthly fudge-factor.

... and there you have it.

I haven't validated it against past data sets: next job, I guess.

Here's the code for computing the moon-phase;


Public Function MoonPhase(ByVal y As Long, ByVal m As Long, ByVal d As Long) As Long

Dim c As Long, e As Long, b As Long
Dim jd As Double

If (m < 3) Then
y = y - 1
m = m + 12
End If

m = m + 1
c = 365.25 * y
e = 30.6 * m
jd = c + e + d - 694039.09
jd = jd / 29.53
b = jd
jd = jd - b
b = jd * 8 + 0.5
b = b And 7

MoonPhase = b

End Function

Adapted from here. The rest of the code is SQL and is quite lengthy - PM me if you want a copy. I can't give you the spreadsheet since it has personal info stored in it's attributes - as soon as I can figure out how to get rid of them, I'll make it public.

Share this post


Link to post
Share on other sites

Fred and I are aware of some signals in the data associated with lunar phase. However, it's not as simple as just taking lunar phase and temperature data. There are clearer signals if you take various categories of the data sorted by other, independent variables.

Let's see if this will upload, it's the signal of lunar events in the data for Toronto, 1841-2010, but filtered to show only those cases where lunar perigee is near this year's position (five to eight days after northern declination maximum).

I see the file did upload. Good. The data so far this year are following that trend line quite well. You can go into the file and find another data set for the first half of the data, that also resembles the long-term average. The events that are labelled are repeating lunar events in our research. While lunar phase is important, there are actually stronger signals from lunar declination, a cycle that overlaps phase in midwinter but runs 2.2 days shorter and begins to decouple so that the declination peaks occur before new and full moon by end of January and into February.

The relevant peaks now are N Max 17-18 Jan before full moon 19-20 Jan, and S Max 30-31 Jan before the 3 Feb new moon.

You can use the excel file I downloaded to view temperature profiles throughout the data set for the analogue year ... however, the labels would need to be adjusted if you take future dates. If you take sets that are 27.3 days or multiples forward (starting from days 35, 62, 90 etc) then the lunar events will still line up but the new and full moon dates will move to the right 2.2 days per time series forward).

LUNARPER.xls

Share this post


Link to post
Share on other sites

Fred and I are aware of some signals in the data associated with lunar phase. However, it's not as simple as just taking lunar phase and temperature data. There are clearer signals if you take various categories of the data sorted by other, independent variables.

Let's see if this will upload, it's the signal of lunar events in the data for Toronto, 1841-2010, but filtered to show only those cases where lunar perigee is near this year's position (five to eight days after northern declination maximum).

Thanks for that, Roger. V. Interesting.

I agree that moon-phases are only part of the story, if it all. I simply wanted to have a stab of my own regarding MB's mythical "method" I guess that it picks out Nov 2010, Xmas day (which didn't verify) and this Feb job, that he must be using some derivation of moon-phases, especially considering stuff you can find on the mis-information highway, like this

If the correlation about the CET and forecast for this moon-phase method seems OK (ie r>0.6,say) I might relook at it, but, as you rightly pointed out, the moon-phase signal really isn't that great (hence the 750 multiplier), so I'd rather leave it to the experts (you + BFTP)

Did you get to look at that (review) paper I posted about atmospheric tides, btw?

Share this post


Link to post
Share on other sites

Interesting. But does it affect other places in a similar way? if not, why would it only affect NW Atlantic synoptics?

Share this post


Link to post
Share on other sites

Here's the entire year ...

post-5986-0-30092600-1295113114_thumb.pn

(EDIT: That chart is wrong - will repost later!)

I must have missed those balmy January days ;) Once I've tightened up the method (it was a bit haphazard) I'll post charts for previous years)

(magnitude is not important, here, as stated before, but rather the correlation in the monthly scores - I've put this up for info - after all, I'm not really expecting snow on the 13th May, either ....)

Interesting. But does it affect other places in a similar way? if not, why would it only affect NW Atlantic synoptics?

This is moon-phases correlated to temperature. If you wanted to do it for another location, I presume you'd have to get a temperature series for that location and go through the whole process for there, too.

Share this post


Link to post
Share on other sites

I'll be surprised if you get a decent PMCC based on that graph of predicted daily temps. However, as I said before, it's probably more interesting to see how well it works over the years, so I'm glad to hear you're going to have a look at past years.

Admittedly you won't get as much praise for correctly predicting the past as you would the future, but I'll be impressed if anything does show up.

Edit: The reason I'll be surprised (putting aside my general scepticism of lunar-based prediction) is that its far too regular - I don't see how it would out like that; cold and mild don't alternate in such a way. But I'll hold back any further comments until I see some more data.

Share this post


Link to post
Share on other sites

Admittedly you won't get as much praise for correctly predicting the past as you would the future, but I'll be impressed if anything does show up.

I'll be gobsmacked.

Share this post


Link to post
Share on other sites

I'll be gobsmacked.

Fair enough! Good to hear that you are approaching this from a neutral/sceptical observation point. I've seen too many people on these forums bending data/observations to suit their beliefs.

Share this post


Link to post
Share on other sites

Fair enough! Good to hear that you are approaching this from a neutral/sceptical observation point. I've seen too many people on these forums bending data/observations to suit their beliefs.

The chances are that if there's anything in it, then someone, probably disaffected, would break ranks and would've published by now. At very best, my suspicion is that it might be used as part of a forecast (say 10%) in a similar way that RJS, and BFTP might use it.

Share this post


Link to post
Share on other sites

I haven't validated it against past data sets: next job, I guess.

I suppose that considering this up and coming storm is a North Sea event, the easiest data to verify would be the 1953 East coast floods?

Just a thought.......

Share this post


Link to post
Share on other sites

Jan CET prediction is looking close ... :cc_confused:

If Feb comes in then, I promise, I will take the method a little more seriously than, ahem, I did before :)

Share this post


Link to post
Share on other sites

Attached are two files:

cet_av.txt

This contains every day of the year and the average cet temperature across the whole series 1772-2010 for each day.

moon_phase.txt

This contains every day of the year and the average anomaly against the CET for each moon phase for the whole series.

Both files are in a CSV format.

So, here's what you can do. If you compute the moon phase for any date (use the code on the previous page, you should be able to create an Excel macros using it) and then take the average CET for that day and modify it according to what the moon phase has historically done.

So, for the next seven days, the CET temperature according to this method should be:

post-5986-0-07388000-1296138068_thumb.pn

... which captures the up and coming cold spell quite well, I think.

This is a different method to that which produced the CET forecast for the whole year.

cet_av.txt

moon_phase.txt

Share this post


Link to post
Share on other sites

Here's the forecast for the entire of February:

post-5986-0-43525800-1296139314_thumb.pn

Here's the anomaly chart - the Man in the Moon says that we are going to get colder until the 22nd which is rock bottom, and then starts the slow climb out to spring.

post-5986-0-22014500-1296139400_thumb.pn

This method gives a Feb average of 3.79C which is colder than the previous method used on page 1 of this thread.

(I can then use the actual data at the end of February to start testing this for skill)

Share this post


Link to post
Share on other sites

If you do as above but multiply the MP anomaly by ten (magnify the moon's "effect") then this is the chart for December last year:

post-5986-0-95744300-1296141072_thumb.pn

Blue is forecast, Red is observed.

Which gives r=-0.45

Not looking promising is it? Best thing I could possibly say is that there is some sort of lag (about two weeks) from the cause (the moon) and the effect (the temperature)

Share this post


Link to post
Share on other sites

  • Some first weekend of Summer - even the pigeons don't fancy it

    Gusting winds, showers and persistent rain - not the type of weather you'd usually associate with the first weekend of summer. But there will be some improvement through the weekend and into next week. Read the full update here

    Netweather forecasts
    Netweather forecasts
    Latest weather updates from Netweather

    UK Storm and Severe Convective Forecast

    UK Severe Convective & Storm Forecast - Issued 2020-06-05 21:46:10 Valid: 06/06/2020 0600 - 07/06/2020 0600 THUNDERSTORM OUTLOOK - SAT 6TH JUNE 2020 Click here for the full forecast

    Nick F
    Nick F
    Latest weather updates from Netweather
  • Recently Browsing   0 members

    No registered users viewing this page.

×
×
  • Create New...