Author Topic: Algorithm for Predicting Weather  (Read 979 times)

0 Members and 1 Guest are viewing this topic.

Offline Inq

  • Member
  • *
  • Posts: 16
    • InqOnThat
Re: Algorithm for Predicting Weather
« Reply #25 on: April 08, 2022, 08:06:42 AM »
Have you seen Dr. KFS research on the Zambretti algorithm? I did a very quick read-thru. It appears the baro range changes depending on whether or not pressure is steady falling or rising. Unless I missed it, I did not see adjustments for latitude. Interesting reading.
https://integritext.net/DrKFS/zambretti.htm

Thank you for your patience.  You have been very helpful.  But... Metrology is WAY down my list of proficiencies (somewhere closer being a midwife than Mathematics and Engineering). 

I actually did read through that link and it was one of the more technical ones I've read on the subject.  And you're right it does not say anything about latitude.

As I didn't write anything about latitude, I'm going to take a WAG and assume you actually went to my website and saw...
 [ You are not allowed to view attachments ]
... and that led to your question. 

Although the your referenced article does not say anything about latitude, he does highlight it being dependent on wind direction.  While I was studying the code, I noted how it handled the wind contribution and how it rightly considered that basically the prevailing winds in the southern hemisphere are mirrored as compared to the northern hemisphere.

Then... you might wonder why I exposed it since this "teaching project" has no wind sensors.  I can only answer that since this project can be very simply reproduced (if you're into Arduino stuff) that someone might want to add a wind sensor AND someone doing so might live in southern hemisphere.  The Zambretti code I incorporated already did this and I saw no reason to rip it out.

VBR,
Inq

Offline Inq

  • Member
  • *
  • Posts: 16
    • InqOnThat
Re: Algorithm for Predicting Weather
« Reply #26 on: April 08, 2022, 08:15:09 AM »
..found a code snippet based on Dr.KFS research
https://w4krl.com/zambretti-forecaster/

Yes, I found that one too.  But, that individual did rip out the wind contribution.  I also have a prejudice against long if/then trees.  In computer science circles that is consider to be sloppy programming especially since Dr.KFS showed that it was linear. 



Offline Bashy

  • brecklandweather.com/pws2021
  • Forecaster
  • *****
  • Posts: 1368
  • brecklandweather.com/index.php
    • Breckland Weather
Re: Algorithm for Predicting Weather
« Reply #27 on: April 09, 2022, 10:35:44 PM »
Quote
My "logic" is telling me that I should maybe start out with defaults (you clued me into [tup]) of our 996 to 1036 and let the sensors expand this range as necessary when they experience a larger extreme.  This way, I would get 100% scaling "tuned" to my area.  I'm wondering if my logic makes meteorological sense???

Sure. If your sensors can expand as necessary - go for it.
I wanted to thank the forum again.

I also wanted to give feedback on the little experiment with adjusting the Zambretti
"local weather range".  I don't know if anyone is interested, but I like to be thorough even if there is only one person I can help. 

I don't know which is the original site/author, but I found this site first http://www.beteljuice.co.uk/zambretti/forecast.html.  Later I found the site copied at several other URL's.  Most sites had the defaults of local weather range set to 950 to 1050 mbars.  However, one site had them set to some numbers that were in in-Hg, but the label for the main entry was still in mbar.  Putting in a legitimate mbar number always gave a bad answer.  This caused me to look more closely at the mathematics.  And that's when I asked the question above about it making meteorological sense.

I have several units worth of components so I made two units and set one using the defaults of 950 to 1050 and a second one using the range I found were my local extremes for all of last year being 996 to 1036.
 [ You are not allowed to view attachments ]
I had both sitting next to each other outside and ran them for two days during which a storm came through.  It lasted about four hours, with thunder and lightning, but only dropped about 1" of rain.  I was working on my computer all day and had the two visual dashboards on my desktop.  Being quite different images between good and bad weather when the Zambretti value changed, it easily caught my eye any time either changed.
 [ You are not allowed to view attachments ]  [ You are not allowed to view attachments ]
It turns out changing the range did significantly change the predicting ability as I would have guessed from looking at the Math.  Unfortunately... maybe too much.  The stock numbers were way too optimistic... it said relatively clear weather until only a couple hours before the storm when it maxed out about 'X'.  The one with my local extremes was way too manic and went above X about 12 hours before and in the last couple of hours went to Z.  I think over time, if I get these tuned to my area it will have very good predicting ability.

Here is the graphed data for one of the units. 
 [ You are not allowed to view attachments ]


Have you setup your zambretti to forecast automatically? i ask because it is best done at 9am local time, as to the difference it makes when running it at any other time, i dont know.
Kind regards
Bashy

Offline Inq

  • Member
  • *
  • Posts: 16
    • InqOnThat
Re: Algorithm for Predicting Weather
« Reply #28 on: April 10, 2022, 05:10:52 AM »
Have you setup your zambretti to forecast automatically? i ask because it is best done at 9am local time, as to the difference it makes when running it at any other time, i dont know.

At the moment, it's setup to do the calculation every minute.  However, I saw that 9:00AM reading in several articles and wonder what its scientific basis is.  Two of the storms that came through my area since I've had it running hit at midnight and 2AM.  At 9AM prior, Zambretti was showing A and B with not a cloud in the sky.  Closer to the storms they reached W and Z.  Another freak storm happened this past Friday/Saturday.  We were getting snow, even though the temperature was 43F (6C).  Higher in the mountains got accumulation.

It's been interesting watching it while I work and especially with the history graphing.  I've added the Zambretti value (0 - 25) to the graph so I can see the sensor readings and the results correlated.  I'm starting to get a better sense for it.  Currently, I only have one unit running using the 950-1050 mbar range and it is not doing a very good job.  I do believe that to setup Zambretti properly, it needs to have the local range.  If you look at the previous graph, even though we had a significant thunderstorm and had a 3-hour pressure differential -10, the actual pressure didn't even drop below 1000 mbar.

What I really need to do is setup an old phone camera on a tripod and have it take a picture every hour and correlate that with the past graphs.

Offline Bashy

  • brecklandweather.com/pws2021
  • Forecaster
  • *****
  • Posts: 1368
  • brecklandweather.com/index.php
    • Breckland Weather
Re: Algorithm for Predicting Weather
« Reply #29 on: April 10, 2022, 07:32:12 AM »
Have you setup your zambretti to forecast automatically? i ask because it is best done at 9am local time, as to the difference it makes when running it at any other time, i dont know.

At the moment, it's setup to do the calculation every minute.  However, I saw that 9:00AM reading in several articles and wonder what its scientific basis is.  Two of the storms that came through my area since I've had it running hit at midnight and 2AM.  At 9AM prior, Zambretti was showing A and B with not a cloud in the sky.  Closer to the storms they reached W and Z.  Another freak storm happened this past Friday/Saturday.  We were getting snow, even though the temperature was 43F (6C).  Higher in the mountains got accumulation.

It's been interesting watching it while I work and especially with the history graphing.  I've added the Zambretti value (0 - 25) to the graph so I can see the sensor readings and the results correlated.  I'm starting to get a better sense for it.  Currently, I only have one unit running using the 950-1050 mbar range and it is not doing a very good job.  I do believe that to setup Zambretti properly, it needs to have the local range.  If you look at the previous graph, even though we had a significant thunderstorm and had a 3-hour pressure differential -10, the actual pressure didn't even drop below 1000 mbar.

What I really need to do is setup an old phone camera on a tripod and have it take a picture every hour and correlate that with the past graphs.

I wonder if you being in the mountains is playing a part in this?
Kind regards
Bashy

Offline Inq

  • Member
  • *
  • Posts: 16
    • InqOnThat
Re: Algorithm for Predicting Weather
« Reply #30 on: April 10, 2022, 07:43:53 AM »
I wonder if you being in the mountains is playing a part in this?

I imagine it does.  But not knowing metrological theory, I couldn't explain how.  I know all our weather comes from the west or north-west.  The only time it comes from the south or east is if a hurricane is in the Gulf of Mexico or Atlantic.  I haven't put much emphasis on adding a weather vane even though the Zambretti would use it because living in a cove, the wind either comes up or down the cove... never across it.  So direction is meaningless unless I send up a weather balloon.   :grin:  Another project... another day.

Offline PaulMy

  • Forecaster
  • *****
  • Posts: 5194
    • KomokaWeather
Re: Algorithm for Predicting Weather
« Reply #31 on: April 10, 2022, 10:41:12 AM »
As an alternate to the Zambretti I have found BT's Global Sager Forecaster to be in more detail
Code: [Select]
Based Upon:
The Sager Weathercaster:  A Scientific Instrument for Accurate Prediction of the Weather
Copyright 1969 by Raymond M. Sager and E. F. Sager
"The Sager Weathercaster predicts the weather quickly and accurately.  It has been in use since 1942.  Not a novelty, not a toy, this is a highly dependable, scientifically designed tool of inestimable value to travelers, farmers, hunters, sailors, yachtsmen, campers, fishermen, students -- in fact, to everyone who needs or wants to know what the weather will be."


378 possible forecasts determined from 4996 dial codes.
*/

Enjoy,
Paul
Davis Vantage Pro 2 Plus 24-FARS Wireless - Komoka, ON Canada
Envoy/USB Logger/Cumulus v1.9.4  www.komokaweather.com  
WFL/CMX www.komokaweather.com/komokaweather-ca2/index.php
WLL/CMX www.komokaweather.com/cumulusmxwll/index.htm
WLL/CMX www.komokaweather.com/cumulusmxwll/index.html
WLL/CMX www.komokaweather.com/cumulusmxwll/index.php
Blitzortung Station #1076


Offline Inq

  • Member
  • *
  • Posts: 16
    • InqOnThat
Re: Algorithm for Predicting Weather
« Reply #32 on: April 11, 2022, 06:58:22 AM »
As an alternate to the Zambretti I have found BT's Global Sager Forecaster to be in more detail
Code: [Select]
Based Upon:
The Sager Weathercaster:  A Scientific Instrument for Accurate Prediction of the Weather
Copyright 1969 by Raymond M. Sager and E. F. Sager
"The Sager Weathercaster predicts the weather quickly and accurately.  It has been in use since 1942.  Not a novelty, not a toy, this is a highly dependable, scientifically designed tool of inestimable value to travelers, farmers, hunters, sailors, yachtsmen, campers, fishermen, students -- in fact, to everyone who needs or wants to know what the weather will be."


378 possible forecasts determined from 4996 dial codes.
*/

Enjoy,
Paul

I looked over several links.  It sounds interesting, but two things off the bat, dissuades me. 
  • I think converting it could be overcome, but it was not clear from my limited exposure how time consuming that would be.
     
  • The second was more terminal - One article touted it's claim to fame over Zambretti was its ability to key-in current conditions and that improved prediction capability greatly.  Without extensive video capturing and significant AI to interpret cloud cover, and wind conditions, it does not appear to be usable.  And... I'm not really interested in having to hand-feed the program on a daily basis to arrive at a prediction.

However, if I'm mistaken about #2 or you can recommend some links that point to ways of overcoming the need to feed it manually, I would be very interested.

Thanks.
VBR,
Inq

 

anything