Forecast Icons

[dane]

OK - hadn't thought of the long Davis forecasts. Don't spend much time on it!

The one I'm using is by Negretti and Zambra

I'd love to see a description of the algorithm - do you have a link, please?

[steve]

http://www.gilai.com/scripts/more/smt48 ... ts-no.html

Beteljuice derived the algorithm (I don't actually know the details of how he did it) and wrote some javascript:

http://www.dickie.webfusion.co.uk/zambr ... ecast.html

I just translated the javascript into Delphi to use in Cumulus.

They also did a larger version with 60 possible forecasts instead of 26, and I believe the Davis stations use something quite similar and come up with about 100 forecasts.

Steve

[dane]

thanks!

[steve]

The next build puts 'Forecast: ' on the front of the forecast, and has an option (default: off) to centre the text.

Steve

[Frank02]

Sorry to be late for replying but I'd like to thank you very much for the addition.

[Frank02]

Just a quick question. If I close Cumulus is the forecast still accurate or it needs some times to adjust.

Thank you.

[steve]

The forecast will be generated again immediately at startup, all the data it needs is available straight away.

Steve

[Frank02]

Hi Steve,

I notice that the algorithm on the Web site uses a "Local Weather Range" for the pressure.

Do you generate that range from the all time records High / Low barometer pressure?

I was wondering since if I am entering the exact same values as in Cumulus within the web site algorithm to generate a forecast, I get different results than I do in Cumulus. The only thing is that I'm not to sure about the values used for High / Low pressure in Cumulus and therefore I use my own high / low records.

It looks like if those high / low values make quite a difference in the forecast. For instance, Cumulus will report Fine Weather even though it's snowing outside and the sky is quite cloudy while the algorithm will generate a "Changeable, some rain" if I use my own high / low (based on all time records) barometer pressure.

My own data are Local low: 987 local high: 1040
Wind direction: NW
Month: November
Current relative hPa: 1012
Trend: Steady

The Web site gives a "Changeable, some rain" (more accurate)

While Cumulus gives "Fine Weather"

Thank you.


Thank you.

[steve]

I'm just using 950 and 1050 for now. I'll add the ability to supply your own values at some point.

For instance, Cumulus will report Fine Weather even though it's snowing outside and the sky is quite cloudy

Don't forget it's supposed to be a forecast.

Steve

[Frank02]

Ok I see. The larger the spread between the high and low, the less sensitive the algorithm is to pressure drop / rising. Which would explain the difference in forcast as here I never saw a pressure as low as 950 or as high as 1050.

That is what makes the difference when I supply the algorithm 987 - 1040 values. If I use the default 950 - 1050 range then I get the same result on the web site than I do in Cumulus (Fine Weather) but that range being to wide to provide accurate weather forecast in my area.

Ok, thanks. I tought that maybe the high / low range would have been settable within the Cumulus.Ini file.

[Frank02]

For instance, Cumulus will report Fine Weather even though it's snowing outside and the sky is quite cloudy

Don't forget it's supposed to be a forecast.

Steve

Yes you are right. However, it didn't forecast the snow and clouds coming up but I am pretty sure that it would have seen it using the 988 - 1040 range as that range is more sensitive to pressure variation.

Thanks again.

[steve]

Does 988-1040 really represent the extremes for your part of the world? 988 seems awfully high.

Anyway, I'll add configuration of the pressure range into the next release.

Steve

[steve]

I'll be uploading a beta with this shortly.

Steve

[Frank02]

Does 988-1040 really represent the extremes for your part of the world? 988 seems awfully high.

Anyway, I'll add configuration of the pressure range into the next release.

Steve

It looks like it (for relative pressure anyway. Absolute pressure is lower). This is the lowest pressure that has been recorded on my 3 stations so far in 1 year and a half 1040 being the highest so far.

Thank you.

[goldrush]

Towards the start of this thread, the possible use of Metars was mentioned.
A couple of small programs may be of interest which can run alonside Cumulus

METARS downloader
http://www.metarsdownloader.com/
Cost around $20. Automatic download, decode and display of realted graphs for 1 Metar station. Auto ftp upload to website.

METAR Weather
http://www.nirsoft.net/

Free. Manual download of multiple Metar stations. auto decode. Manual saving of data for auto upload by Cumulus.