Forecast logic?

[Charlie]

Hi Steve,
I did a search, but maybe messed it up... Can you tell us how the Cumulus forecast works? What's the logic and what's the range? I've been playing with my local pressure extremes to try to get the forecast more acurate, but it might be easier if I knew what knobs I'm actually twiddling.
This sounds like a great wiki entry too since my search did not find it there either.
Thanks!

[steve]

Beteljuice is the expert, I just stole his code

It's based on a physical device called a Zambretti Forecaster. It says there that 91.7925% of forecasts with it are accurate, so it must be good

Beteljuice's original Javascript implementation is here.

[casacota]

Beteljuice is the expert, I just stole his code

It's based on a physical device called a Zambretti Forecaster. It says there that 91.7925% of forecasts with it are accurate, so it must be good

Beteljuice's original Javascript implementation is here.

Yes the Zambretti algorithm is well known and can be fully automated also to mime the original Zambretti Forecaster, I did implement this on the software I did write to run one from my stations:



But there are better ones, eg. that from Raymond F. Sager:



Both combined, and improved with some other things and considerations gave this:

http://www.casacota.net/teranyina?num=1230158655

[casacota]

It turns out that the Sager Forecast is already implemented in the Davis Pro 2 and Davis Vantage Vue stations.

It is probably the best forecast what is to made based only on local variables.

[steve]

It turns out that the Sager Forecast is already implemented in the Davis Pro 2 and Davis Vantage Vue stations.

I don't think it uses the Sager algorithm - how does it know the sky condition, for example? And as far as I can see the wording is different. But I agree that that the forecasts it produces are similar to the Sager ones, so possibly it uses a modified Sager algorithm. I think the Davis algorithm also takes into account latitude and longitude. If someone could work out an algorithm for Sager which didn't require the sky condition, I could consider adding it to Cumulus.

[beteljuice]

Another hiccough is that the Sager technique also (correctly) employs 'trend' eg. change (or not) of wind direction over perhaps last fours hours. These are intuitive to an observer, but without external logging / coding this data is not 'native' output to C1.

The Zambretti forecast could be more verbose, ie. give 'expected' wind direction / force that could be assumed for a given condition, but I shied away from this as I don't know enough about 'world' weather to determine if old (European) weather saws hold true globally (I would think not).

Don't forget, if you want to try adding expected wind etc. , you can know create your own wording for the forecast conditions in the ini file

Please remember that this is a very difficult science that nowadays relies on 'distant' data rather than local observation, also this is NOT part of Cumulus, but a free 'bolt-on' for amusement only