Original release topic

[pernaczy]

v. 1_8_1 installed.
Attached picture shows comparison 1_8_1 to 1_8_0.
Piotr

[HansR]

Nice comparison @pernaczy. Thnx!

[PaulMy]

@PaulMy: Surprise me with a bug.

http://www.komokaweather.com/komokaweather-ca/wxFWI.php

Try to enjoy (I am!),
Paul

[PaulMy]

I did run cumulusutils v1.8.0 (before I read about new release) and noted the Predictions, then v1.8.1 and Humidity and pwsFWI changed for all 5 prediction days.

Enjoy,
Paul

[HansR]

http://www.komokaweather.com/komokaweather-ca/wxFWI.php

Try to enjoy (I am!),
Paul

I am too, I really am
And btw seems to be similar in header of the beteljuice format.

I did run cumulusutils v1.8.0 (before I read about new release) and noted the Predictions, then v1.8.1 and Humidity and pwsFWI changed for all 5 prediction days.

OK. Hope you're not reporting this as a bug... (read the issue meteosangonera reported)

[Phil23]

1.8.1 Just uploaded,

Previous output on last page.

181 Output.

Capture.JPG

[hvd51]

@hvd51: Thanks for the complimentful silence (btw: remarkable that your prediction is generated at midnight rollover and does not have a gap of date)

@HansR :
I had also noticed that I had no gap in the prediction. Maybe to do with the time zone?

Harm

[HansR]

@Phil: looks good, output is better than before, now wait to match prediction and reality. It also confirms you have to run the pwsFWI module after rollover.
@hdv51: No, I don't think it is the timezone (I am in the same zone you are!). Has to do with the API still giving the previous day. Sometimes it does, sometimes it does not. Not sure yet.

@General: there may be a problem with .NET version. I work against version 4.7.2 (most recent version). Until now it had never been a problem (I don't use special language constructs). But with the pwsFWI prediction I use two XML.Linq queries to examine the interface. And that seems to be a problem for Piotr. We're still investigating but I would like you to check - when running Windows - against which version of .NET you are working. If you think it is an issue, please tell me.

[Phil23]

So I now have 2 Predictions for 31's & an actual.

See previous posts.

Predictions were:-
615 @ 7:45am on 31st &
587 @ 9:00am on 31st after I presume a forecast update.

Actual now days reading was 603.

Given my observations during the day of actuals Vs hourly forecasts I'd say the predictions are pretty much right in the ball park.

Loosely speaking that's about +/- 2.5% on the 600 reading.
Temp was forecasted @ 26 at a time where it actually reached 29, but then cloud came in & we got a bit of rain for a while.


Phil.

[HansR]

Hi Phil,
Looks good indeed.

@all:
But could you create a spreadsheet for the prediction vs calculation? A bit like the one I attach.
And use the two days ahead prediction as today is the first, you would enter 2 November in the log and in two days you enter the calculated value.

Thnx, Hans

Oh, And I will definitely publish the analysis results of this on my blog.

[meteosangonera]

Hi Phil,
Looks good indeed.

@all:
But could you create a spreadsheet for the prediction vs calculation? A bit like the one I attach.
And use the two days ahead prediction as today is the first, you would enter 2 November in the log and in two days you enter the calculated value.

Thnx, Hans

Oh, And I will definitely publish the analysis results of this on my blog.

With pleasure https://docs.google.com/spreadsheets/d/ ... edit#gid=0

[PaulMy]

Hi Hans,

@all:
But could you create a spreadsheet for the prediction vs calculation? A bit like the one I attach.

Not that my fire danger is high or important, but I had started my Cumulusutils FWI log a little different as per attached.
It takes a couple of minutes each day to enter the last day's and the next 5 prediction days, but the predictions to actual is automated.

Enjoy,
Paul

[HansR]

Ok guys, thanks. Keep up for a week. Very curious how things work out.

[HansR]

Version 1.8.2 is online

NOTE: This is not a functional release but a maintenance release concerning a problem which occurred in Poland and France where the prediction in pwsFWI was aborted.

Both in Poland and in France the prediction procedure was aborted with the message: The format of the input string is incorrect. I first analysed it as a Win 7 version problem but then I was not so sure. It might also be an error in the API itself (a language problem with a lot of encoding character issues) or it might be a coding problem (also to do with language issues and a date format). I ran a code analyser and worked all the warnings so in my opinion it looks better now. I was not able to get an error on the Polish dayfile/API combination so I think I fixed something (though not a full 100% sure what).

Advice: If you do not have a problem, do not update until both the French and the Polish site give green light. You will find a status message here.

[HansR]

1. I made a blog on the first part if the prediction for pwsFWI as a precursor to the report on prediction behaviour;
2. It seems the French site does not have a problem anymore, waiting for Piotr to confirm.