erratic wind values

[opticsman0127]

I think I follow you here. Tomorrow's plots could be revealing, provided I remember to switch off tonight.

Barry

[opticsman0127]

I came home from work early today and switched off then switched on Cumulus, all within a minute. Attached is a spike which did not actually occur as my office at home has a clear view of any wind disturbance. The spike is massive and must be something very strange causing it in Cumulus. I transfered the data to Weatherlink and its not there. I just did the same experiment again, it did exactly the same thing, second grouping of high averages. The weather is a little breezy, but nowhere near around 24-26mph

Any ideas

Barry

[steve]

Did those high values occur "live", or from logger data?

Can you attach your dec09log.txt and the diags files for those starts (just zip the diags folder up if you're not sure which ones are the right ones).

Have you got Cumulus calculating a 10-minute average, or the station? It's probably best to let the station do it, because it is always running and always has the most data to play with. But that's not actually relevant here anyway, as both graph lines have the high values.

[opticsman0127]

They occur exactly as I switch on cumulus from logged data. It then seems to 'calm down' to achieve the correct level of wind. Cumulus is calculating the 10 min average but, it does it in either setting, not all the time, just randomly.
Diags.zip attached

[opticsman0127]

Had a bit of trouble zipping the files, hope they open OK
Barry

[steve]

All of those high values came from the logger, apart from 1645 and 1655, which were read 'live'. Can you export today's data from Weatherlink and attach it, so I can compare what it got from the logger with what Cumulus got? And could you attach your cumulus.ini file (with any passwords etc taken out), please?

[opticsman0127]

Steve,

Attached Weatherlink file and cumulus.ini file. I can remember adding a password and know what it is but, how do I remove it. Is it a problem to divulge it here, wouldn't have thought so.

Barry

Its just peaked again

[steve]

Thanks, I'll have a look. I've removed your cumulus.ini file from your post. Your ftp login details would allow someone to use your web site for nefarious purposes. Only subscribed members can download attachments, but I can't vouch for all 917 of them.

[steve]

Those speeds are in your weatherlink data.

I need to correct something I said earlier:

With data from logger, plotted at 5-minute intervals:
"Gust" = Peak 3-second speed over previous 5 minutes
"Speed" = Average of 3-second speed over previous 5 minutes.

That's actually what happens when you don't have Cumulus calculating the average speed. If you do, this is what you get:

With data from logger, plotted at 5-minute intervals:
"Gust" = Peak 3-second speed over previous 5 minutes
"Speed" = Average of "gust" values (i.e peak 3-second speed) over previous 10 minutes.

If you look in your weatherlink data, you'll see that your peak speed at 15:05 was 21.0 mph.

Code: Select all

01/12/09	15:05	4.3	4.3	4.3	83	1.7	9.0	SSE	0.75	21.0	SSE	1.0	4.2	0.9	1007.0	0.00	0.0	0.049	0.000	20.0	38	5.2	18.6	115	1	100.0	5

Applying your Cumulus multiplier of 1.2 gives you 25.2 mph, which is what Cumulus logged and plotted:

01/12/09,15:05,3.3,83,0.7,25.2,25.2,142,0.0,0.0,1006.98,103.4,20.0,38,25.2,-3.0,3.3

You seem to have a very high "gust factor" compared to me - your peak gusts are apparently approximately double your mean speeds. I don't know if that is a typical value for an inland location, but it's certainly much higher than I get here. It explains why Cumulus gives you such a difference when working from the logger compared to 'live' data. I would recommend that you switch off the 'calculate 10-minute average' setting and let the station do it, so your average speed will at least be reasonable, but you will continue to get higher gust values (approx double) when Cumulus is using data from the logger.

[opticsman0127]

My wind factor is 1.2 and so is my gust factor.

"but you will continue to get higher gust values (approx double) when Cumulus is using data from the logger." This means that every time Cumulus is switched off for any length of time, when it re-starts, and accepts logged data, it could well plot it incorrectly, possibly giving double the gust value and, somehow, moving the speed value towards the doubled gust value.

Am I getting this right Steve, if I am, it seems that switching off Cumulus is a no no in the wind recording department.

I will switch off the 10 min aver calc in Cumulus and see what happens.

For interest, I have attached the Weatherlink speed graph from the problem period.

Barry

Barry

[steve]

This means that every time Cumulus is switched off for any length of time, when it re-starts, and accepts logged data, it could well plot it incorrectly, possibly giving double the gust value and, somehow, moving the speed value towards the doubled gust value.

It's not plotting it incorrectly, and it's not a doubled gust value, it's the peak gust over the logger interval. I chose to use the peak gust value when using logger data, and it's plotting that. And as you have Cumulus set to calculate the average, it's also using that peak value to calculate the average. I chose to use two wind values in Cumulus, nominally a "gust" and an "average". The problem is that the two wind values that the VP2 provides are a different pair of values depending on whether it's live or from the logger.

For interest, I have attached the Weatherlink speed graph from the problem period.

The difference is that WL is only plotting one value. And because WL only has to work with Davis stations, the way it works is that it periodically downloads the logger data and plots the speed value from that, so the value it always plots is the "average of the 3-second speeds over the previous 5 minutes" in the case of a 5 minute station logger interval.

[steve]

Here's my wind speed graph, having switched to 'plot latest gust' at about 10:00 yesterday. I closed down at about 23:30 and started up at about 07:30.

wind1.png

You can see the change in the "gust" line for the period plotted from logger data, where it's had to change from the 'latest' value to the 'peak' value, because that's all that's available. The speed line continues quite reasonably, because the only difference is that it's now using the 5-minute average rather than the 10-minute average (because that's all that's available in both cases). I have my logger set to 5 minutes and I have the station calculating the 10-minute average.

And here's my graph having unset 'plot latest gust' (and restarted Cumulus, so it re-draws the graph), so it's now plotting 'peak' gust all of the time. This is the way I normally run as I find the display much more useful.

wind2.png

Both lines are now fine, because it's now plotting the same thing regardless of whether it came from the logger or from live data. The only difference is the time period over which the plotted points are calculated, 10 minutes when the data was gathered 'live' and 5 minutes when it came from the logger (because I use a 5 minute logger interval).

[opticsman0127]

I understand what you are saying and, I understand the response in your graphs. Looking at the 2 variants you show, the actual plot of the wind speed in both cases seems virtually identical, so not a problem. Even the Gust plot is not miles away, the plotted Gust from logged data is, on the whole, below the 'real' values.

My set up, for whatever reason, isn't doing that. Please refer back to my graph of 5:00pm yesterday which shows massive anomolies in actual wind speed and plotted wind speed. The 'Speed' in my plots virtually is identical to the 'Gust' speed when plotted from lodded data. If I leave it then running live data, it settles down but, here's the difference, it doesn't re-draw the speed plot, it leaves massive peaks visible which are obviously incorrect in terms of values. your redrawn plots from live data seem fine,mine are not.

From the information you give me, i think my settings are identical to yours so i can't see what is causing this frustrating problem. At work at the moment so unable to offer any plots or values until alater this afternoon. Cumulus currently switched off.

Barry

[steve]

The 'Speed' in my plots virtually is identical to the 'Gust' speed when plotted from lodded data.

You have Cumulus calculating the average, in that case. For a 10-minute average with an interval of 5 minutes between data points (so each average is only calculated from two data points), and with your high ratio of peak gust to average speed, then one high peak is going to skew the average. And when the average is from two peak values, then it's going to come out close to the peak values, as seen in your graph. If you switch to letting the station calculate the average, then the "wind speed" plot on your graphs will be much more consistent between live and logger data. Your "wind gust" plot will still show the higher values from logged data, as does mine when using 'plot latest gust', even more so in your case with your disproportionate peak gusts. Turning off 'plot latest gust' would make your graph look much like my second one (with a bigger separation between the two lines). If you don't want to turn off 'plot latest gust', the only alternative that I see is the proposed "weatherlink mode" option with its dumbed-down plot of just the average speed.

Notes about the settings: If you change the 'plot latest gust' setting, then restarting Cumulus will cause the graphs to be redrawn and the new setting is then apparent. If you change the 'calculate 10-minute average' setting, then the data starts getting calculated differently from that point onwards, so the change is only effective from then onwards, even if you restart Cumulus - the data has already been stored by Cumulus using the previous method.

[opticsman0127]

That sounds like a solution, at least a far better option than I already have Cumulus set up for. Will try it tonight and hopefully report a much more realistic display.

If you did implement the option of the 'weatherlink' format, would Cumulus still record numerically data as it does now or, would it combine gust and speed into one set of figures? Whatever the answer, i think it would be interesting to have that option.

Barry