Incorrectly Low Wind Run measurement ?

[AllyCat]

Hi,

I've only recently put my FO anemometer "on the roof" so hadn't taken much interest in the Wind Run display or records. However, it was quite windy a few days ago but my "Record High" seemed too low. It's given as 140 miles, but summing all the 5 minute logged wind speed entries (and dividing by 12) in a spreadsheet gives 250 miles (which seems about right as the wind was at least 10 mph for much of the day). This is with Cumulus 1.9.2 running on Windows XP, both logger and Cumulus set to a 5 minute interval.

So I've been watching the Cumulus displayed values and graphs and trying a few tests to see what's happening. I think the problem is that Cumulus correctly accumulates the value when it is running (and the value in the today.ini file is correctly recovered) but the data accumulated from the logger is only half of the correct value (so the issue is not going to be seen if Cumulus runs 24/7). I don't run Cumulus (or my PC) overnight and recently have been alternating between various versions during the day).

Here are a few example screenshots where I have overlaid the wind graph next to the Wind Run panel. The first shows a period overnight (Cumulus closed at 1.00 and opened at 10.00) where the reported Run is 20.7 miles yet the average wind is obviously well over 2 mph for 10 hours.

Windrun-20.png

After running Cumulus for about 2 hours the Wind Run has increased by just over 15 miles, which looks consistent with the wind speed graph.

Windrun-20m.png

Finally, this doesn't look like a recent bug, here is the same data imported into Cumulus 1.9.0 where, if anything, the reported Wind Run of 28 miles looks rather less than half of the product of average windspeed (almost 5 mph?) and time (almost 14 hours).

Windrun-190.png

I hope I've understood the definition of Wind Run and how it's implemented in Cumulus; I did search the forum for 'wind run' but got far too many hits on just wind or run.

Cheers, Alan.

EDIT: The Cumulus code has now been chaged so new readers may prefer to skip to my update here.

[steve]

The way to check it is to take the entries from the October log file which were taken from the logger, sum the wind speeds (the 6th column) and then divide that appropriately according to your logger interval. I've just done that (I hadn't run my 'test' version with my Fine Offset so far today), and I got a total of 2001.6 for the wind column. My station logger is set to 5 minutes, so I divide the total by 12, which is 166.8. Cumulus is showing 167.2 (the extra would have come from the first minute of running live before the total is updated on screen).

You posted in the beta section, but there is currently no beta, so which build are you using? I ask this because I do seem to remember noticing a bug fairly recently with wind run from logger data. I'll see if I can find out what it was and when I fixed it.

[steve]

Sorry, I just noticed that you had already doen the calculation I suggested. I've checked for the bug that I thought I remembered, but the code doesn't seem to have changed since at least mid-June. If you zip up the diags folder and attach it, I'll see if I can see anything.

[steve]

I've had a thought - do you have Cumulus set to calculate the average wind speed, and are you using 'gust' in the calculation? It ignores that setting for wind run when using logger data (there isn't really an easy way around it), so the wind run is calculated from the 'raw' wind speed readings from the station.

[AllyCat]

Hi Steve,

Thanks for your reply. I'm running the released 1.9.2 (but not the slightly fixed post-release version). Note I'm also seeing this with 1.9.0 which dates presumably well before June (or even 2011). I posted in the Beta section as I didn't want to clutter up the main forum if my analysis was wrong.

I'm definitely not using the gust value in my spreadsheet (which I've just checked as 37.9 miles for my first graph, to 10.00 this morning), but I do wonder about some of my Cumulus station settings. I have "Calculate 10 min wind average" ticked and "Use 'speed' for avg calculation" cleared. But it seems to be the "raw" data from the logger that is not being calculated correctly.

I'll zip up the diags later if nothing else comes to mind (you've already posted twice whilst I've been typing this reply).

Cheers, Alan.

[steve]

I have "Calculate 10 min wind average" ticked and "Use 'speed' for avg calculation" cleared.

Yes, that's what I believe is the explanation. Wind run from logger data is calculated from the wind speed supplied by the station, so it's not the same wind speed value that you are seeing in the logs. The values you see in the logs are calculated by Cumulus from 10-minutes worth (i.e. two readings) of gust values, because of your settings. This figure will be higher than the station speed value, hence the difference between the reported wind run and the value you get retrospectively by summing the wind speed column.

As I said, it's not easy to change this, but I may think of something.

[AllyCat]

Hi Steve,

Attached is the zipped diags (I moved out .002 which is quite large) but I think I may need to do some more study and tests. By clearing both the check boxes the Graph looks reasonably consistent with the Wind Run, but then both seem only about half of the spreadsheet value.

With my initial "bench tests" I did find that I needed to set the logger to display "gust" and set one of those flags (I think it was "use speed for avg calculation") to produce values consistent with each other (and subsequently Met office wind predictions) but maybe that is wrong? Otherwise, the wind speed readings look ridiculously low even for an urban area such as here. Or perhaps my FO anemometer is only generating one pulse instead of two per rotation (which could take some time to check).

Thanks, Alan.

[steve]

By clearing both the check boxes the Graph looks reasonably consistent with the Wind Run, but then both seem only about half of the spreadsheet value.

Clearing the 'calculate 10-min average' box should mean that the logged wind speed values correlate to the calculated wind run - over the period covered by the logger data. That's how I had mine set which produced the figures above.

[AllyCat]

UPDATE:

My wind sensor is indeed delivering 2 pulses per revolution and I did have only the "average" flag set in Cumulus : configuration : station settings, not the "speed" flag.

However, it now appears that Steve has changed the code sometime between build 1017 and 1022 (i.e. an "update" to v1.9.2) so that the reported wind run does now correlate with the displayed/logged "average" wind speed.

But my "concern" now is whether it is valid for me to use the average of the FO "gust" values to calculate the wind run. Steve has stated that these values correlate far better with his Davis than using the FO "speed" value. I certainly don't dispute this, nor would I expect FO to give more accuate values than a Davis, but my wind conditions here in London are very different to Steve's in the Orkneys.

So, I am attempting to analyse the operation and "calibrate" my FO kit, but for now I am NOT using the "gust" value (i.e. both flags are either set or clear). My reason is that I would prefer to apply a multiplier value to the average wind speed, since this can be "adjusted" by post-processing later. The problem with logging the "gust" average is that Cumulus basically records similar data in both the average and peak columns and "throws away" the FO average value, so it will not be available for analysis/use later, if it becomes possible to determine a valid or better calibration factor.

Cheers, Alan.

[mcrossley]

I do not use the gust value with my FO, I thought that the average was too high. From my reading, the normal expectation is for the average to be about 1/3 - 1/2 the value of the gust which is roughly what I get using the FO average. Using the gust value places the average much higher.

[steve]

I do not use the gust value with my FO, I thought that the average was too high. From my reading, the normal expectation is for the average to be about 1/3 - 1/2 the value of the gust which is roughly what I get using the FO average. Using the gust value places the average much higher.

I think that coastal locations tend to be less 'gusty' (lower ratio of peak gust to mean speed), so the 'gust' calculation probably works better here.