Cumulus Support

I notice that the wind gust ratio appears higher since build 3238. For example, today I have had gusts to 31 knots, and the mean speed has been no higher than 15 knots, whereas through experience I would anticipate gust values being 3-5 knots less than have been recorded.

Lots of subjectivity in that above sentence, but I have been in this game long enough to feel a bit suspicious.

I noticed this difference when upgrading to b3238 in mid April. Since then we have had a fair few occasions when the wind has been between east and south and that direction is heavily influenced by the high ground to my south-east, so I thought the increased gust factor was (probably) due to that. So I have been waiting for a day such as today when there has been a reasonable breeze from the W-WNW, which is an exposed direction for my anemometer - and the increased gust factor is still present.

Has there been any code changes surrounding wind speeds in the recent builds?

I was porting some of the latest updates to V4 yesterday and noticed that there may have been an issue introduced. The affected value is the average speed though. The gust values should be ok. My initial investigate shows that there are two effects, one mainly affects Davis WLL stations (or at least that's where I first noticed a problem), the other any station that uses a calibration factor for the average speed.

Apologies for this, I'll work out a solution for the next patch release...

Thanks for taking a look, Mark, and no need to apologise

It was the ratio between gust speed and mean speed that caught my eye. TBH I wasn't sure if it was mean speed or gusts that was out, but I must've been subconsciously leaning towards gusts - hence the slightly misleading topic title

Just revisiting this topic again with another observation.

The last few times (not sure how many) I have restarted MX after a reboot I have noticed a difference in average wind speed when the data is extracted from the logger archive compared to the average wind speed from live running - as much as 2-3 knots.

I have MX calculate the average wind speed, but obviously this doesn't happen when archive records are used. When viewed on a graph the average wind speed from the logger archive is higher by the difference I mentioned.

As an experiment this afternoon (a breezy day here) I unticked the "calculate average speed" option and saved settings. I left it running like this for 40+ minutes (16:04-16:48 GMT). The result was a notable increase in average speed on the graph while MX was not calculating average speed.

I have MX calculating average speed because the station rounds to the nearest 1 mph and I don't like the resulting "stepped" graph. But until very recently I thought that the MX average was accurate, so was happy to use it.

I attach MX diags file covering the period of the test, with data and debug logging switched on. Apologies but the test unfortunately spans two log files.

I'll take a look, but to examine your data I'll have to decode 240 loop packets, average the wind speed from all of them and compare with the average value MX is calculating and the logger is sending. A job for tomorrow.

I'll fire up my VP2 logger and live debug my values which will be a bit easier.

mcrossley wrote: ↑Mon 18 Sep 2023 8:30 pm I'll take a look, but to examine your data I'll have to decode 240 loop packets, average the wind speed from all of them and compare with the average value MX is calculating and the logger is sending. A job for tomorrow.

I'll fire up my VP2 logger and live debug my values which will be a bit easier.

Thanks Mark, much appreciated. You'll likely see something if you look at just the first 10 minutes - it was immediately apparent.

Before I dive into this today, I had a thought last night... By any chance are the speeds lower by the mph/knots conversion factor?

My first pass at analysing your log files - from the start of the first file, with MX calculating the average. The MX calculation is correct and matches the value in the station for the first 10 minutes.

Next, I'll look at 10 minutes in the period when MX was using the station average.

Easier than I thought as I had the spreadsheet setup

Yep, when using the station average for the 10 minutes 16:10 to 16:20 the MX average is 2.8 mph too high.
I'll investigate that bit of code....

I cannot reproduce the error. With my VP2 set to use the station average, the value in MX matches the value from the station.

Could you check if you have a Wind Speed calibration value set (not gust)? That will only come into play when the station average is being used on VP2 stations.

Thanks for your investigative efforts. Yes i have a multiplier of 1.2. The multiplier^2 is set to 0.

Right, when MX is calculating the average on the VP2 it averages all the instantaneous "gust" values which have their own calibration factor. As the gust calibration is already applied it does not apply a further speed calibration. When using the station average, both the gust and speed calibrations are applied to their respective values.

During catch-up the station average is used because the instantaneous speed readings are obviously not there.

Thanks Mark. Reading between the lines I need the same factor for both gust and speed if I am to have the speed calculated by MX?

In short, yes.

The longer answer I have been thinking about. This will of course also raise your peak gust values by the same factor. Normally people are using a calibration factor because of sub optimal sitting. In my limited experience these locations can also be quite gusty due to the turbulence, so are the peak gusts actually that much lower than they would be clear air? It would require real data from two sensors to answer that question, and of course it would be highly location and wind direction dependent.

My gut feel is that the shaded gusts will probably be lower than they would be in open air. I have no idea if that is correct though.

I can answer that one. Wind gusts are lower in shelter, but not as low proportionally when compared to the reduction in mean speed. Historically with Cumulus I have used a factor for mean speed, but either none or a much smaller factor for gusts due to this.