I’ve read through the various pinned threads on CL, and I’ve done some searching but I still have a few questions.
The short version is simply how long should I wait before I report what I think are calibration issues? The CL threads indicate they run every day and in a couple of the threads they say if there’s an issue after a day or two to notify them. I know the WeatherFlow folks are swamped with all the new Tempest users, so I don’t want to needlessly bug them if the answer is “wait longer”. But I also don’t want to wait too long either, since this data is being broadcast into CWOP and various other places, and local paragliders, surfers and others are using my weather station (in conjunction with my webcams) to get an idea of conditions.
I’m also wondering if data from my two previous Tempests are being used in the calibration for this one. I did the “replace device” so that it would keep data. But now I’m wondering if data from those two failing Tempests might be impacting the calibration of this one? Or is it tracking the deviceId and only calibrating using data from that deviceId?
I put the latest replacement into service on 10/10 around 4pm PDT so it’s been up about 3 days now, and we’ve had some remarkably clear skies during that period.
My temperature seems to be reporting low, but only during the day. At night it seems to be pretty close to “accurate” (best I can tell). But the daytime temperatures are registering a fair bit lower than any neighboring site, and I’ve tried to double-check it with several different temperature gauges by bringing them up near the tempest (but put in the shade). It seems like it’s at least 8-10degF low at the peak of the day (at least on sunny days …). MesoWest has flagged my temperature data as “Caution” the past couple days:
And you can see that temperature is causing the problem when looking at the data quality graph (they flag temps with differences over 15):
The barometer is also reading a bit high - roughly 0.1mmHg as best I can tell based on various readings in the area.
The solar radiation is also reading higher than any station around me. And I don’t know where the Belcher skin gets its “theoretical max” values, but I’ve been well above what it reports as the theoretical max:
My humidity also has been reading well higher than any neighboring sites - anywhere from 20-40% higher. Now of course my location is a bit different than my neighboring sites, as I’m just 70 yards or so from the ocean - most of the other stations are further inland. That definitely has an impact on temperature and humidity. Though I’ve also looked at stations up-and-down the coast which are almost as close to the water, and my numbers are still a bit off compared to those stations.
IIRC, it takes about a dozen clear days during the hour around solar noon to calibrate brightness. You can ask support to do a manual calibration to get it to more sane values.
It took mine less than a week to calibrate the barometer reading. If you haven’t done it, you might cross check the elevation and AGL values.
That temperature seems to be quite aways off from what I recall seeing in the thermal compensation discussions.
Thanks for the info. A dozen clear days? Wow, glad I have it set up now, as this is generally our sunniest time of year (late september into early november). Other times of the year it might take a couple months to get a dozen clear days.
And yeah, the temp certainly seems a bit off (at least during the day), and doesn’t seem to be getting better.
Oh, and I forgot to mention on the pressure … I’m not sure there’s much else I can do in terms of “checking” the elevation and AGL values. When I was setting up the station initially, I compared what the setup said for my elevation to what Google Earth said. Tempest setup said 16’, google Earth said 10’. Based on what I (think I) know, I think 10’ is closer to accurate for the actual spot where my house is located, so I changed it to 10’. For the height above ground, I used the architect’s plans for my house to determine the height of the roofline and then measured from the roofline to the Tempest.
Overall I don’t think I could be much more than about 10’ off. Roughly speaking, I think a difference of 0.1mmHg corresponds to about 100’ at my elevation. And unless I’ve got my brain twisted around, I would have to be physically 100’ lower for the readings to be correct. And at that point I’d be well under water.
Also note that my two previous Tempests were reporting pressures closer to what I think they should be. They were a shade high, but only about 0.03mmHg.
So, the actual hardware is up and running untouched since almost 5 days?
That might, depending on the weather, not be enough for the CL for all parameters.
As gizmoev already mentioned and you remarked, the sun sensor will probably be the last one to get calibration, the pressure sensor should be one of the first, except very stable conditions.
After 5 days one should begin to see an effect. If do not still, I think, you should ask support on Sunday to have a detailed look, even so, they have a lot workload on Mondays, they will come back to you shortly after.
Yes, up and untouched since saturday afternoon, so not quite 5 days yet. I haven’t seen any change in the pressure reports, and the mesowest graph shows it’s been a fairly consistent difference from their expected. If the mesowest graph is actually correct (which it may not be, depending on how many stations they’re using which are as close to the coast as I am) the temp has actually gotten worse every day since I installed it. Here’s the latest:
Strange. Let’s have look, maybe @WFstaff or @WFsupport have taken note of this topic and will have a look into the data, data quality and calibration data.
For comparison, I see, you have relevant data in your profile
Maybe you post an update here on Sunday, hence you could reference the topic in case you have to open a ticket.
I also just added my CWOP/MADIS IDs as well as a link to the weewx weather page I’ve been working on, based on the Belchertown skin but also adding my webcams and some ocean conditions: http://weather.pacificaview.net/
humidity : ran once but not applied, didn’t pass QC
pressure : ran several times and only once applied end september, since then no more changes as they didn’t pass QC (seen the change, next time it passes it will adjust again I think)
UV : ran 3 times but needs at least several more before anything will be applied.
rain events are followed on regular basis for you.
Thanks for checking! The pressure adjustment in September would have been for my previous Tempest - this one went online 10/10.
Is there any adjustment for temperature? And/or is the temp possibly being affected by the UV and humidity? If it’s using a too-high solar radiation to compensate for housing temp, that might contribute to a lower temp reading during the day, right?
I’m not sure I grok’d the “rain events are followed on regular basis for you” comment.
I’m not 100% sure with wind for me because it’s only 5metres up from the ground but I had my first rainfall today and it was spot on with 1.3mm. Will see how it goes on Saturday when we’re expected to get up to 30mm! The barometer is a fair bit off atm but hopefully that sorts itself soon.
I don’t think there is any CL for solar radiation. Only for UV. But I always wondered why they didn’t apply the same calibration they do for UV to the brightness/solar radiation, after all it is almost the same sensor. If the incorrect values are caused by transmission through the window on top, the exact mounting underneath the window, chimney in the way etc, one could assume the same calibration parameters would be valid for UV as for brightness/solar radiation.
That’s interesting. I also would’ve thought both would get adjusted. So my solar radiation is always going to be wrong? Or WF can go in and apply a manual adjustment?
On the temperature discrepancies, it sure seems like something’s going a bit sideways with whatever compensation algorithm they use, and/or it might be related to my humidity and solar radiation readings which seem to be pretty high. Watching how it behaved today was interesting because we swung through several different conditions.
In the morning we had an offshore wind which, as it usually does, leads to lower humidity and higher temperatures. The temp got up to 84F and through all of the morning, the temps from my Tempest roughly corresponded to the closest stations, as well as the Half Moon Bay Airport, which I use as my best sanity check because it’s almost as close to the water as I am, is only about 10mi away and I’m presuming should be using relatively good, calibrated equipment. It also matched the two thermometers I’ve been using to try and compare. So far, so good.
Then the wind shifted to a more on-shore wind around noon which caused the temperature to drop quite dramatically and the humidity jumped up.
Now while the temperature did drop quite a bit, it seems like the Tempest dropped more than it should. Compared to my thermometers and the airport readings, the Tempest was typically between 5F and up to 9F different from my thermometers and the airport readings. And about 20F different from almost all of my closest stations - I suspect because it wasn’t enough of an onshore wind to carry the effect too far inland. So certainly my location is a big part of the discrepancy, but I don’t think it’s all of it.
Anyway, later in the day the onshore wind died off, so the humidity dropped and the temp went up a bit. It was also later in the day, so lower solar radiation. At this point the Tempest was back to reading fairly close to my thermometers and the airport. It was typically within about 1F.
I don’t know if any of this is helpful, or completely unrelated/uncorrelated, but I thought I’d mention it in case it helps.
Hey everyone, good discussion. I’d like to answer a few questions on calibrations
We can apply a calibration adjustment manually to the solar irradiance readings. Usually, a simple offset will suffice. If you have reference data to share, great, clear days with single peaks are helpful. Submit a help ticket and we’ll take care of you!
On the ambient temperature correction…
The Tempest software adjusts the raw temperature reading to an ambient temperature simulating a fan aspirated environment. This correction is calculated using solar radiation and wind speed data. Stronger irradiance values and light or calm winds will yield a higher temperature correction up to a maximum of 1.5 degrees F.
Temperature discrepancies are often explained by location differences. Siting of weather equipment can also play a role, heat sources, or radiant heat from a roof or the side of a house can contribute to discrepancies.
Is there any particular reason why the solar irradiance cannot be calibrated similarly to UV using the Continuous Learning system? On days when the CL system determines that the sky is sufficiently clear at local solar noon that the UV readings should be close to the theoretical upper limit, isn’t the same also true for the solar irradiance? If so, why can’t a calibration be made to the solar irradiance by assuming it should also be close to the theoretical upper limit?
OK, well if the adjustment is a max of 1.5F then it probably doesn’t explain the apparet discrepancies I’m seeing. My numbers are going much further off than just 1.5F. But not all the time. It seems to be worse during the day when humidity is high.
The hardest bit for me has been trying to determine what value is “correct”. Due to our micromiclimates and my proximity to the biggest temperature equalizer in the world (Pacific Ocean), it’s difficult to make comparisons to other stations. And I have a suspicion the MADIS algorithm is placing a lot of weight on SFO (san francisco airport) because it’s an official station and fairly close to me. But it’s on the “other side of the hill”, and the climates often differ quite a bit. I typically try to use the Half Moon Bay Airport as a reference since it’s also on the coast-side, it’s fairly close and it’s also fairly close to the ocean.
But even taking all of that into consideration, it still seems to be off, as best I can tell. But again, it’s not a constant differential. I’ve compared with hand-held thermometer readings, and with my purpleair that I just installed a few days ago just a few feet away from my Tempest, as well as reports from other stations nearby who are similarly close to the water. Ironically the purpleair seems to be coming closer to all of the other thermometers than the Tempest.
Note also that all of the siting influences cited would be causes for the Temest to read high, not low.
Anyway, as I mentioned, there are certainly likely microclimate issues at play when looking at the quality control data from MADIS, but the interesting thing is that the trend is to continuously get worse every day.
As for the solar radiation - I’m not entirely sure where/how to get reference values. But if it helps, here’s a snapshot of the belchertown-based weewx graph, which shows days where we’ve had remarkably clear skies and where my reported solar radiation has been well above what the belchertown skin reports as the theoretical max (though I’m not sure where it’s getting that value). Also note what appears to be some errant spikes.
That’s kind of my question as well. However, I’m not sure if there is a good model for solar radiation available. The UV values I can compare with local stations or national numbers.
But, if lack of good reference values is the problem, why no apply the same correction that is done to the UV measurements, to the Solar Radiation/Brightness values. Seems logical to assume that they are very closely related. @WFstaff
I would be surprised if it were just an offset, as an offset would create non-zero values when there is no light at all. But it could be in a first approximation a multiplication factor.
FWIW, the WeatherFlow staff applied a 0.8 correction to my solar radiation and it appears to be tracking much more closely to the theoretical max. You can see when the correction was applied.
