Strange solar readings

[peterh]

Today, here's what happened:



I can explain the spike between 11:30 and 11:45... there was a long band of cloud, which could have well been lit by the sun, and everyone who has ever handled a photography ambient light meter knows that a partially cloudy sky is a lot brighter than a cloudless sky (as well as a lot prettier to work with if you are snapping piccies).

However.
The max solar line in the graph appears to indicate that sunrise (the start of theoretically possible solar radiation) would be around 09:10.
However, Cumulus itself says (quite accurately) that today, sunrise is at 08:47.
I have been tweaking the values a bit... somewhere this morning, I adjusted the solar 'trans factor', which is default at 0.8, to 0,9, and then to 0.85... since Cumulus said it was sunny, and the cloud in front of the sun said it wasn't.
This explains the dip in the max solar curve. Or so I think.

Funny: the max solar line has sunset at just after 16:30, which corresponds pretty well exactly with what Cumulus itself says: 16:34.

I do not understand why the 'max solar' curve starts half an hour after sunrise.

Ideas, anyone?

[AllyCat]

Hi Peter,

Perhaps you should wait for a complete day when you haven't changed the calibration factor. Does the sensor have a full horizon-to-horizon view?

the max solar line has sunset at just after 16:30, which corresponds pretty well exactly with what Cumulus itself says: 16:34.

It doesn't look like that to me, the "Max solar" appears to "end" at about 16.15 (and start at about 9.15) which seems reasonably consistent with a "noon" at 12.45 (Longitude 4 or 5 degrees East with GMT + 1 hour).

The "max solar" curve used by Cumulus appears to start about 15 minutes after "sunrise" and end about 15 minutes before "sunset". I can't fully explain this, but perhaps the curve used by Cumulus assumes a (horizontal") "flat plate collector" sensor, whilst the FO station uses an "omnidirectional" (diffusing dome) sensor for the Lux measurement. Also bear in mind that the Watts/m2 scale (and Lux) is a linear scale whilst the subjective assessment of light level tends to logarithmic (i.e. 1 watt/m2 may appear as zero on the graph, but it will not seem "dark").

There have been many discussions on this forum about the difficulty of measuring "sunshine" from brightness levels (and also my opinions of the poor accuracy of the FO solar data) which is perhaps why two of the members developed the Sun Recorder.

Cheers, Alan.

[peterh]

Currently, the sensor has a "horizon-to-horizon view" of the sky from where the sun rises to where it sets. In summer, not so much.

And, yes, horizon-to-horizon view is almost literal. If you take a look at the background of my weather site (http://www.dnl-core.net/CothenWeather), the picture used as background is what the sensor location sees to the south. Open enough for you?

I don't know about your FO station, but my FO station (3081) uses a flat-plate collector which is slightly angled southward. There is no diffusor dome. And yes, being a photographer (and owning a Minolta Autometer IV, hehheh), I know about the linearity of light versus light perception

What puzzles me is the START of the Max Solar curve. If sunrise would be at 8:47, why would the Max Solar curve start at 9:10, even though it ends at sunset? Do I have something misconfigured?

[steve]

You can download a spreadsheet from the link below to check the figures calculated by Cumulus. Look for 'Solrad'. I simply extracted the code for the Ryan-Stolzenbach formula from that spreadsheet and translated it into Pascal so that I could include it in Cumulus. I don't pretend to know how it works. It's possible that I've made an error, or it may just be the way the formula works. If you put the same details into the spreadsheet that you've put into Cumulus, you should get approximately the same results.

What logging interval are you using in Cumulus? You need to take that into account when looking at the data in the logs (as with Select-A-Graph).

http://www.ecy.wa.gov/programs/eap/models.html

[mcrossley]

Looks correct to me, the value is below 1 W/m2 until some time after sun rise, and before sun set. Probably if you are in the tropics then the values will coincide. The spreadsheet calculates the correct sun rise time (elevation >0)

[peterh]

Hmmm... the yellow line is the actual w/m2. The pink line is the theoretical max solar line. So the yellow line shouldn't be ahead of the pink line.

I'll post a new graph when I'm done tweaking (tomorrow).

[peterh]

You can download a spreadsheet from the link below to check the figures calculated by Cumulus. Look for 'Solrad'. I simply extracted the code for the Ryan-Stolzenbach formula from that spreadsheet and translated it into Pascal so that I could include it in Cumulus. I don't pretend to know how it works. It's possible that I've made an error, or it may just be the way the formula works. If you put the same details into the spreadsheet that you've put into Cumulus, you should get approximately the same results.

What logging interval are you using in Cumulus? You need to take that into account when looking at the data in the logs (as with Select-A-Graph).

http://www.ecy.wa.gov/programs/eap/models.html

I downloaded the spreadsheet, and put my geographical data into it.
To me, it LOOKS like the spreadsheet thinks that the solar elevation only becomes a positive number after 10 am!

This probably indicates that I am doing something wrong. Could you point out to me which parameters I should pay attention to? This is what I put in:

latitude in decimal degrees (positive in northern hemisphere) 52,004
longitude in decimal degrees (negative for western hemisphere) 5,154
ground surface elevation (m) 4,0
time zone in hours relative to GMT/UTC (PST= -8, MST= -7, CST= -6, EST= -5) 1
daylight savings time (no= 0, yes= 1) 1
start date to calculate solar position and radiation 28-Dec-12
start time 12:00 AM
time step (hours) 0,25
number of days to calculate solar position and radiation 365

[mcrossley]

Hmmm... the yellow line is the actual w/m2. The pink line is the theoretical max solar line. So the yellow line shouldn't be ahead of the pink line.

No, but there again you have a £70 piece of measuring equipment, not £XXk of calibrated specialist device. Adjusting the factors in Cumulus to bring your readings down is the thing to do, but as you say clouds can increase the actual reading above the theoretical max. Fro my readings the FO solar readings are notoriously 'iffy', and it is not clear what they are actually measuring.

[steve]

daylight savings time (no= 0, yes= 1) 1

I think this is probably wrong.

[peterh]

Hmmm... the yellow line is the actual w/m2. The pink line is the theoretical max solar line. So the yellow line shouldn't be ahead of the pink line.

No, but there again you have a £70 piece of measuring equipment, not £XXk of calibrated specialist device. Adjusting the factors in Cumulus to bring your readings down is the thing to do, but as you say clouds can increase the actual reading above the theoretical max. Fro my readings the FO solar readings are notoriously 'iffy', and it is not clear what they are actually measuring.

OK... let me rephrase this.
The pink line shouldn't be behind the yellow line, because the sunrise (according to Cumulus, and according to my observation) is fairly accurately where the yellow line starts.

We're not talking about the FO solar readings being off (they aren't that much off), we're talking about Cumulus thinking that the sun is shining even though the pink curve says it hasn't risen yet - whereas the sunrise time that Cumulus comes up with is correct.

[steve]

We're not talking about the FO solar readings being off (they aren't that much off), we're talking about Cumulus thinking that the sun is shining even though the pink curve says it hasn't risen yet - whereas the sunrise time that Cumulus comes up with is correct.

Note that the sunrise/sunset times are calculated using different code (again, not my code, I have no idea how it works) and aren't related at all to the solar radiation calculations. Cumulus allows a minimum value to be configured below which it will never say that the sun is shining - precisely because of this difficult period at the start and end of the day.

I'm sure I have done this before in the past, when the calculation has been queried, but I put my details into the spreadsheet, and the figures matched the ones calculated by Cumulus for yesterday, to within 1 or 2 W/m2. Note that Cumulus rounds its calculated figure to the nearest whole number.

[peterh]

daylight savings time (no= 0, yes= 1) 1

I think this is probably wrong.

Hmm. If I set that to zero, this model says the sun rises (solar elevation becomes non-negative) somewhere around 08:51. Cumulus says it rises 08:47.

[mcrossley]

The pink line shouldn't be behind the yellow line, because the sunrise (according to Cumulus, and according to my observation) is fairly accurately where the yellow line starts.

We're not talking about the FO solar readings being off (they aren't that much off), we're talking about Cumulus thinking that the sun is shining even though the pink curve says it hasn't risen yet - whereas the sunrise time that Cumulus comes up with is correct.

My point is that the theoretical line 'should' start behind the actual sunrise, as Steve has said Cumulus uses unit values, and the theoretical value only reaches 1 sometime after sunrise for the Ryan-Stolzenbach calculation. The FO registers a reading before this for many reasons, notably it is measuring via a diffusing dome over the sensor (iirc) and not a flat surface. I stick with my assertion that the theoretical value generated by the spreadsheet matches my Cumulus values closely enough and is zero for a while after sunrise at this time of year with the very low sun elevation.

[AllyCat]

my FO station (3081) uses a flat-plate collector which is slightly angled southward. There is no diffusor dome.

Hi Peter,

Unless FO have drastically changed the 3080/1 Solar Pod, the angled PV panel is for (theoretically) charging the batteries, the Lux measurement is made by the small dome to the right and the UV by a tiny (upward facing) sensor to the left. See for example here.

Currently we have no Daylight Saving offset so IMHO the value should be 0.

Cheers, Alan.

[peterh]

Unless FO have drastically changed the 3080/1 Solar Pod, the angled PV panel is for (theoretically) charging the batteries, the Lux measurement is made by the small dome to the right and the UV by a tiny (upward facing) sensor to the left. See for example here.

... aaaand of course you're completely right.

Running the spreadsheet still gives me values that don't correspond with what I see when I look outside... at sunup (which is correct in the Cumulus main screen), I see the sun coming up (duh). Or not, if there are clouds (I guess you saw that coming).
In both cases, Cumulus sees solar radiation, and because this is higher than the max solar radiation (which is still at zero), it assumes that the sun must be shining.

Problem is now solved... or rather worked around, by putting brain in gear. If, at sunup, there is no sun because cloudy sky, the value of W/m2 will be below a threshold, say, 20. If I just plug that value in the Station Configuration, Sun min field, it should not register this time as sunshine. If the sun is actually hanging there, W/m2 should be above that value, and it's ok if Cumulus sees it as sunshine.

From what I've seen in the past 5 days, 20 or 25 seems like a good value. I'll need to keep an eye on this.

Case closed.