Cumulus Support

Not necessarily. To get it reasonably close all you need is a bright clear sunny day then calibrate the sensor to read the same as the Ryan-Stolzebach theoretical maximum for that time and place. In Davis's case they need an accurate known constant to calibrate to.

How do you know what transmission factor to use in the formula, if you don't know what your actual readings should be? I've used 0.9, which is quite high in the stated range. Given where I live, it wouldn't surprise me that if I genuinely do need a high value. On a bright clear sunny day, my readings are close to the calculated maximum, using that value. But I have no way of knowing whether my readings are accurate, or whether I have applied too high a factor to get them to match.

steve wrote:

floggitt wrote:Glad to see this thread as my readings are identical to everybody else's. Would it be possible to add a calibration factor in the setup options in cumulus.

How do you know that your solar readings are wrong? What are you comparing them to?

Like yourself I am more interested in hours of sunshine then w/m2, I have set sun threshold at 100% and trans factor at 95, the result for yesterday is a theoretical maximum of 700 but an actual reading of 900. I have tried "playing" with these settings which is not so easy with limited sunshine. Any less than 100% threshold and the sun lamp is on even when its in cloud. Looking at the graph I would have thought when the spikes hit the red theoretical max the sun is shinning a little over maybe is ok but at 10.30 the reading is 50% over the maximum. Maybe my settings are way off or my sensor is fauty any suggestions please

floggitt wrote:but at 10.30 the reading is 50% over the maximum.

If you want to set your values empirically as I have done, you really need to wait for a good clear sunny day and compare the curve from the sensor with the calculated theoretical max. I understand (but I cannot remember from where I got this information) that a few out of range "spikes" are to be expected with the Davis sensor (as you had at 10:30), so you need to disregard those when doing the comparison.

Ok thanks for the very quick response will wait for a sunny day

Paul- your plot looks perfectly good, as my Cumulus plots are. What I was referring to are the Ly calculations by Davis in Weatherlink, which are sometimes grossly in error, and I think the wind run calcs also suffered the same problem... but its no longer a probem, as I no longer use Weatherlink

The peaks in your graphs are due to cloud edge effect, where the edge of a cloud passes near the sun, acting effectively as a second sun in the sky, albeit a bit dimmer. I've seen peaks of 1700W/m^2 for the ~1min integration periods used by Davis, and 1300-1400W/m^2 are common when there are a few clouds in the sky with the sun well above the horizon. Lots of scattered bright Cu clouds in a clear sky with the sun shining can give extended peaks well above the theoretical max, which only refers to a clear sky.

Steve I've just had a long conversation with a specialist scientist at the NZ government's National Institute of Water and Atmosphere Research. Firstly according to their records the Ryan-Stolzenbach theoretical maximum radiation levels here in Christchurch match your T/Max curve almost exactly using the 0.8 constant. Secondly, the spikes reported from the sensor are most likely to be caused by the pure white radiation from intermittent bright clouds, the radiation from which can often exceed the blue sky radiation. Looking at the data collected here at various times this make complete sense. The suggestion made by NIWA is to read the sensor every minute and report the average every ten minutes. If I do this manually the sensor readings are almost exactly the same as the theoretical maximum on a clear sunny day however it would still be possible to read and average ten high spikes in a row. Another more precise way would be to compare the next reading from the sensor with the last one and if the increase is beyond a reasonable and predetermined amount, declare the last reading invalid and use the previous one. I have used this technique many times to remove spikes from inputs to A-D converters in the past. I'm not sure whether or not you could add this in to the solar reporting code in the next update? Anyway it has provided the answer to the sensor over reading problem. Thanks to everyone concerned.

I'm not sure why you would want to average out the spikes to make the data fit the theoretical curve for a clear sky, since they are real increases in solar radiation levels at the Earth's surface, due to forward scattering and reflection by clouds- the cloud edge effect, as it is known in solar power circles. The peaks certainly coincide with output peaks from my PV solar arrays.

The posted curves showing "proof" of excessively high readings seem all to have been collected in conditions of scattered cloud. Here in Albuquerque, NM, we are blessed with extended cloudless periods, mixed with some scattered cloud on other days.

The issue I think you are missing in firmly holding that your sensor is reading too high is that in conditions of scattered cloud your sensor can correctly see more than maximum solar (or UV) in the case that the direct path to the sun is fully clear, but that light reflected from other clouds adds to the reading.

Here are two day curves from my VP2 solar sensor. On July 22 much of the day was cloudy, but direct sun reached the sensor during the afternoon, with some recorded spikes well above the clear-sky value.


On July 29, the day was nearly cloudless--so you can see the effect of my rather inadequate siting, which puts intermittent shade from a row of Arizona cypress trees on the sensor during the later afternoon. Also the mountain ridge to the east blocks early morning sun, followed by yet more tree action. One can also see a couple of indentations for the five-minute samples compromised by the guy wires on my installation.


Lastly, the shape of the max curve is not perfect, I imagine the atmospheric blocking behavior of "clear" air varies quite a bit from location to location around the world.

Details aside--I suggest that if you want to calibrate your sensor maximum, you wait for a cloudless day (or hour, anyway). Depending on where you live, that might be a long wait.

[edit: I made this post having only read the first of three pages of posts in the thread, and did not realize that several others had mentioned the additive effect of reflection from off-path clouds before me. I'll leave the original text unaltered--and just apologize to those I appear to have ignored.]

I appreciate this is an old topic but I have been 'tweaking' my solar settings. Fortunately we get many clear sky days here.

My settings to give this graph are:

LuxToWM2=0.00832
Sun Threshold 70%
Trans Factor 0.91 (which can probably be lowered).

Hope that may help anyone with the solar settings

Andy

Hi Andy,

Just come across this post from you from last year! Is the graph in your post a "standard" Cumulus 1.9.4 one - I can't find it anywhere and I've just converted to a Watson W8681 solar station so am interested in being able to graph the solar values like this.

Cheers,

Mike

It's a standard graph. It should be selectable on the graphs screen - is it not there?

Hi Steve,
Yes, the graph is there - it was another "D'oh!" moment from me - I'd forgotten to set the relevant cumulus.ini parameters to "1"...
Mike

This is the theoretical maximum possible radiation. The sensor can report less than t/max due to losses through cloud, fog, pollution etc. but cannot report more.

I first ran across this thread about six months ago when I was programming my personal weather program's solar gauge. After a lot of thought on the matter, I believe calling the Ryan-Stolzenbach curve the "theoretical maximum" is inappropriate. Developed in 1972, it was one of the first solar irradiance models that became widely accepted. It was not designed as an attempt to represent a "maximum" curve, but a typical curve based on empirical data. A number of alternative formulas have also been developed ( http://pdxscholar.library.pdx.edu/cgi/v ... cengin_fac ).

The SolRad program, available at http://www.ecy.wa.gov/programs/eap/models.html, includes the R-S model as well as the Bird (1982) and Bras (1990) models. For any given day, each model produces quite different curves, based on assumptions and user inputs defining atmospheric conditions. I find the Bird curve most suitable for the solar data gathered in Sacramento, CA, using a Davis sensor connected to a Wunder Weather Station through a 2.8x precision signal amplifier to make best use of the 5V 10-bit ADC on the WWS.



Solar data is gathered once each second. The prorgam offsets the data by temperature according to Davis specifications. A least-squares calibration routine in the program was used to adjust solar sensor gain as well as determining the additional temperature found in a roof-mounted location. The fit of this curve is excellent. In the morning, trees and mountain haze over the Sierra Nevada to the east diminish the solar readings; in the afternoon, trees and buildings cast shadows starting around 4:40pm.

Here is the same day's data plotted against the Ryan-Stolzenbach equation. In this example, the gain was adjusted so the maximum readings of the observations match the peak of the R-S curve.



The readings before and after local noon are significantly higher than the R-S curve. Clearly, the R-S curve does not calculate the maximum irradiance. It appears to be shaped more like what would be observed in Phoenix or another region with very clear air. The humidity in Sacramento fluctuates between 25 and 80% at this time of year, so there is greater diffuse radiation compared to direct radiation, consistent with the shape of the Bird curve. This is the result of using least-squares calibration against the R-S curve.



Although the standard deviation across the sample data is minimized, the R-S curve underestimates the solar radiation in the morning and afternoon while overestimating the radiation midday. This looks similar to the plot posted by archae86. It appears to me that many or most locations cannot be calibrated to the R-S curve.

Calibrated to the Bird curve, the daily readings are quite revealing. A day with high thin cirrus clouds will show -2.5% (or a similar value) much of the day. Today's solar radiation has been about -4% after the sun rose past the obstructions. Should it change, there is no doubt that the atmospheric conditions are changing.

Here are plots from a few other days. 14 May was punctuated by a few cumulus clouds; the graph shows forward scattering and cloud edge effect quite clearly. 16 May was a very cloudy day. 17 May started with high thin cirrus which grew in density midday, then returned to almost clear skies. I am very satisfied with the performance of the Davis sensor. I can't speak to its absolute calibration, but when calibrated against a representative curve, it provides outstanding, consistent results.

Interesting, from "eyeballing" it I think my homemade (attached to a Davis ISS) solar sensor curve matches the Bird model better than the R-S model too. I'll collect some data today (looks like its going to be sunny most of the day) and compare the models....

Also its -past- time to "adjust" my R-S model transmission factor in Cumulus from its winter value to summer value. I find I have to use a higher transmission value in winter than summer - possibly due to the additional water content in warmer air?