What is your seeing and how do you measure it?

From a number of forum posts, it has struck me that many of those who are new (or relatively new) to AP would benefit from understanding what typical seeing to expect from their site, and how one might measure it.

Despite being a professional astronomer for many years, where I had a array of means at my disposal to measure seeing on a mountain top (observed image quality, dome seeing monitors, external DIMMs, neighbouring telescopes you could call up) as well as a history of seeing measurements against conditions, including wind direction etc. I always know when to expect arc security or sub-arcsecond or to metaphorically pack up and go to bed (although I never did) when the seeing “blew out” for the site in question. Over 2 arcseconds for Hawaii/Chile/La Palma. A little more for Siding Spring.

Now in my AP hobby, I find myself relying on data taken with my own set-up on a non-mountain top site. I am left with the question - is the image quality (FWHM) I measure really indicative of my site, or does my rig need tuned.

I measure seeing based on the image FWHM I get as the bottom of my focus curve using my ZWO EAF with 1 or 2 second exposure.

This is usually in the range 2-2.5pix for my Esprit 100/2400MC combo and 6-8px for my RC8/1600MM. This translates to 4-6arcsec. This is horrible seeing for a mountain top, but may be OK for the mountain valley in which I live.

ALternative, it may be due to the scope (although both scopes are consistent), my mount/pier (but this is based on 1sec images) or locally induced effect (my observatory? - although I have had no better luck in the middle of my garden)

Or it could be the way I am measuring it. Certainly I can get lower FWHM by exposing for a fraction of a second on a bright star, but here we enter lucky imaging territory.

I am convinced that others must get better (in some cases really good) seeing, based on the images presented. Some sharpness can be recovered in post-processing (deconvolution, star reduction, MLT etc) but not that much. So are others just on better sites? Or have managed to set up their equipment better? Or only post images taken in the best of seeing. And how to people measure seeing?

I have also included a poll to gather together data on the seeing Astrobin users enjoy. It’s not meant to be scientific, but it might help inform expectations of those newbies like me.

Once again, thanks in advance for any helpful replies.

Hey Brian,

The first thing one should settle on is a definition of how to define seeing. Obviously there are several ways doing it. One is the FWHM. Although I never cared about measuring seeing, let's think of a way doing it. The effects to smear out a star are:
- Air turbulence (that's what we are actually interested in)
- Aperture diffraction
- tracking/guiding inaccuracies during long term exposures.

The aperture diffraction of course depends on the aperture but is a constant (for small APOs one might be even aperture limited than seeing limited). Therefore, I find the FWHM not very helpful as it measures the effect of three combined effects and you need to address each of them in order to figure out the contribution of the atmosphere.

I think a good way to measure air turbulence would to measure the position of the star's center (with sub pixel accuracy) over time. The three effects are playing on different frequencies. The aperture diffraction at base frequency 0. The tracking/guiding on the 0.001 to 0.1 Hertz range. The seeing is on the Hertz range and above. Filtering for the higher frequencies would give you the data on the seeing (base frequency and amplitude).
I think programs like PHD2 with its guiding assistant would do the trick already. If I remember correctly it reports the "high frequency" oscillations as RMS. It would use exactly this value as the radius of the seeing.
EDIT: of course one shouldn't trust this value blindly. I haven't found detailed information on how PHD2 computes this "high frequency" RMS.

That would be my approach (to be tested).

CS!

Björn

Good point.
When I used to use my Sbig ST8 with the AO8 system, I used to keep the resulting subexposure FWHM as metric to measure the local seeing. In best nights, I've had 1.2-1.3" FWHM on a 900 sec subexposure (C11 at 0.6 arcsec/pix, just a little undersampled).
Today, with my new rig (Newton 250 f/4 and Asi 2600) I've never gone down 2.5 arcsec.
But now there isn't AO8….
What is the real seeing?
Who knows…

If you want to (poor-man) measure your seeing and you have no access to DIMM then point your scope to Polaris (for the northern hemisphere) and take 5s untracked exposures and then measure your FWHM at best focus. That removes all effects of tracking, even jitter, you might have. This works for mid-northern latitudes only, alas. The souther you go the worse that will get. To actually measure the seeing you would need to correctly sample your Airy disk and in most (imaging) situations you don't and then this seeing/FWHM thing becomes a kind of tautology, IOW, there is an atmospheric seeing "out there" and there is the PSF you are measuring "down here". At the end of the day what count is the "down here" thing, not the "out there" one. If you really want to know what is going on "out there", pop in a high-power EP, chose a suitable star and look at the Airy disk. That will tell you all you need to know about the seeing at your location.

Well, IMHO it is more or less an academic question, as long we are not talking about technical problems of your setup causing high values. It is nice to know your seeing, but actually what will change If you know it? Same with SQM.

I have to live with the seeing and light pollution at my location. I know my location can neither compete with the dark skies of a remote location, nor with a mountain top seeing. I have to take what I get. I can complain and obsess about seeing or LP but not change it - except moving to another location or get my equipment remotely hosted, or stop imaging when seeing is too bad, what will lead to max 3 nights per year of observing.
So I have stopped obsessing about things I cannot influence and being happy with that what I can get 🤪 Makes life much easier 😜

I think Brian raised a very valid question here. I also try to optimize my gear and my skills in operating to achieve the best possible results.
So I need to know what the limitations are due to seeing conditions in order to be able to judge my results.
I understand that measuring seeing is a difficult and very complex process and I was quite happy to learn that there is a quite practical approach to things.

Meteoblue, https://www.meteoblue.com, is providing a forecast for the astronomical seeing for your location. They even give you the arc sec value!
See this screen shot from their service below.
There is also an complex explanation of their meathod (in French) here: http://www.meteosurf.com/spastro/seeing/index.html. I did not work through it in detail but it seems to be an scientific approach.

I think using this is a quite pragmatic way to get the seeing conditions, which I would recommend.

CS
Gernot

Hi Gernot,

I think this forecast is nicely suitable for deciding, whether to image in this night or not. But since the values are based on a forecast model, which is definitely not taking care about very local impacts ( e.g. heated house roofs, micro turbulence, many more) it is differing for sure. The value cannot be compared to what you will measure with your setup at all. Hence this data is definitely not suitable to optimize your setup.

If you want to make a valid comparison and optimize you setup you will have to use hardware based and calibrated solutions like this one for sure:
https://www.sbscientific.com/products/new-seeing-monitor/seeing-monitor-2/

This is also used in professional observatories.

Every source will give different measurements. Meteoblue for instantaneous seeing seems accurate when checking for Jovian detail and would be the lower limit for LRGB on shortish subs

0s. Meteoblue often states 1”-1.5” for my location. For AP it is a little different… frame and focus in SGP (5s subs) with a L filter for me will typically give 2.0”-2.5” which can translate to 2.8-3.0” on a 300s sub with 0.5” RMS guiding, flex, diffraction bleed (in SGP). The same sub may show 3-3.5” in DSS or 2.5-3.5 in PI depending on the diffraction model being used. The stacked image of 50 subs will be closer to 4”. But each sub varies by position and altitude in the sky and some subs at 5” need to be culled.

Narrowband will give slightly better measurement because your cutting out any focal shift in wavelength of off-band photons. Best frame and focus shows >1.5” most of the time.

All of these measurements get you into the ballpark… ****However, the best measurement takes a little work. Do a 3x3 drizzle integration on a bright part of your image with 20-30 subs. Zoom in on a small star (not saturated) and measure its diameter in both axis… is it round? Is it still blocky? Is it 3 pixels across (9 with drizzle)? Or more or less. Then check surrounding detail (blurred, sharp?, if software binned and rescaled to the same screen size is there any more detail?). Did you use flats and dither (which improve apparent detail). By doing this exercise it is easy to link internet seeing source, software seeing source and actual detail in your subs together. It will point out tracking or flex issues, maybe focus isn’t perfect, maybe sensor issues show (tilt, microlens…

which you can improve with a little work. After that “imaged” seeing is likely maximized for your location. Otherwise, everyone and every software measures differently.

After doing this exercise, tuning all of the equipment and being in Michigan under the Great Lakes humidity haze, down wind of Chicago and too close to the jet stream I expect a real resolution per 300s sub of 3”-4” which I sample at 3x (1.1”) with nearly round stars (0.4-0.5” RMS guiding) and refocus every 2 degrees F. In the end, the long term repeated recommendation for most people starting AP without a world class site is try targeting a 1”-1.5”/pixel imaging resolution as it is a good balance of best you can get without introducing too much equipment complexity, and avoids the worst of over/under sampled issues.

Excellent discussion. I have also wondered how others attempt to quantify this. My seeing is virtually always terrible (I quickly selected >6) as I am immediately down-wind of Lake Erie and that seems to cause a lot of local turbulence (in addition to Lake Effect snow storms....). My anecdotal data points are two things:

(1) Experimenting with my various cameras of different pixel sizes and getting a qualitative sense of under/over sampling. I have a small camera for imaging planets (178C) that gives me around 1.3"/px. When I tried imaging a DSO with it as an experiment, the images were so grainy and noisy that they were almost unrecognizable - I initially thought something was wrong with the camera (I have an old M33 I tried that's really bad and have never posted...). I now have two cooled cameras that give 1.8"/px (Altair 269C) and 2.0"/px (Altair 26C); the first was often OK, but some nights showed hints of the same noisy image quality. My 26C (2"/px) seems best-suited to my conditions and produces images that I like at the best resolution I think I can obtain. This was found essentially by trial-and-error though.

(2) Imaging Jupiter has been qualitatively instructive for me. Jupiter doesn't vary too much in size over the season and is usually has a radius of about 15-20 arc-seconds or so. If I just put my 178C in live-movie mode (say at ~50ms frame rate), I can watch from frame-to-frame as the seeing aberrates the planet. I can see it distort in shape (jitter) of course but also literally observe it shift in centroid position (tilt). I have seen some nights where it is moving by as much as 0.5-1.0 radius and the shape of the planet is virtually unrecognizable, so that could be as bad as 10-15 arc-second seeing maybe? Even on the best nights, I judge it to be no better than 5 or 6 probably. That's about 3-5X my sampling limit observations, which seems to fall in line with what people say should happen.

Where this discussion probably matters most for people is in buying new equipment and spending their money on good selections for their environment - a bigger telescope or higher resolution camera isn't always better. I've essentially learned that about 2"/px is my practical resolution limit (average "seeing" is probably 3-4x that number?). The nice thing is that my little 80mm refractor is already probably ideal for that, so buying a bigger one (which get both heavy and expensive) wouldn't ever really have any benefit for me. The downside is that it's not possible to take good images of many of the smaller galaxies, planetary nebula, etc. that others are able to do; I simply have to enjoy them on Astrobin. Between the seeing and the light pollution ( Bortle 7 - 8 ), I'll never get the IOTD quality images, but, like @Ruediger said above, my sky conditions are what they are, and it still seems preferable to get out and enjoy the cosmos as best as I can than to sit inside and watch TV.

I'd be interested from @Brian Boyle if there is a single standard or most often used method to measure seeing in the professional astronomy world? I would think it would desirable in a scientific setting to have a common standard used to compare results between different observatories.

I'd be interested from @Brian Boyle if there is a single standard or most often used method to measure seeing in the professional astronomy world? I would think it would desirable in a scientific setting to have a common standard used to compare results between different observatories.

Hi Matthew,

Yes, there is one: Differential Image Motion Monitor (DIMM)

Further information: https://www.eso.org/sci/publications/messenger/archive/no.53-sep88/messenger-no53-8-9.pdf
The method is used for the large one e.g. VLT
You may google for it, and you will get a lot of background information.

But to follow your arguments concerning camera, which are physically correct, would mean in consequence, I have to buy different cameras for each seeing condition or bin at least. I think from the practical point of view and the fact, there is only a bunch of sensors used for all amateur cameras, it is again more or less theoretical. All current CMOS sensor have extremely small pixels. Hence I can decide between heavy or moderate oversampling at medium and longer focal range.

I think due to limited budget and time our gear is always a compromise between theoretical optimum and versatile usage. Especially in non stationary setup.

This is a really Excellent and informative discussion thus far. I am learning a tremendous amount. Most importantly, the reassuring message for me (and other n00bs) that I am not alone.

To @matthew.maclean there is (or rather was) no gold standard measurement for seeing. You just accepted what you got on the night - capturing what photons you could down a slit, fibre or imaging circle which minimised the omni-present sky.

I also agree with @Ruediger Wemhoener that is a somewhat academic question, since there is nothing really that you can do about the seeing.

Nevertheless, I think an collation of seeing measurements from our sites will help frame and guide the expectations for those coming into the hobby. And it may help them in choosing imaging set-ups which are best matched to typical conditions at our sites.

Please keep your advice/input coming - it is all so useful in building a picture of what is possible.

I only just started paying attention to FWHM the past few imaging sessions, so I will need to investigate my subs and see what I've been getting and do some short subs to take out tracking and other variables. I will say that after looking at Meteoblue for the first time today I don't trust it much. Meteoblue range for seeing shows excellent to average worst case. On the Clear Sky Chart for a place just down the road some, I'm lucky to get average seeing and extremely lucky to get any good seeing. I don't think I've ever seen it show excellent. Who knows, maybe the Clear Sky Chart is the inaccurate one, but it seems to line up with my experience to date. Both charts are fairly laughable when it comes to relative humidity. I don't think I've had any nights where it didn't hit 99% save one where it only got up to 94% per the sensor on my equipment.

Anyway, this is a very interesting topic as I delve more into the science and also put more effort in acquiring the best data that I can from my backyard.

I measure my seeing based on FWHM.
Different software and different setting will yield slightly different FWHM.
As noted in one post above, starting at about 100mm and going smaller, seeing becomes less an influence on FWHM, and the optic's diffraction limit plays more the controlling role.
Are you sure you are calculating your FWHM correctly?
For the Esprit you should be around 1.2arc-sec/pix to 1.4 depending on focal reducer – so 1.2x2.5pix = 3arc-sec or 3.5as with 1.4asp. Which would sound more typical.
I found recently on some poor seeing nights (4arc-sec FWHM – I am use to 3arc-sec) – you can still get a very nice picture – but your ability to zoom in is not as great as with good seeing.

Ruediger:
I have to live with the seeing and light pollution at my location. I know my location can neither compete with the dark skies of a remote location, nor with a mountain top seeing. I have to take what I get.

Brian Boyle:
is a somewhat academic question, since there is nothing really that you can do about the seeing.

I am learning a lot from this discussion which, at the end of the day, as both, Ruediger and Brian did mention, there is no much I can do about it, and in my case I have observed also how it changes during one night session, this is particularly true as we get close to winter, where sources of pollution, namely smoke, mist and lights do vary quite rapidly. Sometimes I have to discard 40% of the images due to effect on seeing. The first sign that something goes wrong I noticed not only on the variation on the FWHM but also in the number of stars in the field that a program can count..., but again, this is an interesting thread. Thanks for the training in the topic.

Before discarding too many subs, I would suggest running a good set and a bad set (increase the number of subs you keep) – the value of the SNR might override the poor seeing.
I use a 4/3 size sensor (ASI294) – poor seeing reduces the ability to zoom in – but presenting the image without zooming in I don't think you can tell good FWHM versus poor FWHM.

The MetroBlue seeing prediction is very interesting. It shows a prediction for extraordinary conditions tonight here in central Oregon. Of course, it's pretty much like this all the time! (Not). Now I guess I'll have to pull the 20" out again tonight.

I can generally look at my guide signal to pretty accurately judge the local conditions. Here's the guide screen from just a couple of nights ago when the conditions here in central OR were in the range of 1.0" to 1.5".

The total RMS guide error is only 0.26". That happened for a single nights. The results for then next three nights ranged from 0.40" - 0.70" rms.

Out in NM, I've been comparing how my FWHM measurements in MaximDL compare with the DSW seeing monitor and I've found that it's actually pretty close. I use 1200s exposures so my measurements are made on a much different time scale than the seeing monitor so there are times when wind or other factors may cause the results to diverge.

This brings up an important point. In order to use FWHM as a gauge for seeing, it's important to use a scope with an Airy disk diameter and pixels sampling that's a bit smaller than the local conditions. The Airy diameter for my scopes is 0.8" and 0.5" and the pixels scale is on the order of 0.5"/px, which should be good for estimating seeing down to 0.5" - 0.8". Once you have a stack of data, the probably plot in the SubframeSelector is a good way to look at the probability of seeing conditions over the exposure time. The 50% point is a good way to report the conditions over the entire session. I just happen to have a probability plot for the blue channel for a data set to show what it looks like. In this case, the 50% level sits almost exactly on 2.0".

To more directly answer the OP's question: During the spring, the seeing here in central OR can range from 1" to 4". My guess is that the median might be around 2". I have a LOT more experience (and data) with the conditions out at DSW, where I'd estimate the median conditions to be around 2.2". It can reach 1" but that is extremely rare. Only about 25% of my data shows FWHM less than 2.0". In order to reach a data yield of ~50%, I generally have to set the upper limit for acceptable results to be at 2.4".

John

Hi @ks_observer Please keep also in mind: FWHM is only one among more quality estimators. Eccentricity and SNR are vital points too. It is also a question of target I am going for: A globular cluster needs good HFR, whereas a nebula can bare a high FWHM, but needs good SNR.

CS
Rüdiger

Hi Brian

I am a bit late to the discussion here, but I share your interest in seeing measurements.

In Colorado, east of the Rockies, we suffer from the mid-latitude jet and turbulence in the lee of the mountains. Nevertheless, I have seen some pretty decent nights, and knowing the seeing can very well affect what I am going to do. (Maybe some lucky imaging on a planet vs wide field work.)

Like you, I used to do this for a living, but now am faced with similar problems as an amateur astrophotographer. I am less interested in an absolute measure rather than a consistent differential measure from night to night so I can come up with a decent metric that says "its good" vs. "watch tv."

One could probably do a simple FWHM measurement and call it done, but I think a slightly more consistent and convenient approach is possible. Traditionally you want to take many very short images over some interval and measure the RMS deviation of a stars position.

Older systems used image tubes on a differential image motion monitor (DIMM). A DIMM is used to eliminate the problem of windshake and vibration because large variations will be correlated, vs the effect of individual seeing "cells" around 100 mm across. I'm not willing to dedicate a C-8 or C-10 to the task, but Santa Barbara Scientific has an interesting approach with their commercial seeing monitor:

https://www.sbscientific.com/products/new-seeing-monitor/seeing-monitor-2/

They use a small industrial camera lens and high-sensitivity CCD camera to monitor Polaris, taking images at around 10ms each. The small camera and rigid mounting can help rule out vibration and wind shake.

Polaris is the target , but has the convenience of being able to set up a static mount without tracking. I have cobbled together a prototype system using a QHY 5L-II camera and a 100 mm lens on a dovetail plate. I have confirmed I can get a pretty noisy but serviceable image at 10ms. I think that using something like a local version of astronomy.net software on a RasPi, one could take a long-exposure calibration image, apply an astrometric solution, and then take set of subsequent seeing (fast) images. Sextractor or astrometry.net can supply the centroids for the fast images, and the astrometric solution from the calibration image used to make a correction for the sidereal motion of Polaris in imager (X/Y) space.. In theory, what you will have left is the time-varying position of the centroid due to seeing which could be plotted to get a reasonable for the quality of the night.

I'm hoping to work on this over the summer after some other projects are complete. If I am successful, I would share the code for this (Python).

Ed Beshore

edits. Fixed size of seeing cells from cm to mm

I have both Seeing Monitor and Sky Quality Meter here is a graph from the seeing monitor on a reasonably good night. I am designing a DIMM for Zenith measurements using a synthetic laser guide star

Here is a graph of Sky Quality readings each day from new to 1st quarter moon showing the impact of the moon on sky brightness...

vertical axis : magnitude per Arc Second
Horizontal axis minutes from initial data.

Dark line == New Moon
Green line == First Quarter Moon

I now record these values along with weather conditions, and other data during imaging sessions, so these graphs are from the initial site survey I conducted...

Dave...

Wonderful set-up you have there, Dave. Great data - rivalling/exceeding a lot of professional observatories (at lest in my day). Thank you for sharing
CS and good seeing
Brian

Thanks Brian, its been working great for almost 10 years… I am in the process of upgrading-updating the instruments to win10 from winXp…

Regards,
Dave

Also late to this party, but I can share a couple of comments.

First, I've worked quite a bit with a DIMM (LBT on Mt Graham in southeast AZ).   I also do my own astrophotography a bit further down the hill at 4800ft elevation (vs 10,000 for Mt Graham).    A DIMM must be calibrated, so even with DIMM there is uncertainty about what the "true" seeing is! We often had telescope operators tweaking the inter-pupil distances to try and minimize the reported DIMM seeing values.   Clearly this isn't right, but it was impossible to stop them! I make this comment as consolation for us lowly astro-imagers who think that DIMM is "the" answer; DIMMs have their own issues!   DIMM readings often frustrate the professionals too.    This said, watching Mt Graham's seeing measurement over 6 years, the average seeing on Mt Graham from multiple focal station sources and methods was on the order of 0.7-0.9 arcsecs. Down lower for amateur work, the seeing is on average between (estimated by FWHM and guiding performance) 2.5-3.9 arcsecs.

I keep meticulous archive records of all my image subs, and for over 13K (kept) subs, the average FWHM is 3.5. I'm capturing with an 11" scope, so I have more than one air cell on the mirror, but in general, I consider FWHM values from 2.4 (min for this gear) to 2.9 to be good, and from 3.0 to 3.75 to be average. Once I start experiencing FWHM values above 4, I'm either south/low, or the seeing has turned bad.   I recently took an image of Omega Cen through 5.6 atmospheres (10 degrees max el angle) and had an average FWHM of 5.8.     I'm in the desert with low humidity and some altitude to work with, so lower altitudes and more humid environments I would expect to have higher FWHM.   I won't typically keep subs above 5.5 FWHM unless I'm *very*low south, and I'll start looking for issues (tracking, guiding, focus, etc.) when FWHM and/or HFR is high 4 territory.     Hope this helps….

Hi Brian,

Most amateur sites don't have seeing approaching 1", unlike professional observatories.

To measure seeing using image FWHM is OK, but there are some caveats:

1. The telescope diffraction effect needs to be accounted for. This is particularly true for guide scopes, which itself can contribute more than 1.5" to the FWHM.

2. The tracking needs to be very good, otherwise FWHM would include the effect of tracking errors and oscillations caused by winds, not just seeing.

3. The exposure time needs to be sufficiently long, like more than 10s. I think under short exposure time of a couple of seconds, the low-order tip-tilt effect that slowly move the star around would not be accounted for.

4. The image needs to be better than Nyquist sampled, and the optical aberration needs to be much smaller than seeing.

It's not easy for amateur equipments and imaging workflows to check all these boxes. I think in principle DIMM is better, but not everyone has access to one.

Hello everyone, I am moved by Doug Summers post to speak from the point of view of a backyard non scientist. I have just recently been looking at my FWHM results using Subframe Selector in PI just to learn more about the relationship between my apparent sky, my measured subs and the the overall quality of my final image. We have a year round average humidity of around 70% in my part of the Pacific Northwest and I can tell by the stars I see, my guiding and the appearance of my lights what the probable range of seeing I'm dealing with is, on an individual night. I think Doug's experience with his measurements is spot on and I can certainly tell when, on those rare nights, the sky is more transparent and the darkness is greater than my usual sky. Guiding seems to be the canary in the coal mine and almost certainly for me, the better my guiding numbers, the better my final image.
I have experienced far more FWHM's in the 4-5 range than 2-3 and I can't imagine what a night of 1-2 might look like in my backyard, but If the clouds are more or less not a problem I'm going to be aiming my telescope at something.

Cheers

scott

I do appreciate this discussion very much. There is a lot of value input. Thanks to all!

By reading all the answers, I feel encouraged in my initial standpoint:

FWHM is a suitable quality estimator to identify outliers of your subs taken in one night (aside with eccentricity and SNR) but only comparable with other other setups in a very limited way. The measuring needs to fulfill many strict requirements to be quantitatively comparable - which is actually hard to fulfill as an amateur.

Hence I see no reason to obsess about FWHM too much and use it more or less only for image weighting e.g. PI and maybe selecting a suitable target for my location at a certain point of time. Or it may help to identify an issue in my setup by comparing FWHM with identical imaging train. It may also help me to estimate what quality I can expect at a certain night. But it is not legitimate to compare FWHM between different setups.

What seeing do you measure, on average, at your location?