Description
Seeing describes the quality of atmospheric turbulence above the observation site. The Earth's atmosphere is constantly agitated by thermal convection cells of varying sizes and temperatures, each acting as a small deforming lens along the optical path.
The result is a blurring of the stellar wavefront that broadens the PSF (Point Spread Function) beyond its theoretical diffraction-limited value. When seeing is bad, stars remain round and point-like in the integrated sub, it is not a tracking or focus issue, but their FWHM (Full Width at Half Maximum) explodes, typically to 3-6 pixels instead of the 1.5-2.5 expected under good conditions.
The defect limits the effective resolution of the entire session, caps the detail visible in nebulae and galaxies, and cannot be "fixed" in post-processing: deconvolution helps marginally but does not recreate lost information. Seeing is the primary limiting factor for high-resolution imaging from an amateur site.
Visual signature
Stars are round and symmetric but visibly bloated. Measured via DynamicPSF (PixInsight) or FWHM Analysis (Siril), FWHM exceeds 3.0 pixels in conditions where the sampling and optics would allow better.
In an animation of subs before alignment, stars "shimmer" and change slightly in shape and position from one frame to the next: a direct signature of turbulence.
On targets with fine detail (galaxies, tight clusters), subtle structures (spiral arms, individual stars in globular clusters) are buried in diffuse blur.
In planetary imaging, the signature is even more striking: a mushy lunar surface, Jupiter with poorly defined bands, Saturn without the Cassini Division visible.
Overall image contrast is reduced, giving a "soft" rendering even after careful processing.
Differential diagnosis
To be distinguished from a focus miss (blurry stars but often with a more diffuse signature, sometimes a donut on a defocused Newton, check via Bahtinov mask or autofocus).
Not to be confused with chronic poor guiding (stars elongated in one direction, not round and bloated).
Different from collimation error (stars with asymmetric PSFs, comma or V-shaped, see dedicated defect pages).
Not to be mixed with dewing/fogging on optics (global luminous halo, massive loss of contrast, but FWHM not necessarily blown).
If FWHM is high and varies a lot from one frame to the next (>30%), it is variable seeing. If it is high and stable, suspect a permanent optical defect instead (collimation, spider alignment, tired aluminized optic).
Check seeing via a quick preview on a bright star: visible oscillation = seeing, fixed and fuzzy = focus issue.
Probable causes
- High-altitude atmospheric turbulence (active jet stream above the site)
- Unstable local boundary layer (warm ground under cold tube at session start)
- Poorly exposed site: near heated buildings, asphalt parking lot, neighboring rooftop
- Unthermalized tube (temperature difference between tube and outside air greater than 2 C generates internal turbulence)
- Pointing at low elevation (<30 degrees), traversing a large air mass
- Observing through a window, glazed wall, or enclosed veranda (catastrophic internal turbulence)
- Unstable weather (cold front, convective instability after a storm)
- Unfavorable season for the site (summer in lowlands, winter jet stream over certain regions)
- Sampling too fine relative to the site's typical seeing (pointless oversampling)
Course of action
- Check forecast seeing via Meteoblue (astro forecast), Clear Outside, or Astrospheric before the session
- Thermalize the tube: take the instrument outside 30-60 minutes before the first sub, actively ventilate if the mirror is thick (Newton)
- Prioritize high-elevation targets on nights with poor seeing
- Match sampling to your site: 1.5-2 arcsec/px for average seeing, 0.7-1 arcsec/px only on an exceptional site
- In poor seeing, prioritize SNR (longer subs, more frames) rather than resolution
- Rigorously sort subs with SubframeSelector, keep the best 70-80% by FWHM
- In planetary imaging, exploit lucky imaging: thousands of very short exposures, stack the best 5-20%
- For very exposed sites, consider short individual subs (15-60 s) which partially freeze turbulence
- On the final master, BlurXTerminator or classical deconvolution with a measured PSF partially corrects residual blur
- Document your site's typical seeing (FWHM statistics over 6-12 months) to calibrate your ambitions
The Doc's advice
Seeing is the weather you endure in silence. You can buy the best telescope in the world, but if you shoot from a balcony in summer above a parking lot, you will top out at 3 arcsec/pixel. Conversely, a modest setup under a good winter country sky can produce images that shame observatories. First reflex: know the typical seeing at your site, accept its limits, and match your sampling accordingly. Oversampling under bad seeing is just wasting SNR without gaining resolution.
Think you can see this defect in your image?
Run a diagnosisFrequently asked questions
How do I objectively measure seeing during my session?
The most accessible method is to measure FWHM on an unsaturated star at high elevation, via DynamicPSF (PixInsight) or the FWHM measurement built into NINA/SharpCap. Convert the result from pixels to arcseconds using your plate scale: FWHM_arcsec = FWHM_px x (pixel_size_microns x 206.265 / focal_length_mm). A FWHM of 1.5-2.0 arcsec indicates good seeing, 2.0-3.0 arcsec average seeing, and >3.5 arcsec poor seeing. Professional DIMM seeing monitors exist but remain rare among amateurs; FWHM measurement on image remains the pragmatic reference indicator.
Does seeing really limit resolution even with a large telescope?
Yes, and this is often the bad surprise for beginners who invest in a top-tier instrument. Beyond 200-250 mm aperture, the theoretical resolving power of a telescope becomes better than the typical seeing at an amateur site (1.5-2.5 arcsec on most European sites). In practice, a 200 mm and a 400 mm will deliver equivalent resolution on normal sessions; the 400 mm will only express its potential under exceptional seeing (<1 arcsec). That is precisely why lucky imaging is so effective in planetary work: it selects those brief moments when seeing temporarily collapses.
Can deconvolution compensate for bad seeing?
Only partially. Deconvolution (BlurXTerminator, PixInsight Deconvolution, RL deconvolution) attempts to invert the blur by estimating the PSF, but it cannot reconstruct information that was never recorded. On average seeing (FWHM 2.5-3 arcsec), it delivers a real gain in sharpness. On very bad seeing (>4 arcsec), it tends to amplify noise and generate artifacts (dark halos, see the dedicated defect page). The rule of thumb: deconvolution can recover roughly 30-40% of apparent blur, never fully restore it.
What sampling is right for a typical seeing of 2.5 arcsec?
The theoretical criterion is to sample at roughly half to one third of the FWHM (not the pixel), which for 2.5 arcsec seeing means no finer than about 1.0-1.25 arcsec per pixel. Going finer recovers no additional detail (turbulence has already blurred the image) and divides flux per pixel, degrading signal-to-noise: that is penalizing oversampling. For a site with average seeing, aiming for 1.5-2 arcsec per pixel remains the best trade-off, consistent with the general rule: never sample finer than what your seeing actually allows you to exploit.