Atmospheric Seeing

What "seeing" actually measures

Point a large telescope at a bright star on an ordinary night and you will not see a crisp point of light. You will see a fuzzy, restless blob — boiling at the edges, hopping a fraction of a step from frame to frame, occasionally splitting into bright knots. That blob is atmospheric seeing, and its width is the single number that decides how much detail a ground-based optical telescope can resolve.

The light leaving a star 10, 100 or 10,000 light-years away arrives at the top of our atmosphere as a beautifully flat wavefront — a plane of light with the crests all lined up, because at that distance the star is, geometrically, a point source. The problem is the last 100 kilometers. Our atmosphere is not a uniform sheet of glass. It is a churning fluid of cells at slightly different temperatures, and temperature changes the air's refractive index. A parcel of air 1 °C warmer than its neighbor bends light a tiny bit differently. Stack thousands of these moving cells along the line of sight and the once-flat wavefront arrives at the telescope wrinkled.

A telescope is a machine for adding up a wavefront in phase at its focus. Hand it a flat wavefront and it builds a sharp diffraction-limited point. Hand it a wrinkled one and the crests no longer add coherently — they interfere into a shifting interference pattern, smeared over an angle far wider than the telescope's diffraction limit. Astronomers quantify this as the full width at half maximum (FWHM) of the long-exposure stellar image, measured in arcseconds (1 arcsec = 1/3600 of a degree, about the angular size of a coin seen from 4 km away).

Seeing in numbers: the Fried parameter

The cleanest way to describe turbulence's optical effect is the Fried parameter, written r₀ (pronounced "r-naught"). It is the diameter of a patch of incoming wavefront that is still coherent — distorted by less than about one radian RMS of phase. In good visible-light conditions r₀ is roughly 10–20 cm; on a poor night it shrinks below 5 cm.

Two consequences follow directly. First, the seeing-limited blur angle is approximately

θ ≈ λ / r₀  (radians)

so for green light (λ ≈ 0.5 µm) and r₀ = 10 cm, θ ≈ 5 × 10⁻⁶ rad ≈ 1 arcsec. Second — and this is the brutal part — a telescope larger than r₀ collects more light but gains no extra resolution from a single long exposure. An 8-meter mirror under 10 cm seeing resolves no better than a backyard 10 cm telescope would in space. The atmosphere caps you at the diffraction limit of an r₀-sized aperture.

Because r₀ grows as λ^6/5, seeing is far gentler in the infrared than the visible. That single fact is why most adaptive-optics systems and the sharpest ground-based science operate in the near-IR: r₀ at 2.2 µm can be a meter or more.

Seeing conditions and what they imply (visible light, λ ≈ 0.5 µm)

Condition	Seeing FWHM	Fried r₀	Where / when
Exceptional	0.4″	~25 cm	Best nights, Mauna Kea / Paranal
Excellent	0.6″	~17 cm	Median, top mountaintop sites
Good	1.0″	~10 cm	Typical dark-sky observatory
Mediocre	2–3″	~3–5 cm	Low altitude, humid air
Poor	4″+	<3 cm	Urban rooftop, unstable air

Twinkling, scintillation, and why planets hold steady

Seeing has two visible faces. Image motion and blur (the wandering, swelling blob) come from phase distortions that shift where the light lands. Scintillation — the rapid brightness flicker we call twinkling — comes from those same wrinkles acting like weak lenses high up, focusing and defocusing the beam so the intensity at the ground rises and falls.

This explains the oldest naked-eye observation in the book: stars twinkle, planets generally don't. A star is, to the eye, a true point — its angular diameter (microarcseconds) is thousands of times smaller than a turbulence cell, so the entire image brightens, dims and jumps as one. A planet is an extended disk, several to tens of arcseconds across. It is effectively a grid of many independent points, each twinkling on its own random schedule. Their flickers are out of phase and average out, so the planet glows with a steady light. The same averaging is why twinkling is worst near the horizon, where you look through far more air (high airmass).

The atmosphere reshuffles in milliseconds

Turbulence is not static. High-altitude winds — often 20–40 m/s in the jet stream — blow the turbulent cells across the telescope aperture. Under the frozen-flow (Taylor) approximation, the wavefront pattern simply translates past you, reshuffling on the coherence time τ₀, typically just 2–20 milliseconds in visible light.

This timescale is everything for fighting seeing. Expose for a second and you average over hundreds of independent distortions, baking the blur permanently into the image. Expose for a few milliseconds and you freeze the instantaneous pattern: the blob resolves into a cluster of sharp speckles, each one essentially a tiny diffraction-limited image displaced by the wavefront tilt. Speckle imaging and "lucky imaging" exploit exactly this, recording thousands of short frames and keeping or recombining the sharpest. Adaptive optics exploits it from the other side, correcting the wavefront faster than τ₀ so it never has time to scramble.

How astronomers beat the seeing

Strategies against atmospheric seeing

Approach	How it works	Result
Site selection	Build on high, isolated peaks with smooth laminar airflow above the boundary layer	Median seeing 0.6–0.8″ instead of several arcsec
Go to space	No atmosphere at all (Hubble, JWST, Gaia)	Pure diffraction limit; Hubble ~0.05″ at 0.5 µm
Adaptive optics	Wavefront sensor + deformable mirror correcting >1000×/s, often with a laser guide star	Near-diffraction-limited in the IR; rivals or beats Hubble on 8–10 m mirrors
Speckle / lucky imaging	Thousands of ms-scale exposures; keep or co-add the sharpest frames	High-resolution snapshots of bright targets
Long-baseline interferometry	Combine separated apertures, fringe-track faster than τ₀	Milliarcsecond resolution despite seeing

This is why the world's flagship observatories cluster on a handful of mountaintops. Mauna Kea (4,205 m) in Hawai‘i, Cerro Paranal (2,635 m, the VLT) and Cerro Pachón in Chile's Atacama, and Roque de los Muchachos in the Canaries all share the same recipe: high altitude to rise above the dense boundary layer, a smooth ocean upwind to deliver laminar (non-turbulent) air, and a dry, stable atmosphere. Even there, much of the residual damage comes from dome and ground-layer seeing — heat shed by the telescope, the dome and the warm ground in the first tens of meters — which careful thermal design and ventilation now actively suppress.

Measuring and monitoring seeing

Sites don't guess. A Differential Image Motion Monitor (DIMM) watches a single bright star through two small sub-apertures and measures how much their images jitter relative to each other; the differential motion cancels telescope shake and yields r₀ directly. A SCIDAR or MASS profiler maps how the turbulence is distributed with altitude — crucial for designing adaptive optics, which corrects best for a specific layer. These instruments turn "seeing" from a vague impression into a logged, forecastable number that schedulers use to send the sharpest programs to the best nights.