← All defects

Light pollution gradient

Slow large-scale brightness gradient oriented toward the ground or a nearby town.

Description

The light pollution gradient is a large-scale brightness gradient that sweeps across the image, brighter on one side (toward the ground or the urban area) and darker on the other.

It comes from the luminous halo of cities and from light scattered by the atmosphere: this parasitic background is not uniform across the field -- it fades as the angular distance from the source increases.

Unlike a hardware defect, it is smooth, directional, and stable from one frame to the next. It is one of the most common defects in peri-urban environments and, fortunately, one of the easiest to correct during processing with a gradient-removal tool (DBE, ABE, GraXpert).

It is distinguished from a moon gradient (oriented toward the Moon), from residual vignetting (radial and symmetric darkening), and from the variable haze of cirrus clouds.

Visual signature

The sky background shows a soft, continuous gradient: one corner or edge is noticeably brighter, and the brightness falls off regularly toward the opposite side.

The orientation points toward the pollution source (town, streetlight, industrial zone) and remains consistent across all frames of a session.

The defect is low-frequency: it affects overall brightness without any fine texture or pattern. It becomes most apparent after stretching, when the shadows are pulled up and the background is no longer uniform.

It crushes the contrast of extended nebulosity and skews the color balance, with the background often picking up an orange or greenish cast on the most polluted side.

Differential diagnosis

Not to be confused with a moon gradient: the mechanism is the same (scattered light) but the orientation points toward the Moon and changes with its position; a moonless night settles the question.

Distinct from residual vignetting: vignetting darkens the corners radially and symmetrically around the optical center, whereas a pollution gradient runs in a single direction.

Separate from cirrus haze: cirrus produce a non-uniform background that varies from one frame to the next, whereas a pollution gradient is smooth and stable throughout the session.

Not to be mistaken for dew fogging: dew builds up progressively and typically starts from one corner of the optics, whereas the pollution gradient is present and constant from the very first frame.

Probable causes

  • Luminous halo from cities and public lighting scattered by the atmosphere
  • Observation site in a peri-urban or urban area (high Bortle class)
  • Target low on the horizon in the direction of heavy light pollution
  • No appropriate light-pollution filter in use
  • Local stray light source nearby (streetlamp, lit window, neighbor)
  • Mist or humidity amplifying the scattering of ground-level light

Course of action

  1. Remove the gradient during processing (DBE, ABE, GraXpert), before stretching
  2. Place sample points only on the sky background, never on any nebulosity
  3. Use an appropriate light-pollution filter (broadband or narrowband)
  4. Favor targets high in the sky, away from the direction of the city
  5. Move to a darker site (lower Bortle class) when possible
  6. Eliminate local stray light sources around the setup
  7. Perform color calibration (SPCC, PCC) after gradient removal

The Doc's advice

The light pollution gradient is the daily reality of astrophotography in the city, and the good news is that it is very manageable. At acquisition, an appropriate filter (broadband such as L-Pro, or narrowband for emission targets) reduces the halo that needs correcting. During processing, tackle the gradient BEFORE stretching and before color calibration: DBE or GraXpert places a few sample points on the background and the gradient disappears. Golden rule: never place a sample point on real nebulosity, or you will dig a hole in it. And if you can, orient your framing to move the target away from the direction of the city.

- the Doc, astrophotography defect specialist

Think you can see this defect in your image?

Run a diagnosis

Frequently asked questions

How do I remove a light pollution gradient?

With a gradient-removal tool applied to the linear image, before stretching: DynamicBackgroundExtraction (DBE) or AutomaticBackgroundExtraction (ABE) in PixInsight, or GraXpert as a free alternative. The principle: place sample points exclusively on the sky background, never on a nebula or galaxy, so the software can model and subtract the parasitic gradient. Do this before color calibration (SPCC, PCC), which assumes the background is already neutral. On a strong gradient, two gentle passes are better than one aggressive one.

Does a light-pollution filter eliminate the gradient?

It reduces it; it does not eliminate it. A broadband filter (L-Pro, L-eNhance, etc.) cuts certain emission lines from urban lighting and reduces the amount of halo recorded, which makes processing easier. A narrowband filter goes further for emission targets by passing only narrow bands (Ha, OIII). But no filter makes the background perfectly uniform: a slight residual gradient almost always remains and is removed during processing (DBE, GraXpert). The filter complements the acquisition; it is not a substitute for gradient removal.

Light pollution gradient or vignetting: how to tell them apart?

By geometry. A pollution gradient runs in a single direction (brighter toward the city or ground, darker on the opposite side). Vignetting is radial and symmetric: all four corners darken in the same way around the optical center. Another clue: vignetting is corrected with flat frames (calibration), while a pollution gradient is removed during processing (DBE, GraXpert) because it comes from the sky, not the optical train. The two can coexist; correct vignetting first with flats, then deal with any residual gradient.

Should the gradient be removed before or after stretching?

Before, on the still-linear image. A gradient-removal tool models a low-frequency background, which is far more reliable while pixel values have not yet been compressed by stretching. Once the image is stretched, the gradient becomes non-linear and the subtraction model is less clean, with a higher risk of artifacts. Recommended order: stack, remove gradient (DBE, GraXpert), color calibration (SPCC), then stretch.