Sky coverage, depth and repeat timing

Where, how faint and when are separate questions

A survey’s coverage, or footprint, is the region it samples under stated conditions. Repeated visits to one field increase the number of observations without necessarily increasing unique sky area. Overlaps and masks must be described when deciding which region was actually usable for a measurement.

Depth describes a conditional measurement or detection capability for faint sources. It needs a band, source type, extraction method and decision or completeness convention. A point-source limit under one image response does not automatically describe an extended low-surface-brightness object, a crowded field or another wavelength band.

Cadence is the timing of visits and exposures, including their durations and gaps. A single average interval can conceal long interruptions. The timing needed for a changing source differs from the exposure needed to measure a constant faint source.

The preceding six-hour allocation model showed that more repeats of fewer disjoint fields reduce unique area at a fixed charged budget. It did not specify the observation dates. A survey description needs all three questions, rather than one number advertised as its overall reach.

Exposure gains depend on the noise regime

For a steady faint source in a simplified background-dominated measurement, expected source counts grow in proportion to flux F and exposure time t. If independent background variance grows in proportion to t, the signal-to-noise ratio scales as F√t when the instrument and background rate are fixed.

At the same chosen model signal-to-noise ratio, four times the exposure therefore permits half the flux in that limiting case. This is a scaling relation under assumptions, not a universal catalogue detection rule.

A different limit gives a different gain. If only source Poisson noise matters, the ratio is the square root of the expected source counts and scales as √(Ft). Four times the exposure then permits one quarter of the flux at the same chosen ratio. Read noise, confusion, background estimation and systematic errors can change these limits.

The Subaru Hyper Suprime-Cam survey documentation illustrates why different observing layers and band-specific depth conventions are useful. Its older design values are not needed to learn the principle: report the measurement convention and conditions alongside a claimed depth.

A total exposure can miss a short event

Consider an original noiseless timing model with exposures covering minutes 0–2, 10–12 and 20–22 after a declared start. These are three two-minute integrations separated by gaps. A brief event lasting from minute 14 to minute 16 overlaps none of them.

Now place an event from minute 10.5 to minute 11.5. Let the signal rate be 1 arbitrary unit outside the event and 5 units during it. The 10–12 exposure contains one minute at each rate, so its mean rate is (1 × 1 + 5 × 1) / 2 = 3 units.

The integration dilutes the peak rate in this example. Reporting only the six minutes of total exposure would hide both the missed event and that dilution. The event profiles and rates here are invented; they are not a light curve measured from a real object.

A detection probability would additionally need distributions of event times, durations and amplitudes, a noise and selection model, and the actual observing windows. These three intervals alone justify a statement about overlap, not a guarantee of detection.

Name the physical mechanism before judging mitigation

Artificial satellites can reflect sunlight into an optical or infrared observing system and move across its field during an exposure. The resulting trail depends on illumination, apparent motion, observing geometry, exposure and detector response. It is not a property determined by satellite count alone.

Avoiding a predicted passage, interrupting an exposure or choosing another time can conflict with the astronomical programme. Accurate predictions and suitable observing opportunities matter. Masking a trail can remove contaminated pixels from an analysis, but it does not reconstruct astronomical information hidden or corrupted there. Detector effects and residual contamination require their own assessment.

Radio-frequency interference involves unwanted radio signals entering a measurement system. Transmitted signals, including satellite transmissions, are physically distinct from reflected-sunlight image trails. Strong interference can also affect receiver behaviour and calibration, beyond adding a removable pattern to an otherwise unchanged measurement.

NRAO’s observing guidance describes flagging affected times and frequency channels, with finer sampling sometimes helping identification at a cost in data and processing. Excluding contaminated measurements is different from reliably recovering the astronomical signal within them. Neither flagging nor subtraction can be assumed to restore all desired information.

Report retained information and the science that could not be measured

The International Astronomical Union’s SatCons101 announcement provides an entry to coordinated optical, infrared and radio interference work. Technical optical examples from Hainaut and Williams concern a dated model, not a present-day measured loss rate. Their useful lesson here is the dependence on geometry, instruments and programme requirements.

For a deliberately restricted radio-style counting example, suppose 100 equal-variance independent time-frequency cells measure the same constant signal. Flag 20 cells and average the remaining 80. The retained fraction is 0.8; the mean’s noise standard deviation increases by √(100/80), about 1.118, relative to keeping all 100.

The corresponding fixed-signal ratio falls by √0.8, about 0.894. These values require identical information and variance in every cell, no remaining interference and no calibration change. A narrow spectral feature or brief transient can instead lose essential information in the excluded cells even when most measurements remain.

A useful assessment therefore reports the band and mechanism, affected measurements, exclusion or correction method, residual effects and missed opportunities. A percentage of masked pixels or flagged cells is not automatically the same percentage loss of flux, sensitivity or scientific value.

Worked examples

Try the reasoning

1. Does a limiting magnitude without a band or source convention fully describe survey depth?

   Show the answer and reasoning

   No.

   Depth depends on response, extraction, noise and a stated detection or completeness convention, among other conditions.

2. Can two programmes with identical total exposure sample a transient differently?

   Show the answer and reasoning

   Yes.

   Their individual exposure windows and gaps can differ.

3. Why should reflected optical satellite trails and radio interference be explained separately?

   Show the answer and reasoning

   They enter the measurements through different physical mechanisms.

   Optical reflected sunlight and unwanted radio signals have different propagation, detector or receiver effects and mitigation limits.

4. Does removing a contaminated region from analysis restore the astronomical signal that would have been measured there?

   Show the answer and reasoning

   No.

   Exclusion can improve the reliability of retained measurements while leaving missing information and residual uncertainties.

What to carry forward

• Coverage, conditional depth and cadence describe different aspects of a survey.
• Exposure gains and time sampling require explicit signal and noise assumptions.
• Interference mitigation must be judged by reliable retained information and remaining losses in the specified measurement.