Observational Window Functions in Planet Transit Surveys

The probability that an existing planetary transit is detectable in one’s data is sensitively dependent upon the window function of the observations. We quantitatively characterize and provide visualizations of the dependence of this probability as a function of orbital period upon several observing strategy and astrophysical parameters, such as length of observing run, observing cadence, length of night, transit duration and depth, and the minimum number of sampled transits. The ability to detect a transit is directly related to the intrinsic noise of the observations. In our simulations of observational window functions, we explicitly address non-correlated (gaussian or white) noise and correlated (red) noise and discuss how these two noise components affect transit detectability in fundamentally different manners, especially for long periods and/or small transit depths. We furthermore discuss the consequence of competing effects on transit detectability, elaborate on measures of observing strategies, and examine the projected efficiency of different transit survey scenarios with respect to certain regions of parameter space.

💡 Research Summary

The paper presents a comprehensive quantitative study of how the observational window function governs the probability of detecting a planetary transit in photometric time‑series data. The authors begin by defining the window function W(t) as a binary time‑series that is unity when observations are being taken and zero otherwise. Because real surveys consist of many discrete observing blocks (nightly runs, weather gaps, instrument maintenance, etc.), the window function is inherently non‑continuous, and a transit is only sampled if its mid‑time falls within a block where W(t)=1. The detection probability P_det(P_orb) is therefore expressed as the product of two factors: (i) the probability that at least a prescribed minimum number of transits N_min fall inside the observing blocks, and (ii) the probability that each sampled transit yields a signal‑to‑noise ratio (S/N) above a chosen threshold.

Two distinct noise regimes are modeled. White (uncorrelated) noise σ_white is assumed Gaussian and decreases as √N_obs when more data points are gathered. Red (time‑correlated) noise σ_red is introduced to capture systematic trends such as atmospheric variations, instrumental drifts, and calibration errors; it is modeled with a 1/f power spectrum (α≈1) and is treated as an additional variance term that does not average down with cadence. The total effective noise for a given observation is σ_tot = √(σ_white² + σ_red²). The authors show that while white noise merely scales the detection threshold, red noise can dominate for long orbital periods and shallow transits, dramatically lowering P_det even when the nominal white‑noise S/N would be sufficient.

Four observational strategy parameters are explored in depth:

1. Total observing baseline (T_run) – The longer the baseline relative to the orbital period, the more complete the phase coverage. The simulations reveal a saturation effect: once T_run ≳ 3 × P_orb, additional baseline yields diminishing returns because most possible transit phases have already been sampled.

2. Cadence (Δt) – Shorter cadences improve the sampling of ingress and egress, allowing more accurate depth and duration measurements. However, decreasing Δt increases the per‑point photon noise, raising σ_white. The authors identify an optimal cadence of roughly one‑tenth to one‑fifth of the expected transit duration τ_tran.

3. Night length (L_night) – Seasonal and latitude‑dependent night length determines the size of gaps between observing blocks. Short nights produce larger gaps, which are especially detrimental for short‑period planets (1–3 days) because a given transit may fall entirely in daylight on many consecutive cycles, reducing the chance of meeting N_min.

4. Minimum number of observed transits (N_min) – Raising N_min from 2 to 3 dramatically reduces the false‑positive rate but also cuts detection efficiency for long‑period planets, where only a few transits occur during the survey.

Using Monte‑Carlo simulations of 10⁴ synthetic planetary systems, the authors evaluate five representative survey configurations: (a) a single‑site ground‑based optical survey (30 days, Δt = 10 min, L_night ≈ 8 h), (b) a three‑site coordinated network (60 days total, Δt = 5 min), (c) a space‑based continuous mission akin to TESS (27 days, Δt = 2 min), (d) a long‑duration, low‑cadence ground survey (180 days, Δt = 30 min), and (e) a high‑cadence, single‑night intensive run (10 h, Δt = 1 min).

When only white noise is present, all configurations achieve similar detection efficiencies for deep (>1 %) transits. Introducing a realistic red‑noise level (σ_red ≈ 200 ppm) suppresses the efficiency of the ground‑based, low‑cadence scenarios (a) and (d) by 30–50 %, while the space‑based and multi‑site network retain high performance (≤10 % loss). The impact is most severe for long periods (>15 days) and shallow depths (<0.5 %), where red noise can reduce P_det to near zero even for otherwise favorable baselines.

A key conceptual output is the “detectability ridge” – a curve in the (period, depth) plane that delineates where the combined noise and window function yield a target detection probability (e.g., 90 %). The ridge shifts upward (requiring deeper transits) as the σ_red/σ_white ratio increases, providing a practical design tool for survey planners.

The paper concludes with concrete recommendations: (1) prioritize minimizing red noise through multi‑site coordination or space‑based platforms; (2) schedule baselines at least three times longer than the longest period of interest; (3) adopt a cadence that samples ingress/egress without inflating white noise; (4) set N_min = 3 for robust candidate vetting, but supplement long‑period searches with specialized algorithms that can tolerate fewer observed events; and (5) incorporate the detectability ridge into target‑selection and yield‑prediction models.

Overall, the study demonstrates that the observational window function is a high‑dimensional construct whose influence cannot be captured by a single metric such as total observing time. By explicitly modeling both white and red noise and exploring realistic survey strategies, the authors provide a rigorous framework that can be directly applied to optimize existing and future transit surveys, improve yield forecasts, and guide the allocation of limited telescope resources.