JWST Noise Simulator Overview

• JWST noise simulator is a forward-modeling toolkit predicting signal-to-noise ratios by integrating detailed photon statistics, detector behavior, and systematics.
• It uses modular data flows and underlying models like the Pandeia engine to emulate spectroscopic and imaging time-series for exoplanet transit and eclipse observations.
• Key functions include simulating various noise sources—photon, background, dark, and read noise—and processing detector readout modes to support observation planning.

A James Webb Space Telescope (JWST) noise simulator is a forward-modeling computational toolkit that predicts the signal, noise, and achievable signal-to-noise ratio (S/N) in JWST science observations. It provides a critical theoretical interface between the physical scene (targets and astrophysical context) and JWST’s hardware and operations, integrating detailed models of photon statistics, detector behavior, optical throughput, and instrumental systematics. Such simulators support the design, planning, and interpretation of JWST spectroscopic and imaging time-series, particularly for exoplanet transit and eclipse observations. Multiple simulators have been developed, with notable frameworks including Gen TSO, PandExo, JexoSim, and MIRISim. Their development has leveraged the underlying STScI-maintained Pandeia engine, calibration data products, and empirical characterization of JWST’s detectors and optics.

1. Fundamental Architecture and Data Flows

All leading JWST noise simulators for time-series science are structured around modular data and instrument abstractions. Key components include:

• Pandeia Engine: The foundation for most simulators, Pandeia is a pixel-based, simulation-hybrid exposure time calculator (ETC) that implements realistic point spread functions (PSFs), MULTIACCUM ramp sampling and its correlated noise, throughput calculations, and extraction strategies. Pandeia treats each detector as a 2D grid, annotating pixel-level signal, noise, and covariance, and incorporates external instrument reference files for throughputs, detector gain, dark currents, and noise (Pontoppidan et al., 2017).
• Catalog Integration: Simulators such as Gen TSO aggregate information from the NASA Exoplanet Archive (target properties), TrExoLiSTS (JWST program metadata), and Gaia DR3 (neighboring field stars for target acquisition) (Cubillos, 7 Oct 2024).
• Scene Construction and SED Modeling: The signal path begins with a high-resolution spectral energy distribution (stellar SED from PHOENIX/Kurucz models or user-supplied spectra), normalized to the observed magnitude. The SED is then propagated through the instrument’s throughput and PSF chain.
• Detectors and Readout Modes: Simulators instantiate detector models corresponding to Teledyne HxRG family arrays (NIRSpec, NIRCam, NIRISS) and MIRI’s Si:As detectors, accounting for subarray geometries, sampling schemes (e.g., MULTIACCUM), and readout modes including improvements such as IRS2 (Birkmann et al., 2022).
• Workflow Interfaces: User interaction with the simulation may be via GUIs (Shiny for Gen TSO), Python APIs (Gen TSO, PandExo, JexoSim), or command-line tools (Cubillos, 7 Oct 2024, Batalha et al., 2017, Sarkar et al., 2021).

2. Mathematical Formulation of Noise Components

Noise simulation is mathematically formulated as a combination of several independent (as well as correlated) sources:

• Photon (Shot) Noise: For an expected count $N$ in electrons, the variance is $\sigma_{\rm photon}^2 = N$, implemented as a Poisson process (Pontoppidan et al., 2017, Klaassen et al., 2020, Sarkar et al., 2021).
• Background Noise: This incorporates shot noise from zodiacal light, telescope/instrument thermal emission, and (if included) sky backgrounds. For a background $B$ in electrons, $\sigma_{\rm bkg}^2 = B$ (Batalha et al., 2017, Klaassen et al., 2020).
• Dark Current Noise: Each pixel’s dark count rate $D$ leads to $\sigma_{\rm dark}^2 = D\, t_{\rm eff}$, with $t_{\rm eff}$ the integration time (Birkmann et al., 2022, Klaassen et al., 2020). For mid-IR (MIRI), pixel-dependent dark maps are used (Klaassen et al., 2020).
• Read Noise: Readout noise is characterized per frame (CDS noise) and transformed according to the slope-fitting or ramp-sampling method. For $N_g$ groups and single-frame read noise $\sigma_{\rm frame}$, the variance after slope fitting is $$\sigma_{\rm read}^2 = \sigma_{\rm frame}^2 \cdot 12/(N_g(N_g+1))$$ (Birkmann et al., 2022, Pontoppidan et al., 2017, Sarkar et al., 2021).
• Correlated Noise: Long-timescale "1/f" noise, alternating column noise (ACN), pink noise, and picture-frame artifacts are modeled, often using Fourier-based filtering and empirical PCA patterns (Rauscher, 2015).
• Cosmic Rays: Simulators may inject cosmic ray events according to an empirically or CDP-derived rate, with event shapes selected from pre-compiled libraries (especially for MIRI) (Klaassen et al., 2020). Post-integration data reduction pipelines may further mitigate these events.
• Systematic Noise Floor: To emulate residual systematics not captured by fundamental physical processes, a user-defined (or empirically justified) noise floor can be imposed (e.g., 14 ppm in transmission units for PandExo applications) (Arora et al., 26 Nov 2025, Batalha et al., 2017).

Noise sources are typically combined in quadrature at the level of each spectral or spatial bin, with cross-pixel or intra-pixel correlations treated using covariance matrices based on the instrument’s measured behavior (Pontoppidan et al., 2017, Guilloteau et al., 2020).

3. Simulation Workflow and Signal Extraction

The simulation of a JWST time series typically proceeds through the following steps:

1. Configuration: The user selects the instrument, mode, subarray, and supplies observational parameters (target, magnitude, observing time, transit/eclipse details, etc.).
2. Scene Realization: The spectral scene is propagated through the instrument throughput, PSF, and detector model, forming a pixel-level, time-dependent illumination map (Pontoppidan et al., 2017, Guilloteau et al., 2020).
3. Noise Injection: All relevant noise terms are computed and added at each time step, creating frames that are then grouped according to the MULTIACCUM or other readout scheme.
4. Ramp Processing: Slope fitting or last-minus-first (LMF) processing derives count rates from the raw frames. For time-series, different simulators may propagate the transit/eclipse signal function directly through the scene (Cubillos, 7 Oct 2024, Sarkar et al., 2021). JexoSim and ExoSim support time-domain simulations that preserve correlated noise and systematics (Sarkar et al., 2021, 2002.03739).
5. Spectral Extraction and S/N Calculation: Spectra are extracted via aperture summing or optimal extraction. The statistical uncertainty on the transit/eclipse depth $d(\lambda)$ and thus S/N are propagated using derivatives with respect to in- and out-of-transit fluxes; for example, for $d = 1 - F_{\rm in}/F_{\rm out}$, the uncertainty is

$$\sigma_d^2 = \left( \frac{1}{t_{\rm in} F_{\rm out}} \right)^2 \sigma^2_{\rm in} + \left( \frac{F_{\rm in}}{t_{\rm out} F_{\rm out}^2} \right)^2 \sigma^2_{\rm out}$$

(Cubillos, 7 Oct 2024, Batalha et al., 2017).

1. Output Products: Simulators typically yield 1D S/N spectra, depth spectra with uncertainties, and FITS-format synthetic detector cubes for pipeline testing (Cubillos, 7 Oct 2024, Sarkar et al., 2021, Sarkar et al., 2019).

4. Supported Instruments, Modes, and User Interfaces

Noise simulators vary in their scope of JWST instrument and mode support:

• Gen TSO: Cycle 4 spectroscopic time-series for NIRISS/SOSS, NIRSpec/BOTS, NIRCam SW/LW grism, MIRI LRS/MRS; includes seamless target acquisition simulation leveraging Gaia and ESASky (Cubillos, 7 Oct 2024).
• PandExo: All four JWST instruments for time-series spectroscopy, with interface tightly coupled to Pandeia and supporting batch noise simulations and noise floor imposition (Batalha et al., 2017).
• JexoSim 2.0: Full time-domain frame-by-frame simulation including all time-series supported modes for JWST (NIRISS, NIRSpec, NIRCam, MIRI), with customizable systematic and astrophysical noise injection, and robust pipeline for spectral extraction and error propagation (Sarkar et al., 2021).
• MIRISim: Dedicated to MIRI imaging and spectroscopy, faithfully reproducing detector physics, cosmic rays, and correlated effects using calibration data parameterization (Klaassen et al., 2020).
• Teledyne HxRG Noise Generator (NG): Read-noise and related correlated noise benchmarking for H1RG/H2RG/H4RG-based systems, relevant for simulating instrument detector performance at the pixel level (Rauscher, 2015).

Different simulators offer their functionality via graphical interfaces, Jupyter notebook-optimized APIs, or standalone command-line scripting. For example, Gen TSO’s GUI enables interactive parameter selection with real-time S/N feedback, while its API supports notebook-based survey planning (Cubillos, 7 Oct 2024).

5. Physical and Operational Assumptions

Simulation fidelity depends upon input calibration data, instrument reference files, and assumptions inherited from the underlying noise and throughput models:

• Simulators such as Gen TSO and PandExo bind their noise calculations to versions of the STScI Pandeia library, which is routinely updated to reflect calibration refinements and in-flight commissioning (Cubillos, 7 Oct 2024, Batalha et al., 2017).
• Stellar and planetary input parameters are drawn from static local snapshots of exoplanet archives (updated monthly), which may lag behind the latest discoveries unless manually refreshed (Cubillos, 7 Oct 2024).
• Some simulators do not yet incorporate imaging time-series or certain readout submodes—for instance, early releases of Gen TSO exclude photometric modes and alternative NIRISS extractions (Cubillos, 7 Oct 2024).
• Target acquisition S/N for fainter or off-axis stars is computed using Gaia G-band approximations; empirical confirmation of these predictions is in development (Cubillos, 7 Oct 2024).
• High-fidelity atmospheric spectra must be user-supplied; default models are limited to blackbody or flat depth (Cubillos, 7 Oct 2024).
• Simulators such as JexoSim 2.0 propagate non-Gaussian and time-correlated errors, as well as Fano (photon transfer) noise, into the final S/N budget (Sarkar et al., 2021).

A summary of these assumptions and soft limitations is maintained in documentation for each package.

6. Validation, Limitations, and Future Enhancements

• Benchmarking: Simulators such as PandExo and Gen TSO are validated against independent instrument team simulators, direct comparisons with the JWST ETC, and, where possible, real HST spectra (for cross-checks). Agreement levels are typically within 10% in the photon-limited regime, and error budgets are decomposable into their constituent sources (Batalha et al., 2017, Cubillos, 7 Oct 2024).
• Limitations: Known limitations include incomplete modeling of astrophysical systematics (star spots, unmodeled stellar variability), incomplete handling of correlated electronic noise, and lack of photometric or ground-based support in some toolchains (Batalha et al., 2017, Cubillos, 7 Oct 2024, Sarkar et al., 2021).
• Planned Extensions: Roadmaps for Gen TSO and others include support for photometric imaging time-series, user-supplied arbitrary targets, real-time database refresh, enhanced subarray/mode inclusion, and cloud-based sharing of simulation presets and results (Cubillos, 7 Oct 2024).
• Advances in Time-Domain Modeling: JexoSim 2.0 and ExoSim enable explicit injection of time-correlated noise, astrophysical and instrumental systematics, and Monte Carlo or Allan deviation-based precision estimation—capabilities required for assessing the impact of red noise on retrievals (Sarkar et al., 2019, Sarkar et al., 2021, 2002.03739).

7. Application to Exoplanet Science and Observation Planning

JWST noise simulators are fundamentally enabling tools for exoplanet science with space-based time-series spectroscopy. Their anticipated applications include:

• Design and Scheduling: Rapid assessment of whether a given program will achieve depth precision goals (e.g., detecting spectral features with amplitudes $<100$ ppm at $5\sigma$ significance) (Cubillos, 7 Oct 2024).
• Instrument/Mode Trade Studies: Comparative S/N predictions for different instrument modes, grisms, and subarrays, guiding selection of observational strategy (Sarkar et al., 2021).
• Spectral Retrieval Forecasting: End-to-end data cubes with injected noise and systematics for the systematic evaluation of spectral feature detectability, error budgets, and retrieval robustness (Arora et al., 26 Nov 2025, Sarkar et al., 2021).
• Pipeline and Algorithm Testing: Synthetic detector-level data enable the validation of reduction and extraction pipelines under realistic noise conditions (Klaassen et al., 2020, Pontoppidan et al., 2017).
• Intercomparison of Theoretical Models: By simulating the S/N achieved for the absolute difference between 1D and 3D atmospheric models, JWST simulators allow the quantification of the number of transits and sensitivity needed to empirically distinguish physical scenarios (Arora et al., 26 Nov 2025).

This integration of advanced noise simulation, high-resolution scene specification, and throughput modeling undergirds the operational planning and scientific exploitation of JWST’s exoplanet time-series capabilities.

---

References

• "Gen TSO: A General JWST Simulator for Exoplanet Times-series Observations" (Cubillos, 7 Oct 2024)
• "PandExo: A Community Tool for Transiting Exoplanet Science with JWST & HST" (Batalha et al., 2017)
• "Pandeia: A Multi-mission Exposure Time Calculator for JWST and WFIRST" (Pontoppidan et al., 2017)
• "JexoSim 2.0: End-to-End JWST Simulator for Exoplanet Spectroscopy--Implementation and Case Studies" (Sarkar et al., 2021)
• "Teledyne H1RG, H2RG, and H4RG Noise Generator" (Rauscher, 2015)
• "MIRISim: A Simulator for the Mid-Infrared Instrument on JWST" (Klaassen et al., 2020)
• "The In-Flight Noise Performance of the JWST/NIRSpec Detector System" (Birkmann et al., 2022)
• "Quantifying the differences in transmission and emission spectra for hot irradiated gaseous exoplanet Atmospheres: A comparison of 1D and 3D modeling using JWST" (Arora et al., 26 Nov 2025)
• "Simulated JWST datasets for multispectral and hyperspectral image fusion" (Guilloteau et al., 2020)
• "ExoSim: the Exoplanet Observation Simulator" (2002.03739)
• "JexoSim: A time domain simulator of exoplanet transit spectroscopy with JWST" (Sarkar et al., 2019)