Alexander Cabello, EO Solutions; Claire Acke, EO Solutions; Enrique De Alba, EO Solutions; Jeff Houchard, EO Solutions; Kimmy De Alba, SSC/SYD 85/BI; Zach Gazak, SSC/SYD 85/BI A&AS; Virginia Wright, AFRL
https://doi.org/10.64861/OZOP2845
Keywords: Photometry, Atmospheric Clouds, Simulation, Photometric Calibration, Space Domain Awareness, Electro-Optical, Atmospheric Transmission
Abstract:
Ground-based electro-optical (EO) sensors for Space Domain Awareness (SDA) routinely operate through thin and partial cloud cover, which attenuates target and calibration star signals unevenly across the field of view and degrades photometric accuracy in ways that are difficult to quantify with real data, where the true line-of-sight transmission is unknown. This work introduces a physically parameterized synthetic cloud model in SatSim, an open-source GPUaccelerated space scene simulator, that renders spatially correlated cloud transmission fields together with per-pixel and per-object transmission truth. Using this capability, we generate a large annotated dataset spanning seven sensor models and five cloud morphology presets organized into seven cloudy regimes, together with randomized observing conditions, including lunar geometry and urban sky glow. Every cloudy scene is paired with a clear-sky twin, giving each frame a matched noisy clear-sky reference measurement and enabling end-to-end verification that the exported truth matches the attenuation applied by the renderer. We evaluate forced aperture photometry of synthetic GEO targets and compare the traditional whole-frame zeropoint calibration against spatially local zeropoint models, including cross-validated Chebyshev polynomial surfaces, local grid interpolation, and thin-plate spline fits. On the truth subset, we additionally compare local calibration with a transmission-only reference supplied with perfect cloud knowledge and a source-specific oracle. We derive a truth-assisted per-target reliability estimate from signal-to-noise, zeropoint uncertainty, and local transmission dispersion. On the held-out test split, the cross-validated local polynomial beats the whole-frame baseline in all three transmission bins below 0.8 (reducing untrimmed RMSE from 0.26 to 0.17 mag at aperture transmissions of 0.5 to 0.8 and from 0.90 to 0.62 mag at 0.2 to 0.5) and removes 42 to 56 percent of the cloudinduced excess squared error measured against each scene’s clear-sky twin; above transmission 0.8 the global method is slightly better, so local calibration provides a benefit in the sampled partial-to-deep-cloud bins below transmission 0.8, rather than a uniform improvement. A reliability screen selected on validation data flags 91 percent of gross errors at a 26 percent false-positive rate on the test split; among test targets passing the accuracy filters, the 5-sigma-clipped residual core has 0.07–0.11 mag scatter below transmission 0.5, while untrimmed errors remain heavy-tailed. These results indicate the potential for local zeropoint calibration to extend the usable observing envelope of SDA sensors into partially cloudy conditions, subject to real-cloud testing and validation of a reliability screen based entirely on observable quantities. All results are simulation upper bounds under known astrometry: photometry is forced at truth positions, so detection, astrometric, and association errors are out of scope.
Date of Conference: September 15-18, 2026
Track: Atmospherics/Space Weather