LEO Satellite Track Correction for CSST Multi-Band Imaging Data

TL;DR

This study addresses the contamination of CSST multi-band imaging by LEO satellite trails and presents a trail reconstruction–subtraction method to preserve photometric integrity. By integrating realistic CSST multi-band simulations with an empirically fitted trail template, the approach reconstructs and subtracts trails, reducing photometric biases and noise across bands. The results show substantial reductions in both magnitude errors and trail-induced noise, with effectiveness dependent on band and satellite altitude, demonstrating the method's practical value for CSST data processing. Overall, the work provides a viable pathway to mitigate LEO contamination in large-scale space-based surveys and informs future CSST data processing pipelines.

Abstract

Low Earth Orbit satellite (LEOsat) mega-constellations are considered to be an unavoidable source of contamination for survey observations to be carried out by the China Space Station Telescope (CSST) over the next decade. This study reconstructs satellite trail profiles based on simulated parameters, including brightness levels and orbital altitudes, in combination with multi-band simulated images. Compared to our previous work, the simulated images in this study more accurately replicate the realistic observational conditions of CSST and extend beyond single-band analysis. Variations in LEOsat trail brightness, source brightness, background noise, and source density across different bands result in differing levels of accuracy in trail reconstruction and subsequently affect the reliability of photometric measurements. The reconstructed trail profiles are subsequently applied to correct the contaminated regions. Simulation results reveal varying levels of contamination effects across different bands following LEOsat trail correction, including both reconstruction and subtraction. To evaluate the effectiveness of the correction, we quantified the fraction of affected sources using two metrics: (1) magnitude errors greater than 0.01 mag attributable to LEOsats, and (2) LEOsat-induced noise exceeding 10% of other noise contributions. Following trail repair, the analysis reveals a reduction of over 50% in the fraction of affected sources in the NUV band for both 550 km and 1200 km altitudes, assuming a maximum brightness of 7 in the V band. In the i band, the reduction exceeds 30%. The degree of improvement varies across spectral bands, and depends on both satellite altitude and the adopted brightness model.

Figures (5)

• Figure 1: The figure illustrates the trail of a LEOsat across a simulated $i$-band CCD (9216 $\times$ 9232 pixels), which covers an area of 129.4 square arcminutes.
• Figure 2: The figure illustrates the results summarized in Table 2, comparing the outcomes with and without trail repair. The magnitude error threshold is 0.01; blue denotes the results without trail repair, and red represents the trail-repaired case. The left panels show the results for satellites at an altitude of 550 km, while the right panels correspond to an altitude of 1200 km. The upper panels present the results for V$_{\rm M} = 7$, and the lower panels for V$_{\rm M} = 5$.
• Figure 3: The figure illustrates the relationship in the $i$-band between the distance of sources from LEOsat trails, normalized by the trail’s FWHM, and the decimal logarithm (log10) of the relative photometric error induced by the satellites. The left panels show results for satellites at an altitude of 550 km, while the right panels correspond to an altitude of 1200 km. The upper panels present results for V$_{\rm M} = 7$, and the lower panels for V$_{\rm M} = 5$.
• Figure 4: The figure illustrates the relationship in the NUV-band between the distance of sources from LEOsat trails, normalized by the trail’s FWHM, and the decimal logarithm (log10) of the relative photometric error induced by the satellites. The panel layout is the same as in Figure 2. At the trail center, the distribution in the NUV band is narrower and lower than that in the $i$-band.
• Figure 5: The figure illustrates the results summarized in Table 3, comparing outcomes with and without trail repair, where blue denotes results without repair and red represents the trail-repaired case. It shows the proportion of contaminated sources whose LEOsat trail-induced errors after reconstruction exceed one-tenth of other noise contributions. The left panels correspond to satellites at 550 km altitude, while the right panels represent 1200 km altitude. The upper panels present results for V$_{\rm M} = 7$, and the lower panels for V$_{\rm M} = 5$.