Top: raw NIRISS/SOSS order 1 (0.85–2.83 μm) white-light curve with 1σ error bars (gray), representing the standard deviation on the out-of-eclipse data points and its temporally binned points (black). The best-fit model, including the eclipse and the systematics, is overlapped in red. Second: systematics removed from the white-light curve. The linear systematics model from Equation (2) is shown in turquoise, and the remaining correlated noise modeled with the GP, along with its 1σ confidence interval, is shown in purple. Third: detrended order 1 white-light curve (gray, binned points in black) once the systematics of the second panel are removed from the data of the top panel along with the eclipse model (red). The error bars represent the best-fit scatter parameter (σ = 102.9 ppm). As more easily seen on the binned points, the error bars seem overestimated, as expected from our Allan variance plot from Figure B4, which is explained by potential remaining correlated noise at the integration level. Bottom: residuals to the best-fit model and their rms value (based on the gray unbinned points).