Distribution of training data for the RF classifier targeting interesting planets. Each row represents a different SNR level (with 0 indicating a noise-free spectrum), while each column is labelled with a three-digit binary code indicating the presence (1) or absence (0) of CH|$_4$|⁠, O|$_3$|⁠, and H|$_2$|O, respectively (e.g. ‘010’ means no CH|$_4$|⁠, presence of O|$_3$|⁠, and no H|$_2$|O). The values in each cell (in thousands) denote the total number of noisy spectra generated for that specific atmospheric configuration.