Comparison of the CIGALE-derived line-to-dust luminosities to the observed ones. Left: comparison of observations and models for the spectroscopic sample of SPT galaxies. Most data are from [CII]157.6 μm (Lagache et al. 2018) and [NII]205.2 μm (Cunningham et al. 2020), except for SPT 0418-47 (De Breuck et al. 2019), for which we have several lines. Generally speaking, the brightest [CII]157.6 μm and [OIII]88.3 μm lines are better modelled than the [NII]122. μm and [NII]205.2 μm. The [OI]145. μm is an outlier in this frame. It could be because the former lines are stronger than the latter, or because of physical differences between them. Right: Symbols are colour-coded by intrinsic dust luminosity. The fits suggest that the objects with log₁₀ L_dust < 12.7 and log₁₀ L_dust > 13.5 are found closer to the one-to-one line, or vice versa, meaning that the line predictions for those with 12.7 < log₁₀ L_dust < 13.5 are worse. The objects with a large green circle correspond to the sole object with different lines: SPT0418-47 from De Breuck et al. (2019) (see left panel to identify the lines).