The effect of data reduction pipelines on WASP-39b atmospheric retrievals: a Bayesian analysis.

Since the launch of the James Webb Space Telescope (JWST), big forward steps have been taken in the understanding of exoplanet atmospheric properties. Focusing on the first exoplanet observed with the telescope, WASP-39b, the huge spectral range covered by all its instruments with different spectral resolutions has provided the opportunity to detect many different chemical species. The high quality of the data has been useful even for detecting differences in the planet terminators (Espinoza et al., 2024) or for studying the properties of the atmospheric aerosols (Roy-Perez et al., 2025).

However, it is also important to keep in mind that the differences among the different data collected and their treatment can be relevant when inferring the planet properties. For example, Lueber et al. (2024) studied the differences on the retrieved atmospheric properties using the diverse instrument data acquired as reference. Davey et al. (2025), on the other hand, studied the effect of binning the spectroscopic observations.

In line with this reasoning we study the effect of using different reduction pipelines and calibrations for the same dataset when performing atmospheric retrievals. So far, different pipelines have been used to reduce the WASP-39b observations. For example, Rustamkulov et al. (2023) used four different pipelines to reduce the observed data, all in broad agreement. Due to the presence of a saturated region, the data have been reanalyzed to minimize the issue (Carter et al., 2024).

Figure 1: A comparison of six different reductions of the data from the WASP-39b transit observed during the ERS Program 1366.

 

Following the atmospheric description from Roy-Perez et al. (2025), we explored the effect of using different calibrations in the retrieval process. In order to compare results and quantify the relevance of the reduction process, we extended the analysis including different cloud extinction parameterizations. Three different cloud extinction models were considered: flat extinction, the Angstrom model, and the Mie parametrization using MOPSMAP.

As in our previous work, we obtain a strong evidence to include cloud extinction models other than a flat contribution for all data reduction. However, the retrieved atmospheric parameters in Figure 2 show that the results are even more dependent on data reduction than in the cloud parametrization.

Figure 2: Atmospheric parameter values retrieved when using the different presented data reductions and cloud parametrizations. As reference, some reference values from Faedi et al. (2011) are also plotted.

 

We find notable differences in the physical parameters of the planet, such as size or atmosphere extension, as well as in the stratospheric temperature. The chemical composition may also be affected, leading to significant differences in some cases. These results show the relevance of the data reduction and assess the importance of discussing atmospheric retrievals in the context of the assumed calibration procedure.

 

References

Carter, A. L., May, E. M., Espinoza, N., et al. 2024, Nature Astronomy, 8, 1008

Davey, J. J., Yip, K. H., Al-Refaie, A. F., & Waldmann, I. P. 2025, MNRAS, 536, 2618

Espinoza, N., Steinrueck, M. E., Kirk, J., et al. 2024, Nature, 632, 1017

Faedi, F., Barros, S. C. C., Anderson, D. R., et al. 2011, A&A, 531, A40

Lueber, A., Novais, A., Fisher, C., & Heng, K. 2024, A&A, 687, A110

Roy-Perez, J., Pérez-Hoyos, S., Barrado-Izagirre, N., & Chen-Chen, H. 2025, A&A, 694, A249

Rustamkulov, Z., Sing, D. K., Mukherjee, S., et al. 2023, Nature, 614, 659