Investigating the effect of hadronic models on IACT images

The predictions of hadronic interaction models for cosmic-ray induced air showers contain inherent uncertainties due to limitations of available accelerator data. This leads to differences in shower simulations using each of those models. Many studies have been carried out to track those differences by investigating the shower development or the particle content. In this work, we propose a new approach to search for discrepancies and similarities between the models, via the IACT images resulting from the observations of hadronic air showers. We use simulations of H.E.S.S. as a show-case scenario and, by investigating variables of the camera images, we find potential indicators to highlight differences between models. Number of pixels, Hillas image size, and density showed the largest difference between the models. We then further explore the (in)compatibility of the models by combining all the variables and using Boosted Decision Trees. For protons, a significant difference in the classifier output is found for EPOS-LHC when compared to both QGSJET-II04 and Sybill 2.3d. For helium and nitrogen, QGSJET-II04 is shown to be the outlier case. No significant differences are found for silicon and iron. The distribution of (in)compatibility between the models in the phase space of reconstructed shower parameters shows that a targeted search can be fruitful, with showers with reconstructed energies of a few TeV and reconstructed core closer to the large telescope presenting the largest power of separation. An investigation of the distribution of first interaction parameters has shown that EPOS-LHC and QGSJET-II04 result in significantly different distributions of multiplicity and height of first interaction for protons and elasticity and fraction of energy carried by neutral pions for helium and nitrogen.