**6. Conclusions**

The zero-offset *ZP* and *ZS* volumes were obtained by an isotropic simultaneous inversion and then by two anisotropic methods: statistical modeling and MLFN. The validation of the P-impedance volumes shows that the MLFN model had near-zero residuals and a high R-value (0.94) compared to simultaneous inversion and statistical modeling, which had low R-squared values (0.84 and 0.88, respectively). Similarly, the MLFN resulted in a high correlation for the S-impedance (0.92), while simultaneous inversion and statistical modeling had lesser R-values (0.73 and 0.85, respectively).

The *ZP* and *ZS* volumes, obtained from the MLFN models, were applied to the impedance-domain Thomsen's anisotropy equations to solve for the anisotropy parameters: Epsilon and Delta. The anisotropy magnitude was observed to be low in the shaly wet sand successions, where the contrast of the elastic properties between layers was low, while it showed higher magnitude in the shaly gas sand successions, which had high-contrast fluctuations in the elastic properties.

The *ZP* and *ZS* volumes, obtained from the isotropic inversion and MLFN, were converted into elastic properties from which the sand and gas probabilities were forecasted by logistic regression. The predicted probabilities were inserted into a decision tree to estimate the distribution of gas sand, wet sand, and shale classes within the target zone. The predicted facies distributions were validated by the facies log at the well (×1), where the MLFN showed a high average success rate (0.77) compared to the simultaneous inversion, which had a low success rate (0.56).

Conventional inversion algorithms have proved to be misleading if seismic anisotropy is neglected. On the other hand, statistical methods and ML are more efficient tools for obtaining more accurate rock properties, especially when considering the dependence of these properties on incident angles.

The decision tree of the facies model shows encouraging results due to the contribution of the depth constraint. This proves that the trend factor is crucial for accurate reservoir characterization. Therefore, the availability of geochemical data would be an added value that links other properties, such as elastic properties and anisotropy parameters, to the pressure regime and compaction profile.

The study considers that the subsurface follows the VTI assumption, which applies to horizontal layers. However, the methodology can be extended to include more realistic cases, such as the TTI and HTI assumptions, which can enhance the accuracy of zero-offset parameter modeling. In addition, core data are needed to validate anisotropy measurement and yield confident models.

**Author Contributions:** Conceptualization, M.F.G.; methodology, M.F.G. and S.Y.M.A.; software, M.F.G.; validation, M.F.G., A.H.A.L. and S.Y.M.A.; formal analysis, M.F.G.; investigation, M.F.G.; resources, A.H.A.L.; data curation, M.F.G. and A.H.A.L.; writing—original draft preparation, M.F.G.; writing—review and editing, M.F.G., A.H.A.L. and S.Y.M.A.; visualization, M.F.G.; supervision, A.H.A.L.; project administration, A.H.A.L.; funding acquisition, A.H.A.L. All authors have read and agreed to the published version of the manuscript

**Funding:** This research was funded by YUTP-PRG OF Grant Cost Center: 015PBC-021.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** The data that support the findings of this study are available from Petronas. Restrictions apply to the availability of these data, which were used under license for this study.

**Acknowledgments:** We would like to thank the Department of Petroleum Geosciences at Universiti Teknologi Petronas. In addition, we appreciate all the efforts and support of our colleagues in the Center of Seismic Imaging. A special thanks to Amir Abbas for his valuable help and support.

**Conflicts of Interest:** The authors declare no conflict of interest.
