the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 18 May 2026
Three-Dimensional Geological Modeling based on Dual-Task Stratigraphy-Aware Attention Networks (Geo-SAN v1.0)
Abstract. The current three-dimensional (3D) geological implicit modelling methods are mainly based on interpolation methods, such as Kriging and radial basis functions (RBFs), which struggle to capture the nonlinear characteristics of complex geological structures and are limited in their capacity to integrate multi-source modeling data. To overcome these limitations, we proposed a 3D geological modelling framework, Geo-SAN, which consists of a dual-task stratigraphy-aware attention network. The framework starts with graph neural networks (GNNs) with a multi-scale neighborhood aggregation mechanism which is aimed to identify critical sampled points adjacent to fault planes and aggregate the lithological features. Subsequently, a stratigraphy-aware attention mechanism is introduced to explicitly incorporate similarities in stratigraphic sequence into the framework. A unidirectional stratigraphic scalar field penalty to lithological classification is developed and incorporated into loss functions, thereby denoising lithological classification. Finally, a dual-task prediction head is designed to simultaneously complete lithological classification and scalar field interpolation. Ablation experiment further validates the contributions of the three core components, that is, graph neighborhood aggregation, stratigraphy-aware attention, and dual-task learning. A case study at the Lingnian-Ningping region of Guangxi Zhuang Autonomous Region (GZAR), China, demonstrates that the proposed Geo-SAN framework, with an accuracy of 92.1% in lithological classification and a coefficient of determination (R²) of 0.96 in predicting the scalar field, outperforms the Hermite RBFs (HRBFs). In summary, the proposed framework is an important innovation of intelligent modelling of intricate geological formations, which is promising in the application of concealed mineral exploration.
- Preprint
(4104 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 14 Jul 2026)
- RC1: 'Comment on egusphere-2026-1210', Anonymous Referee #1, 29 May 2026 reply
-
RC2: 'Comment on egusphere-2026-1210', Anonymous Referee #2, 30 Jun 2026
reply
In the manuscript entitled “Three-Dimensional Geological Modeling based on Dual-Task Stratigraphy-Aware Attention Networks (Geo-SAN v1.0)”, the authors propose a three-dimensional geological modeling framework named Geo-SAN. The framework first employs a multi-scale neighborhood aggregation mechanism based on graph neural networks to identify key sampling points near fault planes and aggregate lithological features. It then introduces a stratigraphy-aware attention mechanism to explicitly incorporate stratigraphic sequence similarities into the model. Finally, through a dual-task learning strategy, the framework simultaneously performs lithological classification and scalar field interpolation. The authors use real geological data from the Lingnian–Ningping area in Guangxi Zhuang Autonomous Region, China, including stratigraphic interfaces, lithological data, and attitude points, to construct the geological model. The proposed method is compared with the AdaHRBF method and several ablation variants, namely M1–M4, and the modeling results demonstrate its effectiveness. In addition, by integrating training data from geological maps and cross-sections, together with confusion matrices and class-wise evaluation metrics, the authors further demonstrate the reliability and generalization capability of the model in multi-lithology classification. Overall, the results sufficiently validate the effectiveness and geological plausibility of the proposed method. However, the manuscript still has the following issues:
1. The discussion of the limitations of implicit modeling methods in the Introduction lacks a sufficient summary of the relevant literature. The authors are therefore encouraged to supplement this section with a more comprehensive review of related studies.
2. The discussion of the advantages of deep learning methods in the Introduction lacks a comparative analysis of existing studies. The authors are encouraged to further clarify the limitations of existing methods and explain the specific problems that the proposed method is designed to address.
3. The manuscript further validates the effectiveness of the proposed modules through ablation experiments. However, the current ablation study does not include separate evaluations of the SAN module and the dual-task module. As a result, the performance difference between M3 and M4 cannot be fully attributed to either SAN or dual-task learning alone. This experimental design makes it difficult to clearly distinguish the specific contribution of each component within the proposed network framework. The authors are therefore encouraged to supplement the ablation study by evaluating the SAN module and the dual-task module independently.
4. The Results section should not be limited to describing changes in the evaluation metrics associated with the GAT+GraphSAGE module. Instead, the authors should provide a more in-depth analysis of the specific effects of each core module from the perspective of the overall model architecture, including their contributions to feature extraction, stratigraphic prior constraints, lithological classification, and scalar field interpolation. Such an analysis would more clearly reveal how different modules contribute to the improvement of model performance.
5. The Discussion section lacks sufficient analysis and comparison with existing research methods. The authors are encouraged to supplement this section with more detailed and relevant discussion.
6. The manuscript should ensure terminological consistency throughout the text. For example, “lithological classification” and “lithology classification” are used interchangeably. The authors are encouraged to standardize the use of relevant technical terms throughout the manuscript to improve the clarity, consistency, and professionalism of the writing.Citation: https://doi.org/10.5194/egusphere-2026-1210-RC2
Data sets
Three-Dimensional Geological Modeling based on Dual-Task Stratigraphy-Aware Attention Networks (Geo-SAN v1.0) Zhenxi Fang and Boayi Zhang https://zenodo.org/records/19903694
Model code and software
Geo-SAN v1.0 Zhenxi Fang and Boayi Zhang https://github.com/Geo3D-AI-CSU/Geo-SAN
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 156 | 59 | 13 | 228 | 16 | 15 |
- HTML: 156
- PDF: 59
- XML: 13
- Total: 228
- BibTeX: 16
- EndNote: 15
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
1. The physical basis for the stratigraphic sequence similarity matrix is insufficient. In the learnable lithology similarity matrix defined in Section 2.1.2, the denominator value of σ(x) lacks geological basis and sensitivity analysis, and the impact of this hyperparameter on the distinguishability of the 13 lithology types is not discussed.
2. The weight settings for the dual-task loss lack an adaptive mechanism. In the total loss function in Section 2.2.4, the specific values of the four weight coefficients α, β, γ, and δ are not explicitly given in the text, and a fixed weighting strategy is used instead of a dynamic adaptive balancing strategy. In the early stages of training, the scalar field loss and the lithology classification loss differ significantly in magnitude; fixed weights may lead to one task dominating gradient updates. It is recommended to supplement the experimental basis for weight selection.
3. The baseline settings for the ablation experiments are not comprehensive enough. Table 2 only compares four internal variants in the ablation experiments, lacking a horizontal comparison with current mainstream deep learning geological modeling methods, and therefore cannot demonstrate the advantages of Geo-SAN over other methods. Furthermore, whether the accuracy difference between M3 and M4 is entirely attributable to the SAN module or the contribution of the dual-task structure is unclear, lacking ablation experiments that isolate the dual-task module.
4. The coupling mechanism between scalar field prediction and lithological classification is not sufficiently demonstrated. The paper claims that the dual-task framework establishes an "intrinsic coupling" between the scalar field and lithological categories, but Section 2.1.3 only describes a simple shared backbone + dual-branch structure, without explaining the specific interaction between the two tasks at the feature level. The KL divergence term in L_strat transforms the scalar field into lithological prior probabilities, but the mapping function from scalar values to probability distributions in this transformation process is not defined in detail. The confusion between T₁m and P₁m, and D₂d and D₁y, occurs precisely between adjacent stratigraphic units, indicating that cross-task constraints cannot completely resolve the classification ambiguity problem at stratigraphic boundaries.
5. The research depth on fault handling methods is insufficient. Section 2.1.1 mentions that "faults are encoded as a feature of nodes in relation to fault planes," but only encodes faults as a 0/1 binary feature, without considering key parameters such as fault displacement and fault type (normal fault/reverse fault/strike-slip fault). For left-lateral strike-slip reverse faults like the Nacha fault, it is unclear whether the model can accurately characterize the stratigraphic shift relationship between the two sides of the fault. The magnified view of region A in Figure 15b shows abrupt changes in stratigraphic thickness near the fault, requiring quantitative evaluation of the prediction error near the fault area.
6. The model's generalization ability and sample imbalance issues are not adequately discussed. Figure 13d shows extreme imbalance in the training samples. The paper claims the model has "strong small-sample generalization ability" for sparse classes, but does not provide performance tests under conditions of fewer or zero samples. Furthermore, the experiment uses a fixed 8:2 training-to-test split without cross-validation. This limitation of the modeling method, which often occurs in actual exploration where "new stratigraphic units do not appear in the training set," should be explained in the discussion.
7. Inconsistencies between the claim of "geological interpretability" and empirical evidence. Section 5 mentions that "model interpretability is still limited," but the abstract and conclusion emphasize that the method "reflects prior geological knowledge." It is recommended to supplement the analysis with spatial distribution of attention weights to support the core innovation claim of "stratigraphic awareness."