Review Reports - A 3D Predictive Method for Deep-Seated Gold Deposits in the Northwest Jiaodong Peninsula and Predicted Results of Main Metallogenic Belts

Round 1

Reviewer 1 Report

· The keywords are very long, they are usually 5 to 6 words, here there are 6 long phrases

· “the gold production was 330 t, yet the gold consumption was 1121 t in China” need reference for this assertion

· The first paragraph of the introduction is very long, you must split it to be able to understand what you want to express. There are many statements that need references in this paragraph. must improve

· which included 2,319.08 t of potential resources. ‘Reference”

· 5,000 t, far exceeding previously predicted results “reference”

· At the end of the introduction you have to say how you conducted your study, with what types of samples, what type of analysis you did. The information in the second paragraph in the introduction is very vague, it does not say anything

· Among them, NE-NNE- trending faults are the most developed. “What is the dip of the faults?”

· with gold resources of greater than 5,000 t.

· The gold mineralization types in the Jiaodong Peninsula primarily include altered rock type in fractured zones (Jiaojia type) and quartz vein type (Linglong and Denggezhuang types), followed by a small quantity of altered breccia type (Pengjiakuang type), altered conglomerate type (Fayunkuang type), interlayer decollement-detachment zone type (Dujiaya type), and pyrite-carbonate vein type (Liaoshang type) “ In this paragraph use at least 12 times the word type. Is it necessary to use so many times? Needs improvement”

· The deep prospecting carried out in the peninsula since the beginning of this century has discovered more than 3,000 t of proven gold resources at a depth of 600–2000 m, exceeding the previously proven gold resources at a depth of 500 m and less. “ References”

· 3.3 D predictive method of deep-seated gold deposits “Is it 3.3 or 3D? You must improve the writing for a better understanding of the reader.”

· deposits s in the Zhaoping fault zone. “Delete the letter "s"

· as well as the basic data of these areas “ What are the basic data of the area?”

· In the geological context, it is not clear to me what type of deposits it are. Are they hydrothermal veins? IRG deposits?

· close caption figure 1. You use the word "type" a lot. Why?

· 1) Ore-controlling faults of gold deposits have large scales. (2) The hanging “DO NOT use (1), (2), (3). since it is confused with the references. You can use First, Second, Third”

· 3D cubic blocks “ What would be the dimensions of the 3D cubes?”

· Analysis and processing of data on typical ore deposits and their surrounding mining areas “ What type of data?

· The principles on which this methodology is based are in Chinese, a language I do not know. I would like to read the documents in English to learn about these principles.

· According to what I understand of the principle of inertia, would there not be significant variations in time and space in the deposits? Geology is not static, it is always in constant change, especially in systems where there are constant changes such as mineral deposits.

· I would like to read the principles, especially the one of inertia, in order to understand what the author says. I believe that mineral deposits, during their formation, are in constant evolution and are not static.

· The veins have mineral zones that are not so easy to predict. Very risky to predict what will happen in 3000 meters of depth. Even in veins that are currently being exploited, a detailed geological control is needed to know the behavior of the mineralization, structural controls and other geological parameters. Predict what will happen between -2000 and -5000 meters I see it as very speculative

· Figure 4 has letters in Chinese alphabet, it must be in English

· Cell blocks with favorable information of orebody distribution. Figure 4 shows the extracted information on the distribution range of drilling-controlled orebodies and mineralization enrichment areas. As shown in Figure 4, orebodies show nearly equidistant distribution along fault strikes and dip directions. Accordingly, the orebody concentration areas showing equidistant distribution along fault strikes covered 273,240 cell blocks, which accounted for 41.35% of the total cell blocks in the model. These cell blocks included 5,932 ore-hosting cell blocks, which accounted for 81.67% of the total ore-hosting cell blocks in the model. The orebody concentration areas showing equidistant distribution along fault dip directions included 251,856 cell blocks, which accounted for 38.11% of the total cell blocks in the model. These cell blocks included 5,884 ore-hosting cell blocks, which accounted for 81.01% of the total ore-hosting cell blocks in the model. “ Very confusing paragraph, needs improvement”

· 4.1.2.3. D predictive model and statistics of information amounts “it is 3D?”

· Figure 6 captions need improvement

· 4.2.2.3D predictive model and statistics of information amounts “it is 3D?”

· The statistics of the proportion of various information amount intervals of cell blocks (Table 6) showed that the information amounts tended to stabilize and converge when they were ≥ 9.597 “ What are the statistics that you mention, what analysis was performed?”

· The Sanshandao fault has cumulative proved gold ore reserves of 333.65 × 106 t, gold metal reserves of 1,240,679 kg, an average orebody thickness of 7.93 m, and an average grade of 3.72 g/t. The first mineralization enrichment zone (elevation: 0 ‒ -500 m) has gold ore reserves of 81.50 × 106 t, gold resources of 238,503 kg, an average orebody thickness of 8.10 m, and an average grade of 2.93 g/t. The second enrichment zone (elevation: -500 ‒ - 2,000 m) has gold ore reserves of 252.14 × 106t, gold metal reserves of 1002176 kg, an average orebody thickness of 7.90 m, and an average grade of 3.98 g/t. The ratio of the resources of the second enrichment zone to those of the first enrichment zone is 4.20. “These assertions need to be checked against references. Did you calculate the reserves of the deposits?”

· Figura 10. Exploration lines. This is the position of the lines, or the lines goes to -5000 m deep

· The model and its prediction can work from the mathematical model, but from the point of view of Economic Geology, metals, in this case Au, do not work that way. Deposit of metals depending on factors such as pH, temperature, salinity of the fluid, etc. which at higher temperatures and pressure (-5000m) are different than in shallow areas. What you are doing are speculations that have to be contrasted with the knowledge of Economic geology.

· It is necessary to have a deep profile of the geology, to know what happens with the deep faults, do they change angle?, what would be the possible host rocks through which the hydrothermal fluid could act. What are the mineral zones at these depths (-2000 to -5000). Is it proven that gold is deposited in these conditions? Is there any worldwide example of deep deposits?

· At present, the exploration depths of identified gold deposits in the Jiaodong

· Peninsula are mostly less than 2,000 m, with only a few exploratory boreholes reaching a depth of more than 3,000 m. “references”

· For example, exploratory borehole ZK96-5 in the Xiling mining area in the northern Sanshandao fault has a final hole depth of 4,006.17 m, making it the deepest borehole for gold exploration in China. “references”

· nearly 7 × 10-6 “units”

· have great potential for deep prospecting “ I agree that it has potential, but putting numbers on that potential is very adventurous.”

· Based on the proven gold resources at an elevation of -2,000 m and above, the total gold resources at an elevation of -2,000 ‒ -5,000 m in the Sanshandao, Jiaojia, and Zhaoping ore-controlling faults were predicted to be approximately 3,377‒6,490 t. Therefore, it is believed that the total gold resources in the Jiaodong Peninsula are expected to exceed 10,000 t.

· Conclusion 3 of his work is very risky, it will take many years to know if it was true or not. But I think you have to have measure in the data.

Comments for author File: Comments.pdf

Author Response

Revision Notes

Re: Manuscript ID: minerals-1757759

Responses to the comments by Reviewer #1:

Point 1: The keywords are very long, they are usually 5 to 6 words, here there are 6 long phrases

Reply:

We appreciate the positive feedback and have made corresponding modification.

Point 2: “the gold production was 330 t, yet the gold consumption was 1121 t in China” need reference for this assertion

Reply：

The data comes from the industry statistics of China Gold Association in 2021. No data published in official journals.

Point 3: The first paragraph of the introduction is very long, you must split it to be able to understand what you want to express. There are many statements that need references in this paragraph. must improve

Reply：

Many thanks for the comments. We have revised the first paragraph of the introduction to make it more concise. We have also updated some important references.

Point 4: which included 2,319.08 t of potential resources. ‘Reference”

Reply：

Many thanks for the comments. The data comes from the metallogenic prediction report of the production unit, so there is no reference to add. This data has been deleted.

Point 5: 5,000 t, far exceeding previously predicted results “reference”

Reply：Done as the reviewer suggested. We have updated the important reference.

Point 6: At the end of the introduction you have to say how you conducted your study, with what types of samples, what type of analysis you did. The information in the second paragraph in the introduction is very vague, it does not say anything

Reply:

Thanks for your comment. We have added instructions for performing statistics of known regional resources and 3D information.

Point 7: Among them, NE-NNE- trending faults are the most developed. “What is the dip of the faults?”

Reply:

Thanks for your comment. The dip of the fault is SE or NW.

Point 8: with gold resources of greater than 5,000 t.

Reply：

Thanks for your comment. This data is quoted from the following reference.

Song, M.C.; Ding, Z.J.; Zhang,J.J.; Song, Y.X.; Bo, J.W.; Wang, Y.Q.; Liu, H.B.; Li, S.Y.; Li, J.; Li, R. X.; Wang, B.; Liu, X.D.; Zhang, L.L.; Dong, L.L.; Li, J.; He, C.Y. Geology and mineralization of the Sanshandao supergiant gold deposit (1200 t) in the Jiaodong Peninsula, China: A review. China Geology. 2021, 4: 686-719.

Point 9: The gold mineralization types in the Jiaodong Peninsula primarily include altered rock type in fractured zones (Jiaojia type) and quartz vein type (Linglong and Denggezhuang types), followed by a small quantity of altered breccia type (Pengjiakuang type), altered conglomerate type (Fayunkuang type), interlayer decollement-detachment zone type (Dujiaya type), and pyrite-carbonate vein type (Liaoshang type) “ In this paragraph use at least 12 times the word type. Is it necessary to use so many times? Needs improvement”

Reply：

Done as the reviewer suggested. We have simplified this section.

Point 10: The deep prospecting carried out in the peninsula since the beginning of this century has discovered more than 3,000 t of proven gold resources at a depth of 600–2000 m, exceeding the previously proven gold resources at a depth of 500 m and less. “ References”

Reply:

Done as the reviewer suggested. We have updated some important references.

Point 11: 3.3 D predictive method of deep-seated gold deposits “Is it 3.3 or 3D? You must improve the writing for a better understanding of the reader.”

Reply:

Thanks for your comment. We have conducted a thorough review of this issue to ensure that it no longer exists.

Point 12: deposits s in the Zhaoping fault zone. “Delete the letter "s"

Reply:

Done as the reviewer suggested.

Point 13: as well as the basic data of these areas “ What are the basic data of the area?”

Reply:

Thanks for your comment. We have revised the description to make it clearer. “the basic data used for modeling” has explained in the previous paragraph.

Point 14: In the geological context, it is not clear to me what type of deposits it are. Are they hydrothermal veins? IRG deposits?

Reply:

Thanks for your question. Jiaodong gold deposits roughly correspond to hydrothermal vein type.

Point 15: close caption figure 1. You use the word "type" a lot. Why?

Reply:

Thanks for your question. Because that the Jiaodong gold deposits are composed of many mineralization types with different characteristics.

Point 16: 1) Ore-controlling faults of gold deposits have large scales. (2) The hanging “DO NOT use (1), (2), (3). since it is confused with the references. You can use First, Second, Third”

Reply:

Done as the reviewer suggested.

Point 17: 3D cubic blocks “ What would be the dimensions of the 3D cubes?”

Reply:

Thanks for your question. The dimensions of the 3D cubes are 120m×120m×10m or 120m×120m×15m

Point 18: Analysis and processing of data on typical ore deposits and their surrounding mining areas “ What type of data?

the basic data used for modeling

Reply:

Thanks for your question. It should be the modeling data. We have revised the description to make it clearer.

Point 19: The principles on which this methodology is based are in Chinese, a language I do not know. I would like to read the documents in English to learn about these principles.

Reply:

Thanks for your comment. We have polished it to conform to the reading habits of native English speakers.

Point 20: According to what I understand of the principle of inertia, would there not be significant variations in time and space in the deposits? Geology is not static, it is always in constant change, especially in systems where there are constant changes such as mineral deposits.

Reply:

Thanks for your comment. We have deleted the inertia principle.

Point 21: I would like to read the principles, especially the one of inertia, in order to understand what the author says. I believe that mineral deposits, during their formation, are in constant evolution and are not static.

Reply:

Thanks for your comment. We have deleted the inertia principle.

Point 23: The veins have mineral zones that are not so easy to predict. Very risky to predict what will happen in 3000 meters of depth. Even in veins that are currently being exploited, a detailed geological control is needed to know the behavior of the mineralization, structural controls and other geological parameters. Predict what will happen between -2000 and -5000 meters I see it as very speculative

Reply:

Thanks for your comment. The areas predicted in this paper are the deep parts of Jiaojia and Sanshandao faults. At present, both faults have been confirmed to be ore bearing in depth above -2000m, and have also been verified by individual boreholes at depth between -2000 and -3000m (see references [60] and [61]). Geophysical exploration and 3D modeling confirm that these two faults extend to a depth of at least -5000m, which indicates that mineralization is possible above -5000m. Therefore, the prediction of -2000 ~ -5000m depth in this paper is reliable.

Point 24: Figure 4 has letters in Chinese alphabet, it must be in English

Reply:

Done as the reviewer suggested.

Point 25: Cell blocks with favorable information of orebody distribution. Figure 4 shows the extracted information on the distribution range of drilling-controlled orebodies and mineralization enrichment areas. As shown in Figure 4, orebodies show nearly equidistant distribution along fault strikes and dip directions. Accordingly, the orebody concentration areas showing equidistant distribution along fault strikes covered 273,240 cell blocks, which accounted for 41.35% of the total cell blocks in the model. These cell blocks included 5,932 ore-hosting cell blocks, which accounted for 81.67% of the total ore-hosting cell blocks in the model. The orebody concentration areas showing equidistant distribution along fault dip directions included 251,856 cell blocks, which accounted for 38.11% of the total cell blocks in the model. These cell blocks included 5,884 ore-hosting cell blocks, which accounted for 81.01% of the total ore-hosting cell blocks in the model. “ Very confusing paragraph, needs improvement”

Reply:

Thanks for your comment. We have rewritten this paragraph to make it clearer

Point 26: 4.1.2.3. D predictive model and statistics of information amounts “it is 3D?”

Reply:

Thanks for your comment. We have conducted a thorough review of this issue to ensure that it no longer exists.

Point 27: Figure 6 captions need improvement

Reply:

Done as the reviewer suggested.

Point 28: 4.2.2.3D predictive model and statistics of information amounts “it is 3D?”

Reply:

Thanks for your comment. We have conducted a thorough review of this issue to ensure that it no longer exists.

Point 29: The statistics of the proportion of various information amount intervals of cell blocks (Table 6) showed that the information amounts tended to stabilize and converge when they were ≥ 9.597 “ What are the statistics that you mention, what analysis was performed?”

Reply:

Thanks for your comment. We have revised the description to make it clearer. This sentence has been amended to remove the "statistics of".

Point 30: The Sanshandao fault has cumulative proved gold ore reserves of 333.65 × 106 t, gold metal reserves of 1,240,679 kg, an average orebody thickness of 7.93 m, and an average grade of 3.72 g/t. The first mineralization enrichment zone (elevation: 0 ‒ -500 m) has gold ore reserves of 81.50 × 106 t, gold resources of 238,503 kg, an average orebody thickness of 8.10 m, and an average grade of 2.93 g/t. The second enrichment zone (elevation: -500 ‒ - 2,000 m) has gold ore reserves of 252.14 × 106t, gold metal reserves of 1002176 kg, an average orebody thickness of 7.90 m, and an average grade of 3.98 g/t. The ratio of the resources of the second enrichment zone to those of the first enrichment zone is 4.20. “These assertions need to be checked against references. Did you calculate the reserves of the deposits?”

Reply:

Thanks for your comment. References and the statistics of proved resources have been explained in this paper.

Point 31: Figura 10. Exploration lines. This is the position of the lines, or the lines goes to -5000 m deep

Reply:

Thanks for your question. It means the position of the lines, instead of goes to -5,000 m deep.

Point 32: The model and its prediction can work from the mathematical model, but from the point of view of Economic Geology, metals, in this case Au, do not work that way. Deposit of metals depending on factors such as pH, temperature, salinity of the fluid, etc. which at higher temperatures and pressure (-5000m) are different than in shallow areas. What you are doing are speculations that have to be contrasted with the knowledge of Economic geology.

Reply:

Thanks for your comment. A description of possible future mining problems has been added to conclusion (3).

Point 33: It is necessary to have a deep profile of the geology, to know what happens with the deep faults, do they change angle?, what would be the possible host rocks through which the hydrothermal fluid could act. What are the mineral zones at these depths (-2000 to -5000). Is it proven that gold is deposited in these conditions? Is there any worldwide example of deep deposits?

Reply:

Thanks for your comment. Both references [43] and [59] have geophysical profiles with a depth of more than -5,000 m, showing the extension and dip angle of the fault. Therefore, relevant profiles are not supplemented this time. In references [60] and [61], deep drilling carried out by predecessors revealed the fracture alteration zone with a thickness of tens of meters at a depth of about -3,500m and gold orebodies at a depth of -2810--2854m. This is enough to prove that gold exists objectively in such conditions.

Point 34: At present, the exploration depths of identified gold deposits in the Jiaodong Peninsula are mostly less than 2,000 m, with only a few exploratory boreholes reaching a depth of more than 3,000 m. “references”

Reply:

Thanks for your comment. We have updated some important references:

Point 35: For example, exploratory borehole ZK96-5 in the Xiling mining area in the northern Sanshandao fault has a final hole depth of 4,006.17 m, making it the deepest borehole for gold exploration in China. “references”

Reply:

Thanks for your comment. We have updated some important references:

Chen, S.X.; Zhang, Y.C., Liu, Z.D. Drilling Technology of ZK96-5 Well in Xiling Gold Deposit, Laizhou, Shandong Province [C]//. The 17th National Exploration Engineering (geotechnical drilling engineering) academic exchange conference proceedings. 2013:121-125.

Point 36: nearly 7 × 10^-6 “units”

Reply:

Thanks for your comment. Unit should be g/t.

Point 37: have great potential for deep prospecting “ I agree that it has potential, but putting numbers on that potential is very adventurous.”

Reply:

Thanks for your comment. The prediction in this paper is mainly based on proven gold resources above -2000m and fault-controlled ore-controlling laws, and two drilling holes over -3000m deep have found deep ore-bodies. Therefore, the author believes that the risk of predicted data is not large, but due to the complex mining conditions in deep, the risk of future mining is relatively larger.

Point 38: Based on the proven gold resources at an elevation of -2,000 m and above, the total gold resources at an elevation of -2,000 ‒ -5,000 m in the Sanshandao, Jiaojia, and Zhaoping ore-controlling faults were predicted to be approximately 3,377‒6,490 t. Therefore, it is believed that the total gold resources in the Jiaodong Peninsula are expected to exceed 10,000 t.

Conclusion 3 of his work is very risky, it will take many years to know if it was true or not. But I think you have to have measure in the data.

Reply:

Thanks for your comment. A description of possible future mining risks has been added to conclusion (3).

Author Response File: Author Response.docx

Reviewer 2 Report

Review of the paper entitled

A 3D predictive method for deep-seated gold deposits in the northwest Jiaodong Peninsula and predicted results of main metallogenic belts

By Mingchun Song, Shiyong Li, Jifei Zheng, Bin Wang, Jiameng Fan, Zhenliang Yang, Guijun Wen, Hongbo Liu, Chunyan He, Liangliang Zhang and Xiangdong Liu

This paper addresses important issues concerning the prediction at depth of potential mineralization zones using indirect structural indicators for predicting geophysical anomalies potentially linked to undiscovered mineralized zones (potential maps). This problem has been addressed many times in the literature using various multivariable statistical approaches including weights of evidence (WoE) [1], logistic regression (LR) [8], cluster analysis (CA), etc. The originality of the paper is to use additional predictive structural information to the classical grade values. The potential mineralized zones in unexplored deep areas are then identified from favorable indicator index values (see Erreur ! Source du renvoi introuvable.). The methodology is then applied to estimate the potential ore / metal / average grade resources of unexplored / unrecognized by bore holes deep zones. The methodology used is original, the actual implementation could be improved, and no major errors have been noticed in the derived conclusions.

However, some comments / improvements should be made on some aspects in the implementation issues / typing before publication:

· Some references have been suggested to be added in the list, in particular most of references are from Chinese authors, and many international authors have already contributed to gold deposits deep exploration, in particular in Canada, and Australia. These authors should be cited (see some suggested additional references).

· Basically the proposed methodology consists in characterizing the potential mineralized zones by favorable geophysical indexes coded by 1 (favorable to mineralization) or 0 (not favorable), then to sum them for identifying the most probable mineralized voxet (cell volume), information index with values greater than 4 to 5 being considered as favorable mineralization zones. Most of the indexes are based on indirect geophysical signatures (some of them being redundant such as magneto telluric, conductivity, resistivity, etc. …). No unfavorable indexes are accounted for. The calibration of these indexes on known mineralization zones is not done (such as histogram of favorable on mineralized zones vs histogram of the same index on non-mineralized zones) to define the probability of favorableness. Cross-validation (Jackknife techniques) is not done on known mineralization and non-mineralized zones: i.e. consider a zone with mineralized and non-mineralized zones, apply the proposed predictive method, then count the good guesses (i.e. mineralization), the false positive (predicted as non-mineralized but in fact mineralized), the bad guess or false negative (predicted non-mineralized but in fact mineralized), and the true negative (predict non-mineralized and being non-mineralized), derive conditional probabilities. These probabilities are helpful to derive a probability map of occurrences. The scientific demarche is not enough accurate / precise so the authors could not give a probability of false positive / or false negative. These unsuccessful probabilities are primordial in exploration. Then the authors derive potential reserves at depth but their figures are subject to discussion given the above remarks (in my opinion over estimated, but uncertainty must be evaluated as well using the above probability). This job must be done and calibrated before publication;

· Remark: brut estimation could be: it is said that at Sanshandao fault, the average Au grades are the following: 0-500m-> 81.5Mt @ 2.93 g/t ; 500-2000m-> 252.2Mt @ 3.98g/t (3 x 81.5 » 244.5) thus 2000-5000m (2 times the volume of 500-2000m) -> 504.4Mt. Assuming a relative uncertainty at 10% for ore estimation (should calculated), 8% for Au grades [it is easy to calculate the standard deviation or IQR (interquartile range) for Au grades, be aware of the nugget effect] (i.e. 18% for Au tonnage) it is found the following estimations for the Sanshandao fault: Total Ore 815 ± 81.5 Mt @ 3.96 ± 0.32 g/t = 3.227 ± 0.581 t Au. A similar estimation should be done for each mining areas. How do you explain the enrichment of grades at deeper levels? Is it not an artifact?

· This methodology is not in agreement with the JORG mining procedure requirements for reserve estimations that are split into proven, potential and speculative reserves.

· Notice that potential geophysical methods such as gravity, electromag, resistivity (even magneto telluric) are less accurate as investigation goes deeper, thus calculated indicator indexes based of them may be squeezed at depth compared with those calculated near the surface; this effect is not discussed in the methodology;

· The predicted reserves at depth between 2000 - 5000m are interesting in terms of geological study. However assuming a geotherm gradient of 30°C /km, this means that ambient temperature is between 60-120°C, an unrealistic temperature for mining exploitation by human beings (may be robots) … in addition, the pressures at that depth are very high causing additional exploitation challenges (refer to South African context). The in-situ calculated reserves are not necessarily exploitable, or at least very challenging; this must be said in the conclusion;

· Little attention is payed to the uncertainty of the resulting reserve calculations; obviously some of them have been interpolated. Uncertainty must be accounted for in the calculation of in-situ reserves (see an example above).

· For estimating the uncertainty, It is suggested to apply a cross validation technique (Jackknife) to evaluate the level of robustness of the predictive method: (i) separate randomly a set of samples / cells R from the whole sample set W on which the mineralized and non-mineralized zones are recognized (i.e. levels 0-500m); (ii) perform the prediction on R and derive the statistics (i.e. probabilities of being mineralized, not mineralized, false positive, false negative etc …); (iv) Reiterate the same process several time and then (v) evaluate a kind of probability of miss classification / prediction for each samples /cells.

· In my opinion, given the above remarks (about geophysical indexes, lack of the uncertainty estimation, possible bias at depth for geophysical indexes, challenge for exploiting the deep resources), the conclusive remarks about the potential resources in gold should be more moderate.

No plagiarism has been identified in this paper.

This paper needs major revisions before publication.

Detailed review

See the amended enclosed pdf.

Suggestion for additional references:

[1] Rabeaut O., Legault M., Cheilletz A., Jébrak M., Royer J.J., and Cheng L.Z. (2010) Gold Potential of a Hidden Archean Fault Zone: The Case of the Cadillac–Larder Lake Fault. Exploration and Mining Geology, 19(3-4), 99-116.

[2] Royer, J.J., Cheilletz, A., Rabeau, O & Jébrak, L. Z. (2011) Is Deposit Location Predictable? Example of the Orogenic Gold Deposits in the Abitibi Province. 11^th Biennial SGA meeting, Antofagasta, Chile, 3p.

[3] Rabeau, O, Royer, J J, Jébrak, M, Cheilletz, A (2013) Log-uniform distribution of gold deposits along major Archean fault zones. Mineralium Deposita, 48, 817-824, DOI 10.1007/s00126-013-0470-7

[4] Farahbakhsh E., Hezarkhani A., Eslamkish T., Bahroudi A., Chandra R. (2020) Three-dimensional weights of evidence modeling of a deep-seated porphyry Cu deposit. Geological Society of London for GSL and AAG.26p., doi: https://doi.org/10.1144/geochem2020-038

[5] Goldfarb, R.J., Groves, D.I., and Gardoll, S., 2001, Orogenic gold and geological time: A global synthesis: Ore Geology Reviews, 18, 1-75.

[6] Sibson R.H.1; Scott J., 1998, Stress/fault controls on the containment and release of overpressured fluids: Examples from gold-quartz vein systems in Juneau, Alaska; Victoria, Australia and Otago, New Zealand, Ore Geology Reviews, 13(1), 293-306.

[7] Tripp G.I., and Vearncombe, J.R., 2004, Fault/fracture density and mineralization: A contouring method for targeting in gold exploration: Journal of Structural Geology, 26, 1087-1108.

[8] Mejia-Herrera P., Royer J.J., Caumon G., Cheilletz A. (2014) Curvature Attribute from Surface-Restoration as Predictor Variable in Kupferschiefer Copper Potentials: An Example from the Fore-Sudetic Region. Natural Resources Research, 16p., DOI: 10.1007/s11053-014-9247-7

Detailed comments

p. 1 Abstract, line 23: elevation of -2,000 ‒ -5,000 m were predicted to be approximately 3,377‒6,490 t. Therefore, it is-> you should precise 3,377 – 6,490 of ore @ give the grade (ex 1.5 g/t) or 3,377 – 6,490 of Au.

p. 2 line 15: were 3,026.49 t -> standard deviation on estimated (i.e. estimation variance) values should be given ex 3,026 ± 60 t of Au (I doubted that the digit are significant!) the same for potential resources 2,319 ± 46 t of Au. Give the estimated standard deviation for resources figures given in the rest of the paragraph.

p.2, 2^nd §, line 4: suppress and

p.2, 2^nd §, line 4: precision deep geophysical exploration -> give precision of which geophysical exploration methods are you talking about: micro-gravity, electromag, seismic?

p.4 1^st §, line 3: suppress misprint: deposits s in the -> deposits s in the

P. 4 Figure: boreholes used to build the 3D model should be included in Fig. 2 as line to indicate the relevant depth vs. extrapolated part of the model. Fig 2b indicates that gold deposits are concentrated in the upper 1km not at depth. Could you be more precise: does gold mineralization occur at a depth greater than 1km at Sanshandao. Grades should be indicated as well.

p.5 line2 : high-precision geophysical methods -> you should precise which geophysical methods you talk about.

p. 5 2^nd § line 2: give size of the elementary voxel ex 25x25x10m

p.5 end of 3^rd §, it seems that gold mineralization occurs in fault open space in which mineralized fluids / brines can circulate

p.5 § 3.2.2. Main predictive bases, line 2: could you give an estimate of that equidistance between ore deposits - see the papers by Royer et al . (2012) and Rabeau et al (2013) to see if they are some similitudes with the Abitibi zone. If so the seismic valve mechanism can be evocated to explain that equidistance.

p.5 in § (3) you explain that gold deposits occur in Jurassic Linglong-type granites, but on the map figure 2 some gold deposits occur in Archean formations. So the question is : are gold deposits associated with Jurassic granites not the result of a remobilization of pre-existing Archean gold deposit, have you arguments (dating) in favor or against this?

p.6. Line 2 low /high gravity transition zones could indicate could indicate fossil hydrothermal zone with dissolved rocks (lower density), while magnetic anomaly would be associated to magnetite / pyrrhotite minerals often associated to gold mineralization.

p.6. Line 9: boundaries between low / high resistivity zones could be linked to the presence of conductive minerals such pyrite often pathfinder minerals for gold mineralization. The same explanation for the magnetotelluric methods.

p.7 § Predictive factors: If I understand correctly, basically, you estimate the ore grade for the upper - 2000m along the fault plan, then you extrapolate this grade along the fault plane at depth below -2000m to obtain an estimate. This is a force brut method that can be improved using for instance geostatistical simulation methodology: the advantage is it would give you an idea of the uncertainty on the estimate of additional resources / reserves. At least you can compute the variance together with the average, to give an estimation range of the resource you estimate by your methodology.

p. 7 end of §4.1.1 coding by 0 non-mineralized and by 1 mineralized voxet is an interesting methodology; you can then calculate / estimate / fit the indicator variogram of mineralized zones, it gives you information about the apparent diameter of mineralized zones, or / and spatial periodicity of mineralized zones. You can then use this variogram to simulate the mineralized zone at depth, the result being a probability occurrence map. You can add the secondary favorable mineralization index when estimating potential map

p. 8 1^st § it is not well precise which statistical method you use to determine favorable voxets give the favorable index (weight of evidence method [1], logistic regression). This should be defined.

p.9 figure 4: level (depth) for fig. a and b should be indicated. Fig a and b seem identical comments should be more informative

p. 11 figure 6: since your predictive index can be calculated both on explored zones where you know approximatively the ore grade by borehole, and at depth (with no borehole), you should have used a logistic regression to predict the grade of unknown zones, it should have given you both estimated values and their uncertainty. In particular in figure 6-b the blue areas seems to be mineralized, your index is near 5 along a NNE direction similarly to the unknown zone on the West (in red). Why have you eliminated this blank zone (should be in red as well).

p.11 figure 5 since you have the favorable index both on mineralized and barren zones you should have built the conditional histogram one on the mineralized zones, the other on barren zones. Better by class of ore grade (i.e. Au). Same remarks for figure 7

p.14 line 2: 333.65 Mt

p.16 just above table 8: Assuming a relative uncertainty at 10% for ore estimation (should calculated), 8% for Au grades [it is easy to calculate the standard deviation or IQR (interquartile range) for Au grades] (i.e. 18% for Au tonnage) it is found the following for the Sanshandao fault: Total Ore 815 ± 81.5 Mt @ 3.96 ± 0.32 g/t = 3.227 ± 0.581 t Au. A similar estimation should be done for each mining areas. How do you explain the enrichment of grades at deeper levels? Is it not an artifact?

P. 17 : 7 g/t

p. 18: In my opinion, given the above remarks (about geophysical indexes, lack of the uncertainty estimation, possible bias at depth for geophysical indexes, challenge for exploiting the deep resources), the conclusive remarks about the potential resources in gold should be more moderate.

Comments for author File: Comments.zip

Author Response

Revision Notes

Re: Manuscript ID: minerals-1757759

Responses to the comments by Reviewer #2:

Review of the paper entitled

A 3D predictive method for deep-seated gold deposits in the northwest Jiaodong Peninsula and predicted results of main metallogenic belts

By Mingchun Song, Shiyong Li, Jifei Zheng, Bin Wang, Jiameng Fan, Zhenliang Yang, Guijun Wen, Hongbo Liu, Chunyan He, Liangliang Zhang and Xiangdong Liu

However, some comments / improvements should be made on some aspects in the implementation issues / typing before publication:

Point 1: Some references have been suggested to be added in the list, in particular most of references are from Chinese authors, and many international authors have already contributed to gold deposits deep exploration, in particular in Canada, and Australia. These authors should be cited (see some suggested additional references).

Reply：

Many thanks for the comments. We have revised the first paragraph of the introduction to make it more concise. We have also updated some important references.

Point 2: Basically the proposed methodology consists in characterizing the potential mineralized zones by favorable geophysical indexes coded by 1 (favorable to mineralization) or 0 (not favorable), then to sum them for identifying the most probable mineralized voxet (cell volume), information index with values greater than 4 to 5 being considered as favorable mineralization zones. Most of the indexes are based on indirect geophysical signatures (some of them being redundant such as magneto telluric, conductivity, resistivity, etc. …). No unfavorable indexes are accounted for. The calibration of these indexes on known mineralization zones is not done (such as histogram of favorable on mineralized zones vs histogram of the same index on non-mineralized zones) to define the probability of favorableness. Cross-validation (Jackknife techniques) is not done on known mineralization and non-mineralized zones: i.e. consider a zone with mineralized and non-mineralized zones, apply the proposed predictive method, then count the good guesses (i.e. mineralization), the false positive (predicted as non-mineralized but in fact mineralized), the bad guess or false negative (predicted non-mineralized but in fact mineralized), and the true negative (predict non-mineralized and being non-mineralized), derive conditional probabilities. These probabilities are helpful to derive a probability map of occurrences. The scientific demarche is not enough accurate / precise so the authors could not give a probability of false positive / or false negative. These unsuccessful probabilities are primordial in exploration. Then the authors derive potential reserves at depth but their figures are subject to discussion given the above remarks (in my opinion over estimated, but uncertainty must be evaluated as well using the above probability). This job must be done and calibrated before publication;

Reply：

Thanks for your comment. The information values used in this paper to delineate the predicted target areas are all from the actual data of the statistics of known ore bodies, that is, they have been verified by known ore bodies. If they are used to judge the known mineralized and non-mineralized areas, the probability undoubtedly be 100%. In addition, some of the predicted areas delineated in this paper have been verified by recent drilling. Therefore, this revision does not supplement the information about calibration of known regions. However, "4.1.4 Probability and verification of prediction results" is added to the prediction target area of jiaojia fault zone to explain the probability of prediction and drilling verification. Since the situation of the deep prospecting target area of the Sanshandao fault zone is similar to that of the Jiaojia fault zone, the explanation of the probability and verification of the prediction results of the deep prospecting target area of the Sanshandao fault zone is not supplemented. Magnetotelluric, electrical conductivity, resistivity and other geophysical characteristics are deleted.

Point 3: Remark: brut estimation could be: it is said that at Sanshandao fault, the average Au grades are the following: 0-500m-> 81.5Mt @ 2.93 g/t ; 500-2000m-> 252.2Mt @ 3.98g/t (3 x 81.5 » 244.5) thus 2000-5000m (2 times the volume of 500-2000m) -> 504.4Mt. Assuming a relative uncertainty at 10% for ore estimation (should calculated), 8% for Au grades [it is easy to calculate the standard deviation or IQR (interquartile range) for Au grades, be aware of the nugget effect] (i.e. 18% for Au tonnage) it is found the following estimations for the Sanshandao fault: Total Ore 815 ± 81.5 Mt @ 3.96 ± 0.32 g/t = 3.227 ± 0.581 t Au. A similar estimation should be done for each mining areas. How do you explain the enrichment of grades at deeper levels? Is it not an artifact?

Reply：

Good question. The method of resource prediction in this paper is different from that proposed by the reviewer. In this paper, the resources of -2000 ~ -5000m are predicted according to the actual proved resources of 0 ~ -500m and -500 ~ -2000m. Since the proved resources of 0 ~ -500m are small, and the proved resources of -500 ~ -2000m are large, the prediction data is a numerical range. This range is between the maximum and minimum estimated resources, including the amount of resources and grade deviation. However, according to the method given by the reviewer, the volume of -2000 ~ -5000m is twice that of -500 ~ -2000m, and the predicted resource amount of -2000 ~ -5000m is roughly estimated, without considering the resource amount of 0 ~ -500m. Although the e reviewer's method adopted relative uncertainty to correct, the estimated gold resources of the Sanshandao fault at -2000 ~ -5000m are as follows: total ore 815± 81.5Mt × 3.96± 0.32g /t = 3,227±0.581t Au, far greater than the maximum value estimated in this paper: total ore 560 Mt, gold metal reserves of 2,228 t. It shows that the prediction in this paper is relatively conservative. In Table 7, the gold resources and grade values of 0-500m and -500-2000m of Sanshandao fault in this paper are the results of actual exploration, and there is no human processing problem. The high deep grade may be due to the low mining cost of 0 to -500m and the lower grade ore has been zoned into the deposit during exploration.

Point 4: This methodology is not in agreement with the JORG mining procedure requirements for reserve estimations that are split into proven, potential and speculative reserves.

Reply：

Thanks for your comment. JORC standard is a regulation of mineral resources and ore reserves. This paper is a prediction of deep resources, not a report of exploration results. The gold resources of 0 ~ -2000m described in this paper roughly conform to JORC standard, while the predicted gold resources of -2000 ~ -5000m are not explored according to JORC standard.

Point 5: Notice that potential geophysical methods such as gravity, electromag, resistivity (even magneto telluric) are less accurate as investigation goes deeper, thus calculated indicator indexes based of them may be squeezed at depth compared with those calculated near the surface; this effect is not discussed in the methodology;

Reply：

Thanks for your comment. A note on resource prediction errors has been added in paragraph 3 of section 5.2.

Point 6: The predicted reserves at depth between 2000 - 5000m are interesting in terms of geological study. However assuming a geotherm gradient of 30°C /km, this means that ambient temperature is between 60-120°C, an unrealistic temperature for mining exploitation by human beings (may be robots) … in addition, the pressures at that depth are very high causing additional exploitation challenges (refer to South African context). The in-situ calculated reserves are not necessarily exploitable, or at least very challenging; this must be said in the conclusion;

Reply：

Thanks for your comment. A description of the challenges facing future mining has been added to conclusion (3).

Point 7: Little attention is payed to the uncertainty of the resulting reserve calculations; obviously some of them have been interpolated. Uncertainty must be accounted for in the calculation of in-situ reserves (see an example above).

Reply：

Thanks for your comment. A note on resource prediction errors has been added in paragraph 3 of section 5.2. In addition, this paper is the prediction of deep resources, not the reserve calculation or in-situ reserve results. The uncertainty and error of prediction are very large, and it is difficult to describe with specific probability or error figures.

Point 8: For estimating the uncertainty, It is suggested to apply a cross validation technique (Jackknife) to evaluate the level of robustness of the predictive method: (i) separate randomly a set of samples / cells R from the whole sample set W on which the mineralized and non-mineralized zones are recognized (i.e. levels 0-500m); (ii) perform the prediction on R and derive the statistics (i.e. probabilities of being mineralized, not mineralized, false positive, false negative etc …); (iv) Reiterate the same process several time and then (v) evaluate a kind of probability of miss classification / prediction for each samples /cells.

Reply：

Thanks for your comment. This question is consistent with the Point 2, which is addressed in the Point 2.

Point 9: In my opinion, given the above remarks (about geophysical indexes, lack of the uncertainty estimation, possible bias at depth for geophysical indexes, challenge for exploiting the deep resources), the conclusive remarks about the potential resources in gold should be more moderate.

Reply：

Thanks for your comment. The numbers in the conclusions regarding the predicted gold resources have been revised to be more moderate.

No plagiarism has been identified in this paper.

This paper needs major revisions before publication.

Reply:

Thanks for your comment. We appreciate the positive feedback and have made corresponding modification.

Detailed review

See the amended enclosed pdf.

Suggestion for additional references:

[3] Rabeau, O, Royer, J J, Jébrak, M, Cheilletz, A (2013) Log-uniform distribution of gold deposits along major Archean fault zones. Mineralium Deposita, 48, 817-824, DOI 10.1007/s00126-013-0470-7

[5] Goldfarb, R.J., Groves, D.I., and Gardoll, S., 2001, Orogenic gold and geological time: A global synthesis: Ore Geology Reviews, 18, 1-75.

[7] Tripp G.I., and Vearncombe, J.R., 2004, Fault/fracture density and mineralization: A contouring method for targeting in gold exploration: Journal of Structural Geology, 26, 1087-1108.

Reply:

Many thanks for the comments. We have updated the above important references.

Detailed comments

Point 1: p. 1 Abstract, line 23: elevation of -2,000 ‒ -5,000 m were predicted to be approximately 3,377‒6,490 t. Therefore, it is-> you should precise 3,377 – 6,490 of ore @ give the grade (ex 1.5 g/t) or 3,377 – 6,490 of Au.

Reply:

Done as the reviewer suggested.

Point 2: p. 2 line 15: were 3,026.49 t -> standard deviation on estimated (i.e. estimation variance) values should be given ex 3,026 ± 60 t of Au (I doubted that the digit are significant!) the same for potential resources 2,319 ± 46 t of Au. Give the estimated standard deviation for resources figures given in the rest of the paragraph.

Reply:

Done as the reviewer suggested. The number after the decimal point has been removed and replaced with an integer

Point 3: p.2, 2^nd §, line 4: suppress and

Reply:

Done as the reviewer suggested.

Point 4: p.2, 2^nd §, line 4: precision deep geophysical exploration -> give precision of which geophysical exploration methods are you talking about: micro-gravity, electromag, seismic?

Reply:

Done as the reviewer suggested.

Point 5: p.4 1^st §, line 3: suppress misprint: deposits s in the -> deposits s in the

Reply:

Thanks for your comment. We have conducted a thorough review of this issue to ensure that it no longer exists.

Point 6: P.4 Figure: boreholes used to build the 3D model should be included in Fig. 2 as line to indicate the relevant depth vs. extrapolated part of the model. Fig 2b indicates that gold deposits are concentrated in the upper 1km not at depth. Could you be more precise: does gold mineralization occur at a depth greater than 1km at Sanshandao. Grades should be indicated as well.

Reply:

Thanks for your comment. Since there are 311 borehole data used for 3D modeling, the borehole data are not supplemented because the borehole data will be too dense to be placed in Figure 4, seriously affecting the performance of other information. In order to better display the relationship between fault and ore body, the observation results of rotating fault into a gentle state are shown in Figure 2b, which is not the original state. The depth of the deposit is not true, and its true depth has reached -2000m.

Point 7: p.5 line2 : high-precision geophysical methods -> you should precise which geophysical methods you talk about.

Reply:

Done as the reviewer suggested.

Point 8: p.5 2^nd § line 2: give size of the elementary voxel ex 25x25x10m

Reply:

Thanks for your comment. It has been modified to 3D cubic blocks(120m×120m×10m or 120m×120m×15m)

Point 9: p.5 end of 3^rd §, it seems that gold mineralization occurs in fault open space in which mineralized fluids / brines can circulate

Reply:

Done as the reviewer suggested.

Point 10: p.5 § 3.2.2. Main predictive bases, line 2: could you give an estimate of that equidistance between ore deposits - see the papers by Royer et al . (2012) and Rabeau et al (2013) to see if they are some similitudes with the Abitibi zone. If so the seismic valve mechanism can be evocated to explain that equidistance.

Reply:

Thanks for your comment. It has been added that the roughly equal-spaced distance is 500m and that references have been added.

Point 11: p.5 in § (3) you explain that gold deposits occur in Jurassic Linglong-type granites, but on the map figure 2 some gold deposits occur in Archean formations. So the question is : are gold deposits associated with Jurassic granites not the result of a remobilization of pre-existing Archean gold deposit, have you arguments (dating) in favor or against this?

Reply:

Thanks for your comment. There is further explanation behind “The gold deposits in the Jiaodong Peninsula mainly occur in Jurassic Linglong-type granites”. It’s consistent with the reviewer.

Point 12: p.6. Line 2 low /high gravity transition zones could indicate could indicate fossil hydrothermal zone with dissolved rocks (lower density), while magnetic anomaly would be associated to magnetite / pyrrhotite minerals often associated to gold mineralization.

Reply:

Thanks for your comment. The explanation of the description recommended by reviewer has been added.

Point 13: p.6. Line 9: boundaries between low / high resistivity zones could be linked to the presence of conductive minerals such pyrite often pathfinder minerals for gold mineralization. The same explanation for the magnetotelluric methods.

Reply:

Thanks for your comment. The explanation of the description recommended by reviewer has been added.

Point 14: p.7 § Predictive factors: If I understand correctly, basically, you estimate the ore grade for the upper - 2000m along the fault plan, then you extrapolate this grade along the fault plane at depth below -2000m to obtain an estimate. This is a force brut method that can be improved using for instance geostatistical simulation methodology: the advantage is it would give you an idea of the uncertainty on the estimate of additional resources / reserves. At least you can compute the variance together with the average, to give an estimation range of the resource you estimate by your methodology.

Reply:

Thanks for your comment. Resources above -2000m are not estimated, but actually proved. Based on the statistics of proved resources above -2000 m, the resources below -2000 m are estimated. The results of the estimation give a wide range of resources variation, that is, 3377 ~ 6490 t gold.

Point 15: p.7 end of §4.1.1 coding by 0 non-mineralized and by 1 mineralized voxet is an interesting methodology; you can then calculate / estimate / fit the indicator variogram of mineralized zones, it gives you information about the apparent diameter of mineralized zones, or / and spatial periodicity of mineralized zones. You can then use this variogram to simulate the mineralized zone at depth, the result being a probability occurrence map. You can add the secondary favorable mineralization index when estimating potential map

Reply:

Thanks for your comment. The answer to this question has been integrated into Point 2. The modification is shown in the reply of Point 2.

Point 16: p.8 1^st § it is not well precise which statistical method you use to determine favorable voxets give the favorable index (weight of evidence method [1], logistic regression). This should be defined.

Reply:

Thanks for your comment. It has been added that the roughly equal-spaced distance is 500m and that references have been added

Point 17: p.9 figure 4: level (depth) for fig. a and b should be indicated. Fig a and b seem identical comments should be more informative

Reply:

Thanks for your comment. Figure 4 is the horizontal projection of Jiaojia gold deposit distribution, which has no depth meaning. The base maps (orebody distribution maps) of Figures 2a and b are the same, but they show the equally spaced distribution of orebodies along strike and dip, respectively.

Point 18: p.11 figure 6: since your predictive index can be calculated both on explored zones where you know approximatively the ore grade by borehole, and at depth (with no borehole), you should have used a logistic regression to predict the grade of unknown zones, it should have given you both estimated values and their uncertainty. In particular in figure 6-b the blue areas seems to be mineralized, your index is near 5 along a NNE direction similarly to the unknown zone on the West (in red). Why have you eliminated this blank zone (should be in red as well).

Reply:

Thanks for your comment. The blue area in Figure 6 is the proven ore body area (illustrated in the legend), not the predicted target area, which is indicated in red.

Point 19: p.11 figure 5 since you have the favorable index both on mineralized and barren zones you should have built the conditional histogram one on the mineralized zones, the other on barren zones. Better by class of ore grade (i.e. Au). Same remarks for figure 7

Reply:

Thanks for your comment. We have added the histogram of metallogenic information amount showing the mineralized zones and the barren zones.

Point 20: p.14 line 2: 333.65 Mt

Reply:

Done as the reviewer suggested.

Point 21: p.16 just above table 8: Assuming a relative uncertainty at 10% for ore estimation (should calculated), 8% for Au grades [it is easy to calculate the standard deviation or IQR (interquartile range) for Au grades] (i.e. 18% for Au tonnage) it is found the following for the Sanshandao fault: Total Ore 815 ± 81.5 Mt @ 3.96 ± 0.32 g/t = 3.227 ± 0.581 t Au. A similar estimation should be done for each mining areas. How do you explain the enrichment of grades at deeper levels? Is it not an artifact?

Reply:

Thanks for your comment. The answer to this question has been integrated into Point 3. The modification is shown in the reply of Point 3.

Point 22: P.17: 7 g/t

Reply:

Done as the reviewer suggested.

Point 23: p.18: In my opinion, given the above remarks (about geophysical indexes, lack of the uncertainty estimation, possible bias at depth for geophysical indexes, challenge for exploiting the deep resources), the conclusive remarks about the potential resources in gold should be more moderate.

Reply:

Thanks for your comment. The answer to this question has been integrated into Point 9. The modification is shown in the reply of Point 9.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Keywords are used to broaden the readers search, actually you are using words or phrases that are in the title of the article.

Line 47-48: “In 2021, the gold production was 330 t, yet the gold consumption was 1121 t in China” If this information comes from the industry, there must be reports that support it. This information cannot be published, because it would be purely speculative.

Line 123: The word law is a term used by lawyers. Find the right geological word

Line 200: Avoid the use of numbers in parentheses, as they are confused with references. You could use letters or bullets

Line 216: same Line 200

Line 408: remove chinese characters

Before your study was based on the principle of inertia. Now that you've removed it, on what principle?

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Review of the paper entitled

A 3D predictive method for deep-seated gold deposits in the northwest Jiaodong Peninsula and predicted results of main metallogenic belts

By Mingchun Song, Shiyong Li, Jifei Zheng, Bin Wang, Jiameng Fan, Zhenliang Yang, Guijun Wen, Hongbo Liu, Chunyan He, Liangliang Zhang and Xiangdong Liu

The second version of this paper has been significantly improved despite the fact the authors have not included some suggestions I made in the previous review, especially regarding the Au content and uncertainty on the given resources.

Some improvements should be still made before publication:

· The authors quantify what they call the information amount of five indicators derived from the geological model. Despite it is the cornerstone of the methodology, the authors did not well explain how this information amount is calculated for each indicator. The authors should give a small example in order that the reader can fully understand how this parameter is derived from the statistic of mineralized / non mineralized cells. I suspect that it is defined as

· Then, it is understood that for each the information amounts carried by each indicators I_i are summed and affected to the cell (i.e. I =I₁+I₂+I₃+I₄+I₅). From the statistic on recognized / exploited areas, a cut-off at 4.5 is using to identify mineralized cell (I>4.5) and barren cells (I<4.5). It should be interested to have some statistic on the true positive, true negative, false positive and false negative (see detailed comment);

· The scientific demarche is not enough accurate / precise so the authors could not give a probability of false positive / or false negative. These unsuccessful probabilities are primordial in exploration.

No plagiarism has been identified in this paper.

This paper needs revisions before publication.

Detailed review

See the amended enclosed pdf.

Since the new version is corrected in red, in order to not overload the text, I put my comments on the old pdf version.

Detailed comments

p. 1 Abstract, line 29: “ According to the geological law stating that gold deposits are controlled by faults in a 3D space” This is not always the case see Cu, Au porphyry This is not always the case see Cu, Au porphyry, say instead: Since orogenic gold deposits are often controlled by faulting in the 3D space [16-18], …

elevation of -2,000 ‒ -5,000 m were predicted to be approximately 3,377‒6,490 t. Therefore, it is-> you should precise 3,377 – 6,490 of ore @ give the grade (ex 1.5 g/t) or 3,377 – 6,490 of Au. This gives an idea for comparing / seeing if compatible to similar known gold deposits in the world. I am deappointed that the authors did not account for my previous remarks of the 1^st reviewing stage; suggestions are made to improve the paper.

p.2, line 48: for more readability write 1,121 instead

line 57, 59 give average gold content of each resources (i.e. 2,492t Au @ 2.98?? ± put the statistical standard deviation g/t, 3,963 t Au @ …)

p.7, line 274 The author said “The deep resources were estimated using the equation: deep resources = ore-bearing rate × the area of the deep prediction areas” ore-bearing rate = ore bearing rate meaning the average gold content calculated by projection onto the fault plan. With a similar methodology, it is easy to estimate the standard deviation / IQR (interquartile range) for estimating the gold content variability in each pixel, and therefore to derive an uncertainty range for each estimation.

p.7 line 316 the authors defines an ore indication from “fault parts with a steep-to-gentle transition of fault dip angle “This can be estimated by calculating the Gaussian curvature K_G of the surface fitting the fault(s) surface, and then applying a cut-off of K_G

p.8 line 326 I understand that in exploited zones (0-2000m) it is relatively easy to derive predictive indicators from faulting, but for deeper zones 2000 – 5000m how to estimate such indicator from scarce drilling holes ? This needs precision / comments / discussion.

p.9 Fig 4 What are a and b levels exactly ? Give explanation where they are located. Depth?

p.9 line 346 This predictive model involved five characteristic variables for statistical analysis.-> remind explicitly what are they (Change in fault direction ...) How are you sure of their estimation at 5000m depth? No uncertainty to estimate these indicators in each cell at 4,000m from scarce drill holes? The geology is not so known at that depth. This fact is not accounted for / discussed in the model.

p.9 Table 2 The 1^st and 4^th columns are useless since figures are identical for all indicators. Put instead the number of blocks with favorable factors but not mineralized (false positive) (101 348 for the 1st line ) as 1st column, and the number of mineralized blocks with unfavorable factor (false negative) (133 for the 1st line) as 4^th column.

Indicator:	Nb of blocks
fault buffer zones	ore	No ore	total
1	True positive 7,124 blocs	False positive 101,348 blocs	108,572 blocs
0	False negative 133 blocs	True negative 552,089 blocs	552,228 blocs
Total	7,263 blocs	653,437 blocs	660,800 blocs

Accuracy = 84.62%; disjoint class precision = 15.51%; associated class precision= 98.09%; disjoint class recall = 99.97%; associated class recall = 6.56%

(see https://datascientest.com )

From which you can derive the conditional probability of finding a deposit d when the indicator is favorable , the conditional probability of missing a deposit when the indicator is unfavorable (false negative), the condition probability of having no deposit when the indicator is favorable (false positive), and the condition probability of having no deposit when the indicator is unfavorable (true negative). Reminding that the definition of conditional probability [it comes: , , , ]. It can be done for all indicators or on a combination of them (here the sum).

fault buffer zones	Conditional probability
i\p	ore	No ore
1	True positive 6.56%	False positive 93.35%	16.43%
0	False negative 0.024%	True negative 99.96%	83.57%
	1.10%	98.90%	1

The 5^th column in table 2 (information amount) is unclear and needs additional explanation for the reader. How do you compute the information amount for the first indicator (1^st line in table 2). Please give explanation explicitly with examples so the reader can better understand the used methodology.

I understand that for a given cell you add this information amount for each indicator I=I₁+I₂+I₃+I₄+I₅,(implicitly assuming they are independent); then you select mineralized cells using a cut-off at I> 4.5, and decide that cells with I£ 4.5 are considered as barren.

In table 2 (and in the text bellow) all figures greater than 999 should be written with a coma i.e. 108572-> 108,572 in order to be more readable. I put in yellow across the text (but perhaps forgot some …). Same remarks for table 5.

p. 16 line 472 table 7 the value of the Ore-bearing rate should be given with three maximum digit i.e. 0.043231783 -> 0.0432 (same in Table 8: 3^rd and 4^th columns).

Comments for author File: Comments.zip

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 3

Reviewer 2 Report

Review of the paper entitled

A 3D predictive method for deep-seated gold deposits in the northwest Jiaodong Peninsula and predicted results of main metallogenic belts

By Mingchun Song, Shiyong Li, Jifei Zheng, Bin Wang, Jiameng Fan, Zhenliang Yang, Guijun Wen, Hongbo Liu, Chunyan He, Liangliang Zhang and Xiangdong Liu

The third version of this paper has been significantly improved. However, it still remains some typo, imprecision, spelling …. that require editing before the definitive publishable version (see attached amended pdf).

Moreover, in previous version the authors said that they use five characteristics to compute what they called information amount and to delineate favorable mineralized prospects. Now they spoke of eight characteristics (line 351) but did not update the list of characteristics nor the following text (line 360). This must be fixed clearly before publication.

This paper needs some revisions before publication.

Detailed review

See the amended enclosed pdf.

Detailed comments

p. 1 Keywords, line 43: ;;->;

line 47: The contradiction between the supply and demand of gold is prominent in China.-> This sentence needs changes: The supply of gold in China does not meet the present.

line 48: gold production was 330 t, yet the gold consumption was 1,121 t in China.-> gold production in China was 330 t, while the gold consumption was 1,121 t.

line 52: a complex engineering system-> a complex engineering task

line 57, 160: 21^st

line 79-83 Too long sentence need to be rephrases

line 112-115 suppress the words type (too many repetitions) ; types

line 116 deposits

line 183: corresponded -> corresponds

line 189: such sentences: ‘quantitative prospecting information’ is not precise / scientific enough.

Rephrase as following: the quantitative prospecting information described in § 3.2.2. was assigned to each cell block, i.e. each favorable indicator is coded. Once the coding of the many prospective indicators is finished, an global favorable indicator is computed by combining the favorable coded indicator on each cell. Finally, the areas with higher prospective global scores were identified as prospecting target areas.

Line 225: effect-> efficiency

Line 350: in the previous versions the authors refer to five predictive characteristics, now they are eight but as it is written it is not clear what are these characteristics. Please describe clearly what are these characteristics.

Line 358 )

Line 360: you said eight at line 351.

Comments for author File: Comments.zip

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.