Reservoir-Induced Seismicity in China: Systematic Characterization and Implications for Hazard Assessment

Zhang, Jiashan; Jiang, Changsheng; Yin, Fengling; Zhu, Yiming; Zhai, Hongyu; Bi, Jinmeng; Zhang, Yan

doi:10.5194/egusphere-2026-1124

Preprints

https://doi.org/10.5194/egusphere-2026-1124

Preprints

20 Apr 2026

| 20 Apr 2026

Reservoir-Induced Seismicity in China: Systematic Characterization and Implications for Hazard Assessment

Jiashan Zhang, Changsheng Jiang, Fengling Yin, Yiming Zhu, Hongyu Zhai, Jinmeng Bi, and Yan Zhang

Abstract. Reservoir-Induced Seismicity (RIS) poses a significant geological risk for large-scale hydropower projects. However, its nationwide characteristics remain poorly understood in China, the country with the world's largest hydropower capacity. This study integrates a high-resolution national reservoir dataset (CRD) with a unified earthquake catalog since 1970 to establish the first screening sample of 1,435 large/medium-sized reservoirs (impounded between 1971 and 2015) in China. Based on spatiotemporal clustering criteria, we systematically identified 88 induced-type reservoirs. Analysis of this dataset reveals fundamental characteristics of RIS in China: Spatially, induced-type reservoirs are selectively concentrated in tectonically active regions like the eastern margin of the Tibetan Plateau, showing significant coupling with active faults and historical seismicity. Temporally, 95 % of RIS events occur within 24.8 km of the reservoir shoreline, and their occurrence rate and time lag (Time Lag) follow patterns governed primarily by pore pressure diffusion, leading to the development of a W-score model for probabilistic RIS identification. In terms of sequence characteristics, RIS exhibits statistically distinct parameters (e.g., lower α-value, higher b-value) compared to natural earthquakes, revealing a unique physical process of activating near-critical faults via pore pressure diffusion. The spatial evolution of the b-value further confirms an effective permeability boundary of pore water penetration at approximately 30–40 km, far exceeding the current Chinese national standard (10 km). Correlation analysis shows that static reservoir parameters are interdependent; thus, only one parameter should be selected for hazard analysis. Reservoir capacity (Reservoir Capacity) greater than 3.2 × 10⁸ m³ or reservoir major axis length (Reservoir Major Axis) exceeding 15.9 km could serve as quantitative thresholds for preliminary risk screening during the planning stage. However, correlation matrix analysis reveals only weak correlations between all static/dynamic parameters and the maximum magnitude (M_max), highlighting the urgent need for physics-based mechanistic models. This study provides the first systematic characteristic map, a quantitative identification tool (W-score), and key engineering thresholds for RIS risk assessment in China, pointing towards future directions beyond traditional empirical models.

Received: 27 Feb 2026 – Discussion started: 20 Apr 2026

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Jiashan Zhang, Changsheng Jiang, Fengling Yin, Yiming Zhu, Hongyu Zhai, Jinmeng Bi, and Yan Zhang

Status: final response (author comments only)

RC1:
'Comment on egusphere-2026-1124', Abhineet Gupta, 17 May 2026
Zhang et al have presented compelling statistical analysis of dam reservoirs in China and potentially induced earthquakes from their impoundment. This is a highly relevant manuscript especially because of its large spatiotemporal domain, covering the entire country and over 40 years of reservoir impoundment.
Overall the statical analysis is rigorous and comprehensive. To further improve the manuscript, I suggest explicitly stating the most important assumptions, validating the analysis further by including more examples, and adding more clarity to the methodology. I have described these suggestions below:
For majority of their analysis and for constructing the W-score, the authors have made a crucial assumption: Both the spatial and temporal behavior of induced earthquakes is uniform throughout China, and across all geologies. This assumption of uniformity in the authors’ analysis is a major limitation as their data covers a very large geographical area. However as one of the first studies exploring this relationship, this first order assumption is reasonable, but I suggest listing it clearly in the manuscript.

The authors have demonstrated their W-score for only 1 reservoir, which makes it difficult to determine its effectiveness for other reservoirs. Suggest providing additional examples to support the score effectiveness.

Introduction: Please include references where missing. Please define non-typical terminology prior to first usage.

Introduction: Suggest adding a paragraph briefly describing the mechanism of induced earthquakes from reservoir impoundment, to improve understanding for unfamiliar readers.

Line 195: Please describe the manual selection approach to facilitate understanding, e.g., how was it determined that seismic activity was clearly distinguishable from background activity (and not distinguishable), how was spatial clustering separated from natural spatial clustering, etc. for each of the 1435 reservoirs.

Line 676: Since the analysis of this study is purely statistical, the causal conclusion: spatiotemporal evolution is dominated by the pore pressure diffusion mechanism, cannot be strongly stated. Suggest tempering this conclusion and stating it as indicative from the analysis.

Line 681 and elsewhere: The conclusion is not as strongly demonstrated from the analysis as stated in the manuscript: The W-score model established based on these characteristics effectively quantifies the spatiotemporal coupling effect, providing a new tool for the probabilistic identification of RIS events. Results were shown for only 1 reservoir. There is not enough information in the manuscript to determine whether the analysis was dominated by a handful of reservoirs that caused the most earthquakes, e.g., the Jinping-I reservoir, or the behavior is distributed among all reservoirs.

Line 23: “screening sample” in unclear.

Line 25: “induced-type” is unclear.

Line 29: The sentence begins with “temporally”, but describes a spatial relationship of 24.8 km. Please clarify.

Line 38: Please list examples of “static reservoir parameters” to improve understanding.

Line 39: Is interdependent = correlated?

Line 46: “characteristic map” is unclear.

Line 72: What is meant by PGA > 0.1g, in relation to natural seismicity?

Line 76: Suggest restructuring sentence to improve clarity.

Line 80, 84: Suggest adding Province (region) of Xinfengjiang Reservoir, Three Gorges Reservoir.

Line 80: Suggest adding references to support the case being globally recognized.

Line 90: Suggest adding magnitude type, e.g., M_w, M_L, etc., instead of just representing using M, depending on how the magnitude is specified.

Line 103: Please add reference for China’s current risk assessment standards.

Line 109: Please define “characteristic map” prior to first usage for clarity.

Line 111: Please define “screening sample”.

Line 113: Please define “induced-type reservoirs”.

Line 115: The sentence indicates questions, but the following list only includes statements. suggest restructuring sentence for clarity.

Line 122: Please define “RIS maximum magnitude”.

Line 154: Please list the most relevant advantages.

Line 170: Suggest adding a map showing the locations of all reservoirs and those for which data was not available, to better understand their spatial distribution.

Figure 1: The unfilled circles are difficult to see on the map. Suggest increasing border thickness.

Figure 1a: It is difficult to compare the pre and post impoundment events because their time horizons are different. For example, there appears to be cluster of events prior to impoundment within the purple dashed lines. Did that cluster occur within a short period of time or over the ~40 years pre-impoundment?

Figure 1c: How is reservoir area defined for classifying “inside” and “outside”?

Figure 1c: Why is there a sudden increase in seismicity outside the reservoir area prior to impoundment?

Line 210: What is M_S? I am not familiar with this magnitude scale.

Figure 2a: The map is difficult to read. Suggest enlarging and providing as a separate figure.

Figure 2d: Why is M>4 selected for this figure when Mc=2.5 is used for analysis?

Lines 244: It is unclear how PGA>0.1g is used to reference natural seismicity? Do the authors mean PGA at a particular return value, e.g., 475 years?

Line 320: Can an explanation be added to “nonlinear dependence of triggering probability on pressure changes”?Line 466: How is peak seismicity rate identified?

Line 513: 10^2.5 -> 10^0.5
Citation: https://doi.org/10.5194/egusphere-2026-1124-RC1
- AC1: 'Author response to Referee #1 (Dr. Abhineet Gupta)', Changsheng Jiang, 21 May 2026
  
  We sincerely thank Dr. Abhineet Gupta for the thorough and constructive first review. The detailed comments have been invaluable in strengthening the clarity, rigor, and scope of our manuscript.
  We have carefully revised the manuscript and prepared a comprehensive point-by-point reply addressing each specific comment. Due to system constraints, only one file can be attached to this interactive comment. Therefore, we provide the complete Author Response—including the reproduced text of the new supplementary materials (S1 and S2)—as a single PDF attachment below. The formal supplementary files will be submitted with the revised manuscript through the editorial system in the next step.
  We are grateful for Dr. Gupta's time and insightful guidance, which have significantly improved this work.
  
  Citation: https://doi.org/10.5194/egusphere-2026-1124-AC1
- AC2: 'Author response to Referee #1 (Dr. Abhineet Gupta)', Changsheng Jiang, 23 May 2026
  
  Please find attached the supplementary materials newly added during the revision process.
  
  Citation: https://doi.org/10.5194/egusphere-2026-1124-AC2
RC2:
'Comment on egusphere-2026-1124', Jürgen Mey, 20 Jul 2026

The authors compile datasets of reservoir lakes and seismic events in China and investigate from this the spatial and temporal occurrence of reservoir induced seismicity.
They first identify such induced seismic sequences based on 1) a rapid increase in seismicity after impoundment that is clearly distinguishable from the background seismicity, 2) a clustering of events, 3) the vicinity of these clusters to the reservoir. These criteria were manually applied to all observations within a 50 km distance from reservoir shorelines that occurred in a time frame from one year before to 10 years after impoundment.
Secondly, they look at the regional distribution of reservoirs with induced seismicity, regarding different tectonic settings.
Thirdly, they describe the distance-from-shoreline distribution and time-lag distribution of RIS sequences using empirical functions. These are then combined into a W-score, which basically gives the probability of an earthquake event being induced in the two-dimensional, distance-time lag space. This metric is regarded as a new way to identify RIS events in China.
Fourth, they use the magnitude-frequency distribution and the ETAS model on the RIS events to derive the characteristic parameters b, alpha and p. These are subsequently compared to those from natural earthquake sequences from their database and from the literature.
At last, they do a correlation analysis between static and dynamic reservoir parameters and the maximum magnitude of the induced seismic events.
Their main results are:
88 out of 1435 reservoirs impounded between 1971 and 2015 show induced seismicity and most RIS events are located in tectonically active regions.
The number of RIS sequences decay according to a power-law with distance from the shoreline and 99% of earthquakes occur within 38.2 km from the shoreline. RIS events start with a considerable delay of 1-2 years after impoundment and gradually decay. The W-score successfully describes the probability of the observed events.
The alpha and b value distributions of RIS events show statistically significant differences to those of natural earthquakes. RIS sequences have higher b values and lower alpha values.
Static parameters of the reservoirs show only weak positive correlation with the maximum moment magnitude in a RIS sequence and time-dependent parameters show no correlation. They find that reservoir capacity and major-axis show thresholds, beyond which relationship with RIS is significant.
The interpretation of the investigated RIS pattern is that most events can be attributed to pore pressure diffusion into preexisting fault systems in tectonically stressed crust, which is in line with previous findings from other places (here: France, Canada, Brazil). The broader spatial influence of RIS (40 km) in China with respect to the other examples is explained by the different tectonic setting and higher fault permeability in China. They suggest a revision of the respective national standard used in reservoir risk assessments. The authors propose a “dual-threshold screening process based on tectonic activity and reservoir geometric parameters” in the reservoir planning stage and the adoption of more holistic approaches that integrate more physical factors such as geological structure, in-situ stress field, fault permeability, and hydrogeological conditions.
Overall assessment
The manuscript is well written, and I could follow the line-of-thought. The presented China-wide assessment is a novelty and is without doubt significant in fostering our understanding of the RIS phenomena. However, I feel the manuscript should be restructured.
Major comments
The authors often mix methodological description, results and interpretation. I urge them to clearly define a methods and a results section and separate these from the interpretation/discussion. I also think that some text that is spread throughout several sections could be put into the introduction (see below). After these structural changes the work could be published.
Minor comments
L29: „Temporally, 95% of RIS events occur within 24.8 km of the reservoir shoreline…” suggest delete “Temporally”
L30: Here and elsewhere in the ms you write terms with small letters and in parentheses again with capital letters (like here “time lag (Time Lag)”). I suggest you choose only one form and stick to it.
L33: “distinct parameters” is too unspecific. Maybe better “distinct parameters of their frequency magnitude distribution”.
L67: “<5km” this is not an area but a distance
L72: define abbreviation PGA
L80: “…is a globally recognized typical RIS event” add reference
L137: what is ML?
L142-146: This sentence seems to be unlogic to me. Suggest rephrase
L213: What is “completeness magnitudes” is it a cut-off?
L232-247: This should go to the introduction
L251: “Sichuan-Yunnan rhombic block” could you show this on the map, e.g. in Fig.2
L259-263: How was the mentioned comparison done and how was the significance in the spatial correlation determined?
L266: suggest delete “intensity”
L277-293: this entire paragraph should go to the introduction
L294-305: Here you describe methods and right after that you continue with the results from L306-315, which in turn is followed by a discussion/interpretation from L315-L322.
L325-339: reads more like a discussion/interpretation
L341: “The potential contribution of early elastic loading, triggering very few near-field events…” how were these mechanisms (elastic loading vs. pore pressure) differentiated?
L372-377: So you calculated the alpha, b and p values for the entire RIS dataset and then only the b values of individual reservoirs, right?
Equation 4: please define all variables, i.e. lambda, m, sigma, x
L383: what is G-R relationship?
Equation 5: define lambda, isn’t M0 the same as the completeness magnitude ML?, the variable t was used before as time after impoundment, I suggest staying consistent and use something else for the time after the mainshock, e.g. t*
L397-400: So most of the data (50 out of 88 cases) weren’t actually used. How did you define the number of 40 earthquakes as being the threshold? Could you give more details about the parameter uncertainties? How were these determined?
L431-433: This belongs to the discussion.
L441: define abbreviation KS
L453-455: What do you mean by the “first strategy”. Please be more specific.
L498-502: This is interpretation.
L543: How was the significance of the spatial coupling determined?
L552: What do you mean by “normalized” ? I think “explained” would be a more suitable choice in this context.
L557-559: Where does this citation come from?
L561-563: Could you qualify these statements more? How is the intraplate tectonic setting more complex? Is there any systematic work on the fault permeability in China in comparison to the other regions that you could cite here?
Figures
Fig.1: the key suggests that there are earthquakes above M = 7 shown on the map but I cannot find them. I strongly suggest deleting all symbology from the legend that is not relevant for the map extent. What is meant by “Pickup frame”?
Fig.2: Please add a reference to the work that mapped the faults shown in (d)
Fig.3: Your usage of completeness magnitude and cutoff magnitude are confusing. Couldn’t you just use one of these terms? In panels c) and (d), the colorbar labels are missing.

Citation: https://doi.org/10.5194/egusphere-2026-1124-RC2
- AC3: 'Author response to Referee #2 (Dr. Jürgen Mey)', Changsheng Jiang, 21 Jul 2026
  
  We sincerely thank Dr. Jürgen Mey for the thorough and constructive evaluation of our manuscript. We completely agree with your assessment that the previous version suffered from a mixed structure, where methodologies, results, and interpretations were intertwined. Following your major comment, we have undertaken a comprehensive restructuring of the entire manuscript, strictly separating the methodology, results, and discussion, and relocating background information to the Introduction. We believe this rigorous restructuring has significantly improved the clarity and logical flow of the paper.
  We have carefully addressed all your specific comments and prepared a comprehensive point-by-point reply. Please find the detailed Author Response attached as a PDF below.
  In accordance with the journal's interactive discussion guidelines, the revised manuscript is not prepared or attached at this stage; it will be formally submitted upon the editor's request following the decision on this public discussion. We are grateful for Dr. Mey's time and insightful guidance, which have substantially improved this work.
  
  Citation: https://doi.org/10.5194/egusphere-2026-1124-AC3

Jiashan Zhang, Changsheng Jiang, Fengling Yin, Yiming Zhu, Hongyu Zhai, Jinmeng Bi, and Yan Zhang

Viewed

Total article views: 704 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
550	118	36	704	23	24

HTML: 550
PDF: 118
XML: 36
Total: 704
BibTeX: 23
EndNote: 24

Views and downloads (calculated since 20 Apr 2026)

Month	HTML	PDF	XML	Total
Apr 2026	150	41	14	205
May 2026	174	40	13	227
Jun 2026	32	8	3	43
Jul 2026	189	29	6	224
Aug 2026	5	0	5

Cumulative views and downloads (calculated since 20 Apr 2026)

Month	HTML	PDF	XML	Total
Apr 2026	150	41	14	205
May 2026	174	40	13	227
Jun 2026	32	8	3	43
Jul 2026	189	29	6	224
Aug 2026	5	0	5

Viewed (geographical distribution)

Total article views: 674 (including HTML, PDF, and XML) Thereof 674 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 02 Aug 2026

Short summary

Large hydropower reservoirs can sometimes trigger earthquakes. We combined a detailed national reservoir map with an earthquake record since 1970 to screen 1,435 Chinese reservoirs and identified 88 likely induced cases. Most events happened within about 25 kilometers of the shoreline and soon after filling, especially in areas with active faults. We developed a scoring tool and recommend wider safety zones and simple reservoir size thresholds to flag higher-risk projects early.


Total:	0
HTML:	0
PDF:	0
XML:	0