the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
CRAD: A China Reservoir Attribute Database for Hydrological and Water Resources Applications
Abstract. Dams and reservoirs are essential infrastructure for water security and have wide-ranging climatic and environmental impacts. China possesses the world’s largest number of reservoirs, yet existing databases suffer from a severe lack of attribute information, constraining robust assessments of reservoir functions and their broader resource and environmental impacts. To bridge this gap, we developed the China Reservoir Attribute Database (CRAD) by integrating existing global and regional databases, with attribute information retrieved from government documents, enterprise and professional association materials, academic publications, online encyclopedias, and other web-based sources. Through spatial deduplication, source-prioritized data integration, and multilevel quality control, CRAD compiles more than 20 attributes for 5,142 reservoirs, including geographic location, completion year, dam length and height, reservoir area, storage capacity, catchment area, construction purpose, hydropower capacity, and irrigated area. CRAD captures 87.5 % of China’s total reservoir storage capacity and 62.0 % of its installed hydropower capacity, representing a substantial advance over Global Dam Watch (GDW), currently the most attribute-complete global dam and reservoir database. CRAD supports a wide range of hydrological, water resource, climatic, and environmental modeling applications. It identifies 41.3 % of China’s irrigated grid cells as potentially reservoir-served, compared with only 3.5 % using GDW, and improves streamflow simulations at nearly 70 % of the gauging stations. Overall, CRAD provides a substantially improved data foundation for representing reservoir regulation and assessing its hydrological, agricultural, climatic and environmental impacts in China. The CRAD database is publicly available at https://doi.org/10.5281/zenodo.23116980 (Liang et al., 2026).
- Preprint
(4516 KB) - Metadata XML
-
Supplement
(1962 KB) - BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on essd-2026-694', Anonymous Referee #1, 01 Nov 2026
-
RC2: 'Comment on essd-2026-694', Anonymous Referee #2, 05 Nov 2026
This paper introduces the China Reservoir Attribute Database (CRAD), a compiled dataset containing attribute information for 5,142 reservoirs nationwide. It addresses notable limitations in existing global reservoir datasets, especially for reservoir storage, catchment boundaries and irrigation-related fields. I recognise the substantial work required to gather and harmonise information from disparate sources. Overall, this is a valuable contribution suitable for ESSD, but several points around data quality, methodological choices and result interpretation need to be addressed before acceptance.
1. Many attributes are sourced from online encyclopaedias and web platforms, whose reliability is inconsistent. The authors have implemented quality control steps, yet the absolute accuracy of these entries remains hard to judge. I recommend adding finer source metadata for each attribute field, and validating a representative subset against independent, authoritative records where feasible.
2. The inclusion threshold is set at reservoirs with storage above 5 MCM, but Table 3 shows only 3,709 out of 5,142 reservoirs have valid storage values. The selection logic for the remaining reservoirs needs explanation. Please also detail the calculation behind the stated 87.5% coverage of China’s total reservoir storage.
3. The derived relation between reservoir storage and irrigated area is interesting, but this relationship is likely region-dependent and sensitive to reservoir primary functions. Hydropower or flood-control reservoirs, for instance, often have limited irrigation roles. I suggest cross-validating the regression and discussing whether a single unified relation is appropriate across all reservoirs. Supporting checks for these reservoir–irrigation linkages would strengthen this section.
4. I am cautious about direct comparisons between reservoir storage capacity and annual agricultural water demand. Storage volume does not equal the water available for irrigation in any given year; actual supply depends on inflow regimes, operational rules and competing water allocations. The definition of “potential irrigation supply” should be clarified, and interpretations for Figure 9 toned down accordingly. Please also describe how missing dead storage values were treated.
5. The improved streamflow simulation results are promising. That said, the CRAD simulation incorporates 2,090 reservoirs while the GDW benchmark only includes 28. Better performance may stem primarily from the larger reservoir count rather than improved attribute quality. This confounding effect should be teased apart. Given the simulation period spans 1985–2020, the authors also need to state whether reservoir commissioning dates were considered, as many reservoirs did not exist in the earlier years.
6. Dam coordinates: Table 3 lists surveyed coordinates for 1,954 reservoirs, whereas 2,090 reservoirs enter the hydrological modelling. Please explain how locations were assigned for the extra reservoirs used in simulations.
7. Reservoir surface area: Only 1,088 reservoirs have explicit surface area records, even though polygon boundaries are available for all 5,142 reservoirs. It is unclear why polygon-derived areas were not used to fill these gaps.
8. Reservoir operating parameters: Please briefly describe how CRAD storage attributes were translated into the input parameters required by CaMa-Flood, especially for records with incomplete storage components.
9. Data documentation: I recommend retaining original Chinese reservoir names where available and sharpening definitions for fields such as construction and reconstruction years. This will improve dataset usability and facilitate matching with other domestic datasets.Citation: https://doi.org/10.5194/essd-2026-694-RC2 -
RC3: 'Comment on essd-2026-694', Anonymous Referee #3, 10 Nov 2026
This manuscript presents CRAD, a database that integrates existing data sources to compile more than 20 attributes for 5,142 reservoirs in China. The authors demonstrate its use by identifying potentially reservoir-served irrigation grids and evaluating the effects of reservoir representation on streamflow simulations. The topic is well within the scope of Earth System Science Data. Overall, the paper is well written. Methods are clear and results well support the main conclusions.
However, the manuscript needs to address the following issues before it can be accepted.
Major comments:
1) Source traceability is my main concern. The data integration process involved substantial manual checks and corrections, so the database must include attribute-level source indexing to ensure data traceability and support verification. A reservoir-level Quality category cannot substitute for this information.
2) Greater attribute coverage does not establish reliability. Although the comparisons with GDW demonstrate broader coverage and higher completeness, they do not establish the accuracy of the attribute values. The authors should add a “Data validation” section to assess the reliability of CRAD. For example, they can aggregate key attributes, such as storage capacity, installed hydropower capacity and irrigation area, at least at the provincial level and compare the totals with corresponding records in government statistical reports.
3) The irrigation-grid allocation in Section 3.4.1 need further clarifications. For each selected grid cell, is the entire cell area assumed to be cropland, or is only part of the cell treated as cropland in the calculations? How are supply shares calculated when multiple reservoirs serve the same grid cell? Please provide the calculation details and necessary codes. Without these details, Figure 9 is difficult to assess.
4) Figure 10(a) raises concerns about the simulation results. GDW produces much larger flows than the other displayed series at the representative station. A water-balance explanation is needed. The agreement between CRAD-based simulation and the observations also does not convincingly support the statement in Line 576 that “effectively capturing the attenuation of flood peaks and augmentation of low flows induced by reservoir regulation”. The authors must recheck the simulation settings and plotting details. Only after these discrepancies are resolved can the station-level improvement analysis be considered credible.
5) The author should compare attribute diversity and completeness with more China-specific products, such as Li et al. (2025), Zhou et al. (2025), and Wang et al. (2026). GDW alone is insufficient to support CRAD’s contribution relative to existing Chinese reservoir datasets.
6) Check consistency between figures and text. For example, in Figure 8, the green, blue and red colors described in the caption do not correspond to the displayed bars; in Figure 10a, a No-dam series appears in the legend, but its curve cannot be distinguished. The authors should carefully self-check the manuscript for these and any similar issues.
References
- Li, K. Wang, C. Liu, G. Zhao, Z. Jiang, Q. Luo, G. Wang, D. Zhang, J. Yu, and X. Liu. Exacerbating hydrological extremes in China’s large reservoir drainage areas. Journal of Hydrology, 659:133297, 2025. doi:10.1016/j.jhydrol.2025.133297.
- Zhou, Q. Tang, K. Yang, R. Jiang, W. Pan, Y. Ren, J. Qin, D. Chen, and H. Lu. Improved hydropower simulation for China based on a newly developed hydropower database. Journal of Hydrology, 657:133022, 2025. doi:10.1016/j.jhydrol.2025.133022.
- Wang, G. Zhu, R. Li, J. Yang, Z. Zheng, W. Li, Y. Gun, Z. Liu, Y. Li, and D. Yang. A century-scale dataset of hydropower stations in mainland China (1909–2025). Scientific Data, 2026. doi:10.1038/s41597-026-08093-3.
Citation: https://doi.org/10.5194/essd-2026-694-RC3
Data sets
CRAD: A China Reservoir Attribute Database for Hydrological and Water Resources Applications Hongbin Liang, Shulei Zhang, Omarjan Obulkasim, Qiancheng Ye, Linyi Song, Zhongwang Wei, Yongjiu Dai https://doi.org/10.5281/zenodo.23116980
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 544 | 201 | 48 | 793 | 83 | 44 | 42 |
- HTML: 544
- PDF: 201
- XML: 48
- Total: 793
- Supplement: 83
- BibTeX: 44
- EndNote: 42
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This manuscript presents the China Reservoir Attribute Database (CRAD), comprising more than 20 attributes for 5,142 reservoirs across China. By expanding the range and availability of reservoir attributes, CRAD addresses important gaps in existing international datasets and provides a more complete characterization of China’s reservoirs. The agricultural and hydrological applications illustrate the potential value of the database. The manuscript is generally well written, and the topic is well suited to ESSD. I am supportive of publication, but several aspects require clarification or further evaluation, particularly the handling of conflicting source records, the validation of irrigated-area estimates, and the methods used in the application examples.
1---The regression used to estimate missing irrigated-area values needs stronger validation. How well does storage capacity predict irrigated area for reservoirs with reported values? Please evaluate the regression through cross-validation and report appropriate measures of predictive performance. This would provide a clearer basis for assessing whether the relationship is sufficiently reliable for filling these data gaps.
2---Table 1, it would be better to add a column to present the numbers of dams/reservoirs within China for those global inventories. The authors may consider to add the comparison with a new global inventory (GREI).
3---Section 4.1.1 compares CRAD with existing datasets in terms of storage capacity and reservoir location, but the assessment remains largely descriptive. Please quantify the agreement and discrepancies in shared attributes for matched reservoirs using appropriate statistical measures. This would make the comparison more informative and help readers assess consistency among the datasets.
4---The procedure for reconciling records from different sources needs to be explained more clearly. When sources reported conflicting values for the same reservoir attribute, which value was retained, and on what basis? Please describe how these conflicts were resolved and how the reliability of the selected values was assessed.
5---I found the procedure for linking GDW reservoirs to irrigation areas within the CaMa-Flood river network difficult to follow. Were the same criteria used for CRAD and GDW to identify reservoirs and delineate their irrigation service areas? Please clarify the procedure and any differences in its implementation, as these are important for interpreting the comparison between the two datasets.
6---Table 4 reports median attribute values by river basin, but their interpretation depends on the number of available records. Please include the number of reservoirs contributing to each median, separately for each attribute in each basin. This would help readers judge how representative these summaries are.
7---Attribute information is generally more complete for large reservoirs than for small ones. It would therefore be useful to report attribute coverage by reservoir size category, following the storage-based classification used in China. This breakdown would allow readers to assess how completeness varies with reservoir size and identify where substantial information gaps remain.
8---The substantially improved coverage of installed hydropower capacity in CRAD is a notable strength of the database (Figs. 3 and 5), but its implications receive relatively little attention in the discussion. Which research questions or applications could benefit from these additional records? A more concrete discussion of their potential uses would help readers appreciate the value of this improvement.