the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 07 Sep 2026
A Monthly 30 m Dataset of Lakes Larger than 1 km² on the Tibetan Plateau from 2001 to 2025
Abstract. Long-term monthly records of lake dynamics are critical for understanding hydrological and climatic change on the TP. However, existing products are constrained by factors such as cloud cover, snow and ice contamination, terrain shadow, observational gaps, and scale-related bias. Here we present a new monthly dataset of lakes larger than 1 km² across the TP for 2001–2025 at 30 m spatial resolution. The dataset was generated using TSRFD framework integrating MODIS, Landsat and terrain variables, with refinement steps to improve mapping accuracy and boundary consistency. Technical validation against three reference datasets showed strong spatial agreement (IoU = 0.80–0.92) and very high consistency in lake area estimates (R² > 0.99). Uncertainty analysis indicated that most lake-months were classified as having low to moderate uncertainty. From 2001 to 2025, lakes on the TP expanded in both number increasing by 20.6 % and total area expanding by 22.9 %, despite interannual variability. Small lakes contributed to the increase in lake number, while larger lakes were responsible for most to area expansion. Both metrics showed clear seasonal peaks in late summer. This dataset provides a valuable long-term resource for investigating lake dynamics, hydrological variability, and climate-lake interactions in high-altitude regions.
- Preprint
(1477 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 21 Nov 2026)
- RC1: 'Comment on essd-2026-396', Anonymous Referee #1, 17 Oct 2026 reply
-
RC2: 'Comment on essd-2026-396', Anonymous Referee #2, 18 Oct 2026
reply
This study develops a monthly dataset of lakes larger than 1 km2 across the Tibetan Plateau (TP) for 2001–2025 at 30 m spatial resolution, using a TSRFD framework that integrates MODIS, Landsat, and terrain variables. The dataset has value for investigating long-term lake dynamics, hydrological variability, and climate–lake interactions in high-altitude regions. However, the current manuscript requires further improvement in the description and reproducibility of the methodological framework, the justification of the small-lake enhancement procedures and uncertainty assessment, the rigor of the validation strategy, and the presentation and interpretation of the results.
Specific comments are provided below.
1. Figure 1: NNE first appears in Figure 1 but is not defined in the figure or the accompanying text. Please provide its full name at first mention. In addition, Figure 1 could be improved in terms of visual clarity and layout. The current figure appears overcrowded, and the selection and arrangement of lines and arrows could be refined to better illustrate the workflow and the relationships between individual steps.2. Section 2.1:Please revise the sentences in Lines 77–79 to improve their grammatical correctness and readability.
3. Section 3.1.2:
Several aspects of the preprocessing workflow require further clarification.Lines 105–107: Please provide a more detailed description of the preprocessing procedures applied to different satellite sensors, including cloud and cloud-shadow masking and other relevant steps. The specific datasets, algorithms, and processing procedures should be clearly documented to ensure reproducibility, with appropriate references provided.
Lines 110–112: The rationale for using different masking strategies during model training and prediction needs to be explained. In particular, does the less restrictive mask used during prediction introduce additional unreliable water pixels?4. Section 3.1.3
Lines 114–116: Please specify the spatial resolution of each spectral index and water indicator used in the MODIS-based and Landsat-based workflows. Clarify whether the indices were calculated at their native resolutions or resampled to a common resolution before being used as model predictors.5. Section 3.1.4
The description of the proposed improvements for small lakes is currently too brief to ensure methodological reproducibility. Please provide more detailed explanations of the individual procedures.For each modification, the authors should clarify its implementation, parameter settings, and rationale, as well as explain how it addresses the specific challenges associated with small-lake mapping. A schematic illustration or representative examples comparing the original and improved workflows would also help highlight the methodological contributions and demonstrate the effects of these modifications.
6. Section 3.1.5
The overall two-step downscaling framework is valuable, but the current description is insufficiently detailed to allow readers to fully understand. Both steps require more explicit descriptions of the input data, predictor variables, target variables, model configuration, residual correction, and the rationale behind the proposed procedures. Relevant references should also be provided.6.1. Section 3.1.5.1 Step 1: Downscaling from 500 m to 250 m Using RF Regression with Residual Correction
For example, in Lines 143–145, the term "MODIS monthly water layer" is not sufficiently specific. Please clarify what this layer represents, including its data source, spatial resolution, and how it was generated.Similarly, the terrain variables are described as elevation and slope, but their data sources and spatial resolutions are not specified. Please identify the DEM product used, explain how the slope data were derived, and report the resolutions of the original and processed layers.
Furthermore, terrain variables are used in both Steps 1 and 2. Please clarify whether the same terrain products and spatial resolutions were used in both steps, and explain how these variables were processed and integrated into the respective models.
Overall, the technical implementation and the specific contributions of the individual steps remain unclear. Additional methodological details and relevant references are needed to establish its reproducibility and novelty.
7. Section 3.2.1.4
Please provide a more detailed justification for the coefficients or weights assigned to the individual uncertainty indicators and sub-indices. It is currently unclear how these weights were determined and whether they are based on empirical evidence, statistical analysis, previous studies, or expert judgment.8. Validation:The manuscript contains a section titled "Technical Validation" in Section 3.2, while Section 5 also includes a section with the same title. Please reconsider the section structure to avoid redundancy and establish a clearer distinction between the methodological description of validation procedures and the presentation of validation results.
In particular, Section 3.2.2, "Validation with Reference Datasets", could potentially be integrated with Section 5.2, "Comparison with Reference Datasets". Consolidating these sections may help avoid repetition and improve the overall organization of the manuscript.
9. Accuracy Metrics: The IoU and bias metrics are reported in the discussion of Figure 4, but their mathematical definitions are not explicitly provided in the main text. Please include the corresponding equations and define the relevant variables to ensure that the evaluation metrics are clearly documented and reproducible.
10. Figure Formatting and Language Issues: Please carefully review the formatting of all figures and ensure consistency in coordinate systems, longitude and latitude labels, legends, units, decimal places, symbols, and text formatting in accordance with the journal's requirements.
In addition, the first letter of "Uncertainty" in Line 20 of the Abstract appears to have inconsistent font formatting. Please correct this issue and check the manuscript for similar typographical and formatting errors
Citation: https://doi.org/10.5194/essd-2026-396-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 277 | 81 | 68 | 426 | 81 | 66 |
- HTML: 277
- PDF: 81
- XML: 68
- Total: 426
- BibTeX: 81
- EndNote: 66
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
General Comments:
This manuscript presents a monthly 30 m dataset of Tibetan Plateau (TP) lakes larger than 1 km2 for 2001–2025, derived with a MODIS–Landsat two-step random forest downscaling framework (TSRFD). Such a record would be useful, and I appreciate the authors’ effort in processing a large volume of images and sharing the data openly.
However, I have significant concerns about the added value of the dataset relative to well-established products, and about whether the current workflow and validation can establish the reliability of the monthly signal, which is the main contribution claimed. In particular, several steps that shape the monthly results are not quantified, and the validation is not independent of the production workflow. As addressing these would require substantial additional work, I recommend rejection of the current version.
Specific Comments:
(1) The added value over existing products is not yet demonstrated. Lakes larger than 1 km2 are already covered by existing products, including monthly ones (e.g., Wu et al., 2025; Zhao et al., 2025; Li et al., 2025). The more difficult cases, such as lakes below 1 km2 and especially narrow glacial lakes in shadowed valleys with low signal-to-noise ratio, are beyond the scope of this work. This is understandable, but the manuscript would then need to show quantitatively that it improves on existing monthly products for the same lakes and months. Since the 1–5 km2 class already needs many size-specific adjustments (Sect. 3.1.4) and has the highest uncertainty (Fig. 3), I am also unsure how well the approach would transfer to smaller lakes.
(2) The permanent-water merging step raises questions about the stability of the monthly signal. Step 3 merges an annual permanent-water mask into each monthly map to “improve temporal stability and reduce the omission of stable water bodies” (L173–174). While occurrence-based gap filling is common, this implies that the RF output alone may miss even the most stable water pixels, so I am cautious about its performance on shoreline pixels that carry the monthly variation. Moreover, since the merge is a one-way union based on whole-year information, the monthly maps, particularly in ice-covered months with few valid observations, may partly reflect the annual extent rather than that month's conditions. A related, more minor observation is that in Fig. 5d the lake number decreases from October to December (by ~50) while the total area increases (by ~250 km2). This could arise if smaller lakes are missed under ice while the merged permanent water prevents larger lakes from shrinking. I suggest providing the merge fraction for each lake-month as a quality flag, and comparing raw RF and final time series for representative lakes.
(3) The validation is not fully independent and does not test the monthly dynamics. Lake domains and LakeIDs come from Zhang et al. (2019), and boundaries are refined against Zhou et al. (2024) (L195–197); these same datasets then serve as references. Retaining only matches with IoU > 0.70 may inflate the reported IoU, the comparison is largely limited to October (the month with the lowest uncertainty, Fig. 3a), and R2 > 0.99 is dominated by the largest lakes. An independent, size-stratified assessment across seasons (e.g., using manually delineated high-resolution images or altimetry-derived water levels) would be more convincing.