the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 19 Jan 2026
LI-CCR: Dataset of daily lake ice evolution (2002–2024) across global cold climate regions based on gap-filled MODIS observations
Abstract. Lake ice is an essential component of the terrestrial cryosphere and plays an important role in the socioeconomic and ecological systems of cold regions. However, existing lake ice datasets generally suffer from poor temporal continuity, limited spatial coverage, and a lack of observations for small and medium-sized lakes. Consequently, the global spatial patterns and long-term trends of lake ice have remained insufficiently understood. In this study, using MODIS observations, we developed global datasets of daily lake ice coverage (LIC), annual lake ice-cover status, annual lake ice phenology (LIP), and the probability of complete ice-cover occurrence (PCIO), including 32,800 lakes across global cold climate regions from 2002 through 2024. Validation against multiple remote sensing datasets demonstrated high accuracy and confirmed that our Lake Ice – Cold Climate Regions dataset (LI-CCR) effectively captures the spatiotemporal evolution of the lake ice zone. Quality assessments were conducted for both LIC and LIP, and the information was included as an integral component of the LI-CCR. The results show that most cold regions lakes in the Northern (Southern) Hemisphere freeze before November (May) and melt from May (November) onward, with an average ice cover duration of about 200 days. Presently 89 % of the lakes experience completely frozen annually, and the number of intermittent or even ice-free lakes is increasing under the warming climate. The LI-CCR dataset provides comprehensive records of lake ice evolution in cold climate regions and offers valuable information for studies on lake–climate interactions, cryosphere changes, and ecosystem responses. This LI-CCR dataset v2 is freely available from the Zenodo platform at https://doi.org/10.5281/zenodo.17687698 (Jiang et al., 2025b).
- Preprint
(2632 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 14 Jul 2026)
- RC1: 'Comment on essd-2025-721', Anonymous Referee #1, 14 May 2026 reply
-
RC2: 'Comment on essd-2025-721', Anonymous Referee #2, 12 Jul 2026
reply
Thanks for inviting me to review this manuscript. The authors developed a daily lake ice cover and lake ice phenology dataset for 32,800 lakes in global cold regions from 2002 to 2024 using MODIS data. The main contribution is the use of a cloud-gap-filling method to improve the temporal continuity and spatial coverage of MODIS lake ice observations. This study provides a new remote-sensing lake ice dataset covering a large number of lakes. However, as a data description paper, the current manuscript lacks sufficient validation and description of data accuracy, uncertainty, quality-control information, and applicability. I therefore recommend major revision.
Major comments
1. The statement in Lines 75–76 is inaccurate. Previous studies have not focused only on large lakes, and several recent remote-sensing studies have included large numbers of small and medium-sized lakes. The authors should include the relevant recent literature, revise this statement, and more clearly explain the additional contribution of this study compared with existing large-scale lake ice cover and lake ice phenology products.
2. The validation of LIP mainly relies on a passive microwave product covering 56 large lakes and a model-based product covering 132 lakes on the Tibetan Plateau. This is not sufficient to validate a dataset covering 32,800 lakes. The authors should further compare their results with GLRIPD observations and other large-scale Landsat and MODIS lake ice products, with particular attention to the small and medium-sized lakes emphasized in this study. For example, Wang et al. (2021) (High‐Resolution Mapping of Ice Cover Changes in Over 33,000 Lakes Across the North Temperate Zone) derived lake ice cover for more than 33,000 lakes using Landsat data, while Wang et al. (2022b) (Continuous loss of global lake ice across two centuries revealed by satellite observations and numerical modeling) investigated lake ice phenology for more than 30,000 lakes using MODIS data and numerical simulations. Other global or regional datasets covering small and medium-sized lakes may also be available and should be considered. Comparisons with other large-scale lake ice model products would also be more representative than relying only on a regional model product for the Tibetan Plateau. The validation should not be limited to overall statistical metrics. Spatial patterns, temporal variations, and differences among lake sizes, latitudes, climate regions, and lake types should also be presented.
3. The NDSI threshold and the visible and near-infrared reflectance constraints described in Section 3.2.1 are highly similar to the standard snow and ice classification procedure used in the MOD10A1/MYD10A1 products. The authors must clarify how their classification differs from the existing MOD10A1/MYD10A1 classification and what additional improvement is achieved by reclassifying the MODIS observations. In addition, the Data section only introduces MOD10A1/MYD10A1 and does not identify the source of the original spectral bands required to recalculate NDSI and apply these thresholds. It is therefore unclear whether the procedure described in Section 3.2.1 was actually implemented in this study or merely restates the product algorithm.
4. The cloud-gap-filling assessment in Table 2 includes only several lakes in Asia and may not represent its performance across global regions and lake types. The authors should evaluate the filling proportion and reconstruction performance across representative regions, climate zones, lake sizes, and freeze–thaw stages. The contribution of each processing step should be readily available from the existing results, and such a stratified assessment is necessary to support the global applicability of the dataset.
5. Terra and Aqua use similar sensors, spectral bands, and classification methods and may therefore share common errors. Their comparison can be used to assess cross-sensor consistency and the internal performance of the algorithm, but it cannot independently validate the accuracy of cloud-gap filling. The authors should use different types of sensors or independent reference data for further validation. The sample information and spatiotemporal distribution of the current cloud-gap-filling validation are also insufficiently described.
6. The discussion of algorithm performance and errors relies mainly on the Qinghai Lake case study. Qinghai Lake is a high-altitude plateau lake with distinctive environmental conditions and may not adequately represent lakes in other climate regions or lake types. A reduction in the number of cloudy pixels also does not necessarily indicate that the reconstructed values are accurate. The current analysis therefore does not constitute a complete error assessment.
The authors should evaluate how errors in daily LIC propagate into LIP dates and conduct sensitivity analyses to justify the use of the 10% and 90% LIC thresholds and the three-consecutive-day rule. The current accuracy assessment focuses mainly on cloud cover and cloud-gap filling, while other potential error sources, including thin ice, wet snow, turbid water, sun glint, terrain shadows, and shoreline mixed pixels, are insufficiently discussed.
Stratified validation should be conducted across different climate regions, latitudes, lake sizes and types, and freeze–thaw stages. Accuracy differences among seasons and years should also be evaluated to demonstrate the spatial and temporal consistency and reliability of the dataset.
Minor comments
1. Section 3.5.1 is more appropriately described as an uncertainty estimate based on different data sources and processing steps, while Section 3.5.2 mainly provides data-source information and quality-control flags for LIP. Neither section constitutes a strict quality assessment. More accurate terminology should be used.
2. Line 330 states that the cross-sensor validation used MODIS data from 2008, whereas the caption of Figure 4 states 2015. Please verify and correct this inconsistency.
3. In Line 535, the lake-area threshold should be written as “2 km2” rather than “2 km.” The manuscript also contains several errors in abbreviations and wording and should be carefully checked throughout. For example, Section 3.4.3 lists the four lake ice phenology events as “FUS, FUS, BUS and BUE,” where the second FUS should be FUE. The use of PCIO and PICO should also be made consistent.
Citation: https://doi.org/10.5194/essd-2025-721-RC2
Data sets
LI-CCR: Dataset of daily lake ice evolution (2002-2024) across global cold climate regions based on gap-filled MODIS observations Zhengxin Jiang et al. https://doi.org/10.5281/zenodo.17687698
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 390 | 210 | 34 | 634 | 59 | 99 |
- HTML: 390
- PDF: 210
- XML: 34
- Total: 634
- BibTeX: 59
- EndNote: 99
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
General comments:
The manuscript presents a global dataset, LI-CCR, of daily lake ice coverage (LIC), annual ice status, lake ice phenology (LIP), and the probability of complete ice-cover occurrence (PCIO) for 32,800 lakes across cold-climate regions from 2002 to 2024. Given that lake ice is recognized as an Essential Climate Variable (ECV), this long-term, high-frequency dataset improves the representation of small and medium-sized lakes. The manuscript is generally well organized and clearly written. I only have a few minor comments related to the discussion of uncertainties and the potential future applications of the dataset.
Specific comments:
1. The validation section could further illustrate how lake size influences the performance and uncertainty of different observation methods. For example, methods such as passive remote sensing may show larger uncertainties for small lakes, as noted in the Introduction.
2. The discussion could mention potential biases related to wind speed, which may influence ice formation and/or breakup processes and remote-sensing detection during transitional periods.
3. It would be helpful to further expand the discussion on potential future applications of this dataset, such as its use in climate studies, lake modeling, and lake-atmosphere interaction research.
Technical corrections:
-Line 26: The link of dataset might contain an extra space that should be removed.
-Line 83: It might be helpful to specify the area range of the 32,800 lakes (e.g., >2 km2).
-Line 310: "PICO" should be corrected to "PCIO".
-Eq. (3): "Nremove" should be corrected to "Nremoved".
-Fig. 6 caption: A ":" is missing after “Figure 6”.