| 21 Jan 2026
Daily melt pond and net ice surface fractions in the Arctic Ocean from MODIS visible imagery: 2000–2024
Abstract. We report on a novel data set of the melt-pond fraction on Arctic sea ice. Melt ponds on Arctic sea ice are an important phenomenon of the summer-melt process. They reduce the surface albedo of sea ice substantially, by that influence the net shortwave radiation balance, and with that the amount of solar radiation energy that is received by the sea ice-ocean system in the Arctic during summer. This has also implications for under-ice biogeochemical processes and ice mechanics. Melt ponds have been observed by a number of satellite sensors, mostly in the optical and near-infrared wavelength range. Here we present an updated version of a spectral un-mixing approach published earlier that led to a data set of melt-pond fraction on Arctic sea ice with 8-daily sampling for months May through August from 2000 through 2011. The approach is based on reflectance measurements of channels 1, 3 and 4 of the Moderate Resolution Imaging Spectroradiometer MODIS on board the Earth Observation Satellite (EOS) Terra satellite. We modified the approach and derived the daily melt-pond fraction on Arctic sea ice at 500 m and at 12.5 km grid resolution for months June through August from 2000 through 2024 from MODIS v6.1 observations. In addition, we provide the net ice surface fraction – aka the fraction of sea ice without melt ponds – and the fraction of open water between the ice floes. Our MODIS melt-pond fraction agrees within -3 % to +4 % with independent estimates of the melt-pond fraction from very-high resolution optical satellite imagery and from Operation Ice Bridge Digital Camera System imagery. The MODIS open-water fraction we find to be too small by 2 % to 6 %; the net ice-surface fraction tends to be too large by 2 % to 9 %. The 12.5 km gridded product shows a slightly worse (by 1 %) agreement. While our 12.5 km gridded MODIS product under-estimates the melt-pond fraction from very-high resolution optical satellite imagery by about 2 % in the mean (median: 3 %), the Medium Resolution Imaging Spectrometer (MERIS) product over-estimates these independent estimates by about 8 % (median: 9 %). Our MODIS melt-pond fraction data set is available from Sadikni and Kern (2025): https://doi.org/10.25592/uhhfdm.18069.
This manuscript presents a new dataset of melt-pond fraction on Arctic sea ice, extending the original product time series from 2011 to 2024. While the extension of the data record is potentially useful, there are several major issues that must be addressed before the manuscript can be considered for acceptance in Earth System Science Data.
Currently, the manuscript lacks sufficient verification and discussion regarding data quality and associated uncertainties. The description of algorithmic improvements is not detailed enough, and there is a lack of clarity in the technical descriptions. Furthermore, the terminology used to describe data products is inconsistent, leading to confusion. To meet the standards of ESSD, the authors must provide a more rigorous validation and clearly articulate the added value of this dataset to the research community.
Major Comments
Minor Comments
L18: Please specify the metric used for the "-3% to +4%" range (e.g., bias, error margin).
L17–L24 (Abstract): Please use uniform product names when stating accuracy performance. I recommend replacing "MODIS melt-pond" and "MODIS open-water fraction" with the standard product names defined in the manuscript to avoid confusion. Additionally, spell out "MODIS" upon first use.
L26–L27: Please remove this sentence.
L106–L107: This statement is unclear; please paraphrase for clarity.
L110: This sentence is ambiguous and requires revision.
Section 2.2.1: Please provide more details on the sensors used, including spatial and temporal resolution and the standard product name. Please avoid using authors' names to represent datasets.
Section 2.2.3: Please specify the spatial resolution and time series coverage.
Eq. 1: Please provide a brief description of the condition r = 1.
Table 1: How were the reflectance values for melt ponds, snow/ice, and open water determined? Given that reflectance varies with viewing geometry, please discuss in the Discussion section how this might introduce uncertainties into the estimation results.
Section 3: Please explicitly list the differences between this "novel" method and the method of Rosel et al. (2012). Describe only the modified parts of the method in detail; other parts should be described briefly.
L214–L220: How was the training dataset for the three classes (melt pond, snow/ice, and open water) determined? Please provide more details.
L222–L224: Why was the structure of the ANN changed relative to Rosel et al. (2012) (3-9-27)? Please provide specific reasons for this modification.
Table 2: Please define "MAD" and "mean." Does "mean" refer to mean bias, difference of mean value, or another metric? Please ensure consistent terminology throughout the manuscript.
Fig. 6: Please add units to sub-figures a), c), and e).
L590–L595: Please validate both the Rosel et al. (2012) product and the product presented in this manuscript to support your comments.
L606: The accuracy difference is relevant to product quality, not solely the number of evaluation data points. Please clarify this argument.
L607: Please specify the methodological differences and highlight the pros and cons of each method.
L687: Please ensure terminology accurately reflects the level of innovation. This data is "new" in terms of temporal coverage, but the term "novel" implies methodological innovation, which should be used cautiously.