Open AccessArticle

Spatiotemporal Variations in Sea Ice Albedo: A Study of the Dynamics of Sea Ice Albedo in the Sea of Okhotsk

Yingzhen Zhou

^1,*,†

Wei Li

^1,†

Nan Chen

^1,†

Takenobu Toyota

Yongzhen Fan

Tomonori Tanikawa

⁴

and

Knut Stamnes

^1,†

Department of Physics, Stevens Institute of Technology, Hoboken, NJ 07307, USA

Institute of Low Temperature Science, Hokkaido University, Sapporo 060-0808, Japan

Cooperative Institute for Satellite Earth System Studies (CISESS), Earth System Science Interdisciplinary Center (ESSIC), University of Maryland, College Park, MD 20740, USA

⁴

Meteorological Research Institute, Japan Meteorological Agency, Tsukuba 305-0052, Japan

Author to whom correspondence should be addressed.

^†

Current address: Light and Life Laboratory, Burchard Building, Stevens Institute of Technology, Hoboken, NJ 07307, USA.

Remote Sens. 2025, 17(5), 772; https://doi.org/10.3390/rs17050772

Submission received: 22 December 2024 / Revised: 11 February 2025 / Accepted: 14 February 2025 / Published: 23 February 2025

(This article belongs to the Special Issue Monitoring Sea Ice Loss with Remote Sensing Techniques)

Download

Browse Figures

Versions Notes

Abstract

This study utilizes a novel albedo retrieval framework combining radiative transfer modeling with scientific machine learning (RTM-SciML) to investigate sea ice dynamics in the Sea of Okhotsk. By validating albedo estimates derived from the MODIS sensor against in situ pyranometer measurements near the Hokkaido coast, we achieved a robust Pearson coefficient of 0.86 and an RMSE of 0.089 for all sea ice types, with even higher correlations for specific surfaces like snow-covered ice (Pearson-r = 0.89) and meltwater/open water (Pearson-r = 0.90). This confirms the framework’s efficacy across varying surface conditions. Cross-sensor comparisons between MODIS and the Second-Generation Global Imager (SGLI) further demonstrated its consistency, achieving an overall Pearson-r of 0.883 and RMSE of 0.036. Integrating these albedo estimates with sea ice concentration data from the Advanced Microwave Scanning Radiometer 2 (AMSR-2), we analyzed the complex role of the Sea of Okhotsk’s polynya systems and ice interactions in regional climate processes. Our results highlight the synergistic advantage of pairing optical sensors, like MODIS and SGLI, with microwave sensors, offering a more comprehensive understanding of evolving sea ice conditions and paving the way for future climate and cryosphere studies.

Keywords:

sea ice; albedo; sea ice contentration; Sea of Okhotsk

1. Introduction

The Sea of Okhotsk, shown in Figure 1, is the southernmost marginal sea in the northern hemisphere with seasonal ice cover. Unlike polar regions, its lower latitude and warmer climate give rise to unique sea ice properties. Sea ice primarily forms in coastal polynyas due to significant air–sea temperature differences and offshore winds [1], before drifting southward where it melts [2]. Understanding the impact of ice advection and interactions within the polynya region remains challenging.

Surface albedo is a pivotal physical parameter for the cryosphere, governing the reflectivity of surfaces comprising snow, sea ice, meltwater, open ocean, and their mixtures. While primarily determined by surface type, albedo can also be affected by external influences such as biological activity (e.g., algal blooms) and anthropogenic pollution (e.g., black carbon and soot deposition), which further modulate surface reflectivity. It plays a crucial role in controlling the radiative energy budget, especially in regions like the southern Sea of Okhotsk [3]. Albedo measurements are central to estimating heat fluxes and sea ice production. However, due to the limited amount of observational data [4], it was difficult to extend these results to the whole region and other seasons since the measurements were obtained only in the ice growth season, limiting comprehensive year-round understanding.

Existing satellite-based retrieval products further exacerbate this challenge. Land-focused albedo retrievals, such as the MODIS MCD43 product [5,6], are restricted to coastal regions due to their reliance on bi-directional reflectance distribution functions (BRDFs) designed for land surfaces, leaving the Sea of Okhotsk blank. Similarly, the melt-pond detection algorithm [7,8] applied to OLCI and MERIS sensors is designed for summer melt conditions in the Arctic and does not provide retrieval values for seasonal ice regions at lower latitudes. The direct-estimation GLASS product (which ceased operation in 2019) and its successor, the VIIRS Land Surface Albedo product [9,10], also fail to provide retrievals over sea ice in the Okhotsk region due to methodological limitations.

In our previous work [11], we developed an albedo retrieval framework combining radiative transfer modeling with scientific machine learning (RTM-SciML), which proved highly effective in Arctic sea ice regions. This study addresses the gap in marginal seas by applying the RTM-SciML framework to the Sea of Okhotsk. The framework’s coupled radiative transfer model is uniquely suited to handle the region’s complex, mixed surfaces, including snow-covered ice, bare ice, meltwater, and open ocean.

Incorporating unique validation data from irradiance measurements aboard the PV Soya icebreaker, our study overcomes the scarcity of ground truth data in the Sea of Okhotsk. By comparing retrievals from two satellite sensors, MODIS and the relatively underutilized Second-Generation Global Imager (SGLI), we provide insights into cross-sensor performance and robustness. Through the analysis of the spatial and temporal evolution of surface albedo combined with sea ice concentration, this work demonstrates how multi-sensor fusion and coupled modeling can address longstanding gaps in albedo estimation for dynamic, seasonal sea ice environments.

2. Data and Models

2.1. Albedo Retrievals and Surface Classification Using SGLI and MODIS Radiance Data

In our previous research, we introduced the RTM-SciML framework—a fusion of Radiative Transfer Model (RTM) simulations and Scientific Machine Learning (SciML) techniques [11]. This algorithm focuses on retrieving the albedo of cryospheric surfaces by processing top of the atmosphere (TOA) radiances captured from specific satellite channels. The RTM-SciML framework can be briefly summarized as follows:

Using a coupled atmosphere–surface RTM [12] a synthetic dataset (SD) was constructed, consisting of TOA radiances simulated in suitable satellite channels as a function sun-satellite geometry angles, as well as the associated broadband albedo at the surface. This SD encompasses a range of different surface types, including bare ice, snow-covered ice, melt water, as well as their mixtures.
All optical properties of surface and atmospheric constituents, as well as radiative processes within the coupled atmosphere–snow–sea ice–water system, are deduced from first principles. Information about both the surface BRDF and the inherent optical properties (IOPs) of the atmosphere as well as the underlying sea ice and water is implicitly taken into account. Hence, there is no need to perform explicit atmospheric corrections.
Based on the physically consistent SD, a scientific machine learning (SciML) model was used to approximate the solution to the “inverse problem”, which answers the following question through an implicit iterative process (i.e., by minimizing a loss function repeatedly during the training of the SciML model): “given the observed TOA radiance and the sun-satellite geometry angles, what is the most likely albedo of the sea ice surface?”
The SciML model does not rely on predefined spectral reflectance threshold values for individual surface types, which eliminates errors caused by incorrect surface condition assumptions. When deployed in practice, surface classification and albedo retrieval are separate processes. An independent machine learning classification mask (MLCM) performs both cloud screening and surface classification.

Our SciML model was trained on a dataset representing diverse surface types such as snow-covered sea ice, bare ice, melt water, and open ocean water. The sea ice thickness was also taken into account. Although other sea ice types, such as pancake ice, exist in this area, we think it is reasonable to assume that the optical properties and physical thicknesses of any “missing” ice types will be covered by the “bare ice” category considered in this study. Therefore, we believe that our simplified approach is sufficient to provide an adequate description of the radiative characteristics of this complex mixture of sea ice (with and without melt water and snow cover) and open ocean water encountered in the Sea of Okhotsk. Hence, our conjecture is that the universality of these physical and optical properties implies that the algorithm’s application is not restricted by regional constraints. Instead, the main consideration was the range of sun-satellite geometric configurations and atmospheric parameters, ensuring the algorithm’s robustness across a variety of observational conditions. The comprehensive range of solar zenith angles, sensor polar angles, and relative azimuth angles considered makes the dataset inclusive of conditions seen in both the Arctic and the Sea of Okhotsk (refer to parameters in Table A4 in [11]).

Traditional retrieval algorithms falter in the face of the complex and dynamic landscapes of polynyas, which feature an ever-shifting mix of snow-covered ice, ice covered with surface meltwater, and open ocean water. The conventional approaches (e.g., [7]) require surface-type classification as a precursor to the retrieval process. In stark contrast, our RTM-SciML methodology negates this preliminary step. We employ a Machine Learning Classification Mask (MLCM) for cloud screening and surface classification, where its primary role is to refine accuracy by focusing on clear-sky observations and thereby discarding data marred by cloud-induced distortions. Notably, the albedo retrieval process is executed independently of the surface-type identified by the MLCM. This distinction proves invaluable when navigating the Sea of Okhotsk’s polynyas. Here, clear demarcations between thin ice and refrozen melt-water or slush blur into obscurity, especially since both sets exhibit near-identical albedo values.

In this paper, the albedo retrievals utilize radiance data from the Second-Generation Global Imager (SGLI) on the Global Change Observation Mission-Climate (GCOM-C) satellite and the Moderate Resolution Imaging Spectroradiometer (MODIS) on the Terra and Aqua satellites. SGLI and MODIS radiance data from January to May 2021 were analyzed, covering key episodes of the sea ice growth and melt phases.

Table 1 lists the central wavelengths used for albedo retrievals in this work. While MODIS (Collection 6.1) provides native 1 km resolution for these bands, SGLI channels have varying resolutions (250 m to 1 km). For consistency, the 250 m SGLI channels were aggregated to match the 1 km grid. Terra’s MODIS sensor, with overpass times approximately 15 min apart from GCOM-C, complements the SGLI observations, enabling enhanced temporal coverage for monitoring rapid sea ice changes under clear-sky conditions.

For a detailed description of the methodology, including the specifics of channel selection, step-by-step process and visual flowchart, please refer to [11].

2.2. Validation Data

The validation data for this study were obtained from irradiance measurements onboard the PV “Soya”, facilitated by an EKO MR-40 four-component radiometer (manufactured by ECO company in Tokyo, Japan). These measurements were taken in February during the time period between 2002 and 2015 (except for 2010 and 2011). The pyranometer was strategically positioned at the tip of a 3 m long ladder at the bow of the vessel. A visual representation of the Soya Icebreaker’s navigational routes, and the location of the pyranometer on the ship is provided in Figure 2.

The collected data includes both upward and downward longwave (3–50

μ

m) and shortwave (305–2800 nm) irradiance data. This study primarily emphasizes the shortwave irradiance data, from which the shortwave albedo is derived. These measurements were consistently recorded at one-minute intervals.

Differing from the localized footprint of the pyranometer measurements, the satellite-derived albedo has a spatial resolution of 1 km. Using a K-dimensional Tree (KD-Tree) algorithm, each day’s pyranometer readings were mapped to their nearest corresponding MODIS pixels. For each identified MODIS pixel, the KD-Tree algorithm located the 30 closest measurement points. Their mean value was computed and then associated with the respective MODIS pixel.

Fog is a prevalent phenomenon in many coastal regions, especially in areas like the Sea of Okhotsk where polynya or sea ice is present. Given that the Soya voyages occurred in the coastal region of Hokkaido, this area was no exception to this common occurrence. The heat and moisture exchange between the ocean and the atmosphere in the area leads to the formation of localized clouds or fog. By utilizing the MLCM, cloud and fog-affected MODIS pixels were identified and excluded. The matching validation data of these pixels were also removed. This approach ensures that only uncontaminated satellite retrievals are considered for validation. It is worth noting that due to potential time differences between the MODIS overpass and on-site measurements, the precise conditions at the time of the pyranometer measurements (e.g., overcast or foggy) are not always known.

Ultimately, the validation dataset was refined to focus on three surface categories: bare sea ice, snow-covered sea ice, and water (including melt-water), resulting in a total of

N =

911 validation data points.

3. Validation Results

In our previous work [11], validation of MODIS-retrieved sea ice albedo using the RTM-SciML method against pyranometer measurements indicated a strong correlation with

\hat{α} = 0.91 α + 0.06

(

\hat{α}

denotes the satellite retrieval and

α

denotes the pyranometer measurements), Pearson-r = 0.9, and RMSE = 0.099, evaluated on

N =

7964 data points. Such results exemplified the effectiveness of the RTM-SciML method in retrieving sea ice surface albedo.

However, in the Sea of Okhotsk, the correlation, with

\hat{α} = 0.79 α + 0.04

, Pearson-r = 0.8, RMSE = 0.11 (Figure 3a), is not as robust as observed in the Arctic. The Sea of Okhotsk, distinguished by its dynamic coastal polynya and young, thin ice, presents unique challenges. Rapid alterations in young ice within polynya, influenced by atmospheric conditions, tidal activity, and ocean currents, often lead to swift ice formation and melting.

To elucidate the potential discrepancies between the MODIS retrievals and Soya measurements, we turn to specific case studies as illustrated in Figure 4 and Figure 5. On 10 February 2004 (Figure 4), around midday local time (UTC + 9 h), the Soya measurements show a low value (<0.2) in contrast to the MODIS retrieval, which is >0.2. This discrepancy could be due to sea ice observed by MODIS at an earlier time that might have melted or drifted away by the time Soya measured the same location. Another possible reason would be the impact of a change in solar zenith angle on the albedo due to the time lag: a higher solar zenith angle implies a higher albedo. Similarly, the difference seen on the top-right is also due to this time lag. The MODIS retrievals are capturing regions where the sea ice has already melted into slush or meltwater, which naturally has a lower albedo. This results in measured albedo values ranging from 0 to 0.25 but satellite-retrieved values around 0.07∼0.08; hence, the apparent cut-off line around 0.07∼0.08 albedo in Figure 3.

On 11 February 2008 (bottom of Figure 5), a significant difference is evident during the local morning. While Soya measurements suggest a snow-covered sea ice surface (with measurements

α \geq 0.75

), the MODIS data identifies the area as bare sea ice (

\hat{α} \approx 0.5

). It is possible that the intervening time between Soya measurements and MODIS retrievals allowed for the surface snow to melt. Additionally, measurement inaccuracies could come into play. Given the pyranometer’s positioning on the vessel, at certain times, especially around 23:30–00:30 UTC, it could fall within the ship’s shadow, causing an inaccurate measurement of irradiance. This trend, where in situ measurements consistently show higher albedo values than MODIS retrievals during this time window, can be observed across multiple panels in Figure 6.

Moreover, as evident in Figure 6, the closer the measurements are to the MODIS overpass time (represented by the dotted lines), the smaller the discrepancy between in situ measurements and satellite retrievals. As the time difference increases, potential gaps between the two sets of measurements can also occur.

In light of these observations, for a more precise validation, we confined the time gap between the MODIS transit and Soya measurements to a maximum of one hour. This decision helps mitigate, to some extent, potential discrepancies arising from swift ice dynamics and other transient factors, though such uncertainties are not fully eliminated. The refinement resulted in the exclusion of approximately 570 data points, retaining 335 for a more dependable validation. The statistics for this refined dataset, segmented by different surface types, are presented in Table 2.

After imposing the time constraint, the improved correlation, particularly in the “All sea ice” row, confirms the retrieval algorithm’s applicability to not only Arctic sea ice but also polynyas and sea ice in the Sea of Okhotsk. This achievement is notable given the distinct characteristics of polynyas and the inherent challenges they pose for satellite retrievals.

4. Comparison Between MODIS and SGLI Retrievals

The SGLI and MODIS sensors utilize different center wavelength channels for radiance measurements, as well as different channel widths and different response functions [13]. As a result, despite applying the same RTM-SciML computational framework for albedo retrieval, differences in outcomes are inevitable. This section seeks to elucidate these variations, taking the results from the MODIS product, which was validated in Section 3, as a reference. The overarching goal is to evaluate the consistency of the RTM-SciML retrieval algorithms when applied to different sensors, and to quantify the observed discrepancies.

Albedo values from both sensors, spanning January to May 2021 over the Sea of Okhotsk region, were analyzed. To ensure a consistent comparison, the data were standardized onto a 1 km grid, allowing for pixel-to-pixel evaluation. Surface classifications from MLCM were used to extract relevant data subsets (open water, sea ice, and snow on sea ice) from both MODIS and SGLI imagery. Within the unified grid, if either SGLI or MODIS MLCM flagged a grid value as ‘cloud’, it was excluded from both datasets, ensuring only clear data comparisons.

For a comprehensive view, albedo retrieval and surface classification data were averaged on a weekly basis. Periods with limited clear-sky coverage were omitted, resulting in eight aggregate weekly images from MODIS and SGLI retrievals. Metrics including Pearson-r and RMSE for valid sea ice pixels during these intervals are provided in Table 3.

It is noteworthy that both the SGLI and MODIS sensors, due to their orbital paths, capture the Okhotsk region in three snapshots daily. However, there is a potential time differential of up to two hours between their captures. This time gap can lead to noticeable variations, especially during rapid events like ice movement. As shown in Figure 7, during colder periods (panels a–c), the correlation between the two is near the 0–1 line. In contrast, during warmer periods (panels d–h), the correlation tends to decrease.

An overall analysis across all eight periods, further segmented by surface type, is depicted in Figure 8. The data indicates a Pearson-r of 0.883 and an RMSE of 0.036, suggesting a close agreement between the two sensors. However, it is observed that SGLI’s albedo values for snow-covered sea ice are consistently higher than those of MODIS. This discrepancy could be attributed to the channel response functions, particularly evident for the short-wave near-infrared channels [13].

5. Spatial and Temporal Distribution of Sea Ice in the Sea of Okhotsk

The Sea of Okhotsk is renowned for its dynamic polynya systems and complex sea ice interactions, which significantly influence both regional and global climate patterns. Albedo, which determines the reflectivity of ice or snow surfaces, and sea ice concentration (SIC) are intertwined, revealing critical details about these dynamics. For instance, while albedo can offer insights into the state of the polynyas, highlighting changes in snow grain size, the presence of contaminants, or melt water forming on ice surfaces, the SIC quantifies the fraction of the ocean area covered by sea ice, illustrating the distribution and evolution of ice within the polynya. Consequently, combining these two parameters allows for a holistic understanding of the dynamics of sea ice in the Sea of Okhotsk. Additionally, since microwave sensors, like the Advanced Microwave Scanning Radiometer 2 (AMSR-2) from which SIC is derived, remain unaffected by cloud coverage, they offer consistent and reliable data complementing the albedo values derived from optical sensors that are compromised under cloudy conditions.

In this section, we leverage the SIC data from AMSR-2 to delve into the spatial and temporal progression of sea ice within the Sea of Okhotsk during the eight-week span detailed in Section 4. The ability of microwave sensors to operate unhindered by cloud coverage enabled daily SIC derivation. These SIC data, in tandem with the albedo retrieved using the SGLI sensor, pave the way for evaluating the potential interplay between these two pivotal physical properties. For simplicity and clarity in representation, a single image was chosen to depict the sea ice condition for each week (Table 4).

As high-resolution (5 km) data were available only since March 2021, the figures in Section 5.3 used the low-resolution (15 km) product as an illustration, but the statistical analysis on SIC was conducted based on the high-resolution data (Section 5.2). Note that the high-resolution product does not provide data for pixels along the coastline.

5.1. Spatial Coverage of Sea Ice in the Sea of Okhotsk

In 2021, the maximum extent of ice coverage was in mid-March. Except for new thin ice, the sea ice is mostly covered with snow (Figure 9(b-4)). In March, the ice shifts from the ‘growth’ phase to ‘melt’ phase. The southern region of the Sea of Okhotsk (at latitudes lower than 55°N) responds faster to the spring warming.

In the east bank of Sakhalin, melting started in late March. Pack ice broke apart and some ice drifted further to the east (and south) by wind and ocean currents, resulting in a melt water area surrounded by sea ice (Sakhalin polynya). In early April, the drifted ice melted away, and the sea ice area shrunk to around a third of that during the time of the peak value. In the higher latitudes (≥57.5°N), despite the increased brightness temperature (see Figure 9) in late March, the sea ice in the northern shelf polynya and Shelikhov Bay continued growing, and the ice concentration in the northeast corner increased to the 90∼100% level. The ice concentration in the northwest polynya above Shantar Islands increased by 10∼20%, which indicates that melt water formed in early March in this region was partly refrozen or experienced drainage (Figure 9(c-4,c-5)).

The ice drift, melting and refreezing were also reflected in the albedo values of the corresponding pixels (Figure 9(a-4,a-6)). In early April, the northwest polynya (NWP, labelled on Figure 1) is a mixture of open water, melt water, bare sea ice with different thicknesses, and snow-covered sea ice. The thin line along the coastline, which was identified as ‘open water’ by the MLCM and first appeared in late March (Figure 9(b-5)) has evolved to a larger area in early April. From the coastline to the southeast direction, the surface albedo values increase from

α \leq

0.3 (melt water or dark nilas), to 0.3

\leq α \leq

0.5 (a mixture of melt-water covered sea ice and new young ice, with 60% ≤ SIC ≤ 65%), and to 0.5

\leq α \leq

0.65 (thicker bare-ice, SIC ≈ 90%). The areas with high albedo (

α \geq

0.7) and high ice concentration (SIC ≥ 90%) are mostly snow-covered sea ice. Snow is in general a better insulator than sea ice, and its high reflectivity also prevents the underlying sea ice from melting.

5.2. Relation Between Sea Ice Albedo and Concentration in the Sea of Okhotsk

In this study, the ice edge for the seasonal ice zone (SIZ) is defined as the 0% ice concentration contour. For a quantitative analysis of the albedo and ice concentration in the SIZ, the 5 km resolution SIC [14] and 1 km resolution retrievals (albedo and the corresponding classification) were all mapped to a 0.2° × 0.2° mesh grid. The SIC was then divided into ten categories with 10% intervals. The ranges of each SIC interval are exclusive on the left and inclusive on the right; SIC = 0% data were not included. Figure 10 shows the average (black line) and standard deviation (error bar) of albedo of the pixels belonging to each SIC category for the 1 April∼7 April period. The bar plots in the backdrop are the surface subtype fractions of snow-covered ice, bare ice, and water (melt water on sea ice surface and open water) in each SIC category.

Extending this procedure to the other dates, we found that the albedo values within each SIC category fluctuate week-by-week. For SIC = 100%, the mean albedo ranged from 0.8 in early March to 0.6 in late April and early May. The difference can be attributed to the properties and conditions of each surface type. Thicker and smooth snow cover as well as thick ice all have higher albedo than thin, bare ice. A larger albedo difference was found for SIC < 80%, which can be attributed to either the thinning (thickening) of bare-ice or a larger (smaller) open-water fraction, or the combined effect of both factors (the ‘ice albedo feedback’ described in [15]).

To analyze the albedo fluctuations of each surface type, data from March through May 2021 listed in Table 4 were all ‘bucketed’ into ten categories based on the SIC level. At each SIC level, the coverage percentage as well as albedo of each subtype were calculated. The albedo of snow-covered sea ice pixels remains stable in the 0.7∼0.8 range, with standard deviation ≲ 0.1 (Figure 11a), whereas the standard deviations of bare ice and melt water albedo at each SIC level (Figure 11b) are a lot higher. Albedo of sea ice without snow cover also has a much wider data range (

0.22 \leq α \leq 0.6

The explanation is two-fold. First, the coastal polynya experiences frequent phase changes from melt water to sea ice. Without the insulation effect from snow, bare ice, being a poor thermal barrier, would more easily become thinner and break. In addition to the deformation, advection and ice drifts also complicate the surface heat flux and influence the albedo.

Second, the variation in ice thickness and solar zenith angle (SZA) in the area of coverage. Ice thickness plays a crucial role in determining the albedo of bare ice. Toyota et al. [4] established a regression relationship between ice concentration and albedo (blue shadow in Figure 11b) based on in situ measurements in the southwestern Sea of Okhotsk. Their study found that incorporating solar zenith angle and ice thickness into the regression model reduced the RMSE from 0.062 (blue shaded region) to 0.05, underscoring the importance of these factors.

As shown in Figure 11b, our retrieved sea ice albedo values align reasonably well with Toyota et al.’s [4] measurements. Furthermore, radiative transfer model simulations indicate that albedo is particularly sensitive to ice thickness when the ice is thin (

h < 0.3

m) and at high solar zenith angles (SZA

> 60^{\circ}

). This finding is consistent with the trends observed in Toyota et al.’s [4] study, reinforcing the role of ice thickness and SZA in shaping the albedo variability of bare ice.

5.3. Spatiotemporal Analysis of Surface Albedo

Two areas in the Sea of Okhotsk were selected to conduct temporal analysis of surface albedo: (1) Tatar Strait and the NWP adjoining the Japan Sea, (2) northern polynya. Sea ice in the two regions last the longest periods and is the last to melt. The coordinates of the selected regions are marked with orange and green boxes in Figure 1.

Figure 12 shows probability density function (PDF) plots of sea ice albedo in the two regions, estimated using a Gaussian kernel probability density [16]. The probability of albedo values falling within the interval

P (a < x \leq b)

is measured by subtracting the two integral values

P (a < x \leq b) = F_{x} (b) - F_{x} (a) = \int_{a}^{b} f (x) d x

, where

F (x)

is the cumulative density function (CDF) and

f (x)

is the PDF. The entire area under the PDF-curve and above the x-axis is equal to one [17]. The distribution of albedo values changes from unimodal in January and February, with peak on the high-value side (indicating a higher percentage of snow and sea ice pixels), to bimodal in March and April (indicating partially melting snow/sea ice). When generating Figure 12, the missing data corresponding to pixels with persistent cloud coverage throughout the week were filled with the albedo values of the nearby pixels that have the same surface type.

Due to large heat loss, coastal polynyas are high sea ice production areas in the Okhotsk Sea. When reaching maximum coverage, the sea ice area in the Tatar Strait is around 3% of the entire Japan Sea [18]. Figure 13 shows the temporal variation in albedo, surface type, ice concentration, and brightness temperature from ice growth (January to early March) to ice melt (mid-March to May) in 2021. The dynamic nature of sea ice coverage, including the rapid transition from the growth phase in early March to the melt phase by mid-March, aligns with the known variability of sea ice area and thickness in this region [19].

Previous studies [20,21,22] indicate that ice south of Shantar Island is generally thicker due to convergent ice motion induced by wind and winter turbulent cooling. During ice formation months (January to March 2021), the V-shaped area between Russia and Sakhalin (left column of Figure 13) exhibited the lowest brightness temperature in the Sea of Okhotsk. The topography and cold water influx from the Amur River promote ice growth. The northwesterly monsoon from Siberia forces newly formed ice to drift towards the southern bank, where it accumulates into thick, ridged ice. Blown snow insulates the ice, increasing surface albedo. These factors extend ice retention, making this region the last to melt each summer. On 10 May 2021 (see Figure 12), sea ice albedo in the northern shelf was approximately 0.5, while around north Sakhalin it was greater than 0.6.

A stark contrast in albedo values exists above and below the 52°N line (first row in Figure 13). The reason is that the narrow Nevelskoy Strait tends to prevent ice exchange between Tatar Strait and the Sea of Okhotsk. In April, when sea ice in the Tatar Strait nearly disappears, albedo north of the Nevelskoy Strait remains high.

The albedo increase in the V-shaped area in early March (area with

α > 0.85

in Figure 13c) was likely due to new snowfall. In our RTM simulation, with SZA = 60° and an optically thick snow layer (

h_{s} = 20

cm), an albedo of this high value requires pure white snow with an effective grain size of

r_{e} \approx 50 μ

m and a black carbon impurity volume fraction of

f_{imp} = 1 \times 10^{- 8}

6. Discussion

For the first time, this study provides a comprehensive evaluation of the spatial and temporal variation in surface albedo throughout a complete seasonal ice cycle in the Sea of Okhotsk, from January during initial ice formation to May when the ice melts. This was made possible through the application of the RTM-SciML albedo retrieval framework, which overcomes the limitations of existing operational satellite-retrieval methods [5,8,10], that are either poorly suited or inapplicable to marginal seas like the Sea of Okhotsk. The inclusion of in situ validation data from the PV Soya icebreaker addresses the historical lack of ground truth data, providing reliable reference points across different surface types.

Surface heat flux estimation in regions like the Sea of Okhotsk typically relies on SIC, atmospheric variables such as air temperature and humidity, and assumptions about ice surface conditions [15,23]. However, SIC alone quantifies only the extent of ice cover and does not account for surface-type variability—whether it is snow-covered ice, bare ice, or ice with meltwater—or for critical physical properties like ice and snow thickness, which strongly influence heat absorption. Ref. [24] demonstrated that variability in these surface properties can cause significant differences in heat flux, even when SIC remains high, emphasizing the need to include these factors for accurate energy balance estimations. Figure 10 and Figure 11 illustrate how albedo values can vary substantially within the same SIC range due to differences in surface composition, underscoring the importance of incorporating surface property (albedo) information into heat flux models.

This limitation is particularly critical during transitional periods, such as in coastal polynyas and regions experiencing ice advection, where surface conditions change rapidly, and the composition of snow-covered and bare ice can shift within short time frames. Dynamic processes like melt pond formation, refreezing, and snow accumulation further complicate the estimation of heat flux, making the inclusion of albedo a necessary component for capturing these rapid changes.

Integrating albedo retrievals into SIC-based models enables more accurate identification of mixed or heterogeneous surface conditions, which are characteristic of the Sea of Okhotsk’s polynya systems. This advancement enhances flux model accuracy and better captures the ice-ocean feedback mechanisms, which are essential for climate studies focusing on energy transfer and long-term monitoring.

The seasonal variation in albedo presented in this study establishes a baseline for understanding how surface reflectivity evolves during the ice growth and melt cycles in the polynya. This knowledge can inform future studies investigating regional heat flux estimates and their impact on seasonal ice dynamics in the Sea of Okhotsk, where observational data have historically been scarce.

7. Concluding Remarks and Future Perspectives

In this study, we applied the RTM-SciML albedo retrieval framework to the Sea of Okhotsk, a region with dynamic seasonal ice cover and complex polynya systems. By validating MODIS retrievals using decadal in situ irradiance data from the PV Soya icebreaker and comparing them to SGLI retrievals, we demonstrated the framework’s accuracy and adaptability across different sensors and ice surface conditions. Validation efforts yielded a Pearson coefficient of 0.83 and an RMSE of 0.097 over 325 data points, underscoring its robustness in monitoring spatially heterogeneous sea ice.

An important contribution of this work is the integration of surface albedo and sea ice concentration (SIC), which provides a more detailed view of reflectivity variations across different ice types and seasonal transitions. Incorporating this variability into energy balance models is particularly important during transitional periods when rapid changes in surface conditions occur. Our results demonstrate the potential for improving heat flux estimates by capturing these critical variations. Furthermore, with the albedo retrieval product applicable to the Sea of Okhotsk, it is now possible to conduct in-depth analyses of the ice melt and refreeze phases—an area where SIC data alone has historically fallen short. The observed effects of parameters such as ice thickness and solar zenith angle on albedo values are consistent with previous findings, reinforcing the robustness and scientific significance of this work within the broader field of climate studies.

We also demonstrated the potential for multi-sensor integration by combining data from “optical” sensors (MODIS and SGLI) with microwave-based SIC retrievals (AMSR-2). The RTM-SciML framework’s adaptability across multiple sensors highlights its potential as a valuable tool for monitoring sea ice variation across different timescales, leveraging the varying overpass times of each satellite under clear-sky conditions. This capability is critical for understanding ocean–ice–atmosphere interactions in regions with dynamic ice behavior.

Despite these contributions, several limitations remain:

1. Sensor-Specific Training and Optimization: Each new satellite sensor requires the careful selection of spectral channels to construct a suitable synthetic training dataset using radiative transfer models. Channels prone to saturation or atmospheric interference must be avoided, and identifying optimal channels is not always straightforward. Improvements in channel characterization through pre-launch calibration and inter-sensor cross-validation can help mitigate this limitation.

2. Consistency Across Sensors: Although the RTM-SciML framework demonstrated adaptability across MODIS and SGLI, the retrieval results from different sensors are generated using sensor-specific models due to variations in spectral response functions. This introduces potential discrepancies when comparing results across sensors. Developing harmonized training datasets and incorporating inter-sensor correction mechanisms could improve consistency in multi-sensor integration.

3. Uncertainty in Bare Ice Retrievals and Limited Validation Data in Sea of Okhotsk: Our validation efforts primarily rely on geographically and temporally limited in situ data from the PV Soya icebreaker, covering the Hokkaido region during February. While the retrievals generally show reasonable agreement under stable surface conditions, bare sea ice retrievals exhibit higher variability compared to other surface types due to dynamic processes within polynya regions. Factors such as rapidly changing ice thickness and surface conditions, along with shifts in the solar zenith angle and melt-refreeze cycles, contribute to the observed discrepancies. Expanding in situ datasets—particularly for bare ice in more stable conditions—through collaborations with field campaigns or autonomous measurement platforms would help refine retrieval accuracy and improve our understanding of retrieval performance across varying ice conditions.

4. Potential Future Integration with VIIRS Data: As MODIS nears the end of its operational lifetime, VIIRS represents a key resource for future albedo retrievals. Since the current VIIRS Land Surface Albedo product focuses on land surfaces and lacks retrievals for sea ice in the Okhotsk region, extending the RTM-SciML framework to this sensor would address a major gap in current albedo monitoring capabilities. This will require careful optimization of spectral channel selection and the development of a new training dataset. Given its improved spatial resolution and global coverage, a VIIRS-based sea ice albedo product could significantly improve heat flux estimates and climate modeling for marginal seas.

By addressing these limitations, future research can enhance the generalizability and accuracy of albedo retrievals across a broader range of ice conditions and sensor platforms. The combination of optical and microwave sensors, as demonstrated in this study, offers a promising path toward comprehensive monitoring of seasonal ice dynamics and their impact on regional and global climate systems.

Author Contributions

Conceptualization, W.L. and Y.Z.; methodology, software, validation, formal analysis, investigation, Y.Z. and N.C.; data curation, T.T. (Takenobu Toyota) and T.T. (Tomonori Tanikawa); visualization, writing—original draft preparation, Y.Z.; writing—review and editing, K.S., T.T. (Tomonori Tanikawa) and Y.F.; supervision, project administration, and funding acquisition, K.S. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the Japan Aerospace Exploration Agency (JAXA) under grant number JX-PSPC-544506.

Data Availability Statement

Data underlying the results presented in this paper are available in PANGAEA [25,26,27].

Conflicts of Interest

The authors declare no conflicts of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.

Abbreviations

The following abbreviations are used in this manuscript:

AMSR	Advanced Microwave Scanning Radiometer
BRDF	Bidirectional Reflectance Distribution Function
CDF	Cumulative Density Function
KD-Tree	K-dimensional Tree
IOPs	Inherent Optical Properties
MLCM	Machine Learning Classification Mask
MODIS	Moderate Resolution Imaging Spectroradiometer
MPD	Melt Pond Detection
PDF	Probability Density Function
RTM	Radiative Transfer Model
RTM-SciML	Radiative Transfer Model Calculations with Scientific Machine Learning Estimates
SD	Synthetic Dataset
SGLI	Second Generation Global Imager
SIC	Sea Ice Concentration
SIZ	Seasonal Ice Zone
SZA	Solar Zenith Angle
TOA	Top of Atmosphere

References

Martin, S.; Drucker, R.; Yamashita, K. The production of ice and dense shelf water in the Okhotsk Sea polynyas. J. Geophys. Res. Ocean. 1998, 103, 27771–27782. [Google Scholar] [CrossRef]
Simizu, D.; Ohshima, K.I.; Ono, J.; Fukamachi, Y.; Mizuta, G. What drives the southward drift of sea ice in the Sea of Okhotsk? Prog. Oceanogr. 2014, 126, 33–43. [Google Scholar] [CrossRef]
Toyota, T.; Wakatsuchi, M. Characteristics of the surface heat budget during the ice-growth season in the southern Sea of Okhotsk. Ann. Glaciol. 2001, 33, 230–236. [Google Scholar] [CrossRef]
Toyota, T.; Ukita, J.; Ohshima, K.I.; Wakatsuchi, M.; Muramoto, K. A Measurement of Sea Ice Albedo over the Southwestern Okhotsk Sea. J. Meteorol. Soc. Jpn. Ser. II 1999, 77, 117–133. [Google Scholar] [CrossRef]
Lucht, W.; Schaaf, C.; Strahler, A. An Algorithm for the Retrieval of Albedo from Space Using Semiempirical BRDF Models. IEEE Trans. Geosci. Remote Sens. 2000, 38, 977–998. [Google Scholar] [CrossRef]
Schaaf, C.B.; Gao, F.; Strahler, A.H.; Lucht, W.; Li, X.; Tsang, T.; Strugnell, N.C.; Zhang, X.; Jin, Y.; Muller, J.P.; et al. First Operational BRDF, Albedo Nadir Reflectance Products from MODIS. Remote Sens. Environ. 2002, 83, 135–148. [Google Scholar] [CrossRef]
Zege, E.; Malinka, A.; Katsev, I.; Prikhach, A.; Heygster, G.; Istomina, L.; Birnbaum, G.; Schwarz, P. Algorithm to retrieve the melt pond fraction and the spectral albedo of Arctic summer ice from satellite optical data. Remote Sens. Environ. 2015, 163, 153–164. [Google Scholar] [CrossRef]
Istomina, L. Retrieval of Sea Ice Surface Melt Using OLCI Data Onboard Sentinel-3. Am. Geophys. Union Fall Meet. 2020, 2020, C017-07. [Google Scholar]
Qu, Y.; Liang, S.; Liu, Q.; Li, X.; Feng, Y.; Liu, S. Estimating Arctic Sea-Ice Shortwave Albedo from MODIS Data. Remote Sens. Environ. 2016, 186, 32–46. [Google Scholar] [CrossRef]
Peng, J.; Yu, Y.; Yu, P.; Liang, S. The VIIRS Sea-Ice Albedo Product Generation and Preliminary Validation. Remote Sens. 2018, 10, 1826. [Google Scholar] [CrossRef]
Zhou, Y.; Li, W.; Chen, N.; Fan, Y.; Stamnes, K. A sensor-agnostic albedo retrieval method for realistic sea ice surfaces: Model and validation. Cryosphere 2023, 17, 1053–1087. [Google Scholar] [CrossRef]
Stamnes, K.; Hamre, B.; Stamnes, S.; Chen, N.; Fan, Y.; Li, W.; Lin, Z.; Stamnes, J. Progress in forward-inverse modeling based on radiative transfer tools for coupled atmosphere-snow/ice-ocean systems: A review and description of the accurt model. Appl. Sci. 2018, 8, 2682. [Google Scholar] [CrossRef]
Chen, N.; Li, W.; Tanikawa, T.; Hori, M.; Shimada, R.; Aoki, T.; Stamnes, K. Fast yet accurate computation of radiances in shortwave infrared satellite remote sensing channels. Opt. Express 2017, 25, A649. [Google Scholar] [CrossRef] [PubMed]
Lu, J.; Heygster, G.; Spreen, G. Atmospheric Correction of Sea Ice Concentration Retrieval for 89 GHz AMSR-E Observations. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2018, 11, 1442–1457. [Google Scholar] [CrossRef]
Nihashi, S.; Ohshima, K.I.; Nakasato, H. Sea-ice retreat in the Sea of Okhotsk and the ice-ocean albedo feedback effect on it. J. Oceanogr. 2011, 67, 551–562. [Google Scholar] [CrossRef]
Turlach, B.A. Bandwidth selection in kernel density estimation: A review. In CORE and Institut de Statistique; Citeseer; Université Catholique de Louvain: Louvain-la-Neuve, Belgium, 1993. [Google Scholar]
Scott, D.W. Multivariate Density Estimation: Theory, Practice, and Visualization; John Wiley & Sons: Hoboken, NJ, USA, 2015; pp. i–xviii, 1–384. [Google Scholar]
Nihashi, S.; Ohshima, K.I.; Saitoh, S.I. Sea-ice production in the northern Japan Sea. Null 2017, 127, 65–76. [Google Scholar] [CrossRef]
Toyota, T.; Kimura, N.; Nishioka, J.; Ito, M.; Nomura, D.; Mitsudera, H. The Interannual Variability of Sea Ice Area, Thickness, and Volume in the Southern Sea of Okhotsk and Its Likely Factors. J. Geophys. Res. Ocean. 2022, 127, e2022JC019069. [Google Scholar] [CrossRef]
Shcherbina, A.Y.; Talley, L.D.; Rudnick, D.L. Direct observations of North Pacific ventilation: Brine rejection in the Okhotsk Sea. Science 2003, 302, 1952–1955. [Google Scholar] [CrossRef]
Ito, M.; Ohshima, K.I.; Fukamachi, Y.; Mizuta, G.; Kusumoto, Y.; Nishioka, J. Observations of frazil ice formation and upward sediment transport in the Sea of Okhotsk: A possible mechanism of iron supply to sea ice. J. Geophys. Res. 2017, 122, 788–802. [Google Scholar] [CrossRef]
Nihashi, S.; Kurtz, N.T.; Markus, T.; Ohshima, K.I.; Tateyama, K.; Toyota, T. Estimation of sea-ice thickness and volume in the Sea of Okhotsk based on ICESat data. Ann. Glaciol. 2018, 59, 101–111. [Google Scholar] [CrossRef] [PubMed]
Ohshima, K.I.; Watanabe, T.; Nihashi, S. Surface Heat Budget of the Sea of Okhotsk during 1987–2001 and the Role of Sea Ice on it. J. Meteorol. Soc. Jpn. 2003, 81, 653–677. [Google Scholar] [CrossRef]
Landrum, L.L.; Holland, M.M. Influences of changing sea ice and snow thicknesses on simulated Arctic winter heat fluxes. Cryosphere 2022, 16, 1483–1495. [Google Scholar] [CrossRef]
Zhou, Y. Comparison of MODIS and SGLI Albedo Retrievals Over the Sea of Okhotsk (January–May 2021). PANGAEA 2024. [Google Scholar] [CrossRef]
Zhou, Y. Albedo Retrievals from MODIS Over the Sea of Okhotsk (Validation Data, 2002–2014). PANGAEA 2024. [Google Scholar] [CrossRef]
Zhou, Y. Spatial and Temporal Distribution of Sea Ice in the Sea of Okhotsk (SIC, SGLI-retrieved Albedo, and Brightness Temperature Data, 2021). Figshare 2024. [Google Scholar] [CrossRef]

Figure 1. Map of the Sea of Okhotsk. The 200 m and 1000 m isobaths are indicated by red and blue dashed lines. The red box delineates the specific region where the PV Soya Icebreaker predominantly operated between 2002 and 2015 (see Section 2.2). The orange box (i) represents the Tartar Strait region, and the green box (ii) corresponds to the Northern Polynya, indicating areas selected for specific studies (see Section 5.1, Section 5.2 and Section 5.3).

Figure 2. (a) Close-up map of the region in the Sea of Okhotsk (the red box in Figure 1). The trace colors indicate the different navigational paths of the Soya Icebreaker in the years between 2002 and 2015, with the corresponding voyage years indicated in the colorbar placed at the bottom. (b) A detailed view of the PV Soya navigating through the Sea of Okhotsk, surrounded by sea ice and polynyas. The red circle on the tip of the ship highlights the location of the EKO MR-40 pyranometer that was used for the irradiance measurements.

Figure 3. Correlation between shortwave albedo measurements from the Soya Icebreaker and RTM-SciML retrievals. Panels (a,b) display the results with maximum time differences of three hours and one hour, respectively, between the measurement time and the MODIS overpass time. The color indicates the time interval between pyranometer measurements and satellite overpass. On the top left, the correlation equation (

y = a \cdot x + b

), Pearson r coefficient, root mean square error (RMSE), and the number of pixels (N) used to calculate the statistics are provided. The solid black lines represent (0,0)–(1,1), and the dashed black lines represent the 15% error range.

y = a \cdot x + b

Figure 4. Representative visualizations from 10 February 2004 (a) and 9 February 2008 (b). From top to bottom, each column displays: (1) true color RGB maps constructed using MODIS channels 645 nm, 555 nm, and 469 nm as the (R, G, B) bands, respectively; (2) surface classification maps with the spatial overlay of the Soya voyage shown in purple; (3) albedo retrieval maps with the spatial overlay of pyranometer measured values on RTM-SciML retrieved albedo; and (4) scatter–dot comparisons between the measurements and retrievals. On (a-3,b-3), the boxed A and B indicate the start and end points of the ship.

Figure 5. Representative visualizations from 14 February 2002 (a) and 11 February 2008 (b). From top to bottom, each column displays: (1) true color RGB maps constructed using MODIS channels 645 nm, 555 nm, and 469 nm as the (R, G, B) bands, respectively; (2) surface classification maps with the spatial overlay of the Soya voyage shown in purple; (3) albedo retrieval maps with the spatial overlay of pyranometer measured values on RTM-SciML retrieved albedo; and (4) scatter–dot comparisons between the measurements and retrievals. On (a-3,b-3), the boxed A and B indicate the start and end points of the ship.

Figure 6. Comparison of shortwave albedo measurements between the Soya Icebreaker and RTM-SciML retrievals represented as scatter plots. Each panel indicates the Pearson r coefficient and the number of pixels (N) at the top left. The dotted black lines delineate the MODIS overpass time. The x-axis across all panels displays the pyranometer measurement time (UTC).

Figure 7. Density plots illustrating the correlation between MODIS and SGLI albedo retrievals. Top two rows (subfigures 1): Bare sea ice; bottom two rows (subfigures 2): Melt-water/Water. (a–h) represents the eight time periods discussed in the main text. Each plot provides the Pearson correlation coefficient (r) and the root mean square error (RMSE) for the respective time periods on the top left.

Figure 8. Density plots comparing albedo retrievals from MODIS and SGLI sensors for different surface types over the total observation period from January to May 2021. (a) from bare ice surface, (b) from snow-covered sea ice surface, (c) from meltwater or open water and (d) from all valid sea-ice/water surfaces combined.

Figure 9. Comprehensive visualization of various parameters over the Sea of Okhotsk. From top to bottom, the rows depict (a) shortwave albedo, (b) surface classification, (c) brightness temperature as captured by SGLI’s

10.8 μ

m channel, and (d) high-resolution sea ice concentration from AMSR-2. Each column (numbers 1–7) corresponds to distinct retrieval periods as detailed in Table 4, showcasing the evolution of sea ice conditions.

10.8 μ

Figure 10. Average surface albedo of sea ice (left axis) and pixel percentage of the sea ice with different subtypes (right axis) during 1 April 2021∼7 April 2021. The colored bar plots and the labelled texts show the composition of sea ice. The black line is the relation between the average albedo (of all subtypes) and the SIC level. Error bars are the standard deviations of albedo.

Figure 11. (a) Snow pixel albedo values and their coverage percentages. (b) Sea ice (bare ice and ice with melt-water coverage) pixel albedo values and their percentages. The blue line and shadings are the relation and RMSE of 0.062 derived by [4].

Figure 12. Probability density curves of albedo in (a) Tatar Strait and the NWP adjoining the Japan Sea and (b) northern polynya. The number of pixels used to generate the curves in each panel are labelled at the top.

Figure 13. A detailed examination of the Tatar Strait region, derived from the broader overview provided in Figure 9. The rows from top to bottom feature (a–f) shortwave albedo, (g–l) surface classification, and (m–r) AMSR-2 sea ice concentration at 15 km resolution, alongside (s–x) brightness temperature data from SGLI’s 10.8 µm channel. Color-bars for each parameter are included for reference at the bottom of their respective rows.

Table 1. Central wavelengths and spatial resolutions of the MODIS and SGLI channels used for albedo retrievals.

Sensor	Central Wavelengths (nm)	Spatial Resolution
MODIS	469, 555, 645, 858.5, 1240, 1640, 2130	1 km
SGLI	443, 530, 673.5, 868.5, 1050, 1630, 2210	250 m∼1 km (aggregated to 1 km)

Table 2. Statistics comparing the satellite-derived albedo with pyranometer measurements for different surface types in the Sea of Okhotsk, constrained by a maximum 1 h time difference between satellite and Soya measurements.

Surface Type	Pearson-r	RMSE	Correlation	N
Snow-covered Ice	0.89	0.081	$\hat{α} = 0.93 α + 0.05$	57
Bare sea ice	0.76	0.101	$\hat{α} = 0.81 α + 0.02$	175
Melt-water/open water	0.90	0.071	$\hat{α} = 1.01 α - 0.03$	103
All sea ice	0.86	0.089	$\hat{α} = 0.94 α - 0.01$	335

Table 3. Comparison metrics, including Pearson-r coefficients and RMSE, for all valid sea ice data (including bare, melt-water covered, and snow-covered sea ice) across various time periods between MODIS and SGLI retrievals.

Time Period for Retrieval	Pearson-r	RMSE
2021-01-08∼2021-01-13	0.803	0.086
2021-01-31∼2021-02-06	0.900	0.051
2021-02-11∼2021-02-17	0.814	0.048
2021-03-05∼2021-03-12	0.822	0.036
2021-03-15∼2021-03-20	0.716	0.046
2021-04-01∼2021-04-07	0.736	0.02
2021-04-22∼2021-04-30	0.775	0.016
2021-05-07∼2021-05-12	0.664	0.011
All time	0.883	0.036

Table 4. The date ranges of the SGLI images that were used for retrievals (Albedo/MLCM), and the dates of AMSR-2 sea ice concentration data.

Time Period for Retrieval	Dates for Sea Ice Concentration
Time Period for Retrieval	15 km Resolution	5 km Resolution
2021-01-08∼2021-01-13	2021-01-10	/
2021-01-31∼2021-02-06	2021-02-04	/
2021-02-11∼2021-02-17	2021-02-14	/
2021-03-05∼2021-03-12	2021-03-08	2021-03-06
2021-03-15∼2021-03-20	2021-03-17	2021-03-15
2021-04-01∼2021-04-07	2021-04-04	2021-04-02
2021-04-22∼2021-04-30	2021-04-29	2021-04-29
2021-05-07∼2021-05-12	2021-05-10	2021-05-10

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

© 2025 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).

Share and Cite

MDPI and ACS Style

Zhou, Y.; Li, W.; Chen, N.; Toyota, T.; Fan, Y.; Tanikawa, T.; Stamnes, K. Spatiotemporal Variations in Sea Ice Albedo: A Study of the Dynamics of Sea Ice Albedo in the Sea of Okhotsk. Remote Sens. 2025, 17, 772. https://doi.org/10.3390/rs17050772

AMA Style

Zhou Y, Li W, Chen N, Toyota T, Fan Y, Tanikawa T, Stamnes K. Spatiotemporal Variations in Sea Ice Albedo: A Study of the Dynamics of Sea Ice Albedo in the Sea of Okhotsk. Remote Sensing. 2025; 17(5):772. https://doi.org/10.3390/rs17050772

Chicago/Turabian Style

Zhou, Yingzhen, Wei Li, Nan Chen, Takenobu Toyota, Yongzhen Fan, Tomonori Tanikawa, and Knut Stamnes. 2025. "Spatiotemporal Variations in Sea Ice Albedo: A Study of the Dynamics of Sea Ice Albedo in the Sea of Okhotsk" Remote Sensing 17, no. 5: 772. https://doi.org/10.3390/rs17050772

APA Style

Zhou, Y., Li, W., Chen, N., Toyota, T., Fan, Y., Tanikawa, T., & Stamnes, K. (2025). Spatiotemporal Variations in Sea Ice Albedo: A Study of the Dynamics of Sea Ice Albedo in the Sea of Okhotsk. Remote Sensing, 17(5), 772. https://doi.org/10.3390/rs17050772

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Menu

Spatiotemporal Variations in Sea Ice Albedo: A Study of the Dynamics of Sea Ice Albedo in the Sea of Okhotsk

Abstract

1. Introduction

2. Data and Models

2.1. Albedo Retrievals and Surface Classification Using SGLI and MODIS Radiance Data

2.2. Validation Data

3. Validation Results

4. Comparison Between MODIS and SGLI Retrievals

5. Spatial and Temporal Distribution of Sea Ice in the Sea of Okhotsk

5.1. Spatial Coverage of Sea Ice in the Sea of Okhotsk

5.2. Relation Between Sea Ice Albedo and Concentration in the Sea of Okhotsk

5.3. Spatiotemporal Analysis of Surface Albedo

6. Discussion

7. Concluding Remarks and Future Perspectives

Author Contributions

Funding

Data Availability Statement

Conflicts of Interest

Abbreviations

References

Share and Cite

Article Metrics

Article Access Statistics

Further Information

Guidelines

MDPI Initiatives

Follow MDPI