Wintertime Emissivities of the Arctic Sea Ice Types at the AMSR2 Frequencies

Abstract

The surface effective emissivities of Arctic sea ice are calculated using Advanced Microwave Scanning Radiometer 2 (AMSR2) measurements. These emissivities are analyzed for stable winter conditions during the months of January–May and November and December of 2020 for several main sea ice types defined with the sea ice maps of the Arctic and Antarctic Research Institute (AARI). The sea ice emissivities are derived from the AMSR2 data using the radiation transfer model for a non-scattering atmosphere and ERA5 reanalysis data. The emissivities are analyzed only for areas of totally consolidated sea ice of definite types. Probability distribution functions are built for the emissivities and their functions for such sea ice types as nilas, young ice, thin first-year (FY) ice, medium FY ice, thick FY ice and multi-year ice. The emissivity variations with frequency are estimated for each of the considered sea ice type for all seven months. The variations are calculated both for the emissivities and for their gradients at the AMSR2 channel frequencies. Obtained emissivities turned out to be generally lower than reported previously in scientific studies, whereas the emissivity variability values proved to be much larger than was known before. For all FY ice types, at all the frequencies, an increase in the emissivity at the beginning of winter and its decrease by the end of May are observed. The emissivity gradients demonstrate noticeable decreases with sea ice age, and their values may be used in sea ice classification algorithms based on the AMSR2 data.

1. Introduction

Satellite remote sensing is of particular importance for monitoring sea ice cover in remote areas, where the use of traditional contact measurements is complicated or impossible. Monitoring Arctic sea ice cover parameters with the data from long-term satellite measurements has been making it possible to quantify the climate changes that have occurred in recent years [1]. The study of the seasonal and long-term variability of Arctic sea ice is one of the most important tasks, the solution of which allows correct assessment of these changes [2].The Arctic sea ice area and volume decrease in recent decades has been confirmed by numerous studies exploring various sea ice concentration and thickness retrieval methods [3,4,5]. Regular sea ice concentration (SIC) mapping is possible only with the measurements of scanning multichannel microwave radiometers. Their data allow quantitative SIC retrievals regardless of cloudiness and time of day [6]. In spite of a large number of existing SIC retrieval methods and approaches, the task of improving SIC retrieval accuracy remains actual to this day [7,8].

Numerous studies devoted to the analysis of SIC retrieval accuracy have shown that the results of SIC retrieval algorithm application over sea ice edge areas (with SIC ranging from 15 to 70%), as well as over consolidated sea ice during melting season, can differ by 50–100% [7,8,9,10,11,12]. The reasons for such differences lie in the difficulty of taking into account the whole variety of conditions developing in the sea ice–ocean–atmosphere system and inevitable errors accompanying retrieval algorithms, which are ever based on certain assumptions and simplifications [13]. Every SIC retrieval algorithm uses so-called tie points, representing some microwave radiation characteristics for predefined surface types. These can be both brightness temperatures (Tb)—and Tb functions at different measurement channels—or emissivities of sea ice and an ice-free sea. Exploration of fixed tie point values leads to retrieval errors, as in reality their variability may be quite large.

The sea ice microwave properties depend on the radiation frequency and ice physical characteristics, which in turn are determined by the sea ice type, the history of the ice surface formation (its crystal structure and salinity), its roughness and its temperature. Theoretical modeling of the sea ice emissivity shows that first-year (FY) ice is practically opaque for microwave radiation at a depth of the order of λ/6, where λ is the radiation wavelength [14]. This means that the microwave radiation of FY ice is formed by a thin surface layer no larger than 1 cm and does not depend on the thickness. On the other hand, experimental data provide evidence of brightness temperature contrasts between newly formed, young and FY ice. [15,16]. Young ice (YI) is characterized by the presence of brine on its surface, the radiative properties of which are those of salt water. This complicates identification of YI with microwave radiometer data [17]. On the other hand, the dielectric properties of this surface brine correlate with sea ice thickness, as the brine is formed during the ice formation. This opens possibilities to determine YI thickness using microwave methods [18,19,20]. However, because the freezing/melting of the brine surface layer occurs at low air temperatures, the radiative properties of the YI with the brine on the surface are very variable and influenced by meteorological conditions [20,21].

Effective emissivities of the sea ice χ are also affected by the parameters of the snow cover [22]. If snow falls on the sea surface with a temperature below zero, it is saturated with sea water and contributes to the sea’s freezing (slush ice is formed). The snow on the sea ice surface scatters the sea ice radiation and forms its own radiation. These processes strongly depend on the snow moisture, its density and its radiation frequency [23]. Dry snow on the sea ice surface scatters radiation from the sea ice, leading to a decrease in effective values of χ. The higher the frequency, the stronger the scattering properties of dry snow (due to the typical sizes of snow crystals). This consideration is typically used when the thickness of dry snow is determined on the basis of the gradient ratios in measurements in the Ku and Ka bands [24]. In a number of studies, theoretical modeling of the radiative characteristics of a multilayer system of sea ice–snow cover is carried out, and the influence of snow cover properties on the effective emissivity for different sea ice types is analyzed [25,26]. In [7] it is shown that the variations in the SIC, retrieved from satellite microwave measurements over central Arctic areas with consolidated ice, are caused exactly by variations in snow characteristics. The work [27] describes an electrodynamic model of Arctic sea ice radiation, taking into account physical and structural characteristics of snow and ice. This model has been used in the SIC retrieval algorithm developed in [28,29]. Still, most often in developing SIC retrieval methods, sea ice emissivity experimental data are used, despite the fact that they are not numerous and always obtained for specific conditions: certain sea ice types, snow cover conditions, season, region, etc.

The first measurements of sea ice emissivities were carried out as a part of specially organized aircraft experiments. Airborne measurements of radiometers in May 1967 and June 1970 in the Alaska region confirmed the strong radio brightness contrasts between sea ice and sea water and the presence of two sea ice types in the Arctic with different radiometric properties at the frequencies of 19–37 GHz [30]. For the first time, the difference in the electromagnetic properties of FY and multi-year (MY) sea ice was formulated: at frequencies f below 40 GHz, FY ice radiation is practically independent of f, while MY ice radiation decreases with f. In the spring of 1977, the Naval Research Laboratory conducted a series of experiments to measure the emissivities of various sea ice types at nadir in the Greenland Sea for a wide range of f from 14 to 90 GHz [31]. An analysis of measurement results allowed formulating a number of emission properties for the FY and MY ice, some of which were confirmed in subsequent studies. Airborne radiometer measurements as a part of the Norwegian Remote Sensing Experiment (NORSEX) program in September–October 1979 over the sea ice edge northwest of Svalbard made it possible to calculate the χ values of vertically (V) and horizontally (H) polarized radiation from the FY and MY ice at an angle of 50° for the frequencies of 4.9, 10.4, 21, 36 and 94 GHz [32].

Later, an extensive series of sea ice characteristic measurements at the same angle were carried out in different seasons as a part of the Marginal Ice Zone Experiment (MIZEX) program in the Fram Strait region [33,34]. The results of these measurements generally agree with the earlier studies, except that the authors of [33] consider a new type of MY ice—flooded multi-year ice, not present in the international sea ice nomenclature, whose properties in the marginal ice zone are close to FY ice properties due to flooding and penetration of sea water into the upper ice layers. When such ice is present, it is not possible to separate FY and MY ice. The authors also reported lower emissivity values for the initial ice types. Summer conditions, characterized by the appearance of a wet layer of snow or ice on the surface, bring the electromagnetic properties of MY ice closer to those of FY ice. This is confirmed by independent summer measurements of χ in in different Arctic regions [35,36].

Later experiments in the 1990s partially confirmed the main conclusions of earlier studies: the emissivity of thick FY ice is close to 1 and decreases as the ice ages (the brine in the surface sea ice layers flows out of salt pockets) at frequencies near 90 GHz [37].

Airborne measurements of sea ice emissivity are extremely limited in terms of spatial and temporal coverage, but the atmospheric correction is typically highly accurate, as it is based on direct measurements of meteorological parameters [31,35,37]. Meanwhile, estimation of χ from satellite microwave radiometer measurement data is always accompanied by the errors associated with the lack of information on the atmospheric parameters needed to correctly determine its contribution to the resulting radiation of the sea ice–atmosphere system.

The study [38] describes the analysis of χ for the channel characteristics of the Advanced Microwave Scanning Radiometer–Earth Observing System (AMSR-E). Surface emissivities were calculated for FY and MY ice test patches based on AMSR-E measurements for a whole year and data from the European Center for Medium-Range Weather Forecasts (ECMWF) for atmospheric correction. The authors mapped the Arctic sea ice emissivity and analyzed the seasonal features of χ values for selected areas. The source of errors in [38] is the assumption of an unchanged-in-time sea ice type and its 100% concentration in the areas under consideration. Moreover, summer interpretation of measurements is questionable due to an increase in atmospheric humidity and cloud water content. In summer months the use of ECMWF data for atmospheric correction leads to significant errors in χ estimation.

In this study we analyzed the emissivity of different Arctic sea ice types in stable winter conditions (in the absence of melting) for the characteristics of the Advanced Microwave Scanning Radiometer 2 (AMSR2) measurement channels for vertically (V) and horizontally (H) polarized radiation at the frequencies of 6.9, 10.65, 18.7, 23.8, 36.5 and 89 GHz. Surface emissivity values were calculated using the AMSR2 measurements for the entire Arctic (for latitudes above 60°) for the year 2020. To estimate the atmospheric parameters of radiation (upwelling and downwelling atmospheric radiation (T_a) and atmospheric absorption (τ)), the radiation transfer equation was solved in the approximation of a non-scattering atmosphere [39]. ERA5 reanalysis data were used as input data for the calculation of T_a and τ. Sea ice emissivities were analyzed for the periods of January–May and November–December of 2020. The AMSR2 measurements were classified by the sea ice type using the Arctic and Antarctic Research Institute (AARI) sea ice maps.

2. Materials and Methods

2.1. Methodology

The brightness temperature (Tb) of the microwave radiation of the surface–atmosphere system, measured by a microwave radiometer, is a function of the atmospheric parameters and the effective surface emissivity. In the approximation of the absence of scattering and without taking into account refraction, the solution of the radiative transfer equation (RTE) is the sum of the upwelling and downwelling atmospheric radiation, the oceanic radiation and the background cosmic microwave radiation:

$$T_b{}^{H,V}=T_a{}^{\uparrow }+{\chi }^{H,V}·T_s·e^{-\tau }+(T_a{}^{\downarrow }+e^{-\tau }·T_c)·e^{-\tau }·(1-{\chi }^{H,V})$$

(1)

where

$$T_a^{\uparrow }=\frac{1}{\cos \theta }{\displaystyle \underset{0}{\overset{\infty }{\int }}T(h)\alpha (h)}·\exp (-\frac{1}{\cos \theta }{\displaystyle \underset{h}{\overset{\infty }{\int }}\alpha (h^{\prime })}d{h}^{\prime })dh$$

(2)

$$T_a^{\downarrow }=\frac{1}{\cos \theta }{\displaystyle \underset{0}{\overset{\infty }{\int }}T(h)\alpha (h)·\exp (-\frac{1}{\cos \theta }{\displaystyle \underset{0}{\overset{h}{\int }}\alpha (h^{\prime })}d{h}^{\prime })dh}$$

(3)

$$\tau =\frac{1}{\cos \theta }{\displaystyle \underset{0}{\overset{\infty }{\int }}\alpha (h^{\prime })}d{h}^{\prime }$$

(4)

$T_a^{\uparrow }$, $T_a^{\downarrow }$ and τ are the upwelling and downwelling atmospheric radiation and the atmospheric optical thickness along the observation direction, respectively; α(h) is the atmospheric absorption coefficient along the atmospheric height h, depending mainly (at the AMSR2 frequencies) on the atmospheric water vapor, liquid water and oxygen amount; θ is the zenith observation angle; χ is the surface effective emission coefficient; T_c is the background cosmic microwave radiation: T_c = 2.7 K. Equation (1) provides the means to calculate T_b^H,V for horizontally (H) and vertically (V) polarized radiation at given effective emission coefficients χ^H,V, atmospheric parameter profiles and surface temperature T_s for a frequency f and observation angle θ. Vice versa, at known values T_b^H,V, measured by the AMSR2, and given (measured or modeled) atmospheric parameters and T_s, we can calculate the values of χ^H,V as:

$${\chi }^{H,V}=\frac{T_b{}^{H,V}-T_a{}^{\uparrow }-e^{-\tau }·(T_a{}^{\downarrow }+e^{-\tau }·2.7)}{T_s-(T_a{}^{\downarrow }+e^{-\tau }·2.7)}·e^{\tau }.$$

(5)

$T_a^{\uparrow }$, $T_a^{\downarrow }$ and τ depend only on the atmospheric meteorological parameters—profiles of pressure p(h), temperature T(h), air humidity ρ(h) and cloud liquid water content w(h). To calculate $T_a^{\uparrow }$, $T_a^{\downarrow }$ and τ using Equations (2)–(4), we used the atmospheric microwave radiation absorption model [40] and atmospheric meteorological parameter profiles taken from the ERA5 reanalysis. ERA5 data on T_s were used in χ^H,V calculations.

Daily averaged AMSR2 measurements of the microwave radiation brightness temperatures at 6.9, 10.65, 18.7, 23.8, 36.5 and 89 GHz on H and V polarization (will be referred to as T06H, T06V, T10H, T10V, T18H, T18V, T23H, T23V, T36H, T36V, T89H and T89V) were interpolated onto the ERA5 grid, and the daily effective surface emission coefficients χ₀₆^H, χ₀₆^V, χ₁₀^H, χ₁₀^V, χ₁₈^H, χ₁₈^V, χ₂₃^H, χ₂₃^V, χ₃₆^H, χ₃₆^V, χ₈₉^H and χ₈₉^V were calculated with Equation (5) for the whole Arctic area (for latitudes higher than 60°) for the year 2020.

An analysis of the χ fields for the whole of 2020 led to the conclusion that from June to October, water content in the Arctic atmosphere (in liquid and vapor forms) is so high that the errors in the reanalysis data and the radiative transfer model do not allow using the proposed methodology for calculating χ. This conclusion is based on the observed variability of the calculated fields of sea ice emission coefficients on diurnal time scales. This variability can only be explained by errors associated with atmospheric variability. Therefore, the analysis of the sea ice χ was carried out only for the periods from January to May and from November to December.

The Sea Ice Concentration (SIC) data from the University of Bremen, interpolated on the ERA5 grid, were used to select the areas of consolidated sea ice [41]. Areas with SIC > 95% were classified as sea ice (SI), while areas with SIC = 0 were classified as sea water (OW).

For the analysis of χ of different sea ice types, AARI sea ice maps were used. These data include sea ice concentration ranges, age gradations and forms [42]. The sea ice age correlates with its thickness and is often replaced by the term “sea ice type”, integrating into the same type the ranges of sea ice thicknesses. For the analysis of χ for a specific sea ice type, polygons with 100% partial concentration of this type were selected. Thus, the regions with mixed sea ice types were excluded from consideration. Six Arctic sea ice types from the AARI classification nomenclature were used: nilas, young ice (including gray and white-grey), thin FY, medium FY, thick FY and MY ice. The PDFs were constructed and analyzed both for the values of χ at different AMSR2 frequencies and polarizations, and for their differences (Δχ) for some measurement channels. For each sea ice type, monthly and winter averages of χ and Δχ and their standard deviations (σ_χ and σ_Δχ) were obtained for the entire Arctic region.

2.2. Data

Several types of data were used in the work: (1) daily averaged brightness temperature measurements of the AMSR2 instrument from the GCOM-W1 satellite, (2) ERA5 reanalysis data, (3) AARI sea ice maps and (4) SIC fields retrieved from the AMSR2 data. All data were interpolated on the ERA5 grid with a resolution of 0.25 × 0.25°.

2.2.1. AMSR2 Data

The satellite microwave radiometer AMSR2 from the GCOM-W1 satellite measures vertically and horizontally polarized microwave radiation in conical scanning mode at an observation angle of 55° at 7 frequencies: 6.9, 7.3, 10.65, 18.7, 23.8, 36.5 and 89 [43]. Calibrated data of different levels—from Level 1R brightness temperatures to geophysical parameters, including various levels of time averaging—are processed and distributed through the service https://gportal.jaxa.jp (accessed on 31 January 2021). We used daily averaged calibrated Level 3 Tb measurements at all the frequencies, except for 7.3 GHz, on vertical and horizontal polarization for the Arctic region with a spatial resolution of 25 km. The measurements at 7.3 GHz do not provide new information on sea ice as compared with the measurements at 6.9 GHz and are typically used for the identification of radio frequency interference over open water areas. The initial data period was from 1 January to 31 December 2020. Only the data for January–May, November and December were used for the sea ice type emission coefficient analysis.

2.2.2. ERA5 Reanalysis Data

ERA5 data of pressure levels (https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-pressure-levels (accessed on 1 February 2021)) for 2020 on a 0.25 × 0.25° grid for atmospheric pressure, humidity, temperature and cloud water content were used to calculate frequency dependent the following atmospheric radiation parameters: $T_a^{\uparrow }$, $T_a^{\downarrow }$ (atmospheric upwelling and downwelling radiation) and optical thickness τ along the observation direction. Hourly data were averaged to obtain daily averages; the values of $T_a^{\uparrow }$, $T_a^{\downarrow }$ and τ were calculated using Formulas (2)–(4). To calculate the surface emission coefficients, the ERA5 data of the surface level (https://cds.climate.copernicus.eu/cdsapp#!/dataset/reanalysis-era5-single-levels (accessed on 1 February 2021)) on the sea ice and the sea water surface temperature T_s were also used.

2.2.3. AARI Sea Ice Maps

AARI sea ice maps are one of the most reliable sources of information on Arctic sea ice age composition and forms, as they are prepared by highly qualified experts based on the analysis of all available satellite data. The basis for the construction of such maps consists of satellite remote sensing data, supplemented by ship, sometimes airplane and station observations, integrated and analyzed by the sea ice experts [44]. The sea ice polygon boundaries are determined by analyzing not only current satellite images, but also the images from previous days. In addition, the accumulated historical knowledge about the sea ice age composition in a particular geographic region is used. To determine the age of sea ice fields, direct and indirect features visually revealed in the images are used: brightness, shape, size and texture. Areas of homogeneous ice characteristics are distinguished if they have approximately the same age composition of ice with the same partial concentration and distinctive sea ice field sizes and shapes. Detailed AARI sea ice maps for individual Arctic seas (http://wdc.aari.ru/datasets/d0004/ (accessed on 10 February 2021)) were used, as they contain the most detailed information about the sea ice cover parameters.

The sea ice maps are issued in a standard Sea Ice Grid (SIGRID-3) format, developed by the International Ice Charting Working Group (IICWG) for the World Meteorological Organization (WMO) [45]. They contain a set of polygons with uniform sea ice characteristics for the three sea ice types prevailing in the area—A, B and C in descending order of thickness (age). Types A, B and C may be any of the age gradations. For analysis, only areas with a single specific type of sea ice with 100% concentration were selected. Thus, the polygons with mixed sea ice types were excluded from consideration. The analysis was carried out for the 6 main age gradations in the AARI classification: nilas, young ice (including gray and white-grey), thin FY, medium FY, thick FY and MY ice.

2.2.4. SIC Data Retrieved from the AMSR2 Measurements

Because the AARI sea ice maps have low time resolution (they are issued once every 7 days), SIC values retrieved from the AMSR2 data were analyzed to separate the data into sea ice and sea water data. We used the daily average SIC satellite product created at the University of Bremen from the AMSR2 data using the Artist Sea Ice (ASI) algorithm (http://www.iup.uni-bremen.de:8084/amsr2data/asi_daygrid_swath/ (accessed on 15 February 2021)) [41]. Although satellite microwave radiometer measurements have low spatial resolution, they are currently the only source of regular sea ice concentration fields [10]. SIC data with a spatial resolution of 6.25 km × 6.25 km were interpolated onto the ERA5 coordinate grid. Areas with SIC > 90% were classified as sea ice; areas with SIC = 0 were classified as sea water.

3. Results

Daily values of the effective surface emission coefficients χ₀₆^H, χ₀₆^V, χ₁₀^H, χ₁₀^V, χ₁₈^H, χ₁₈^V, χ₂₃^H, χ₂₃^V, χ₃₆^H, χ₃₆^V, χ₈₉^H and χ₈₉^V were calculated and analyzed for the periods of January–May, November and December of the year 2020. Figure 1 illustrates an example of daily fields of χ on 1 January 2020 at 6.9, 36.5 and 89 GHz with H and V polarization.

After the data were classified as SI or OW with SIC data (SI if SIC > 95%, OW if SIC = 0), probability density functions (PDFs) of the χ values for SI and OW were built both for monthly datasets (for all SI and OW data for each of the 7 months) and for the whole 7-month dataset. The PDFs of χ_SI and χ_OW, calculated for the whole OW and SI datasets, are shown in Figure 2.

The most general features of SI and OW χ at the considered frequencies are well studied and described in the literature. Figure 2 is shown to illustrate the wide range of sea ice emissivities. The higher the radiation frequency, the wider this range. With increasing frequency the sizes of inhomogeneities, which effectively scatter microwave radiation, decrease. These inhomogeneities, such as air bubbles and salt pockets inside sea ice or snow crystals, have dielectric permittivities different than the dielectric permittivity of ice crystals. The variability of these effective scatterers’ distribution determines the variability of χ at different frequencies.

To analyze the emission of different sea ice types, the subsets of χ for each of the predefined six main sea ice types, based on the collocations with the AARI sea ice maps, were organized. To avoid misinterpretation in subjectively classified maps, only the polygons with 100% partial sea ice concentration were included in the corresponding χ subsets. PDFs of χ values were built, and monthly and winter χ values and their standard deviations σ_χ were calculated for the entire Arctic region for six AMSR2 measurement frequencies with vertical and horizontal polarizations. For monthly χ values (χ^month) we mean the values of χ, averaged over the sea ice type subsets, collected for each month. For winter χ values we mean the values of χ, averaged over the total sea ice type subsets, collected for all 7 months. The dependence of winter χ on frequency for the six Arctic sea ice types is shown in Figure 3.

It can be seen that the frequency dependence of χ for sea ice with a thickness up to 30 cm (nilas and young ice) is different from the dependence of χ for thicker ice (FY ice of different thicknesses, MY ice). Young ice is characterized by a noticeable increase in χ with frequency f up to 37 GHz, while for thicker, FY ice, the values of χ practically do not change with frequency with increasing f up to 23.8 GHz, and they decrease with increasing f from 23.8 to 89 GHz. The emission from MY ice decreases sharply from 6.9 to 37 GHz and with a smaller gradient, similar to that of FY ice with a thickness exceeding 30 cm (thin, medium and thick FY ice).

The standard deviation (σ_χ) characterizes the range of χ values and shows in fact how many different sea ice surfaces are actually aggregated into the same sea ice type. Figure 4 shows the change in χ and σ_χ with the sea ice type (age).

The sea ice emission coefficient χ increases with increasing frequency for nilas and young ice. For FY ice types, in the frequency range of 6.9–36.5 GHz, it does not change significantly, and then it decreases. For MY ice χ decreases for the entire frequency range. The higher the frequency, the smaller the standard deviation of χ values (σ_χ) for ice with a thickness less than 30 cm. For older ice, σ_χ is practically independent of frequency and remains low (the average σ_χ is ~0.07 for horizontal polarization and 0.03–0.05 for vertical polarization). Thus we see that from the very beginning of sea ice formation, sea ice emission increases for all frequencies until the scattering processes become significant. At 89 GHz, these processes are not negligible already in young ice with a thickness of 10–30 cm, for lower frequencies—starting only from the stage of thin FY ice.

Vertically polarized radiation is always higher than horizontally polarized radiation: the polarization difference Δχ_f = χ^V − χ^H for all frequencies f for all sea ice types is much less than that for open water. This is due to the low values of the dielectric permittivity of sea ice compared to sea water (3–4 compared to 80). Figure 5a shows the dependence of Δχ_f on frequency for different sea ice types, and Figure 5b illustrates the dependence of Δχ_f on the sea ice type for the AMSR2 measurement frequencies.

For the thin sea ice types (with a thickness up to 70 cm, corresponding to thin FY ice) the older the ice, the lower the Δχ_f values. At the same time we observe no differences in the polarization properties either for FY ice of any thickness or for MY ice. This means that there is no significant fraction of roughness on the ice surface changing the radiation polarization. Moreover, scattering by brine/air pockets does not make a significant contribution to the variability of the radiation polarization.

Monthly averaged values of χ (χ^month) and the standard deviations σ_χ, calculated for the χ subsets for every month for all AMSR2 frequencies for H and V polarized microwave radiation are presented in

Table 1. Monthly averaged values of χ and σ_χ, calculated for the χ subsets for every month at 6.9, 10.65, 18.7, 23.8, 36.5 and 89 GHz with H and V polarization (06H, 10H, 18H, 23H, 36H, 89H, 06V, 10V, 18V, 23V, 36V, 89V correspondingly) for the six sea ice types.

Nilas
Month	χ06H	σχ06H	χ10H	σχ10H	χ18H	σχ18H	χ23H	σχ23H	χ36H	σχ36H	χ89H
1	0.54	0.26	0.56	0.26	0.59	0.24	0.61	0.23	0.63	0.20	0.66
2	0.48	0.22	0.49	0.22	0.51	0.22	0.53	0.20	0.56	0.18	0.62
3	0.60	0.20	0.62	0.20	0.64	0.19	0.66	0.18	0.67	0.16	0.70
4	0.60	0.17	0.62	0.18	0.63	0.18	0.65	0.16	0.66	0.15	0.66
5	0.66	0.13	0.67	0.14	0.68	0.15	0.68	0.14	0.68	0.12	0.63
11	0.49	0.19	0.51	0.20	0.55	0.20	0.57	0.19	0.61	0.17	0.67
12	0.53	0.24	0.55	0.24	0.58	0.23	0.61	0.22	0.64	0.20	0.69
Month	χ06V	σχ06V	χ10V	σχ10V	χ18V	σχ18V	χ23V	σχ23V	χ36V	σχ36V	χ89V
1	0.74	0.18	0.76	0.17	0.79	0.15	0.80	0.13	0.80	0.11	0.82
2	0.70	0.15	0.71	0.14	0.74	0.13	0.75	0.12	0.77	0.09	0.81
3	0.79	0.14	0.80	0.13	0.82	0.12	0.83	0.11	0.83	0.09	0.84
4	0.78	0.11	0.80	0.11	0.82	0.10	0.82	0.09	0.82	0.07	0.82
5	0.82	0.09	0.84	0.09	0.84	0.08	0.84	0.08	0.83	0.07	0.77
11	0.72	0.13	0.75	0.13	0.78	0.11	0.79	0.11	0.81	0.09	0.84
12	0.75	0.16	0.77	0.15	0.79	0.13	0.80	0.12	0.82	0.10	0.85
Young Ice
Month	χ06H	σχ06H	χ10H	σχ10H	χ18H	σχ18H	χ23H	σχ23H	χ36H	σχ36H	χ89H
1	0.70	0.19	0.72	0.19	0.74	0.19	0.75	0.18	0.77	0.16	0.77
2	0.60	0.21	0.62	0.21	0.64	0.21	0.66	0.20	0.68	0.18	0.71
3	0.61	0.20	0.63	0.20	0.65	0.20	0.67	0.19	0.69	0.17	0.71
4	0.62	0.20	0.64	0.20	0.65	0.20	0.67	0.19	0.68	0.17	0.69
5	0.53	0.18	0.55	0.18	0.56	0.18	0.58	0.17	0.60	0.15	0.63
11	0.73	0.14	0.75	0.14	0.78	0.14	0.79	0.13	0.80	0.12	0.79
12	0.70	0.17	0.72	0.17	0.74	0.17	0.76	0.16	0.78	0.14	0.79
Month	χ06V	σχ06V	χ10V	σχ10V	χ18V	σχ18V	χ23V	σχ23V	χ36V	σχ36V	χ89V
1	0.85	0.12	0.87	0.12	0.88	0.11	0.88	0.10	0.88	0.09	0.87
2	0.79	0.14	0.80	0.13	0.82	0.12	0.83	0.11	0.84	0.09	0.85
3	0.80	0.14	0.81	0.13	0.83	0.12	0.84	0.11	0.84	0.09	0.85
4	0.79	0.13	0.81	0.13	0.83	0.11	0.83	0.11	0.84	0.09	0.83
5	0.74	0.12	0.75	0.12	0.78	0.10	0.79	0.10	0.80	0.08	0.81
11	0.87	0.08	0.89	0.08	0.90	0.07	0.90	0.07	0.90	0.06	0.88
12	0.86	0.11	0.87	0.11	0.89	0.09	0.89	0.09	0.89	0.07	0.89
Thin FY Ice
Month	χ06H	σχ06H	χ10H	σχ10H	χ18H	σχ18H	χ23H	σχ23H	χ36H	σχ36H	χ89H
1	0.82	0.09	0.84	0.09	0.84	0.09	0.85	0.09	0.84	0.09	0.78
2	0.80	0.10	0.81	0.11	0.82	0.11	0.82	0.11	0.82	0.10	0.77
3	0.81	0.10	0.82	0.11	0.83	0.10	0.84	0.10	0.83	0.09	0.78
4	0.75	0.10	0.76	0.11	0.77	0.12	0.77	0.11	0.77	0.11	0.73
5	0.66	0.19	0.67	0.19	0.67	0.19	0.68	0.17	0.68	0.15	0.65
11	0.86	0.04	0.87	0.04	0.88	0.04	0.88	0.04	0.87	0.04	0.76
12	0.87	0.04	0.88	0.04	0.89	0.04	0.89	0.04	0.88	0.05	0.75
Month	χ06V	σχ06V	χ10V	σχ10V	χ18V	σχ18V	χ23V	σχ23V	χ36V	σχ36V	χ89V
1	0.92	0.06	0.93	0.06	0.93	0.06	0.93	0.06	0.91	0.06	0.85
2	0.91	0.07	0.92	0.07	0.92	0.07	0.92	0.06	0.90	0.06	0.84
3	0.91	0.07	0.92	0.06	0.92	0.06	0.92	0.05	0.91	0.05	0.85
4	0.87	0.07	0.88	0.07	0.89	0.06	0.89	0.06	0.87	0.06	0.82
5	0.81	0.12	0.83	0.12	0.84	0.11	0.84	0.10	0.83	0.08	0.78
11	0.94	0.02	0.95	0.02	0.95	0.02	0.95	0.02	0.93	0.03	0.82
12	0.96	0.03	0.97	0.03	0.97	0.02	0.97	0.02	0.94	0.04	0.80
Medium FY Ice
Month	χ06H	σχ06H	χ10H	σχ10H	χ18H	σχ18H	χ23H	σχ23H	χ36H	σχ36H	χ89H
1	0.87	0.03	0.88	0.03	0.89	0.03	0.89	0.04	0.87	0.05	0.75
2	0.85	0.04	0.87	0.04	0.87	0.04	0.87	0.05	0.85	0.06	0.75
3	0.84	0.05	0.85	0.05	0.86	0.05	0.86	0.05	0.84	0.05	0.75
4	0.83	0.07	0.84	0.08	0.84	0.08	0.84	0.08	0.82	0.08	0.73
5	0.79	0.11	0.79	0.12	0.78	0.12	0.79	0.11	0.76	0.11	0.66
11	0.84	0.03	0.86	0.03	0.86	0.04	0.85	0.04	0.81	0.08	0.67
12	0.86	0.05	0.87	0.05	0.88	0.06	0.88	0.06	0.86	0.09	0.75
Month	χ06V	σχ06V	χ10V	σχ10V	χ18V	σχ18V	χ23V	σχ23V	χ36V	σχ36V	χ89V
1	0.96	0.02	0.97	0.02	0.97	0.02	0.96	0.02	0.93	0.04	0.81
2	0.95	0.03	0.96	0.03	0.96	0.03	0.95	0.03	0.92	0.04	0.81
3	0.94	0.03	0.94	0.03	0.94	0.03	0.94	0.03	0.92	0.03	0.82
4	0.92	0.04	0.93	0.04	0.93	0.04	0.93	0.04	0.90	0.05	0.81
5	0.89	0.07	0.90	0.07	0.90	0.07	0.89	0.06	0.87	0.06	0.75
11	0.94	0.02	0.94	0.02	0.94	0.03	0.93	0.03	0.87	0.07	0.72
12	0.96	0.03	0.96	0.03	0.96	0.04	0.95	0.04	0.92	0.07	0.80
Thick FY Ice
Month	χ06H	σχ06H	χ10H	σχ10H	χ18H	σχ18H	χ23H	σχ23H	χ36H	σχ36H	χ89H
1	0.84	0.02	0.85	0.03	0.84	0.03	0.83	0.04	0.79	0.07	0.69
2	0.84	0.03	0.85	0.03	0.84	0.04	0.84	0.05	0.80	0.07	0.70
3	0.86	0.03	0.87	0.03	0.87	0.03	0.87	0.04	0.84	0.05	0.74
4	0.85	0.03	0.86	0.03	0.86	0.04	0.85	0.04	0.83	0.05	0.74
5	0.83	0.05	0.84	0.06	0.83	0.07	0.83	0.07	0.81	0.08	0.70
Month	χ06V	σχ06V	χ10V	σχ10V	χ18V	σχ18V	χ23V	σχ23V	χ36V	σχ36V	χ89V
1	0.94	0.02	0.94	0.02	0.93	0.03	0.91	0.03	0.86	0.06	0.75
2	0.94	0.02	0.94	0.02	0.93	0.03	0.92	0.04	0.87	0.06	0.76
3	0.95	0.02	0.95	0.02	0.95	0.02	0.95	0.03	0.91	0.04	0.80
4	0.93	0.02	0.94	0.02	0.94	0.02	0.93	0.03	0.91	0.04	0.81
5	0.92	0.03	0.93	0.03	0.92	0.04	0.92	0.04	0.89	0.05	0.78
MY Ice
Month	χ06H	σχ06H	χ10H	σχ10H	χ18H	σχ18H	χ23H	σχ23H	χ36H	σχ36H	χ89H
1	0.85	0.03	0.84	0.03	0.81	0.04	0.78	0.04	0.70	0.06	0.62
2	0.85	0.03	0.85	0.03	0.81	0.04	0.78	0.04	0.70	0.06	0.63
3	0.86	0.03	0.85	0.03	0.81	0.04	0.77	0.04	0.69	0.05	0.62
4	0.83	0.03	0.83	0.03	0.79	0.03	0.76	0.03	0.68	0.04	0.65
5	0.83	0.03	0.82	0.03	0.78	0.04	0.76	0.04	0.70	0.05	0.66
11	0.83	0.05	0.82	0.05	0.79	0.05	0.78	0.05	0.73	0.05	0.63
12	0.86	0.04	0.85	0.04	0.83	0.04	0.82	0.05	0.77	0.06	0.68
Month	χ06V	σχ06V	χ10V	σχ10V	χ18V	σχ18V	χ23V	σχ23V	χ36V	σχ36V	χ89V
1	0.96	0.02	0.95	0.02	0.91	0.02	0.86	0.03	0.77	0.06	0.67
2	0.96	0.02	0.95	0.02	0.90	0.03	0.86	0.04	0.76	0.06	0.68
3	0.96	0.02	0.95	0.02	0.90	0.03	0.85	0.04	0.75	0.06	0.67
4	0.93	0.02	0.92	0.01	0.87	0.02	0.83	0.03	0.75	0.04	0.71
5	0.91	0.02	0.91	0.02	0.87	0.02	0.84	0.03	0.77	0.05	0.72
11	0.95	0.03	0.94	0.03	0.91	0.03	0.88	0.03	0.81	0.05	0.69
12	0.97	0.03	0.96	0.02	0.93	0.03	0.91	0.04	0.84	0.06	0.73

. To illustrate the winter variability of χ^month, its dependence on month in 2020 with H polarization for the six sea ice types is presented in Figure 6.

Table 1 and Figure 6 show that the seasonal changes in χ are very dependent on the sea ice type. Monthly averaged emissivity of nilas varies in the range of 0.48–0.66 for 6.9 GHz, H, and in the range of 0.70–0.82 for 6.9 GHz,V. With the frequency increase these seasonal changes tend to decrease. At 36.5 GHz, χ of nilas varies in the range of 0.56–0.68 with H polarization and in the range of 0.77–0.83 with V polarization. At the beginning of winter and in February, we have the lowest χ of nilas at any of the AMSR2 frequencies. At the end of winter, χ of nilas grows at all the frequencies except 89 GHz and starts to be almost frequency independent in May. Moreover, the emissivity variability σ_χ for nilas is the lowest in May at all the frequencies. Young ice emissivities decrease at all the frequencies from November to February and then slightly increase up to April (except at 89 GHz); they then decrease almost by 0.1 to the end of freezing season. The emissivities of thin and medium FY ice grow slightly at the beginning of winter (except χ of thin FY ice at 89 GHz). For thick FY sea ice, the maximum χ is observed in March. Then, χ decreases by the end of winter at all the frequencies. The thinner the FY ice, the greater the decrease in χ by May and the lesser the frequency dependence of χ. The results for MY sea ice indicate the decrease in χ at the end of winter for the frequencies not exceeding 23.8 GHz and the increase in χ at 36.5 and 89 GHz. We may see also that the older the ice, the lower its seasonal variability.

For each sea ice type, the average emissivity gradients δχ and their standard deviations σ_δχ (standard deviations of δχ from their average values) were calculated for all 7 months and separately for each month: δχ1 = χ_10V − χ_06V, δχ2 = χ_23V − χ_18V and δχ3 = χ_36V − χ_18V. Figure 7 illustrates the dependence of the average δχ and σ_δχ on the sea ice type, whereas Figure 8 shows the monthly averaged values of the gradients for each of the sea ice types.

Figure 7 shows that the values of δχ decrease with the sea ice age for all gradient types. The gradient δχ3 seems to be the most sea-ice-age-dependent function. Both δχ3 and δχ2 are negative for the sea ice types thicker than 30 cm. The δχ3 of MY ice is almost 4 times higher in its absolute value than the δχ3 of thick FY ice. The gradient δχ1 is positive for all sea ice types except MY ice. The gradient variability is the largest for δχ3 for nilas and MY ice. The lowest gradient variability is observed for FY sea ice of different thicknesses.

Figure 8 shows that all the gradients increase by the end of winter for all sea ice types except nilas and medium FY ice. Nilas δχ1 and δχ2 decrease during all months; δχ3 has a local maximum in February and then also decreases by the end of winter. The changes in the monthly averaged values of δχ for medium FY ice are too small to be of any significance. The most significant changes of about 0.05 (more than 10 K in the AMSR2 measurements) are observed in δχ3 from March to May.

4. Discussion

The emissivities of different Arctic sea ice types, given in the previous section, were estimated for stable winter conditions for different Arctic regions for the characteristics of the AMSR2 measurement channels: for vertically and horizontally polarized radiation at the frequencies of 6.9, 10.65, 18.7, 23.8, 36.5 and 89 GHz. Though the results were obtained with the help of accurate Tb physical modeling and one of the best reanalysis methods to account for the atmospheric contribution to Tb, we did not risk analyzing the data from June to October to make any valued conclusions. The amount of atmospheric water is too large during these months, and the fields of emissivities obviously have atmospheric signatures. The same source of errors—atmospheric data or/and Tb model uncertainties—leads to emissivity errors during winter months. The higher the frequency, the greater the errors. Tb calculations over sea ice with the atmospheric parameters varying within the range of their climatic monthly variability show that the errors of χ may be as large as 13% for the frequencies of the C- and X- band and up to 10–20 % for 89 GHz depending on the month and Arctic region. Thus, generally all the conclusions are more justified for lower frequencies and more questionable for higher ones, especially for 89 GHz, even during winter months.

It should also be underlined that the statistical analysis of calculated sea ice emissivities has been carried out for the sea ice types of the same age gradations, but these are in reality different sea ice samples formed in various Arctic regions under different conditions. At considered frequencies the electromagnetic properties of sea ice depend on the state of the topmost sea ice layer and its roughness. Therefore, the radiative properties of sea ice in a certain month are determined by the conditions of its growth, deformations and changes in the snow cover during some time period which is long for thick sea ice types and short for thin ice samples. For example, a March thin FY ice sample most probably forms during the period of March–April, whereas a March thick FY ice sample grows starting from the beginning of the freezing period [21]. This circumstance complicates the interpretation of seasonal features of the emissions of different sea ice types. Having in mind these sources of errors, we suppose that some of the results are new and worth reporting.

The obtained results differ from the existing understanding of the FY frequency dependence. It has always been supposed that the emissivity of FY ice with a thickness exceeding 30 cm is close to 1 and decreases with frequency only for MY ice due to brine flowing out of salt pockets [37]. Our results provide evidence of an almost linear decrease in χ of FY ice with frequency above 23.8 GHz in the range of 0.93–0.95 for V polarization and in the range of 0.85–0.87 for H polarization for the ranges of 0.80–0.82 (V) and 0.72–0.75 (H). The effective emissivity of nilas and young ice increases with frequency.

A decrease in χ with frequency for all sea ice types with a thickness exceeding 30 cm is most probably associated with the beginning of brine flowing out and the appearance of air pockets with the characteristic scales of effective scatterers. At the frequencies of 6.9–23.8 GHz (wavelength λ of 1–4 cm), there are no such scatterers either in the sea ice crystal structure or in the snow. With a λ decrease, it begins to correspond to the scatterer sizes. Microwave radiation is most effectively scattered on air pockets (the higher the difference in dielectric permittivities at the media interface, the more effective the scattering). This is one of the reasons for low values of χ for frequencies higher than 18.7 GHz for MY ice which survived the period of summer melting.

The largest dispersion of σ_χ values, observed in Figure 4 (up to 0.2 for horizontal polarization and up to 0.15 for vertical polarization for frequencies up to 23.8 GHz) takes place for the ice with a thickness not exceeding 30 cm (nilas and young ice). This dispersion can be explained not only by the variability of the sea ice upper layer salinity, but also by the presence of different scale roughness on nilas and YI surfaces—a mostly variable parameter within the same sea ice type—with scales up to 5 cm. Under “surface roughness” we understand here the sea ice undulations in the range of millimeters to centimeters, strongly influencing the intensity of microwave scattering [46]. Generally, an increase in scattering coefficients leads to a decrease in emission coefficients. The older the ice, the more accumulations of ice, hummocks and other roughness types on its surface appear, leading to an increase in scattering. However, large scale roughness on the order of decimeters to a few meters seems to be of no importance for emission: we see that σ_χ either decreases with sea ice age or is almost constant at all the AMSR2 frequencies. The older the ice (FY ice, older than YI, and MY ice), the more brine flows out of the salt pockets of the upper brine layer and the greater the scattering by air pockets. This explains the observed decreases in sea ice emission for all the frequencies except 6.9 and 10.65 GHz. This means that the sizes of pockets in the sea ice crystal structure do not exceed 1.5 cm.

Nilas and young ice are characterized by a significant variability of χ even within one month, while the older the ice, the less the emissivity changes. By the end of winter, the growing nilas has a greater χ than at its beginning, which can be explained by a decrease in scattering due to a decrease in average wind speeds in the Arctic. Typically, nilas does not keep the form of a continuous layer but breaks into fragments under wind and wave impacts. These fragments are partially accumulated, forming stripes of layered ice at the boundaries of thicker ice, with a high scattering coefficient [22]. An increase in nilas χ values indirectly indicates wind weakening.

The obtained χ values for young ice (10–30 cm) have proved to be lower than those in published data [31,33]. For all considered frequencies, we obtained χ values not exceeding 0.89 with V polarization and even smaller values with H polarization. During winter χ decreases by about 0.2 on average with H polarization and by 0.1 with V polarization. The variability (σ_χ) is 0.15–0.2 depending on the month (with V polarization, it is always less than with H polarization). At 89 GHz the emission from young ice exceeds that of any other sea ice type. Young ice has the highest salinity of the upper layer (except nilas, whose radiation is strongly affected by its roughness), which means that the radiation is formed in the very topmost ice layer. Therefore, the brine pockets with scales of a few mm, scattering radiation at 89 GHz, have the least effect on the young ice χ at this frequency.

We see also that the polarization difference Δχ is much higher for young ice and nilas than for all the other sea ice types (Figure 5). This result is very important for those SIC retrieval algorithms which are based on the polarization difference (PD) near 90 GHz [41]. Underestimating the sea ice tie point leads to SIC underestimation. An example of such underestimation is shown in Figure 9a, where a MODIS optical image provides evidence of consolidated grey sea ice fields in the Bay of Buor-Khaya in the Laptev Sea on 1 November 2022 (grey ice is undoubtedly classified in optical images). A simultaneous SIC product, produced in the University of Bremen with [41], shows large regions of open water instead of sea ice in this area (Figure 9b). Our results indicate that Δχ at 89 GHz is about 0.06–0.07 for FY and MY ice types, 0.11 for young ice and 0.17 for nilas. Supposing T_s~250 K and τ~0.2 [39], we have PD = Δχ·T_s·exp(−τ) ~12 K, 22 K and 35 K for FY and MY ice, young ice and nilas, respectively. At higher atmospheric thickness, PD decreases to even lower values. Obviously, too low a PD for the sea ice tie point (9.4 K in [41]) results in high underestimation of SIC when sea ice is represented by young ice types and nilas.

FY ice of different thicknesses is characterized by a slight increase in emissivity at the beginning of winter and its decrease by the end of May at all frequencies. It would have been possible to relate the observed changes in χ to an increase in the snow cover thickness if we had not seen the growth of χ of MY ice by the end of winter at the frequencies above 36.5 GHz (Figure 6f). Meanwhile, for FY ice of different thicknesses the intra-annual variability of χ does not depend on frequency; for MY ice we observe an increase in χ at the beginning of winter for all the frequencies and a decrease in χ at the end of winter for the frequencies not exceeding 23.8 GHz. For 36.5 and 89 GHz, χ continues growing at the end of winter. Such a frequency-dependent change in the radiative properties can be explained by the smoothing of surface inhomogeneities with scales from several mm to cm due to snow cover metamorphism. The decrease in FY ice χ is most likely associated with an increase in surface scattering, which, in turn, is explained by the fact that by the end of winter, as a result of repeated hummocking, a large number of ridges of hummocks of various spatial scales as well as zones of grated ice are formed on the sea ice surface [22]. A decrease in the total sea ice salinity within the same sea ice type may also contribute to the emission decrease through an increase in volumetric scattering. Because the age gradation contains a wide range of thicknesses, by the end of winter sea ice of the same nomenclature type becomes generally thicker, with a less saline surface layer.

The gradients of the emission coefficients δχ1, δχ2 and δχ3 differ for different types of ice. The older the ice, the lower the values of δχ. Differences in δχ of FY ice of different thicknesses are not very large, but young, FY and MY ice have different, distinctive gradients δχ2 and δχ3 and rather narrow PDFs for these functions. δχ3 has the largest σ_δχ values for all sea ice types, which can be explained by the snow cover variability. δχ3 is traditionally used to estimate the snow cover depth, and recent studies report the possibility of using δχ2 to retrieve snow cover parameters [47].

The results of our work show that the sea ice type (or rather the salinity of its upper layer, correlating with the sea ice age) has a significant impact on δχ3: the older the ice, the lower the salinity and the lower the value of δχ3. It is generally accepted that dry snow cover on the sea ice surface leads to a decrease in the effective emissivity due to an increase in radiation scattering. Snow is characterized by a negative spectral gradient at frequencies of about 19 and 37 GHz. This negative gradient is caused by snow crystals having typical sizes of 5–10 mm, which leads to greater scattering of the radiation at 36.5 GHz than at 18.7 GHz due to the crystals’ characteristic sizes (5–10 mm). The microwave radiation at 18–19 GHz is effectively scattered by particles with sizes of ~1–2 cm, at 36–37 GHz by particles with sizes of ~5–10 mm and at 89 GHz by ~3 mm particles (air/brine pockets in sea ice). A negative value of the difference between the radiation at 37 and at 19 GHz is generally considered as an indicator of snow on the surface, and a decrease in δχ3 is traditionally associated with an increase in the snow cover thickness [47].

However, we observed a similar decrease with increasing sea ice thickness not only in δχ3, but also in δχ2 and δχ1 (Figure 7a). If the variability of δχ3 had been caused only by the variability of the snow cover parameters, there would have not been such a dramatic difference between δχ3 for FY and MY ice. Despite the fact that in general the thickness of the snow cover on MY ice is greater than on FY ice, the thicknesses of dry snow that has not survived the summer melting on FY and MY ice are comparable. Moreover, in such a case we could have observed a decrease in the average values of δχ3 for both FY and MY ice by the end of winter, associated with snow accumulation. Yet we observed the opposite picture: the values of δχ3 increased for MY ice starting from March and for thick FY ice from February (Figure 8). This growth can be explained by the metamorphism of the snow cover under the solar radiation, leading to a change in the snow crystal structure

One of the most interesting obtained results is the positive values of the gradient δχ1 = χ_10V − χ_06V for all sea ice types, except for MY ice. Taking into account the low variability of this gradient—both seasonal and regional—and the low impact of the atmosphere on Tb at the frequencies of 6.9 and 10.65 GHz, this parameter can be used to confidently identify the areas of MY ice. Figure 10 shows an example of the distribution of MY and FY ice given by the OSI SAF Sea Ice Type product (Figure 10a) and using the threshold δχ1 = 0.001 to separate MY and FY ice (Figure 10b). An AARI sea ice map is presented in Figure 10c as an ice expert classification product. This map shows the polygons of similar sea ice characteristics (sea ice types, forms and concentration). Up to three sea ice types are selected for every polygon, whereas the polygon color is defined by the predominant sea ice type. This example is given to illustrate that for most areas, the δχ1-based classification is similar to that of the OSI SAF product. More justified conclusions need further work.

5. Conclusions

The paper analyzes the coefficients of vertically and horizontally polarized microwave radiation χ of the main Arctic sea ice types—nilas, young ice, first-year ice of different thicknesses and multi-year ice—for winter stable conditions without melting at the measurement frequencies of the AMSR2 instrument: 6.9, 10.65, 18.7, 23.8, 36.5 and 89 GHz. Statistical estimates of the average and monthly average values of χ and their standard deviations for the considered sea ice types for the periods of January–May, November and December of 2020 have been obtained using the radiation transfer model for a non-scattering atmosphere and ERA5 reanalysis data for surface temperature and atmospheric meteorological parameter profiles. The obtained values of χ have been related to the sea ice types using the AARI sea ice maps collocated with the AMSR2 measurements.

Despite the high level of spatial and temporal averaging of the sea ice maps, the obtained results have revealed a number of regularities in the microwave emissions of nilas, young ice, FY ice of different thicknesses and MY ice at the frequencies of 6.9–89 GHz in the absence of melting. Some of these regularities are already known from the results of published studies, while others appear to be new. Thus, the increase in χ with frequency f for young ice, its weak dependence on f in the range of 6.9–36.5 GHz for FY ice and its decrease with f for MY ice are well-known facts which have only found additional confirmation in the analysis of a large amount of data. The older the ice, the lower the variability of the emissivity. In general, for almost all types of ice (especially for young ice) for all the frequencies, we obtained lower χ values than reported in published research. During winter, the average monthly values of emission coefficients of the considered sea ice types decreased on average by approximately 0.2 with horizontal polarization and by 0.1 with vertical polarization. The variability of χ within one month is ~0.15–0.20 depending on the month (the value with V polarization is always less than that with H polarization). For FY ice of different thicknesses, at all the frequencies, an increase in the emissivity at the beginning of winter and a decrease by the end of May is observed, which can be explained both by an increase in sea ice roughness by the end of winter and a decrease in the salinity of the FY ice upper layer. MY ice shows an increase in χ at the beginning of winter for all the frequencies and a decrease in the end of winter for the frequencies not exceeding 23.8 GHz. At the frequencies of 36.5 and 89 GHz, χ continues to grow at the end of winter, probably due to the smoothing of the surface roughness with scales from several mm to cm as a result of snow cover metamorphism.

Our results show that the differences (gradients) of the emissivities at different frequencies significantly depend on the sea ice type: the older the ice, the lower the salinity of its upper layer and the smaller the values of the gradients δχ1 = χ_10V − χ_06V, δχ2 = χ_23V − χ_18V and δχ3 = χ_36V − χ_18V. A significant difference in δχ3 of FY and MY ice (> 0.1), as well as an increase in δχ3 by the end of winter, cannot be explained by the dependence of δχ3 on the snow cover thickness, but rather indicate the presence in the crystal structure of ice of a significant proportion of pockets with characteristic scales of the order of cm, on which scattering at 36.5 GHz is stronger than at 18.7 GHz. The outflow of brine by the end of winter and the transformation of salt pockets into air ones leads to an increase in scattering and larger gradient contrasts.

The values of the gradient δχ1 are positive for all sea ice types, except for MY ice. Given that the variability of this gradient, both intra-annual and regional, is minimal, and the errors in the calculation of χ due to the atmosphere at the frequencies of 6.9 and 10.65 GHz are low, this parameter can be successfully used to determine the areas of MY ice in the absence of melting.

Unfortunately, the presented approach does not allow determining the surface radiation under conditions of optically thick atmospheres. Therefore, the months from June to October were excluded from consideration. Interpretation of the results in the period from November to May also seems to be superficial due to complicated nature of the sea ice–snow system and overlaying factors which cannot be correctly accounted for due to the absence of measurement data on the sea ice and snow parameters. Still, the results provide data on the general sea ice type emissivity averages and their variability and may be used in various studies related to satellite microwave measurement data analysis in the Arctic.