THIN ICE AREA EXTRACTION IN THE SEASONAL SEA ICE ZONES OF THE NORTHERN HEMISPHERE USING MODIS DATA

Sea ice has an important role of reflecting the solar radiation back into space. However, once the sea ice area melts, the area starts to absorb the solar radiation which accelerates the global warming. This means that the trend of global warming is likely to be enhanced in sea ice areas. In this study, the authors have developed a method to extract thin ice area using reflectance data of MODIS onboard Terra and Aqua satellites of NASA. The reflectance of thin sea ice in the visible region is rather low. Moreover, since the surface of thin sea ice is likely to be wet, the reflectance of thin sea ice in the near infrared region is much lower than that of visible region. Considering these characteristics, the authors have developed a method to extract thin sea ice areas by using the reflectance data of MODIS (NASA MYD09 product, 2017) derived from MODIS L1B. By using the scatter plots of the reflectance of Band 1(620nm-670nm) and Band 2(841nm-876nm )) of MODIS, equations for extracting thin ice area were derived. By using those equations, most of the thin ice areas which could be recognized from MODIS images were well extracted in the seasonal sea ice zones in the Northern Hemisphere, namely the Sea of Okhotsk, the Bering Sea and the Gulf of Saint Lawrence. For some limited areas, Landsat-8 OLI images were also used for validation

Since 1978, passive microwave radiometers, including AMSR2 onboard GCOM the earth for 40 years. The long the passive microwave observation showed clear decline trend of the Arctic sea ice cover 2018 etc.). The warming in the Fifth Assessment Report of IPCC (2014). Usually, sea ice extent is calculated from sea ice concentration derived from brightness temperatures measured by passive microwave radiometers. fundamental parameter of sea ice which can be calculated from brightness temper radiometers. There are number of sea ice concentration algorithms including NASA Team Algorithm (Cavarieli et al., 1984), Bootstrap Algorithm  (Svendsen et al. Since the heat thickness  parameter of sea ice. However, the sea ice thickness information cannot be estimated from the sea ice concentration data. on estimating ice thi acquired from passive microwave radiometers onboard satellites have been done in the past inc (2002), Martin et al. (2005), and Tamura et al. the detailed validatio thickness is still on the way. Estimating ice thickness from passive microwave radiometer is not easy. been develop brightness temperature scatter plots o polarization difference (V Sea of Okhotsk "thin ice" is defined as the ice which thickness is around less than 30cm. Cho et al

INTRODUCTION
passive microwave radiometers, including AMSR2 onboard GCOM-W satellite, have been continuously observing the earth for 40 years. The long the passive microwave observation showed clear decline trend of the Arctic sea ice cover , JAXA, 2012 The result is referred warming in the Fifth Assessment Report of IPCC (2014). Usually, sea ice extent is calculated from sea ice concentration derived from brightness temperatures measured by passive microwave radiometers. Ice concentration is the most fundamental parameter of sea ice which can be calculated from brightness temperatures measured by passive microwave radiometers. There are number of sea ice concentration algorithms including NASA Team Algorithm (Cavarieli et al., Bootstrap Algorithm (Comiso, ) vendsen et al. .1987. Since the heat flux of ice is strongly affected by the ice , i parameter of sea ice. However, the sea ice thickness information cannot be estimated from the sea ice concentration data. on estimating ice thickness from the acquired from passive microwave radiometers onboard satellites done in the past inc (2002), Martin et al. (2005), and Tamura et al. the detailed validation of the accuracy of the estimated sea ice thickness is still on the way. Estimating ice thickness from passive microwave radiometer is not easy.
developing a method to brightness temperature scatter plots o polarization difference (V-H) vs 19GHz V polarization Sea of Okhotsk , "thin ice" is defined as the ice which thickness is around less Cho et al. (2011 has done detailed stud

INTRODUCTION
passive microwave radiometers, including AMSR2 W satellite, have been continuously observing the earth for 40 years. The long-term sea ice e the passive microwave observation showed clear decline trend , JAXA, 2012 referred as an evidence of global warming in the Fifth Assessment Report of IPCC (2014). Usually, sea ice extent is calculated from sea ice concentration derived from brightness temperatures measured by passive Ice concentration is the most fundamental parameter of sea ice which can be calculated from atures measured by passive microwave radiometers. There are number of sea ice concentration algorithms including NASA Team Algorithm (Cavarieli et al., Bootstrap Algorithm  flux of ice is strongly affected by the ice ice thickness is another important parameter of sea ice. However, the sea ice thickness information cannot be estimated from the sea ice concentration data.
ckness from the brightness temperature acquired from passive microwave radiometers onboard satellites done in the past including those of Tateyama et al. (2002), Martin et al. (2005, and Tamura et al. n of the accuracy of the estimated sea ice thickness is still on the way. Estimating ice thickness from passive microwave radiometer is not easy. a method to extract thin ice area using brightness temperature scatter plots of AMSR2 19GHz H) vs 19GHz V polarization , 2014, 2015 "thin ice" is defined as the ice which thickness is around less 2011, 2012) has done detailed stud Sea ice has an important role of reflecting the solar radiation back into space. In addition, the heat strongly affected by the ice thickness difference. Therefore, ice developed a thin ice area extraction algorithm using passive microwave radiometer AMSR2 Sea of Okhotsk. The basic idea of the algorithm is to use the brightness temperature scatter plots of H) vs 19GHz V polarization. The algorithm was also have become clear. One was that some the thin ice areas were not well extracted, and the other that some of the consolidated ice were mis-extracted as thin ice areas. In this study, the authors extraction algorithm to solve these problems. By adjusting the polarization difference (V extracted in the Bering Sea. The authors also introduced an equation using the brightness and horizontal polarization to reject the thin ice area misextracted thin ice areas in the Bering Sea were well extracted. The algorithm was also applied to the Gulf of Lawrence with good result. The thin ice area extracted data are planed to be approved by JAXA

INTRODUCTION
passive microwave radiometers, including AMSR2 W satellite, have been continuously observing term sea ice extent derived from the passive microwave observation showed clear decline trend , JAXA, 2012 as an evidence of global warming in the Fifth Assessment Report of IPCC (2014). Usually, sea ice extent is calculated from sea ice concentration derived from brightness temperatures measured by passive Ice concentration is the most fundamental parameter of sea ice which can be calculated from atures measured by passive microwave radiometers. There are number of sea ice concentration algorithms including NASA Team Algorithm (Cavarieli et al., Bootstrap Algorithm (Comiso, 1995) and ASI Algorithm flux of ice is strongly affected by the ice ce thickness is another important parameter of sea ice. However, the sea ice thickness information cannot be estimated from the sea ice concentration data. Studies brightness temperature acquired from passive microwave radiometers onboard satellites uding those of Tateyama et al. (2002), Martin et al. (2005, and . However, n of the accuracy of the estimated sea ice thickness is still on the way. Estimating ice thickness from passive microwave radiometer is not easy. The authors have thin ice area using f AMSR2 19GHz H) vs 19GHz V polarization for the , 2015). In this study, "thin ice" is defined as the ice which thickness is around less 2011,2012)  Sea ice has an important role of reflecting the solar radiation back into space. In addition, the heat strongly affected by the ice thickness difference. Therefore, ice thickness is one of the most important parameters of sea ice. In our developed a thin ice area extraction algorithm using passive microwave radiometer AMSR2 Sea of Okhotsk. The basic idea of the algorithm is to use the brightness temperature scatter plots of H) vs 19GHz V polarization. The algorithm was also applicable to the Bering Sea, and could extract most of have become clear. One was that some the thin ice areas were not well extracted, and the other extracted as thin ice areas. In this study, the authors extraction algorithm to solve these problems. By adjusting the parameters of the algorithm applied to the brightness temperature polarization difference (V-H) vs 19GHz V polarization, most of the thin ice areas extracted in the Bering Sea. The authors also introduced an equation using the brightness and horizontal polarization to reject the thin ice area misextracted over consolidated ice. By applying the thin ice areas in the Bering Sea were well extracted. The algorithm was also applied to the Gulf of Lawrence with good result. The thin ice area extracted data are planed to be approved by JAXA passive microwave radiometers, including AMSR2 W satellite, have been continuously observing xtent derived from the passive microwave observation showed clear decline trend , NSIDC, as an evidence of global warming in the Fifth Assessment Report of IPCC (2014). Usually, sea ice extent is calculated from sea ice concentration derived from brightness temperatures measured by passive Ice concentration is the most fundamental parameter of sea ice which can be calculated from atures measured by passive microwave radiometers. There are number of sea ice concentration algorithms including NASA Team Algorithm (Cavarieli et al., and ASI Algorithm flux of ice is strongly affected by the ice ce thickness is another important parameter of sea ice. However, the sea ice thickness information Studies brightness temperature data acquired from passive microwave radiometers onboard satellites uding those of Tateyama  Sea ice has an important role of reflecting the solar radiation back into space. In addition, the heat thickness is one of the most important parameters of sea ice. In our developed a thin ice area extraction algorithm using passive microwave radiometer AMSR2 Sea of Okhotsk. The basic idea of the algorithm is to use the brightness temperature scatter plots of applicable to the Bering Sea, and could extract most of have become clear. One was that some the thin ice areas were not well extracted, and the other extracted as thin ice areas. In this study, the authors parameters of the algorithm applied to the brightness temperature H) vs 19GHz V polarization, most of the thin ice areas extracted in the Bering Sea. The authors also introduced an equation using the brightness over consolidated ice. By applying the thin ice areas in the Bering Sea were well extracted. The algorithm was also applied to the Gulf of Lawrence with good result. The thin ice area extracted data are planed to be approved by JAXA paring the in situ data observed by optical sensor and MODIS on Aqua/Terra thickness is less than the ice thickness difference can be detected with optical sensors such as RSI and MODIS. used as the reference of identifying thin ice areas, of extract thin ice comparing with the MODIS images. The result of applying the thin ice area extraction algorithm Thin Ice Algorithm, zones of the northern hemisphere including the Sea of Okhotsk, Bering Sea, and Gulf Figure 1 show the map of the test sites analyzed study which are the Sea of Okhotsk, the Bering sea and Gulf of Saint Lawrence All three are seasonal se zones of the northern hemisphere. The Sea of Okhotsk is located north side of Hokkaido, Japan, surrounded by the Island of Sakhalin and eastern Siberian coast, Kamchatka Peninsula and Kuril Islands. The sea of the most southern seasonal sea ice zones in the northern hemisphere, and many thin

THIN ICE AREA EXTRACTION IN THE SEASONAL SEA ICE ZONES OF THE NORTHERN HEMISPHERE USING ASMR2 DATA
kohei.cho@tokai sea ice, passive microwave radiometer, global warming, GCOM Sea ice has an important role of reflecting the solar radiation back into space. In addition, the heat flux of ice in thin ice areas is thickness is one of the most important parameters of sea ice. In our developed a thin ice area extraction algorithm using passive microwave radiometer AMSR2 Sea of Okhotsk. The basic idea of the algorithm is to use the brightness temperature scatter plots of applicable to the Bering Sea, and could extract most of have become clear. One was that some the thin ice areas were not well extracted, and the other extracted as thin ice areas. In this study, the authors have improved t parameters of the algorithm applied to the brightness temperature H) vs 19GHz V polarization, most of the thin ice areas extracted in the Bering Sea. The authors also introduced an equation using the brightness temperatures difference of 89GHz vertical over consolidated ice. By applying the thin ice areas in the Bering Sea were well extracted. The algorithm was also applied to the Gulf of Lawrence with good result. The thin ice area extracted data are planed to be approved by JAXA as a AMSR2 re in situ ice thickness measurement optical sensors such as on Aqua/Terra. The result suggested that if the ice thickness is less than 30cm, under the less snow cover condition, thickness difference can be detected with optical sensors MODIS. In this study, as the reference of identifying thin ice areas, of extract thin ice area with AMSR2 data are validated by comparing with the MODIS images. The result of applying the thin ice area extraction algorithm Thin Ice Algorithm, to AMSR2 data in the seasonal sea ice zones of the northern hemisphere including the Sea of Okhotsk, Bering Sea, and Gulf of St. Lawrence are presented in this paper.

SITES
map of the ed in this the Sea of Bering sea and Gulf of Saint Lawrence. All three are seasonal sea ice zones of the northern he Sea of located at the north side of Hokkaido, Japan, surrounded by the khalin and eastern Siberian coast, Kamchatka Peninsula and Kuril Islands. The sea is one of the most southern seasonal sea ice zones in the northern many thin Figure. 1

THIN ICE AREA EXTRACTION IN THE SEASONAL SEA ICE ZONES OF THE NORTHERN HEMISPHERE USING ASMR2 DATA
kohei.cho@tokai-u.jp sea ice, passive microwave radiometer, global warming, GCOM-W flux of ice in thin ice areas is thickness is one of the most important parameters of sea ice. In our developed a thin ice area extraction algorithm using passive microwave radiometer AMSR2 Sea of Okhotsk. The basic idea of the algorithm is to use the brightness temperature scatter plots of AMSR2 19GHz polarization applicable to the Bering Sea, and could extract most of have become clear. One was that some the thin ice areas were not well extracted, and the other have improved the thin ice area parameters of the algorithm applied to the brightness temperature H) vs 19GHz V polarization, most of the thin ice areas temperatures difference of 89GHz vertical over consolidated ice. By applying the above two methods to thin ice areas in the Bering Sea were well extracted. The algorithm was also applied to the Gulf of as a AMSR2 research product.
ckness measurement s such as RSI on FORMOSAT . The result suggested that if the ice under the less snow cover condition, thickness difference can be detected with optical sensors In this study, the MODIS images as the reference of identifying thin ice areas, area with AMSR2 data are validated by comparing with the MODIS images. The result of applying the thin ice area extraction algorithm, hereafter refe to AMSR2 data in the seasonal sea ice zones of the northern hemisphere including the Sea of Okhotsk, of St. Lawrence are presented in this paper.

Figure. 1 Map of the t THIN ICE AREA EXTRACTION IN THE SEASONAL SEA ICE ZONES OF
flux of ice in thin ice areas is thickness is one of the most important parameters of sea ice. In our developed a thin ice area extraction algorithm using passive microwave radiometer AMSR2 for the AMSR2 19GHz polarization applicable to the Bering Sea, and could extract most of the thin have become clear. One was that some the thin ice areas were not well extracted, and the other he thin ice area parameters of the algorithm applied to the brightness temperature were also well temperatures difference of 89GHz vertical above two methods to thin ice areas in the Bering Sea were well extracted. The algorithm was also applied to the Gulf of St.
search product.
ckness measurement result with the on FORMOSAT-2 . The result suggested that if the ice under the less snow cover condition, thickness difference can be detected with optical sensors MODIS images are as the reference of identifying thin ice areas, the possibility area with AMSR2 data are validated by comparing with the MODIS images. The result of applying the ferred to as the to AMSR2 data in the seasonal sea ice zones of the northern hemisphere including the Sea of Okhotsk, of St. Lawrence are presented in this paper.

Map of the test sites
flux of ice in thin ice areas is thickness is one of the most important parameters of sea ice. In our AMSR2 19GHz polarization the thin have become clear. One was that some the thin ice areas were not well extracted, and the other he thin ice area parameters of the algorithm applied to the brightness temperature l temperatures difference of 89GHz vertical above two methods to St. The brightness temperature data microwave radiometer were used in this study 2012 and AMSR2 has been observing the earth for over 6 ye Table 1 show concentration data derived from AMSR2 data using Bootstrap Algorithm (Comiso, 2009) order to identify thin i MODIS onboard Aqua satellite were used show the specifications of MODIS. As for MODIS, only the Band 1 and 2 which have the were used in this study. distribution of sea ice can be observed from MODIS images. Since Aqua and GCOM the frame work of constellation of same area four minu onboard GCOM effective validation data for AMSR2 data.   However, it is difficult to identify ice thickness differences or thin ice areas from the image. On the other hand, sea ice distributions can be observed image of MODIS shown on Figure 2(b) sample areas of thin sea ice, big ice sea ice are selected can be found in the the northernmost part of the Pacific Ocean, which is surrounded by the Siberia, the Kamchatka Peninsula, the Alaska Peninsula and the Aleutian Islands. The Bering Sea is connected to Arctic Ocean by the Bering Strait. The Gulf of St. Lawrence is a kind of inland sea located in eastern Canada. It is the outlet of North America's Great Lakes via the Atlantic Ocean.

ANALY
The brightness temperature data radiometer AMSR2 in this study. GCOM 2012 and AMSR2 has been observing the earth for over 6 ye Table 1 shows the specifications of AMSR2. concentration data derived from AMSR2 data using Bootstrap (Comiso, 2009) order to identify thin ice area MODIS onboard Aqua satellite were used show the specifications of MODIS. As for MODIS, only the Band 1 and 2 which have the were used in this study. Under the cloud free condi distribution of sea ice can be observed from MODIS images.
Aqua and GCOM-W are in the same orbital "track" under the frame work of the NASA's A of satellites, MODIS onboard Aqua observe same area four minutes after the observation of AMSR2 onboard GCOM-W. Therefore, MODI effective validation data for AMSR2 data.  However, it is difficult to identify ice thickness differences or thin ice areas from the image. On the other hand, ributions can be observed MODIS(Band 1 to blue and red, Band 2 to green) Figure 2(b). In ou of thin sea ice, big ice are selected in this study as sown on Figure   can be found in the sea. The Bering Sea is part of the Pacific Ocean, which is surrounded by the Siberia, the Kamchatka Peninsula, the Alaska Peninsula The Bering Sea is connected to Arctic Ocean by the Bering Strait. The Gulf of St. Lawrence is a kind d in eastern Canada. It is the outlet of North America's Great Lakes via the Saint Lawrence

ANALYZED DATA
The brightness temperature data acquired fr AMSR2 onboard GCOM-W was launched by JAXA in 2012 and AMSR2 has been observing the earth for over 6 ye the specifications of AMSR2. concentration data derived from AMSR2 data using Bootstrap were also used in this study. I ce areas, data collected by optical sensor MODIS onboard Aqua satellite were used as reference show the specifications of MODIS. As for MODIS, only the Band 1 and 2 which have the highest spatial Under the cloud free condi distribution of sea ice can be observed from MODIS images. are in the same orbital "track" under the NASA's A-Train , MODIS onboard Aqua observe tes after the observation of AMSR2 W. Therefore, MODIS data is one of the most effective validation data for AMSR2 data. However, it is difficult to identify ice thickness differences or thin ice areas from the image. On the other hand, ributions can be observed in the (Band 1 to blue and red, Band 2 to green) our thin ice algorithm, we first of thin sea ice, big ice floe, open water and mixed in this study as sown on Figure   The Bering Sea is located in part of the Pacific Ocean, which is surrounded by the Siberia, the Kamchatka Peninsula, the Alaska Peninsula

IFOV
The Bering Sea is connected to Arctic Ocean by the Bering Strait. The Gulf of St. Lawrence is a kind d in eastern Canada. It is the outlet of North Saint Lawrence River into the ED DATA acquired from passive onboard GCOM-W satellite W was launched by JAXA in 2012 and AMSR2 has been observing the earth for over 6 ye the specifications of AMSR2. The ice concentration data derived from AMSR2 data using Bootstrap also used in this study. I , data collected by optical sensor as reference. However, it is difficult to identify ice thickness differences or thin ice areas from the image. On the other hand, more detailed in the color composite (Band 1 to blue and red, Band 2 to green) thin ice algorithm, we first select floe, open water and mixed in this study as sown on Figure 2 and 3. located in part of the Pacific Ocean, which is surrounded by the Siberia, the Kamchatka Peninsula, the Alaska Peninsula The Bering Sea is connected to Arctic Ocean by the Bering Strait. The Gulf of St. Lawrence is a kind d in eastern Canada. It is the outlet of North River into the passive W satellite W was launched by JAXA in 2012 and AMSR2 has been observing the earth for over 6 years.
The ice concentration data derived from AMSR2 data using Bootstrap also used in this study. In , data collected by optical sensor  shows the scatter plot of AMSR2 19GHz V versus H) of the Sea of Okhotsk observed on February27, thin ice and ■ our Thin Ice Algorithm by applying the following two Tb) of 19GHz AMSR2.

19GHzH) Vs 19GHzV 27, 2013) (b) AMSR2 ice concentration . Comparison of AMSR2 and MODIS images. (d)Open water of different ice types extracted from
shows the scatter plot of AMSR2 19GHz V versus February27, ■ m by applying the following two 19GHz The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XLII-3/W7, 2019 TC III WG III/2,10 Joint Workshop "Multidisciplinary Remote Sensing for Environmental Monitoring", 12-14 March 2019, Kyoto, Japan It is impossible to identify ice concentration areas. ice area with 80% or higher sea ice concentration. words, our target is thin ice area which concentra than 80%/ The microwave brightness temperature of water is much lower in H polarization than in that of V polarization. Since thin ice areas are rather wet temperature of thin ice area polarization than in that of V polarization. the microwave brightness temperature of consolidated ice does not show big difference between V and H polari Considering these characteristics, the authors have introduced equation (2) for extracting thin ice area.

Sea of Okhotsk
Firstly, the authors have applied the AMSR2 data of the Sea of Okhotsk. Figure  AMSR2 sea ice concentration image of February 27, 2013. The cyan areas in the image show the "thin ice areas" extracted using AMSR2 data using equations (1) and (2)

Bering Sea
The authors have applied the scenes of AMSR2 data 19, 2016. Figure  versus 19GHz (V 2016. The blue as thin ice area with equation (1) and (2). extracted sea ice areas( It is impossible to identify ice thickness difference ice concentration areas. The equation (1) is used ice area with 80% or higher sea ice concentration. rget is thin ice area which concentra The microwave brightness temperature of water is much lower in H polarization than in that of V polarization. thin ice areas are rather wet of thin ice area polarization than in that of V polarization. the microwave brightness temperature of consolidated ice does not show big difference between V and H polari Considering these characteristics, the authors have introduced for extracting thin ice area.

Sea of Okhotsk
he authors have applied the AMSR2 data of the Sea of Okhotsk. Figure  AMSR2 sea ice concentration image of February 27, 2013. The cyan areas in the image show the "thin ice areas" extracted sing AMSR2 data using equations (1) and (2). are specified for the Sea of Okhotsk. areas were overlaid on the simultaneously collected MODIS image for evaluation as shown on Figure  all

Bering Sea
The authors have applied the AMSR2 data for the Figure 6 shows the scatter plot of AMSR2 19GHz versus 19GHz (V-H) of the Bering Sea observed on blue meshed area represents the thin ice area with equation (1) and (2). extracted sea ice areas(■) are distributed outside of the blue ice thickness difference The equation (1) is used ice area with 80% or higher sea ice concentration.
rget is thin ice area which concentra The microwave brightness temperature of water is much lower in H polarization than in that of V polarization. thin ice areas are rather wet, the microwave brightness of thin ice areas become much lower in H polarization than in that of V polarization. the microwave brightness temperature of consolidated ice does not show big difference between V and H polari Considering these characteristics, the authors have introduced for extracting thin ice area.

EXTRACTED RESULT
he authors have applied the Thin Ice Algorithm AMSR2 data of the Sea of Okhotsk. Figure  AMSR2 sea ice concentration image of February 27, 2013. The cyan areas in the image show the "thin ice areas" extracted sing AMSR2 data using equations (1) and (2). specified for the Sea of Okhotsk. areas were overlaid on the simultaneously collected MODIS image for evaluation as shown on Figure 5  The authors have applied the Thin Ice Algorithm the Bering Sea shows the scatter plot of AMSR2 19GHz Bering Sea observed on meshed area represents the thin ice area with equation (1) and (2). It is clear that the un are distributed outside of the blue ice thickness difference in the low The equation (1) is used to extract ice area with 80% or higher sea ice concentration. In other rget is thin ice area which concentration is higher The microwave brightness temperature of water is much lower in H polarization than in that of V polarization. the microwave brightness become much lower in H polarization than in that of V polarization. On the other ha the microwave brightness temperature of consolidated ice does not show big difference between V and H polarization. Considering these characteristics, the authors have introduced

EXTRACTED RESULT
Thin Ice Algorithm AMSR2 data of the Sea of Okhotsk. Figure 5(a) shows the AMSR2 sea ice concentration image of February 27, 2013. The cyan areas in the image show the "thin ice areas" extracted sing AMSR2 data using equations (1) and (2)   In this study, authors have applied the AMSR2 Algorithm which was Okhotsk also The extracted thin sea ice areas were validated by comparing simultaneously collected MODIS images. around 10 scenes for the Gulf of St. Laurence. ice areas identified in MODIS images were well extracted from AMSR2 data by applying the algorithm some tuning of the parameters may algorithm when applying the algorithm of the Northern Hemisphere. thin ice product from AMSR2 data the research product of AMSR2.