Comparative Study of Sea Ice Comparative Study of Sea Ice
Concentration by Using DMSP SSM/I, Aqua Concentration by Using DMSP SSM/I, Aqua
AMSR-E and KOMPSAT-1 EOCAMSR-E and KOMPSAT-1 EOCHyangsun Han and Hoonyol LeeHyangsun Han and Hoonyol Lee
Department of Geophysics, Kangwon National University, KoreaDepartment of Geophysics, Kangwon National University, Korea
[email protected], [email protected] [email protected], [email protected]
Orbit Date Total Scenes Cloud-free Scenes
1 2005/09/25 60 14
2 2005/10/05 61 40
3 2005/10/08 61 5
4 2005/11/04 63 9
Total 245 68
70%
80%
90%
100%
70% 80% 90% 100%
EOC SIC (W+G+D)
SSM
/I S
IC
For all points---------------------------y = 1.59x - 0.64
R2 = 0.5145Avg_diff = 7.0%STD_diff = 5.6%
For black points------------------------y = 0.50x + 0.44
R2 = 0.1487Avg_diff = 4.5%STD_diff = 2.3%
(a)
70%
80%
90%
100%
70% 80% 90% 100%
EOC SIC (W)
SSM
/I S
IC
For black points------------------------y = 0.24x + 0.72
R2 = 0.3417Avg_diff = -4.2%STD_diff = 4.6%
For all points--------------------------y = 0.31x + 0.64
R2 = 0.4068Avg_diff = -8.1%STD_diff = 11.9%
(b)
70%
80%
90%
100%
70% 80% 90% 100%
EOC SIC (W+G)
SSM
/I S
IC
For black points-----------------------y = 0.55x + 0.41
R2 = 0.2661Avg_diff = 2.3%STD_diff = 2.2%
For all points-------------------------y = 0.90x + 0.07
R2 = 0.6877Avg_diff = 2.5%STD_diff = 4.2%
(c)
70%
80%
90%
100%
70% 80% 90% 100%
EOC SIC (W+G+D)
AM
SR-E
SIC
For all points--------------------------y = 1.60x - 0.6016
R2 = 0.4465Avg_diff = 2.0%STD_diff = 6.4%
For black points------------------------y = -0.33x + 1.32
R2 = 0.0507Avg_diff = -0.2%STD_diff = 1.3%
(d)
70%
80%
90%
100%
70% 80% 90% 100%
EOC SIC (W)
AM
SR-E
SIC
For black points-------------------------y = -0.02x + 1.01
R2 = 0.0709Avg_diff = -11.7%STD_diff = 13.4%
For all points---------------------------y = 0.25x + 0.7479
R2 = 0.2175Avg_diff = -13.0%STD_diff = 13.5%
(e)
70%
80%
90%
100%
70% 80% 90% 100%
EOC SIC (W+G)
AM
SR-E
SIC
For black points------------------------y = -0.08x + 1.07
R2 = 0.0729Avg_diff = -2.9%STD_diff = 3.5%
For all points---------------------------y = 0.87x + 0.1454
R2 = 0.5480Avg_diff = -2.4%STD_diff = 5.5%
(f)
70%
80%
90%
100%
70% 80% 90% 100%
SSM/I SIC
AM
SR-E
SIC
For black points------------------------y = 0.25x + 0.76
R2 = 0.2678Avg_diff = 4.5%STD_diff = 1.7%
For all points--------------------------y = 0.9901x + 0.0585
R2 = 0.8385Avg_diff = 5.0%STD_diff = 3.3%
0%
5%
10%
15%
20%
0% 5% 10% 15% 20%
EOC SIC (D)
(AM
SR-E
SIC
)-(S
SM/I
SIC
)
For black points------------------------y = 0.98x + 0.03
R2 = 0.2556Avg_diff = -2.4%STD_diff = 1.5%
For all points--------------------------y = 0.48x + 0.04
R2 = 0.4333Avg_diff = -1.7%STD_diff = 2.7%
(b)
SSM/I and AMSR-E have produced the daily sea ice concentration (SIC) products since 1987 and 2002, respectively. Many studies have focused on calibration and evaluation of sea ice algorithms and reported the disagreement of SICs between SSM/I, AMSR-E, and other sensors especially for thin ice areas. Most of them, however, used low-to mid-resolution satellite images such as SAR, Landsat TM/ETM+, Terra/Aqua MODIS, and NOAA AVHRR. Higher resolution sensors would be more favorable to observe various ice types, snow cover, cracks and leads mixed in a relatively small region, but few studies have used them for polar research due to their high cost, low success rate of obtaining cloud-free images, and low sun elevation.
In this paper, we report the evaluation results of SSM/I SIC derived from NASA Team Algorithm (NTA) and AMSR-E SIC from NASA Team2 Algorithm (NT2A) by using 6 m resolution optical images of the Antarctic sea ice obtained by EOC (Electro-Optical Camera) onboard Kompsat-1 (Korea Multi-Purpose SATellite-1). We classified various ice types in EOC images, calculated SIC of each ice types, and compared with SSM/I and AMSR-E SIC.
As we have no in situ ice thickness data, we setup new-sea ice classes such as White ice (W), Grey ice (G), Dark-grey ice (D), and Open water (O), according to grey-level of sea ice in EOC images. For the classification of ice types in EOC images, we applied supervised classification (Fig. 1(c)). EOC SIC of each ice types were compared with the results by SSM/I and AMSR-E SICs.
We have extracted SSM/I and AMSR-E SICs of the same observation date and location as Kompsat-1 EOC to compare with. We also calculated the Spatio-Temporal STandard Deviation (ST-STD) of SIC in each pixel using 3×3×3 cubic pixel window that comprises the days before and after, and neighboring pixels to see the temporal stability and spatial homogeneity of the SIC.
IntroductionIntroduction
Data and ProcessingData and ProcessingTable I. KOMPSAT-1 EOC images used in this study.
Fig 1. (a) KOMPSAT-1 EOC orbits overlying SSM/I sea ice concentration image of the Antarctic on 25 Sep. 2005 (7600 km×8300 km). (b) An example of EOC scene of the Antarctic sea ice taken on 5 Oct. 2005 (18 km×18 km). (c) Classification example of EOC ice types into W, G, D, and O (2 km×2 km).
Comparison of EOC and PM SICsComparison of EOC and PM SICs
During September to November 2005, we obtained 676 scenes from 11 orbits along the Antarctic continental edges. Among them, only 68 cloud-free scenes from 4 orbits could be used for further analysis (Table 1, Fig. 1a). Fig. 1(b) shows an example of sea ice by EOC taken on 5 October 2005. It was springtime in the Antarctic when sea ice began to melt slowly after peak expansion of ice regime. Sea ice blocks that were thickened during winter were mostly covered with snow. Leads were filled by refrozen thin ice with little snow ice cover while narrow cracks exposed open water.
We have compared various combinations of EOC SIC of W+G+D, W, and W+G with SSM/I (Fig. 2(a)-(c)) and AMSR-E SIC (Fig. 2(d)-(f)). Vertical error bar of each data point indicates the ST-STD of SIC from PM sensors. The black points were the SICs with ST-STD less than 2.5% while white dots otherwise.
When we compared SSM/I SIC with EOC (W+G; excluding D) SIC, all statistics indicated the best match among Fig. 2(a)-(c). Therefore, we could conclude that SSM/I SIC responds to W and G, but not to D.
For all points, AMSR-E SIC seemed to best correlate with EOC SIC of W+G, excluding D (Fig. 2(f)), the closest to one among Fig. 2(d)-(f). However, some black points were spread in a spurious way distorting the regression model for black points in Fig. 2(f). For the comparison of black points only, all ice types of EOC (W+G+D) SIC was better matched with AMSR-E SIC, as shown in Fig. 2(d). As the statistics obtained with the black data points only was considered more reliable than those with the all point case, it is postulated that AMSR-E SIC responds not only to white ice (W) and grey ice (G) but also to dark-grey ice (D).
Effect of Dark-Grey Ice on Passive Effect of Dark-Grey Ice on Passive Microwave SensorsMicrowave Sensors
Fig 2. The relationship between EOC SIC and SSM/I SIC (a-c) and EOC SIC and AMSR-E SIC (d-f) in Antarctic springtime, 2005. Black dots are SICs with standard deviation (vertical error bar) less than 2.5%.
Fig 3. (a) Comparison of SSM/I and AMSR-E SIC used for the comparisons with EOC SIC. AMSR-E SIC shows higher value than SSM/I SIC by approximately 5% due to the detection of ice type C for the Antarctic sea ice in the NT2A in addition to ice type A and B for the SSM/I NTA. (b) Comparison of the difference of AMSR-E SIC and SSM/I SIC with EOC SICs of dark-grey ice. It is indicating that dark-grey ice is the major source of the difference between AMSR-E SIC and SSM/I SIC.
Fig 4. Examples of EOC scenes (18 km×18 km) with different ice concentrations. (a) EOC image with large portion of dark-grey ice (W: 51.8%, G: 29.0%, D: 13.0%, O: 6.2%) taken on 25 September 2005 (AMSR-E: 89%, SSM/I: 80%). (b) EOC image with little dark-grey ice (W: 97.4%, G: 1.0%, D: 0.5%, O: 1.1%) taken on 5 October 2005 (AMSR-E: 98%, SSM/I: 96%). It supports the finding in Fig. 3(b) that the dark-grey ice concentration positively correlates with the difference of AMSR-E and SSM/I SICs.
ConclusionConclusionComparison of EOC with passive microwave sensors showed that SSM/I SIC calculated by NASA Team Algorithm (NTA) responds not only to W but also to G, but not to D, while AMSR-E SIC calculated by NASA Team2 Algorithm (NT2A) can see D as well. Sea ice types can be defined differently from different observation methods one may have used, and could not be homologous to each other. However, this study suggests that SSM/I SIC responds not only to multi-year ice and first-year ice but also to young ice, while AMSR-E SIC includes new ice as well. It is also suggested that new ice is the major source of difference between SSM/I and AMSR-E SIC, and is analogous to the ice type C defined in AMSR-E NT2A.
1, 4
2, 3
W
OG
D
(a) (c)(b)
(a) (b)
(b)(a)
W
GD
O
W
G
O
Recommended
