Revised Sea Ice Extent Data
The first version of the Sea Ice Extent was published online in 2007 on IJIS (IARC-JAXA Information System) homepage.
This version was based on only observational data from the Advanced Microwave Scanning Radiometer-EOS (AMSR-E), which was developed by JAXA and launched on 2002.
After AMSR-E observations were stopped in Oct. 2011, sea ice extent was updated using data from WindSat developed by the US.
Since May 2012, decadal averages for the 1980's, 1990's, and 2000's were published based on SMMR and SSM/I, which were developed by the US.
Sea ice extent was reprocessed in keeping with the upgrade of AMSR-E brightness temperature data, which was completed in May 2012.
In addition, the observational data obtained from AMSR2 launched in May 2012 was incorporated after July 2012, and the long-term sea ice extent data were reprocessed.
In order to distinguish the revised sea ice extent, we define the new data as Version 2, and the earlier version as Version 1.
In the following section, we explain about the modifications in data processing from Version 1 and the influence of these modifications on sea ice extent.
(1) Modifications based on the upgrade of AMSR-E
With the upgrade of AMSR-E Level 1 brightness temperature data, geolocation errors were improved from ±10 km to ±1 km.
The Version 2 sea ice extent was calculated after the analyzing the Arctic sea ice concentration derived from the upgraded AMSR-E brightness temperature data.
In addition, observational data from other satellites (1980, 1990, 2000 and 2010's average of SMMR, SSM/I, and WindSat data) were used to calculate sea ice concentration after adjusting the brightness temperature of each sensor using AMSR-E data as standard data;
the sea ice concentration threshold, which determines the sea ice extent, was adjusted to maintain consistency with the AMSR-E sea ice extent.
The modified processing resulting from the improvement in the geometric precision of AMSR-E Level 1 brightness temperature data is described below.
(i) Cancellation of Land Expanded Mask
In Version 1, sea ice may be falsely detected along coasts because of contamination of ocean pixels by the passive microwave emission of land (the false sea ice).
To reduce the extent of false sea ice, we applied the “land expanded mask” (See Fig.1).
By improving the geometric precision of AMSR-E and reducing the false sea ice, we stopped applying the mask during the processing in Version 2.
Version 2 sea ice extent derived from SSM (or SMMR) and AMSR-E (or WindSat) increased by 13-14%, and 8-9%, respectively, compared to the sea ice extent in Version 1.
(Reference) The principle of the land expanded mask
For eliminating false sea ice near the coast, the land expanded mask is applied;
this mask considers horizontally and vertically adjacent pixels are considered as land when a 3×3 box is centered on the land pixel.
Fig.1 The principle of the land expanded mask
* Numerical values show sea ice concentration (%)
(ii) Modified Land-Ocean Mask
Version 1 employed the land-ocean mask, which is provided for SMMR and SSM/I.
However, for Version 2, we developed a new land-ocean mask to be agreeable with the improvement in AMSR-E geometric precision.
This new mask is adjusted for the footprint size of the 18-GHz band of AMSR-E (IFOV: 16 × 27 km) and is applied to the analysis of sea ice concentration.
Compared to the sea ice extent in Version 1, the sea ice extent in Version 2 has decreased by 5.0-6.0%.
Fig.2 Differences in the sea ice concentration before and after modifying the land-ocean mask
(Fig. 2 depicts values for Ver. 1 minus those for Ver. 2.)
[Click and Show Large Image]
(iii) Utilization of the Land Filter
In Version 2, the false sea ice detected near the coast decreased owing to the geometric precision improvement of the AMSR-E.
However, the false sea ice cannot be removed completely.
Hence, we applied the land filter proposed by Cho (1996).
When at least one of 3 × 3 pixels was identified as coastal area, the central pixel will be replaced with the minimum value within the 3 × 3 pixels on the basis of the assumption that the central pixel is affected by land spill over and has led to an increase in sea ice concentration.
By applying this land filter process, the sea ice extent in Version 2 was decreased by 3.5-4.5% in the summer and 1.5-2.5% in the winter as compared to the corresponding values in Version1.
Fig.3 The principle of the land filter
* Numerical values show sea ice concentration (%).
Cho, K., N. Sasaki, H. Shimoda, T. Sakata and F. Nishio, Evaluation and improvement of SSM/I sea ice concentration algorithms for the Sea of Okhotsk, J. Remote Sensing of Japan, 16(2), 47-58, 1996.
(2) Release of AMSR2 Sea Ice Extent
After the AMSR-E observations were stopped, sea ice extent was calculated based on WindSat data in Version1;
however, in Version 2, WindSat data were replaced by AMSR2 data since July 2012.
In September 2012, the Arctic sea ice extent was the smallest recorded in observation history;
however, since Version 2 employed AMSR2 data in 2012, the minimum sea ice extent was 3.18 × 106 km2, which was smaller than the corresponding value obtained using WindSat data in Version 1 by 0.3 × 106 km2.
Furthermore, there is no modification in the ranking of the successive sea ice extent owing to the latest upgrade.
Fig.4 Arctic sea ice extent during the minimum period
Last Upate:Sep. 6, 2013