Sea ice concentration over the Antarctic Ocean from satellite pulse altimetry

Earth Sciences

© Science China Press and Springer-Verlag Berlin Heidelberg 2010 earth.scichina.com www.springerlink.com

*Corresponding author (email: yuande_yang@yahoo.com.cn)

• RESEARCH PAPER • January 2011 Vol.54 No.1: 113–118

doi: 10.1007/s11430-010-4108-7

Sea ice concentration over the Antarctic Ocean from satellite

pulse altimetry

YANG YuanDe

, E DongChen

, WANG HaiHong

, CHAO DingBo

,

HWANG CheinWay

, LI Fei

& AI SongTao

1 _{Chinese Antarctic Center of Surveying and Mapping, Wuhan University, Wuhan 430079, China;} 2 _{School of Geodesy and Geomatics, Wuhan University, Wuhan 430079, China;} 3

Key Laboratory of Surveying and Mapping Technology on Island and Reef, State Bureau of Surveying and Mapping, Qingdao 266510, China;

4 _{Department of Civil Engineering, “National” Chiao Tung University, Hsinchu, Taiwan, China} Received October 14, 2009; accepted August 23, 2010; published online October 16, 2010

Sea ice concentration (SIC) is an important parameter in characterizing sea ice. Limited by the environment and the spatial ex-tent of observation, it is difficult for field work to meet the needs of a large-scale SIC study. However, with its many advan-tages, such as the ability to make large-scale, high-resolution and long-duration observations, the altimeter can be used to de-termine SIC on a large scale. Using the correspondence between the satellite pulse altimeter waveform and reflector property, waveform classification is employed. Moreover, this paper develops an algorithm to obtain the SIC from altimeter waveforms. In an actual computation, Pyrz Bay in the Antarctic is taken as an experimental region, and one-year and seasonal SICs are de-rived from ERS-1/GM waveforms over this study area. Furthermore, altimetric SICs are compared with those of SSMR SSM/I. The results show that the spatial distribution and the regions of maximum SIC determined employing these two methods are consistent. This demonstrates that altimeter data can be used to monitor sea ice.

sea ice concentration, Antarctic, altimeter waveform, SSMR SSM/I, ERS-1/GM

Citation: Yang Y D, E Dong-Chen, Wang H H, et al. Sea ice concentration over the Antarctic Ocean from satellite pulse altimetry. Sci China Earth Sci, 2011, 54: 113–118, doi: 10.1007/s11430-010-4108-7

Sea ice concentration (SIC) is the proportion of the ocean area actually covered by ice in an area. Because SIC reflects the dynamic change in ice sheets and shelves, and affects the interaction between the ocean and atmosphere, thus it has a crucial role in the ocean and atmosphere circulation. Therefore, information on the SIC is useful in the research on global climate change and shipping safety. However, it is difficult for a conventional research ship to pass through a region of sea ice. Moreover, because of technique difficul-ties, poor accuracy, high cost and a limited observation

range, field work employing an icebreaker can not even meet the needs of a large-scale study on sea ice. Remote sensing, especially passive microwave sensing, is usually employed to obtain the SIC. For example, Cao and Jin [1] summarized the advantages and disadvantages of different remote sensing techniques. In addition, Wang et al. [2] used AMSR-E data to analyze the multi-year SICs of the Arctic Ocean.

One remote sensing technique uses a satellite microwave radar altimeter, which was initially designed for accurate measurements of marine gravity, sea surface height and geodynamics parameters [3]. Consequently, this technique records a large number of data to improve our

understand-A waveform is a direct measurement by a satellite radar altimeter. Because its shape is determined by the properties of the reflecting surface property, there is correspondence between the altimeter radar waveform and reflector. When two reflectors have a similar reflecting property, they gen-erate similar waveforms, and vice versa. On the base of this relationship, many experiments have been carried out over different regions. For example, by analyzing waveforms reflected by polar regions, Laxon [8] demonstrated that sea ice mainly produces specular waveforms, which are very different from diffuse waveforms produced by open oceans.

Waveform classification can be employed to determine the waveform type corresponding to a certain reflector. Ba-sic waveforms can be classified into categories of typical diffusion with only one leading edge over open-ocean, typical diffusion with multiple leading edges over coastal waters and diffuse-like pulses over sea ice. Some authors have carried out waveform retracking after obtaining the specular waveform from sea ice with the aim to improve the accuracy of the sea surface height and sea ice thickness [9–11]. This paper develops an index of the sea ice wave-form to obtain the SIC and its spatial distribution over the Antarctic Ocean. Firstly, the waveform index is calculated for every altimeter waveform, with this index describing an average over the illuminated area. The index is then com-pared with a threshold value. A reflector with an index greater than the threshold value is identified as sea ice, and that with an index less than the value as ocean. Finally, the gridded SICs on a certain scale are precisely computed over the studied ocean.

1 Waveform data and SSMR SSM/I

Among different satellite altimeters, the coverage of the ERS series is almost global, up to 81.5°S–81.5°N, covering the entire Antarctic Ocean and part of the Arctic Ocean. Therefore, these data can be used to monitor sea ice over the Antarctic Ocean. ERS-1, launched in 1991, was the first European Remote Sensing Mission. Its objective is repeat measurements of the global environment. Typical observa-tion cycles have periods of 3, 35 and 168 days, with the density of the 168-day geodetic mission (GM) phase being the highest among these phases. The altimeter data used in this paper are ERS.ALT.WAP obtained by AVISO in the

tion over this region is about 4 km.

To assess the effectiveness of the altimeter-derived SICs, SSMR SSM/I data, provided by the National Snow and Ice Data Center, are used as the remote sensing data in this pa-per [13]. These gridded data were obtained by Nimbus 7 SSMR and DMSP SSM/I sensors. Data were derived em-ploying the NASA method, which is based on the polariza-tion ratio and frequency gradient, and had 25 km×25 km spatial resolution, as well as daily and monthly temporal resolutions. Furthermore, the data are a polar stereographic projection, with a range of [0, 250]. Thus, the remotely sensed SCIs can be obtained from gridded data by multi-plying by 0.4.

2 Altimeter waveform and waveform

classifica-tion

An altimeter waveform is a curve of the return power versus time, sampled in setting time by an altimeter with automatic gain. According to the workings of the altimeter, the al-timeter waveform consists of three parts: thermal noise, a rapidly rising leading edge and a decaying trailing edge. As the reflection of the ocean has a Gaussian distribution, the mean returned waveform over open-oceans can be ex-pressed as [14]: 1, 0, ( ) 1 exp( ), 0, 2 2 t A t P t erf t α t σ < ⎡ ⎛ ⎞ ⎤ ⎧ = _{⎢ ⎜ ⎟+ ⎨⎥} − ≥ ⎝ ⎠ ⎩ ⎣ ⎦ (2)

where A is the amplitude of the waveform, σ controls the slope of the leading edge, t is the time difference between the sample time and the center of the leading edge, α is the exponential decay of the trailing edge, and erf is an error function. This type of waveform is referred to as an ocean waveform, and has a steep leading edge and a slowly de-caying trailing edge.

The properties of the reflectance of sea ice differ from the assumptions of the Brown model, meaning that the sea ice waveform usually deviates from the ocean waveform. Figure 1 shows the typical waveforms for open ocean and sea ice. The ocean waveform with diffuse reflection is re-ferred to as a diffuse waveform, while the typical sea ice waveform is referred to as a specular waveform, with specular or near-specular reflection. Both the leading and

Figure 1 Typical waveform. (a) Open-ocean; (b) sea ice.

trailing edge of the specular waveform are steep, with the waveform being needle-like.

The key to monitoring sea ice via satellite altimeter is to detect a sea ice waveform or specular waveform. The aim of waveform classification is to identify the two waveform types. Dwyer and Godin [15] attempted to qualitatively analyze sea ice using waveforms obtained by satellite. Re-search on waveform classification began in 1994 when Laxon [8] found that the sea ice waveform differs from the open ocean waveform. Peacock and Laxon [9] developed the concept of pulse peakiness (PP) and successfully ex-tracted the specular waveform over the Arctic Ocean. Pulse peakiness for the ERS-1 waveform is given by

max 64 5 31.5 ( ) i P PP P i = × =

∑

where Pmax is the waveform peak power and P(i) is the power of the ith gate. The parameter PP is employed as the index of waveform classification in this paper. A threshold value of 1.8 is used to differentiate diffuse and specular waveforms. A waveform with PP less than 1.8 is deter-mined to be diffuse and that with PP greater than 1.8 to be specular. Recently, PP has been widely used in waveform classification [6].

3 Sea ice concentration

The SIC, denoted I, on a grid can be obtained as

I

100S

I

S

= , (4)

where SI and S are the area covered by ice and the entire area of the grid.

It is known that the altimeter waveform represents the mean over the illuminated area. In estimating the SIC, we make the following assumptions.

1) The altimeter waveform represents the mean SIC over the illuminated area, and the same radius R of the wave-forms of the grid is considered.

2) The SIC I is equal to the average along-track SIC

Itrack.

Considering the resolution of SSMR SSM/I data, the SIC is calculated on a 12′×12′ grid. We suppose there are N tracks in a grid, with Pi waveforms corresponding to the ith

track. The SIC of this grid is then estimated as follows. 1) We compute PP of the jth waveform corresponding to the ith track using eq. (3) and compare PP with 1.8 to iden-tify the waveform type, with Wij=1 for a specular waveform

and Wij=0 for a diffuse waveform.

2) We obtain the along-track sea ice area SItrack and the entire along-track area Strack for latitude ϕij and waveform

type Wij: 2 Itrack _{1 1} 2 track 1 1 π cos , π cos . i i N P ij ij i j N P ij i j S R W S R ϕ ϕ = = = = = ⋅ ⋅ = ⋅

∑ ∑

∑ ∑

3) We substitute SItrack and Strack for SI and S in eq. (4),

2 2

track ₁ i₁π cos ₁ i₁π cos .

N P N P

ij ij ij

i j i j

I =

∑ ∑

∑ ∑

I=100. However, when all waveforms are diffuse, SItrack=0 and I=0. The actual situation is between these the two cases, and hence, the range of I is 0–100.

4 One-year and seasonal SIC

Pyrz Bay is defined by 55°–95°E and 55°–72°S. Here, IERS and ISSM represent the SICs obtained from the ERS-1/GM waveform data and SSMR SSM/I data. In this section, yearly and seasonal SICs and the distributions of IERS and

ISSM are estimated. These two results are compared and analyzed. Finally, the differences between the two methods are explained.

4.1 One-year SIC

1) IERS is greater than ISSM overall.

2) The highest SICs are for high-latitude waters near land. However, the locations in Figure 2(a) are coastal waters near land, and the corresponding positions in Figure 2(b) are waters near land with a land boundary.

These two differences can be attributed to the following factors.

1) The satellite altimeter is more sensitive than SSMR SSM/I. Because the reflectivity of sea ice is greater than that of ocean, the waveform is specular even if there is less than 1% sea ice coverage in the illuminated area [9]. The grid is identified as sea ice if the PP is greater than the threshold value. However, Cao and Jin [1] showed that the pixel of a remote sensing image is very coarse for SSMR SSM/I. Therefore, when the waveform for a pixel with sea ice information is weak, the corresponding result may be corrupted. The NASA method was employed to obtain ISSM; the result for spring is less accurate than that for autumn, and even less for summer as the ice melts. Furthermore, the results underestimate values over snowy and other primary sea ice and high-density sea ice. Hence, IERS is higher than

ISSM, such as in the case of maximum values.

2) The satellite radar waveform may be contaminated by reflectance from land as the altimeter approaches land. In this case, a complex waveform with multiple leading edges is received, and this waveform differs from a specular or diffuse waveform, making PP an invalid index of waveform classification. Therefore, ISSM is more accurate than IERS

SSMR SSM/I 90 6 40.7 ±19

Idiff 39.9 −27.8 12.5 ±13.2

4.2 Seasonal change of sea ice concentration

In the Antarctic, spring is from August to October, summer from November to January, autumn from February to April and winter from May to June. The seasonal IERS values are calculated from 1994-04–1995-03 altimeter data, as shown in Figure 3(a), while the corresponding SSMR SSM/I result is shown in Figure 3(b).

Figure 3 shows that the seasonal SICs obtained employ-ing the two different approaches have the followemploy-ing characteristics.

1) The SICs are the same for different seasons. For ex-ample, there is no sea ice around 68°, 80° and 93°E. How-ever, the results show that altimetry can detect floating sea ice.

2) The derived SICs are similar.

3) Among the four seasons, the minimum sea ice distri-bution is in summer, with there even being regions without sea ice The extent of sea ice increases from autumn to spring, with the maximum being in spring.

Figure 3 shows that the seasonal IERS is greater than the overall ISSM. Thus, the seasonal SICs are similar to the one-year result. The statistics are then derived for the four seasons (Table 2). The mean Idiff is about 14.8. The seasonal results differ, indicating different sea ice distributions, and

Figure 3 Seasonal changes in the SICs from 1994-04 to 1995-03 for Pyrz Bay. (a) ERS-1/GM; (b) SSMR SSM/I. Table 2 Statistics of of differences in the seasonal SICs

Season Max Min Mean Standard deviation Spring 39.9 −29.8 17.7 ±14.7

Summer 39.9 −36.2 15.1 ±16.9

Autumn 39.9 −37.7 16.3 ±18.0

Winter 39.9 −36.0 10.3 ±14.9

thus affect the result. For example, the summer sea ice dis-tribution is less, which influences not only ISSM in the NASA method but also the altimeter waveform and IERS. However, the statistics show that satellite altimetry is an effective tool to monitor sea ice.

5 Conclusions and discussions

Using the correspondence between an altimeter radar waveform and the reflector, this paper discussed the deter-mination of the SICs and their distribution, and introduces the principle of determination and computation procedure. Pyrz Bay was taken as an experimental area, and the one-year SICs were obtained from ERS-1/GM waveforms. To assess the effectiveness of this method, the altimetric re-sult was compared with SSMR SSM/I remote sensing data. The comparison showed that the two methods give similar spatial distributions of the sea ice and its concentration. However, it also indicated that the former result is overall

larger than the latter. The factors relating to the difference were analyzed, and difference statistics indicated that the method of determining the SIC via altimetry works well.

Moreover, seasonal SICs were derived from the same data. The results showed that the extent of the sea ice dis-tribution is a minimum in summer, increases from autumn to spring and is at maximum in spring. Comparing the al-timetry and remote sensing results also indicated that altim-etry can be employed to monitor SICs. The results indicated that differences between the results of the two methods de-pend on the season.

To assess the method of determining the SIC effectively, the results of field work and altimetry need to be quantita-tively compared. Factors affecting the difference, such as the sensitivity of the altimetric results to environmental change and the effects of assumptions, should then be miti-gated. Finally, further research should involve complemen-tary analysis of the altimetry and remote sensing results and determination of the SIC.

We thank the anonymous reviewers for suggesting improvements to the manuscript and AVISO for providing the ERS-1/GM waveform data. This work was supported by National Key Technology R & D Program (Grant No. 2006BAB18B01), the National Natural Science Foundation of China (Grant No. 40806076), Antarctic Exploration Fundamental Project (Grant No. 14699907111091), Chinese Polar Strategic Research Foundation (Grant No. 20080203), and Key Laboratory of Surveying and Mapping Technology on Island and Reef of the State Bureau of Surveying and Map-ping (Grant No. 2009B04).

6 Lee H K. Radar altimetry methods for solid earth geodynamics stud-ies. Doctoral Dissertation. Columbus: Ohio State University, USA, 2008

7 Frédérique R, Soazig P. Antarctic ice sheet and radar altimetry: A Review. Remote Sens, 2009, 1: 1212–1239

8 Laxon S. Sea ice altimeter processing scheme at the EODC. Int J Remote Sens, 1994, 15: 915–924

Boulder, Colorado USA: National Snow and Ice Data Center, 1996, updated 2008

14 Brown G S. The average impulse response of a rough surface and its applications. IEEE Trans Anttennas Propag, 1977, AP-25: 67–74 15 Dwyer R E, Godin R H. Determining sea-ice boundaries and ice

roughness using GEOS-3 altimeter data. NASA Contractor Report, NASA, 1980