Ocean and Ice Features Detection using the ERS SAR Browse Images

S.T. Dokken, H. Laur
European Space Agency, ESRIN
via G. Galilei, 00044 Frascati, Italy
sverre.dokken esrin.esa.it , henri.laur esrin.esa.it

Abstract

A very small proportion (about 10%) of the ERS SAR raw data has actually been converted into SAR images. This is because of the lengthy processing time associated with the generation of high resolution products. In order to overcome this weakness, ESA has developed a SAR browse processor capable of generating low resolution images (resolution around 200 m) corresponding to the totality of an acquired SAR segment (up to 4000 km).

The first purpose of the browse images is to promote the use of high resolution data (by identification of scenes to be processed at high resolution). However this product will also be of use in applications such as the detection of large scale phenomena in oceanography and sea ice studies.

The paper provides examples of feature detection using the SAR browse images such as:
. oil slicks, eddies and bathymetric signatures in Northern Europe seas,
. ice edge determination in the Arctic ocean.

Several mosaics of browse images covering Northern Europe are also presented.

1. Introduction

The SAR instruments on-board the ERS-1 and ERS-2 satellites have generated an impressive amount of results [Ref. 1], [Ref. 2]. However most of the results achieved so far have been obtained with high resolution products because the original ESA list of ERS SAR products did not contain medium or low resolution SAR products (i.e. products with pixel spacing between 50 and 500 m). However this family of products would be beneficial for several reasons:

1. visibility of the ERS SAR acquisitions, i.e. the possibility to `see' all acquired raw data: it is interesting to note that within the ESA ground segment (Kiruna, Fucino and Maspalomas acquisition stations), only 10% of the ERS-1 raw data acquired have been processed to high resolution products. This may seem a small value but it corresponds to 38 000 high resolution products. By consequent about 90% of the ERS-1 SAR raw data have never been "visualized", i.e. transformed to image products.
2. promotion for the use of high resolution data (browse function, i.e. identification of scenes to be processed at high resolution),
3. monitoring of the SAR instrument performances,
4. scientific studies and surveillance applications of large scale phenomena (as ocean and ice features detection),
5. small data volume and thereby reduced time of distribution to users.

In order to meet these needs, two groups of products can be foreseen:

• medium resolution products (spatial resolution around 100 m), with an high quality (i.e. high radiometric resolution), used as a digital product for studies of large scale phenomena.

• browse images (spatial resolution around 200 m), with a low quality (i.e. low radiometric resolution), mainly used to assist users in the selection of high resolution products.

2. Characteristics and Handling of the Browse Images

The image generated by the ERS SAR browse processor [Ref. 3] corresponds to the full acquired stripe, has a pixel spacing of 200 m, is not calibrated and is Jpeg compressed (one stripe of 2000 km corresponds to about 1.5 Mbytes, one scene is about 70 kbytes). The ERS SAR browse processor is currently installed at UK-PAF and ESRIN. The generated browse images mainly correspond to the Kiruna acquisition station (Europe + Arctic). The browse product is used as basic SAR product to populate image servers such as the Image Browser of "earthnet online", the information service for Earth Observation users provided by ESA.

In order to show the potentialities of the browse images over land, a mosaic of Europe (limited to Kiruna acquisition visibility) is produced. The mosaic of Europe (Figure 1) is created by searching the ESA Earth Observation image browser and by stitching the SAR browse stripes generated by the SAR Browse Processor. Figure 2 shows a mosaic extract over the Alps and the spatial resolution achievable. The mosaic can also be used to show some of the ocean features commonly observed in the ERS products (paragraph 3).

Figure 1: This ERS mosaic of Europe is composed of jpeg browse images and covers an area
of about 2200 km by 2200 km. The acquisition segments were selected according to the wind speed in order to have a mosaic with relatively similar oceanographic conditions.
The ERS stripes were merged together with a contrast matching algorithm provided by ERDAS Imagine software. The mosaic is composed of 41 orbit tracks. If the SAR instrument was always switched on, the mosaic would correspond to a time range of slightly more than one month.

Figure 2: Extract of the mosaic over the Alps covering an area of about 820 km by 600 km.

A similar mosaic of the Arctic ice-cap (Figure 3) is used to study the ice detection performances. The ice-edge location on the Yermak Plateau in the Svalbard region is preferred for mosaicing because of the many reference points in the island in the Svalbard archipelago.

3. Ocean Feature Detection

Over the ocean, the SAR images the capillary waves which are formed primarily in response of the wind speed. Oceanographic phenomenon may change these capillary waves and thereby the radar backscatter. The oceanographic phenomenon detectable with SAR ranges from scales of some few meters to several kilometres. As the typical ERS SAR scene covers an area of 100 x 100 km, the browse images (100 x maximum 4000 km) enlarge the visibility of the oceanographic phenomena.

Oil slicks

Several events have confirmed the excellent ERS oil-slick detection capabilities in suitable wind conditions [Ref. 4]. Despite its pixel size of 200 m, browse images easily allow to detect oil-slicks. For example (Figure 4) probable oil-slicks are visible around oil-ridges in the North Sea. Even when the wind speed is rather strong for oil-slick detection, raw-oil spills could still be observed.

In the mosaic of Europe a large vessel North-East of Aberdeen, Scotland has black patches in its wakes. This might be a spill from cleaning the oil-tanks. Other examples are the Norwegian trench (an area with several patches) and an area North-East of Bornholm in the Baltic Sea. If they would be distributed in near real time to pollution control authorities, browse images would be definitely a valuable tool for oil-slick detection and monitoring.

Atmospheric waves

Pressure waves create over large areas when e.g. wind blows over mountain formations. The waves dampen the capillary waves on the ocean and hence are visible in SAR images. The area of Cardigan bay, West of Wales (Figure 1) has atmospheric wave formations that would not be completely displayed in a single ERS scene. The area South-East of Isle of Man (UK) has atmospheric (lee-) waves propagating to mainland Wales.

Bathymetric signatures

The structure of the ocean floor and the influence of water motion sets up ocean features such as eddies, current shears and internal waves. A typical example of a bathymetric signature is the Norwegian Trench (Figure 5) where (tidal-) water circulate into the more static water masses around. Relative motion of water masses with different properties (temperature, salinity and/or pressure) is important for climatic studies, ocean forecasting and spreading of pollutants. Typical depth of this trench is about 300-400m while the area around is mostly 200m. Eddies, current shears and internal waves are thus common signs around the edges in this trench [Ref. 5].

In the mosaic (Figure 1) several current shears are visible in the Norwegian Trench. Another current shear is visible in the Celtic sea, South of Ireland, with a length of approximately 110 km.

Ocean eddies

Ocean eddies may be visible in ERS SAR images because of surface roughness variations. One type of SAR signature frequently occurring in coastal water is the dark spiral slick while another look more like curved current shear [Ref. 5]. An excellent example from North of Gdansk, Poland Figure 6 (left) shows a dark spiral slick with a diameter of 30 km turning above a bathymetric spectacular structure, the Soedra Midsjoe Bank (Figure 6 right). A NOAA AVHRR image (Figure 6 middel), showing sea surface temperature, all three together describe the relationship between water-mass distribution and roughness field giving a "3D-look" of the ocean circulation in the area. The spatial resolution of the IR-image is 1 km and the SAR image ~200 m. It is further possible to follow and study such eddies by making multi temporal browse images.

Other examples in the mosaic are South of Plymouth, England and North in the Norwegian Trench.

Figure 6: These ERS browse image (left) and NOAA AVHRR IR image (middel) were obtained quasi simultaneously on 5 June 1995. They cover the same area, North of Poland (54-55N, 17-18E). The Sea Surface Temperature image shows the same whirling structure as in the SAR browse image. The bathymetric map (right) indicates the Soedra Midsjo Bank where the eddy is located.

Internal waves

Many examples of internal waves are seen in the mosaic (Figure 1), like North of Cap Arkona, Germany, and in the South-East part of the Norwegian Trench. It is interesting to follow the evolution of wave trains from their generation point, but this requires good spatial coverage. Tidal current generate internal waves in a 12 hour cycle and in the North Sea internal waves propagate in an average speed of 0.5 m/s [Ref. 6]. Because of this low speed it is possible to compute their evolution and history from SAR images.

Train sets are observed to propagate over more than 100 km, which make the use of browse images valuable. However, in the North Sea, the resolution in these images (~200m) limits the number of wave crests visible to a maximum of 3 crests [Ref. 6].

4. Sea Ice Feature

Remote sensing data are the primary source of information on the high-frequency temporal and spatial variability of sea ice in the polar regions. Sea ice plays an important role in the climate system and there is a strong need of historical information for sea-ice database for input to environmental forecasting models. The type of information required includes location of the ice-edge, definition of ice types, estimation of total ice concentration and concentration per type, and measurement of ice-drift direction and speed.

A key objective defined for the ERS missions was to improve sea-ice and icebergs monitoring for offshore activities and ship routing in polar regions. For that purpose, fast delivery of the data to users operating near the ice-edge or in thick pack ice is necessary. However, the limitation of the transmission system is cited as the primary constraint [Ref. 2]. Another constraint might be the difficulties identifying ice type and the border between open water and ice. The ERS SAR browse image size (in bytes) is 2000 times smaller than the corresponding PRI image (i.e. maximum of 3 Mbytes for a 4000 km stripe), and therefore its transmission performances are no more an obstacle. The capability of the browse images for ice-edge movement and static ice properties should be further examined.

Ice-edge movement

The mosaic of the Svalbard region (Figure 3) shows a distinct ice-edge dividing the open sea from the sea ice. The ice-edge is determined in the browse images in most weather conditions. The area North of Spitsbergen shows a ice-edge easy to distinguish from the more or less open water, while in the area North of Nordaustlandet it is possible to see two ice-edges (Figure 3). This is due to the movement within the 10 days separating the two acquisitions (7 and 17 March). This area also contains some slush (thin) ice and ice-floes connecting the ice pack to the islands.

Several automatic algorithms are developed to derive sea ice motion by detecting the displacement of features in pairs of satellite images. By mosaicing acquisition segments we get an easy way of viewing the ice-edge movement in the period of interest. As an example, the 120 km large area North of Spitsbergen also has a distinct ice-edge in January 1996 (Figure 7). The movement of the ice-pack is 15 km in 10 weeks.

Figure 3: Mosaic of ERS SAR browse images covering the North-West part of Svalbard. The mosaic is created out of 6 ERS-1 stripes within 10 days in March 1996 (16, 13, 10, 7, 17, 14 March `96). The North-East part of the ice-edge has unclear ice-edge because of the 10 days difference in the merged ERS stripes.
Figure 7: Two ERS browse stripes from ERS-1 in January 1996 show clearly the ice-edge. This is a valuable tool for vessels operating close to the active Artic ice-cap.

Static ice properties

The fresh new formed and sludge ice area close to the ice-edge is observed in suitable wind conditions. Drifting ice floes are detected depending on size and wind speed.

First year ice and multi-year ice is distinguished in most of the studied cases. Lanes through the ice and meltponds are observed depending on size and wind speed.

The reduced resolution compared to PRI images makes it more difficult to detect the smaller ice-floes, but do not considerably reduce the ice-edge detection capabilities. The browse image is therefore suitable for offshore activities and ship routing in polar regions. For climatic studies, the browse images associated with the ESA image browser is excellent for searching areas and large scale studies.

4. Conclusions

The new SAR browse image is a valuable supplement to the series of ERS products already available. Several comparisons to the high resolution PRI image corroborate the usefulness of the browse image for ocean and ice feature detection. Most of the browse images are available on the ESA Earth Observation information service "earthnet online".

This opens for real time browse images over Europe and in particular over the Mediterranean area. If coupled with real time information extraction such as oil slicks detection, this would certainly enlarge the use of ERS SAR data.

The Envisat satellite will be equipped with a more advanced C-band SAR. Browse and medium-resolution products (resolution ~150 m) will be systematically generated and will certainly strengthen the scientific studies and surveillance applications of large scale phenomena, in particular over the oceans and the polar regions.

5. References

[2] "Applications Achievements of ERS-1", ESA SP-1176/II, February 1996.

[3] H. Laur, P. Bally, "ERS SAR Browse Product", ESA Technical Note ES-TN-DPE-HL06, November 1995.

[4] "SAR ocean feature catalogue", ESA SP-1174, October 1994.

[5] S. T. Dokken, T. Wahl, "Observations of spiral eddies along the Norwegian Coast in ERS SAR images", FFI Rapport 96/01463, 1996.

[6] T. Wahl, S. T. Dokken, M. K. Vinje, "Statistical characterization of ocean internal wave trains observed by ERS-1 on the Norwegian Continental Shelf", submitted to Int. Journal of Remote Sensing, Remote Sensing Letters, April 1996.

Note:

Acknowledgement:

The authors would like to thank Phillipe Bally for providing valuable contributions in the starting phase of this project.

Keywords: ESA European Space Agency - Agence spatiale europeenne, observation de la terre, earth observation, satellite remote sensing, teledetection, geophysique, altimetrie, radar, chimique atmospherique, geophysics, altimetry, radar, atmospheric chemistry