SEARCH

SEARCH BY CITATION

Keywords:

  • Gulf Stream;
  • SEASAT SAR image;
  • internal waves;
  • seamount wakes;
  • solitary wave packet

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Study Area
  5. 3. SAR Image Interpretation
  6. 4. Dynamical Analysis
  7. 5. New Observations
  8. 6. Conclusions and Discussion
  9. Acknowledgments
  10. References

[1] A SEASAT synthetic aperture radar (SAR) image taken over the Gulf Stream region shows streak-like patterns. The physics of their generation and interaction with the Gulf Stream are disputed. This study seeks a convincing interpretation for the SAR imagery patterns. Bathymetric maps show that the sea floor area beneath the streaks is the northeast Hoyt Hills, where isolated seamounts with the heights of 20 to 140 m above the background sea floor are distributed. All the SAR imagery streaks originate from these seamounts and extend downstream. Thus the SAR imagery streaks are interpreted as surface roughness imprints of the seamount wakes. Hydrostatic flow dynamics of the generation of wakes on the lee side of a solid obstacle is used to explain the generation mechanism and internal structure of the seamount wakes. The analysis indicates that boundary conditions and hydrodynamic conditions are favorable for the generation and vertical propagation of the seamount wakes to the upper layer.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Study Area
  5. 3. SAR Image Interpretation
  6. 4. Dynamical Analysis
  7. 5. New Observations
  8. 6. Conclusions and Discussion
  9. Acknowledgments
  10. References

[2] The SEASAT satellite was the first dedicated ocean remote sensing satellite and was launched by NASA on June 28, 1978. One of the microwave sensors carried onboard SEASAT was an L-band (1.275 GHz), HH polarization synthetic aperture radar (SAR), which was designed for monitoring of the global ocean surface wave field and polar sea ice conditions [Lame and Born, 1982]. During its 105 mission days, approximately 42 hours of SAR data were collected. Further research indicated that in addition to ocean surface waves and polar sea ice, a variety of important atmospheric and oceanic processes were also manifested in the SEASAT SAR images [Vesecky and Stewart, 1982]. Surface manifestation of the ocean bottom topographic features in the SAR imagery was of particular interest, which was not intuitive since radar pulses of any frequency are not capable of penetrating into seawater. A SEASAT SAR image taken over the Gulf Stream region as shown in Figure 3aattracted many scientist's attention, as the grouped, streak-like patterns appear to be related to ocean bottom features lying at water depths as deep as 600 m.Hayes [1981] and Fu and Holt [1982] interpreted the streaks as imagery of ocean bottom topographic features, while Mollo-Christensen [1981] proposed that the streaks were signals of hydrodynamic processes. The physics of the feature generation and interactions with the Gulf Stream have remained essentially unknown [National Aeronautics and Space Administration, 1989].

[3] SAR imaging of ocean bottom topography is an important research area, because of its potential significance to sea floor mapping and submarine object detection [Alpers et al., 2004; Li et al., 2010]. On the other hand, the SAR image is a single parameter (radar backscattering coefficient) product. Different atmospheric and oceanic phenomena, such as atmospheric boundary layer waves, windrows, rainfall at the ocean surface, ocean bottom topography, ocean vortexes and eddies, as well as ocean internal waves, all form imprints or patterns, many of them similar to each other, on SAR images [Fu and Holt, 1982; Jackson and Apel, 2004]. To identify these phenomena physically is a fundamental effort for SAR image interpretation and application. Here we re-examine this SEASAT SAR image of the Gulf Stream region and try to develop a convincing interpretation of this unique feature.

2. Study Area

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Study Area
  5. 3. SAR Image Interpretation
  6. 4. Dynamical Analysis
  7. 5. New Observations
  8. 6. Conclusions and Discussion
  9. Acknowledgments
  10. References

[4] Figure 1 shows a National Oceanic and Atmospheric Administration (NOAA) Advanced Very High Resolution Radiometer (AVHRR) sea surface temperature (SST) image of the Gulf Stream region off the southeastern coast of the United States. One can see that the Gulf Stream is shown as a high SST water belt (in red) flowing northeastward along the continental slope between the Florida peninsula and Cape Hatteras, North Carolina. A black box shows the coverage of the SEASAT SAR image of July 25, 1978 (shown in Figure 3a), which is a segment of the Gulf Stream over the Charleston Rise region. Typically, the Gulf Stream is 100 km wide and 800 to 1200 m deep. The current velocity is fastest near the surface, with the maximum velocity of about 2.5 ms−1 [Stommel, 1976]. On the continental shelf, the Gulf Stream flow may reach to near the ocean bottom and still keep a strong current parallel to the isobaths [Hayes, 1981].

image

Figure 1. An AVHRR SST image of the Gulf Stream region off the US southeastern coast taken by the NOAA-18 satellite at 07:21:05 UTC on February 28, 2006. The black box represents the SEASAT SAR imaged area. Isobaths are in m.

Download figure to PowerPoint

[5] Figure 2shows a resolution-reduced bathymetric map of the study area, covering the Charleston Rise region from 31°00′ to 32°00′N and from 77°00′ to 78°30′W. The water depth is between 500 and 800 m in the middle segment of the continental slope. The ocean bottom topography is characterized by densely distributed seamounts named the Hoyt Hills. The original high-resolution bathymetric map shows that the seamounts' heights above the background sea floor range from 20 to 140 m. All these seamounts are approximately elliptical. The mean major and minor axis lengths are 8.4 ± 4.9 km and 4.1 ± 2.2 km, respectively.

image

Figure 2. Resolution-reduced bathymetric map of the study area. Original data were downloaded from websitehttp://www.ngdc.noaa.gov/mgg/bathymetry/. Two dashed lines represent boundaries of the southern portion of the SEASAT SAR image taken on July 25, 1978 (Figure 3a).

Download figure to PowerPoint

3. SAR Image Interpretation

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Study Area
  5. 3. SAR Image Interpretation
  6. 4. Dynamical Analysis
  7. 5. New Observations
  8. 6. Conclusions and Discussion
  9. Acknowledgments
  10. References

[6] The SEASAT SAR image of the Gulf Stream region taken on July 25, 1978 is shown in Figure 3a. A map of seamounts extracted from the original high-resolution version of the bathymetric map is overlaid on the SAR image. We mark the Gulf Stream boundaries and direction based on the infrared image shown inFigure 1. From Figure 3a, we find the following key points.

image

Figure 3. (a) A SEASAT SAR image of the Gulf Stream region on July 25, 1978. The swath width is 100 km, and the original spatial resolution is 25 m by 25 m. Coded black ellipses represent locations, sizes and orientations of seamounts of the Hoyt Hills. Dotted lines and arrow represent the boundaries and direction of the Gulf Stream. (b) Details of seamount wakes on zoomed SAR image.

Download figure to PowerPoint

[7] All the SAR imagery streaks originate from seamounts, without exception. The streaks are not uniformly distributed, but in general fall into two groups. One group originates from a seamount chain consisting of seamounts 1–4 and another from seamounts 8 and 9. Starting from the seamounts, the streaks extend downstream, and become enhanced when they pass over a large seamount, such as the case of seamounts 5 and 7. This implies that dynamically, seamounts are generation sources for the disturbance signals represented by the SAR imagery streaks.

[8] Inside the grouped streaks, there are secondary structures. Along the cross-stream direction, the width of individual streaks varies. Along the downstream direction, the streaks are not smooth lines or bands, but contain the fine structure as shown inFigure 3b. This implies that there is similarity among the seamount streaks, island wakes [Zheng et al., 2008] and the mountain wakes in the atmosphere [Tsuchiya, 1969].

[9] The streaks are concentrated in the central Gulf Stream flow and generally follow the dominant current flow direction as marked by a large arrow in Figure 3a. The total length of the streaks reaches about 150 km. This implies that the generation and development of the streaks are associated with the Gulf Stream.

[10] In summary, the above imagery features of the streaks on the SEASAT SAR image indicate that they are imprints of seamount wakes on the sea surface. The seamount wakes are generated by the interaction between seamounts and the Gulf Stream flow. In other words, seamounts serve as a disturbance source, and the Gulf Stream provides a dynamical source.

4. Dynamical Analysis

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Study Area
  5. 3. SAR Image Interpretation
  6. 4. Dynamical Analysis
  7. 5. New Observations
  8. 6. Conclusions and Discussion
  9. Acknowledgments
  10. References

[11] For underwater linear topography, previous investigators have developed SAR imaging theories [Alpers and Hennings, 1984; Shuchman et al., 1985; Romeiser and Alpers, 1997]. This study deals with a new type of ocean bottom topographic feature, seamounts. The above interpretation results indicate that the SAR imagery streaks are imprints of the seamount wakes. In order to verify this interpretation, dynamically we should answer three questions. 1) How does a seamount generate the disturbance signals in the bottom ocean layer? 2) How do the disturbance signals propagate from the bottom layer vertically to the upper ocean? 3) How do the upper ocean disturbance signals generate the satellite sensor signals? For the third question, previous investigators have provided possible answers, i.e., change in sea surface roughness or Bragg wave spectral density due to currents passing over bathymetric features [Plant, 1990; Yuan, 1997; Zheng et al., 2001]. In this section, we look for the answers for the first two questions.

4.1. Generation Mechanisms of Seamount Wakes

[12] The SAR image interpretation in Section 3indicates that seamounts are generation sources for the disturbance signals represented by the SAR imagery streaks. This implies that the disturbance signals are generated in the bottom layer of the ocean. In this layer, the Brunt-Väisälä frequencyN (N1) is about 4 cph as shown in Figure 4. The length scale (diameter) of seamounts is a = O(5 km). The current velocity of Gulf Stream at 500 m is U = O(1 ms−1) [Stommel, 1976]. We have Na/U = 35»1. Thus the flow is approximately hydrostatic [Baines, 1995]. According to Baines [1995], the generation and behavior of the wakes depend on the Reynolds number Re (= UD/ν for a circular cylinder, here U is the upstream velocity, D is the cylinder diameter, and ν is the viscosity), and another parameter, Nh/U, here Nis the Brunt-Väisälä frequency of stratified flow, andh is the seamount height. The viscosity should be within a reasonable range from 1 × 105 cm2s−1 to 5 × 105 cm2s−1. The Reynolds number, Re, should be in a range of 100 < Re < 500. To estimate N, we choose the nearest NOAA Station 1770962 (31°52.8′N 79°10.8′W) to represent the study area. We use the conductivity-temperature-depth (CTD) data measured on July 5, 2002 and August 24, 2003 to estimate the vertical stratification in the same season at the SEASAT SAR imaging time, July 25, 1978, as no simultaneous measurements are available. From the vertical profiles of density (Figure 4a), the Brunt-Väisälä frequencies are calculated, and give an averageN (N1) of about 4 cph for the bottom ocean layer as shown in Figure 4b. Thus we estimate that N at the SEASAT SAR imaging time on July 25, 1978 was about 4 cph. The scale of h is 100 m. In this case Nh/U is about 0.7. Under the conditions of 100 < Re < 500 and Nh/U ≈ 0.7, the wakes may appear in three possible forms: symmetric vortex shedding (SVS), asymmetric vortex shedding (AVS) and lee waves with attached vortices (WV) [see Baines, 1995, Figure 6.30].

image

Figure 4. (a) Vertical profiles of seawater density derived from CTD data measured at NOAA Station 1770962 (31°52.8′N 79°10.8′W) on July 5, 2002 and August 24, 2003, and (b) corresponding profiles of the Brunt-Väisälä frequency. CTD data were downloaded from websitehttp://www.nodc.noaa.gov/GTSPP/access_data/.

Download figure to PowerPoint

[13] On the other hand, the details of SAR-imaged seamount wakes are shown inFigure 3b, which contains a zoomed portion of the SEASAT SAR image. As interpreted in Section 3, the seamount wake signatures are not uniformly distributed, but grouped. Inside the grouped signatures, there are secondary structures. Along the cross-stream direction, the width of an individual wake feature varies. Along the downstream direction, the signatures are not smooth lines or bands, but contain fine, vortex-like structure. One example is a wave-like feature containing a grouping of 9 features starting from the upstream slope at seamount 4 and extending downstream. The average wavelength, which is defined as the distance between two successive waves, is 1.2 km. The wave packet is uniformly distributed on the left hand of seamount, while on the right hand side a discontinuously linear pattern is shown, which also extends downstream as far as the wave packet extends. This complex internal structure seems to be consistent with the predicted asymmetric vortex shedding. Other seamount wake imageries have similar, but finer structure to that of seamount 4, implying the same generation mechanism.

4.2. Vertical Propagation Mechanism

[14] For a stratified ocean, Zheng et al. [2006]derived dynamical solutions for the vertical propagation of disturbance signals from the ocean bottom to the upper layer. For a three-layer ocean, the pass frequency band isN3 < σ < N2, where N3 and N2are the Brunt-Väisälä frequencies of the upper and middle layers, andσ is the angular frequency of the disturbance signal. From Figure 4b, one can see that the ocean can be treated as a three-layer ocean, where the 1) bottom layer (N1) is depth > 150 m, 2) middle layer (N2) is from 20 to 150 m, and 3) upper layer (N3) is from the surface to 20 m. The Brunt-Väisälä frequencies areN3 = 7 cph, N2 = 10 cph and N1 = 4 cph on July 5, 2002, and N3 = 3.5 cph, N2 = 8 cph and N1 = 4 cph on August 25, 2003. Reasonably taking the averages to estimate the case at SEASAT SAR imaging time, we obtain N3 = 5 cph, N2 = 9 cph and N1 = 4 cph. In the above section, we have estimated that the average wavelength of the seamount wakes is 1.2 km for the typical case of the wave packet associated with seamount 4. The phase speed can be estimated by inline image, in which g′(=δρ/ρg) is the reduced gravitational acceleration, c is estimated as 2.0 ms−1 for δρ/ρ = 2 × 10−3, and d = 200 m. Therefore, the seamount wakes frequency σm (=c/λ) is about 6 cph, which satisfies the relation N3 (5 cph< σ(6 cph< N2 (9 cph). This implies that the seamount wakes may penetrate through the stratified ocean vertically to the upper layer without any dynamical constraint.

5. New Observations

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Study Area
  5. 3. SAR Image Interpretation
  6. 4. Dynamical Analysis
  7. 5. New Observations
  8. 6. Conclusions and Discussion
  9. Acknowledgments
  10. References

[15] In order to confirm the phenomena shown on the SEASAT SAR image, we examined more recent C-band RADARSAT-1 ScanSAR images of the same Gulf Stream region. We found that the phenomena show up frequently on high resolution SAR-derived wind field images. From July to October 2006, we found that at least 9 images contain the same or similar wake patterns in the same area of the SEASAT SAR coverage.Figure 5shows four examples, which were acquired on July 21, August 21, September 14 and October 1, 2006, respectively. One can see the grouped, streak-like patterns are similar to the SEASAT SAR image. The difference is that SEASAT SAR image contains more details including the appearance of fine structure. The explanations for the resulting differences may include differing spatial resolutions (25 m for SEASAT SAR and 100 m for Radarsat-1 ScanSAR), signal-to-noise ratios, and radar wavelengths between the two instruments.

image

Figure 5. High-resolution SAR wind field images of the Gulf Stream region derived from C-band RADARSAT-1 ScanSAR data acquired on July 21 (2006202), August 21 (2006233), September 14 (2006257) and October 1, 2006 (2006274). The white-circled area represents the coverage of seamount wakes on SEASAT SAR image inFigure 3a. The large arrows represent wind direction present at the time of each SAR image.

Download figure to PowerPoint

6. Conclusions and Discussion

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Study Area
  5. 3. SAR Image Interpretation
  6. 4. Dynamical Analysis
  7. 5. New Observations
  8. 6. Conclusions and Discussion
  9. Acknowledgments
  10. References

[16] We examined the connection of the SEASAT SAR imagery to seamounts, analyzed the boundary and dynamical conditions, compared the imagery to new observations, and came to the following findings and conclusions. 1) In the area covered by streak-like patterns, there exist isolated seamounts of the Hoyt Hills with the size of 2–15 km and the top height of 20 to 140 m above the background sea floor (∼600 m). The area is located within the main stem of the Gulf Stream. The current velocity is as high as 2.5 ms−1 at the surface and 1.0 ms−1at the depth of 500 m. This is a necessary dynamical condition for the occurrence of seamount wakes. The SAR image shows that all the imagery streaks originate from seamount chains, extend downstream, and are enhanced after passing over the large seamounts. 2) The boundary and dynamical conditions are favorable for the generation of the seamount wakes. The internal structure of seamount wakes generally agrees with the behavior of wakes generated by the hydrostatic flow passing over a solid obstacle as predicted by classic fluid dynamics. These results confirm the seamount origin of the wakes as shown on the SEASAT SAR image. 3) The ocean stratification structure completely supports the seamount wakes to propagate vertically to the upper layer. Thus it is physically sound to interpret the grouped, streak-like patterns on the SEASAT SAR image taken over the Gulf Stream region on July 25, 1978 to be imprints of seamount wakes.

[17] Previous investigators have revealed the detailed behavior of the flow past isolated topography using laboratory experiments. For example, Richards et al. [1992] found that a wave packet was generated over and behind a Gaussian hill when the Rossby number, R (= U/fL, here U is the flow speed, f the Colioris parameter and L a horizontal length scale of the topography), was equal to 1.3. This is just the same as the case we observed over and behind seamount 4 (R = 1.3 for U = 1 ms−1, f = 0.76 × 10−4 s−1 and L = 10 km).

[18] It is worth noting the following two facts. 1) The SAR detects sea surface roughness, which may be the sea surface manifestations of ocean phenomena and surface imprints of atmospheric phenomena. 2) The stratified ocean plays a role of band-pass-filter in the vertical propagation process of ocean bottom disturbance signals [Zheng et al., 2006], thus the SAR imagery primarily contains signal components within a frequency band of N3 < σ < N2. This implies that SAR cannot always image the seamount wakes, and the imagery patterns may vary with the seasonal variability of wind conditions, ocean stratification and the Gulf Stream current field as shown in Figure 5.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Study Area
  5. 3. SAR Image Interpretation
  6. 4. Dynamical Analysis
  7. 5. New Observations
  8. 6. Conclusions and Discussion
  9. Acknowledgments
  10. References

[19] This work is partially supported by US National Science Foundation Award 0962107 and Academician Foundation of China. The views, opinions, and findings contained in this report are those of the authors and should not be construed as an official NOAA or U.S. Government position, policy, or decision. B.H. carried out this work at the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration. The bathymetric map is downloaded from www.ngdc.noaa.gov/mgg/bathymetry/. This work is also supported by Shanghai Science and Technology Committee Program – Special for EXPO under Grant No.10DZ0581600, and a grant (SHUES2011A07) from Shanghai Institute of Urban Ecology and Sustainability.

[20] The Editor thanks the anonymous reviewer for assisting in the evaluation of this paper.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Study Area
  5. 3. SAR Image Interpretation
  6. 4. Dynamical Analysis
  7. 5. New Observations
  8. 6. Conclusions and Discussion
  9. Acknowledgments
  10. References
  • Alpers, W., and I. Hennings (1984), A theory of imaging mechanism of underwater bottom topography by real and synthetic aperture radar, J. Geophys. Res., 89, 10,52910,546, doi:10.1029/JC089iC06p10529.
  • Alpers, W., G. Campbell, H. Wensink, and Q. Zheng (2004), Underwater topography, in Synthetic Aperture Radar Marine User's Manual, edited by C. R. Jackson, and J. R. Apel, chap. 10, pp. 245262, U.S. Dep. of Commer., Camp Springs, Maryland.
  • Baines, P. G. (1995), Topographic Effects in Stratified Flows, 482 pp., Cambridge Univ. Press, New York.
  • Fu, L.-L., and B. Holt (1982), SEASAT views oceans and sea ice with Synthetic Aperture Radar, JPL Publ., 81–120, 200 pp.
  • Hayes, R. M. (1981), Detection of the Gulf Stream, in Spaceborne Synthetic Aperture Radar for Oceanography, Johns Hopkins Oceanogr. Stud., vol. 7, edited by R. C. Beal, P. S. DeLeonibus, and I. Katz, pp. 146160, Johns Hopkins Univ. Press, Baltimore, Md.
  • Jackson, C. R., and J. R. Apel (2004), Synthetic Aperture Radar Marine User's Manual, 464 pp., U.S. Dep. of Commer., Camp Springs, Md.
  • Lame, D. B., and G. H. Born (1982), SEASAT measurement system evaluation: Achievements and limitations, J. Geophys. Res., 87, 31753178, doi:10.1029/JC087iC05p03175.
  • Li, X., X. Yang, Q. Zheng, L. J. Pietrafesa, W. G. Pichel, Z. Li, and X. Li (2010), Deep-water bathymetric features imaged by spaceborne SAR in the Gulf Stream region, Geophys. Res. Lett., 37, L19603, doi:10.1029/2010GL044406.
  • Mollo-Christensen, E. (1981), Surface signs of internal ocean dynamics, in Spaceborne Synthetic Aperture Radar for Oceanography, Johns Hopkins Oceanogr. Stud., vol. 7, edited by R. C. Beal, P. S. DeLeonibus, and I. Katz, pp. 140145, Johns Hopkins Univ. Press, Baltimore, Md.
  • National Aeronautics and Space Administration (1989), Earth Observing System, vol. 11f, SAR: Synthetic Aperture Radar Instrument Panel Report, 2 pp., Natl. Aeronaut. and Space Admin., Washington, D. C.
  • Plant, W. J. (1990), Bragg scattering of electromagnetic waves from the air/sea interface, in Surface Waves and Fluxes, vol. 2, Remote Sensing, edited by G. L. Geernaert and W. J. Plant, pp. 41108, Kluwer Acad., London.
  • Richards, K. J., D. A. Smeed, E. J. Hopfinger, and G. Chabert D'Hières (1992), Boundary-layer separation of rotating flows past surface-mounted obstacles, J. Fluid Mech., 237, 343371, doi:10.1017/S0022112092003446.
  • Romeiser, R., and W. Alpers (1997), An improved composite surface model for the radar backscattering cross section of the ocean surface: 2. Model response to surface roughness variations and the radar imaging of underwater bottom topography, J. Geophys. Res., 102, 25,25125,267, doi:10.1029/97JC00191.
  • Shuchman, R. A., D. R. Lyzenga, and G. A. Meadows (1985), Synthetic aperture radar imaging of ocean-bottom topography via tidal-current interactions, theory and observations, Int. J. Remote Sens., 6(7), 11791200, doi:10.1080/01431168508948271.
  • Stommel, H. (1976), The Gulf Stream, 2nd ed., pp. 4474, Univ. of Calif. Press, Berkeley.
  • Tsuchiya, K. (1969), The clouds with the shape of Kármán vortex street in the wake of Cheju Island, Korea, J. Meteorol. Soc. Jpn., 47, 457464.
  • Vesecky, J. F., and R. H. Stewart (1982), The observation of ocean surface phenomena using imagery from the SEASAT synthetic aperture radar: An assessment, J. Geophys. Res., 87, 33973430, doi:10.1029/JC087iC05p03397.
  • Yuan, Y. (1997), Representation of high frequency spectra of ocean waves and the basis for analyzing SAR images, Chin. J. Oceanol. Limnol., 28, Suppl., 15.
  • Zheng, Q., Y. Yuan, V. Klemas, and X.-H. Yan (2001), Theoretical expression for an ocean internal soliton SAR image and determination of the soliton characteristic half width, J. Geophys. Res., 106, 31,41531,423, doi:10.1029/2000JC000726.
  • Zheng, Q., L. Li, X. Guo, Y. Ge, D. Zhu, and C. Li (2006), SAR Imaging and hydrodynamic analysis of ocean bottom topographic waves, J. Geophys. Res., 111, C09028, doi:10.1029/2006JC003586.
  • Zheng, Q., H. Lin, J. Meng, X. Hu, Y. T. Song, Y. Zhang, and C. Li (2008), Sub-mesoscale ocean vortex trains in the Luzon Strait, J. Geophys. Res., 113, C04032, doi:10.1029/2007JC004362.