IGARSS2011presen_final.ppt
Upcoming SlideShare
Loading in...5
×
 

Like this? Share it with your network

Share

IGARSS2011presen_final.ppt

on

  • 505 views

 

Statistics

Views

Total Views
505
Views on SlideShare
496
Embed Views
9

Actions

Likes
0
Downloads
10
Comments
0

1 Embed 9

http://www.grss-ieee.org 9

Accessibility

Categories

Upload Details

Uploaded via as Microsoft PowerPoint

Usage Rights

© All Rights Reserved

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Processing…
Post Comment
Edit your comment
  • Thank you Mr. chairman. My name is Hiroyuki Wakabayashi. I am with Nihon University. Today, I would like to talk about “ SEA ICE CHARACTERISTICS IN THE SOUTHERN REGION OF OKHOTSK SEA OBSERVED BY X- AND L- BAND SAR ” My co-authors are shown here Dr. Nakamura is in charge of backscattering model calculation. Prof. Nishio is responsible for ground truth data analysis.
  • This is a contents of my talk today. First, I would like to explain about study background and objective of this research. Then I ’ ll show you our test site and the data acquired in this research We did a regression analysis of TerraSAR-X and PALSAR data with ground truth data. Finally, we summarize the result and give our future work.
  • As you know, Sea ice extent and volume are related to local as well as global climate change, Because sea ice acts as an insulator between air and water. There was a sign that sea ice is decreasing especially in the Arctic Ocean. It is very important to monitor sea ice frequently by using remote sensing data. Microwave remote sensing is a powerful tool to get the data even in bad weather conditions. We expect SAR data will be an operational tool in sea ice monitoring, because of its high spatial resolution and polarimetric capabilities. We had a chance to get both TerraSAR-X and PALSAR data over our test site in 2010.
  • The objective of this research is into possible use of SAR data to monitor sea ice, especially in thin sea ice area, such as the southern region of Okhotsk Sea. We want to clarify backscattering characteristics, such as frequency and polarization dependence, Finally, we want to develop a method to extract sea ice physical parameters, such as thickness and surface roughness. Our past experience related to this research is given in here, We have been conducting field experiments at Lake Saroma for almost 20 years now, We did single-pol SAR analysis, such as ERS-1/2, JERS-1, and RADARSAT, in 90s. We also analyzed polarimetric SAR data from airborne Pi-SAR and PALSAR data recently. We would like to apply our methods to dual-pol TerraSAR-X data in this research.
  • This viewgraph shows our test sites. Our test site is located southern edge of Sea of Okhotsk. North-east of Hokkaido island. The Sea of Okhotsk is surrounded by Sakhalin, Kamchatka, and Kuril Islands. This area is known as a seasonal ice zone, So sea ice exists only winter time. Most sea ice found in this area is relatively thin sea ice, less than 1 m. We use Lake Saroma as an educational sites for field work for 20 years. Lake Saroma is the third biggest lake in Japan and is a salt water lake which is connected to the Okhotsk Sea. So, the salinity of lake water is almost the same as sea water, more than 30 ppt. Lake ice grows to 40 cm thick in mid Feb. and stable enough to get the ground truth data. In 2010, we set more than 60 sampling points on the east part of the lake. We could set approximately 500m interval between the sampling points.
  • This table shows a list of satellite data used in this analysis. TerraSAR-X acquired the data in dual-pol mode in two different incidence angle. PALSAR acquired the data in fully polarimetric mode. The field observation on Lake Saroma was conducted in the week started on Feb. 16 simultaneously with the satellite data observation.
  • I ’ ll show you satellite data in this viewgraph. All data were distributed in slant range complex data. Slant range data were transformed onto UTM coordinate. Lake Saroma is located here. You can recognize the east part of Lake Saroma on all images. You can recognize the difference of backscattering characteristics from these figures. It depends on observation frequencies and incidence angles
  • These figures show the sampling points on Lake Saroma overlaid on satellite data. We set more than 60 sampling points in the east part of Lake Saroma. Distances between points were about 500 m.
  • Ground truth data acquired at sampling points on the lake are summarized and shown in this viewgraph. Each figure shows the histogram of snow depth on ice, ice thickness and ice surface salinity. Mean of these values are given here. We found these values are considered to be typical values on the lake in mid February.
  • We also measured ice surface roughness by using roughness comb. At least 4 ice surfaces are taken by camera and RMS height as well as correlation length were averaged. Mean values were 4.2 mm and 36 mm respectively.
  • This is a relation between ice thickness and ice surface salinity. Blue diamond-shaped dots show the relation in 2010 experiment. The other circles show the summary in 1996 to 1998 and 2003 to 2004 taken from sampled cores acquired by vessel observations. As you can recognize that plots for the 2010 experiment show almost same characteristics for sea ice of the thickness less than 40 cm in terms of surface salinity.
  • These are relations between ice thickness and other ice physical data, RMS height and snow depth. Snow depth and surface roughness were weakly correlated with ice thickness.
  • Before a regression analysis between SAR and ground truth data, a backscattering model analysis using IEM model was conducted for finding sensitive parameters to ice physical data. This is a flow of our backscattering model analysis. Input data to this model were coming from the summarized ground truth data .
  • These are example of model output of each backscattering coefficient and co-pol ratio. We found that backscattering coefficient and co-pol ratio in high incidence angle decrease as ice thickens.
  • Next I ’ ll give you some results found in our regression analysis between SAR data and ground truth data. The first example is coming from TerraSAR-X data analysis. We found that the co-pol back scattering coefficients of sea ice monotonically decrease as ice thickens. A relatively higher correlation was found in lower incidence angle data. Especially in the small snow depth area, the correlation increased. This characteristics are coincident with the backscattering model simulation.
  • We also found that co-pol ratio decreased with ice thickness in higher incidence angle, when you take a look at small snow depth area.
  • Since the contribution of snow layer to backscattering coefficient is larger in X-band compare to the other lower frequencies, we investigated the sensitive parameters to snow depth on the ice, Both co-pol correlation and dual-pol entropy were found, which might be related to the surface scattering contribution, are related to snow depth in the same level. This is the result found in lower incidence angle observation. We need the further investigation about snow layer in the next experiment to prove.
  • Next we will show you the result of PALSAR data analysis. This graph shows the relation between ice thickness and backscattering coefficient, PALSAR co-pol backscattering coefficient has almost no correlation with ice thickness.
  • However, when we apply Freeman and Durden three component decomposition technique to PALSAR polarimetric data. We found that Bragg scattering is related to RMS height of ice surface. Volume scattering is also weakly correlated with snow depth. I believe that FD decomposition can be applied to estimate ice surface roughness and snow contribution to backscattering coefficient.
  • I would like summarize regression analysis at Lake Saroma. For TerraSAR-X data, We found that lower incidence angle observation is superior to higher incidence angle data in terms of ice thickness and snow depth observation. For PALSAR data, Although we could not found any parameter which has a relation with ice thickness, RMS height and snow thickness are related to the bragg scattering and volume scattering derived by Freeman and Durden decomposition.
  • We also analyzed the backscattering characteristics of sea ice in the off-shore area in TerraSAR-X high incidence angle data. Since there was no ground truth data available, we used the MODIS data as reference for TerraSAR-X data. We followed Sea ice classification using MODIS albedo proposed by this paper. MODIS albedo was calculated based on this equation and classification rule is applied. The backscattering range were derived in the area of open water and new ice.
  • For PALSAR data, we used scattering entropy to detect sea ice in the off-shore area. The scattering entropy is calculated from eigenvalues of covariance or coherence matrix of the observed scattering matrix Based on our past research using Airborne Pi-SAR data, we could make classification rule for Open water, New ice and Young and FY ice.
  • This table shows average backscattering coefficients for three categories including open water and two sea ice types. The classification using MODIS albedo was consistent in two observations.
  • This viewgraph summarized the results for sea ice backscattering characteristics in off-shore area. For both new ice and young ice area, Backscattering range of TerraSAR-X is higher than PALSAR range.
  • We conducted ground truth experiment at Lake Saroma simultaneously with satellite observations. From regression analysis between SAR and ground truth data at Lake Saroma, We found these characteristics.
  • Since this is the first time for us to analyze TerraSAR-X data, we deployed corner reflectors on the lake to derive absolute calibration coefficients. Four square trihedral corner reflectors were deployed. The distance between corner reflectors were approximately set 500 m. The size of reflectors were 50 cm and 70 cm.
  • These figures show corner reflectors imaged on the TerraSAR-X data. Your can recognize four bright points on lake ice. We used the integration method to derive absolute calibration coefficients.
  • This table shows the result of estimated calibration coefficients and calibration data provided with the products. The maximum difference was found in VV-pol in higher incidence angle data. However, the difference was very small. In terms of HH and VV phase differnce, we need to correct the value of phase difference between HH and VV in lower incidence angle data. This equation should be used in order to convert DN to backscattering coefficient.

IGARSS2011presen_final.ppt Presentation Transcript

  • 1. SEA ICE CHARACTERISTICS IN THE SOUTHERN REGION OF OKHOTSK SEA OBSERVED BY X- AND L- BAND SAR
    • Hiroyuki Wakabayashi (Nihon university)
    • Shoji Sakai (Nihon university)
    • Kazuki Nakamura (AIST)
    • Fumihiko Nishio (CEReS/Chiba university)
    • IGARSS2011 仙台 in Vancouver
    • presented on Jul.28,2011
  • 2. Outline
    • Study background and research objectives
    • Test site and SAR data
    • Ground truth experiment
    • Data analysis
    • Summary and future work
  • 3. Study background
    • Role of sea ice monitoring
      • Sea ice extent and volume are related to local as well as global climate change
      • Sea ice acts as an insulator between air and water
      • Sign of decreasing sea ice extent in the Arctic Ocean
    • Importance of SAR data
      • Microwave remote sensing plays an important role in monitoring sea ice in cryosphere due to its all weather capabilities
      • SAR data from TerraSAR-X and ALOS were available during wintering period in 2010
  • 4. Research objectives
    • Objectives
      • Possible use of SAR data to monitor sea ice in the southern region of Okhotsk Sea
        • Backscattering characteristics(Frequency, Polarization)
        • Develop a method to extract sea ice physical parameters
    • Our past experience
      • Field experiments from 1992 to 2011(Lake Saroma)
      • Single-pol SAR analysis (ERS-1/2,JERS-1,RADARSAT)
      • Polarimetric SAR analysis (Pi-SAR, PALSAR)
    Dual-pol. TerraSAR-X data
  • 5. Test sites
    • Southern region of the Sea of Okhotsk
        • Seasonal Ice Zone : sea ice exists only in wintertime
        • Most sea ice in this area has the thickness less than 1 m
    • Lake Saroma ( 150km 2 The third biggest lake in Japan )
        • Salt water lake connected to the Sea of Okhotsk with 2 channels
        • Salinity of lake water is almost the same as that of sea water
        • ( > 30 ppt )
        • Lake ice grows till 40 cm thick in winter time and is stable enough to get the ground truth data in winter time
        • More than 60 sampling points with 500m interval were set in 2010
  • 6. Satellite and Ground Observations
    • Ground truth experiment :2010/02/16-02/26
    • Satellite observation
    List of satellite data used in this analysis satellite sensor observation date observation time (UT) polarization incidence angle (scene center) TerraSAR-X 2010/02/18 08:12 HH+VV 36.8° ALOS PALSAR 2010/02/20 12:37 HH+VV+ HV+VH 24.0° TerraSAR-X 2010/02/24 08:04 HH+VV 20.8°
  • 7. Satellite images (HH-pol) TerraSAR-X(2010.02.18) High incidence angle TerraSAR-X(2010.02.24) Low incidence angle ALOS/PALSAR (2010.02.20) Lake Saroma -Slant range complex data provided by Pasco and JAXA were used in this research 31.7km 39.6km 27.4km 54.5km 63.4km 68.9km
  • 8. Ground truth sampling points TerraSAR-X(2010.02.18) High incidence angle (62 points) TerraSAR-X(2010.02.24) Low incidence angle (62 points) ALOS/PALSAR (2010.02.20) (36 points) -Sampling points were approximately set 500 m interval on the east part of Lake Saroma
  • 9. Summary of ground truth data 1
    • Mean of measured data
    • snow depth : 8.0 cm
    • ice thickness : 31.5 cm
    • ice surface salinity : 10.8 ppt
    -Typical ice on the lake found in mid February
  • 10. Summary of ground truth data 2
    • Mean of measured surface roughness
      • (Roughness comb measurements)
    • RMS height: 4.2mm
    • Correlation length: 35.8mm
    - At least four measurements were averaged at each sampling point
  • 11. Relation between ice thickness and ice surface salinity comparison of 2010 and previous experiments - Plots in Saroma 2010 showed almost the same characteristics for sea ice of the thickness less than 40 cm.
  • 12. Relation between ice thickness and other parameters - Snow depth and surface roughness were weakly correlated with ice thickness
  • 13. Microwave scattering model analysis Surface-scattering model (Integral Equation Method Model) Dielectric constant model Ice salinity model • Frequency • Incidence angle • Air temperature • Water temperature • Roughness parameter (RMS height & Cor. length) • Ice thickness • Snow depth Model parameters Backscattering coefficients VV to HH backscattering ratio etc.
  • 14. Model simulation results(X-band) - Co-pol backscattering coefficients decrease as ice thickens - Co-pol ratio at higher incidence angle is sensitive to ice thickness
  • 15. Ice thickness vs backscattering coefficient(TerraSAR-X) - Backscattering coefficient at lower incidence angle is correlated with ice thickness, especially at small snow depth area R=0.43 R=0.57 (Ts<10cm)
  • 16. Ice thickness vs co-pol ratio (TerraSAR-X) - Co-pol ratio in higher incidence angle at small snow depth has some relation with ice thickness (no strong correlation) R=0.18 R=0.50 (Ts<10cm)
  • 17. Snow layer related parameters (TerraSAR-X) - Co-pol correlation and dual-pol entropy in lower incidence angle have weak relations with snow depth
  • 18. Ice thickness vs backscattering coefficient(PALSAR) - PALSAR backscattering coefficient has almost no correlation with ice thickness
  • 19. FD decomposition vs. truth data (PALSAR) - Freeman and Durden three component decomposition gives reasonable relation to RMS height and snow depth
  • 20. Summary of regression analysis at Lake Saroma
    • Relation between TerraSAR-X and ground truth data
      • Ice thickness : relatively higher correlation found in lower incidence angle at small snow depth area
      • Ice surface roughness : no significant relation was found
      • Lower incidence angle observation is better
      • Contribution of snow layer to backscattering coefficient cannot be ignored
    • Relation between PALSAR and ground truth data
      • Ice thickness : no significant relation was found
      • Snow depth and RMS height : 3 component decomposition result shows reasonable relations
      • Scattering decomposition technique is useful to extract information of ice physical data
  • 21. TerraSAR-X and MODIS albedo 2010.02.18 Al=0.3265*B1+0.4364*B3+0.2366*B4
    • Classification rule based on MODIS albedo
      • Open water ( Al < 0.1 )
      • New ice ( 0.1 ≦ Al < 0.4 )
      • Young ice ( 0.4 ≦ Al < 0.6 )
      • First-year ice ( 0.6 ≦ Al )
    where Al: albedo B1,B3 and B4: reflectances observed in Band 1,3,and 4 - MODIS albedo used for sea ice detection is calculated as follows, Reference D.K.Hall, D.J.Cavalieri, T.Markus: Assessment of AMSR-E Antarctic Winter Sea-Ice Concentrations Using Aqua MODIS, IEEE Trans. on Geo-science and Remote Sensing. Vol.48, No.9, pp.3331-3339, 2010. Offshore area
  • 22. PALSAR backscattering and entropy 2010.02.20
    • Classification rule based on s cattering entropy (H)
      • Open water ( H < 0.15 )
      • New ice ( 0.4 ≦ H )
      • Young ice & First-year ice ( 0.15 ≦ H <0.4)
    Reference H. Wakabayashi, T. Matsuoka, K. Nakamura and F. Nishio: Polarimetric characteristics of sea ice in the Sea of Okhotsk observed by airborne L-band SAR, IEEE Trans. on Geo-science and Remote Sensing, Vol. 42, No.11, pp. 2412-2425, 2004. Offshore area Scattering entropy used for sea ice detection
  • 23. Backscattering characteristics of sea ice in the offshore area TerraSAR-X(2010/02/18) PALSAR(2010/02/20) HH(dB) VV(dB) VV-HH(deg.) MODIS Albedo HH(dB) VV(dB) HV(dB) Scattering entropy MODIS Albedo New ice -16.4 -15.0 6.5 0.15 -21.7 -21.0 -28.4 0.73 0.17 Young ice FY ice -8.6 -9.4 8.0 0.25 -13.3 -12.5 -25.7 0.30 0.26 Open water -19.3 -18.1 0.6 0.10 -9.8 -8.6 -26.1 0.13 0.095
  • 24. Summary of backscattering characteristics of sea ice in the off-shore region
    • New ice area
      • PALSAR : -21 dB(VV) -21.7dB(HH)
      • TerraSAR-X : -15.0dB(VV) -16.4dB(HH)
          • TerraSAR-X : 5 to 6 dB higher than PALSAR
    • Young ice area
      • PALSAR : -12.5 dB(VV) -13.3dB(HH)
      • TerraSAR-X : -9.4dB(VV) -8.6dB(HH)
          • TerraSAR-X : 3 to 5 dB higher than PALSAR
    • Considering TerraSAR-X and PALSAR incidence angles, the difference of backscattering range would be much larger at the same incidence angle
    • TerraSAR-X is better than PALSAR in detecting thin sea ice (e.g. New ice)
  • 25. Summary
    • Ground truth experiment was conducted (Feb. 16 to 26, 2010)
      • In-Situ data at more than 60 sampling points were acquired
      • Backscattering calibration by reflectors was conducted
        • Absolute calibration coefficients were consisted with the provided cal. coefficients.
        • Phase difference between HH and VV should be corrected at lower incidence angle.
    • TerraSAR-X and PALSAR regression analysis on Lake Saroma
      • TerraSAR-X
        • Lower incidence angle observation is preferable for ice physical data extraction.
        • Contribution of snow layer to backscattering coefficient cannot be ignored.
      • PALSAR
        • Scattering decomposition is useful to extract information of ice physical data.
    • Backscattering characteristics of sea ice in the offshore area
      • Backscattering coefficients for new ice and young ice were higher in X-band than that in L-band.
      • X-band sea ice detection is better than L-band in thin sea ice area.
  • 26. Acknowledgement
    • This research was supported by Grant-in-Aid for Exploratory Research of MEXT (No. 20651004).
    • The PALSAR data were distributed under the agreement of JAXA Research Announcement. The research was titled &quot;Sea ice study and its application using PALSAR polarimetric data in the Sea of Okhotsk (JAXA-PI: 205)&quot; .
    • TerraSAR-X data were distributed under the support of SAR technical application research committee organized by Pasco cooperation.
    Future work
    • Develop a backscattering model of sea ice in X-band to include snow layer on the ice.
    • Investigate an inversion technique to extract ice physical data, such as snow depth on ice, ice surface roughness and ice thickness.
  • 27. Weather data(AMEDAS at Tokoro)
  • 28. CR deployment N 500 m 50 cm Trihedral CR 70 cm Trihedral CR 50 cm Trihedral CR 70 cm Trihedral CR 500 m 500 m
  • 29. Image of corner reflectors TerraSAR-X(2010.02.18) High incidence angle TerraSAR-X(2010.02.24) Low incidence angle Range Azimuth
  • 30. Result of radiometric calibration CR#1 CR#2 CR#3 CR#4 CR#1 CR#2 CR#3 CR#4 2/18 2/24 HH(dB) VV(dB) VV-HH(deg.) HH(dB) VV(dB) VV-HH(deg.) CR#1 49.387 49.336 -1.1 47.288 47.050 81.2 CR#2 49.324 49.261 -1.5 46.988 47.140 82.1 CR#3 48.971 48.620 -1.8 46.889 46.895 80.9 CR#4 49.701 49.367 -1.2 47.341 47.452 82.0 Average 49.346 49.146 -1.4 47.127 47.134 81.6 Cal. data Provided 49.583 47.308
  • 31. Correlation matrix(high incidence angle) σ 0 HH σ 0 VV σ 0 VV /σ 0 HH ρ HHVV T s T i σ H l σ 0 HH 1 σ 0 VV 0.966 1 σ 0 VV /σ 0 HH 0.175 0.423 1 ρ HHVV 0.773 0.803 0.353 1 T s -0.167 -0.123 -0.203 -0.239 1 T i -0.066 -0.110 -0.187 -0.118 0.657 1 σ H -0.039 -0.033 0.013 0.013 0.318 0.420 1 l 0.200 0.252 0.261 0.340 -0.190 0.103 0.201 1
  • 32. Correlation matrix(low incidence angle) σ 0 HH σ 0 VV σ 0 VV /σ 0 HH ρ HHVV T s T i σ H l σ 0 HH 1 σ 0 VV 0.991 1 σ 0 VV /σ 0 HH 0.039 0.170 1 ρ HHVV 0.842 0.878 0.341 1 T s -0.547 -0.532 0.056 -0.510 1 T i -0.430 -0.425 0.026 -0.349 0.657 1 σ H -0.116 -0.096 0.143 -0.033 0.318 0.420 1 l 0.367 0.368 0.038 0.353 -0.190 0.103 0.201 1
  • 33. Correlation matrix(combined high and low incidence angle) σ 0 HH (θ L )-σ 0 HH (θ H ) σ 0 VV (θ L )-σ 0 VV (θ H ) T s T i σ H l σ 0 HH (θ L )-σ 0 HH (θ H ) 1 σ 0 VV (θ L )-σ 0 VV (θ H ) 0.972 1 T s -0.628 -0.620 1 T i -0.513 -0.514 0.657 1 σ H -0.125 -0.110 0.318 0.420 1 l 0.193 0.190 -0.190 0.103 0.201 1