This document discusses using radiometric terrain correction (RTC) to improve polarimetric SAR (PolSAR) land cover classification in boreal environments like interior Alaska. It introduces a study using ALOS PALSAR data and describes challenges with topography. RTC normalizes backscatter values using a terrain model to account for local terrain effects. The document demonstrates applying RTC to PALSAR coherency matrices and parameters, and integrating it with PolSARpro and MapReady tools for classification. Results show RTC improves classification accuracy, particularly for deciduous forests, by reducing radiometric variability from topography. However, RTC does not address differences in scattering mechanisms' sensitivity to incidence angle.

1. USE OF RADIOMETRIC TERRAIN CORRECTION TO IMPROVE POLSAR LAND COVER CLASSIFICATION Don Atwood1 and David Small2University of Alaska FairbanksUniversity of Zurich, Switzerland

2. Presentation OverviewIntroduce Boreal Land Cover Classification project

3. Focus on species differentiation in boreal environment

4. Introduce reference data for land cover classification

5. Introduce method of Radiometric Terrain Correction (RTC)

6. Terrain-flattened Gamma Naught Backscatter

7. Perform RTC on polarimetric parameters to address topography

8. Demonstrate synergy of PolSARpro and MapReady Tools

9. Compare results for RTC-corrected and non-corrected classification

10. Characterize optimal classification approach for Interior AlaskaStudy RegionBoreal environment of Interior Alaska Characterized by: rivers

11. wetlands

12. herbaceous tundra

13. black spruce forests (north facing)

14. birch forests (south facing)

15. low intensity urban areasLand Cover Reference

16. Study DataQuad-Pol data selected: ALOS L-band PALSAR

17. 21.5 degree look angle

18. Of April, May, July, and Nov dates, July 12 2009 selected Post-thaw

19. Leaf-on

20. Coverage includes Fairbanks and regional roadsPauli Image

21. Problem of TopographySpan (Trace of T3 Matrix) Wishart Segmentation

22. Aγ & γ0Aβ & β0Aσ & σ0Normalized Radar Cross SectionsSensorNadirLet’s compute Normalized Radar Cross Sectionsfor an Ellipsoidal Earth

23. for TopographyNearFar

24. Backscatter Reference AreasFor an Ellipsoidal EarthRelationships between cross sections for ellipsoidal surfaces

25. Terrain-flattening We need to move beyond the ellipsoidal Earth to the hills and valleys of the Fairbanks region:Address the layover and foreshortening of geometric distortions

26. Correct the radiometric variations associated with topography.To improve our radiometry:use local area contributing to backscatter at each location in the SAR sceneTerrain-flattening

27. Terrain-flattening Solution: Use simulated image to Normalize β0XRef.: Small, D., Flattening Gamma: Radiometric Terrain Correction for SAR Imagery, IEEE Transactions on Geoscience and Remote Sensing, 13p (in press).

28. Terrain Correctionin Coastal BC VancouverGTC (Sept 2008) Integrated contributing area ENVISAT ASAR WSM data courtesy ESA(based on SRTM3)

29. Terrain Correctionin Coastal BC GTC (Sept 2008) Integrated contributing area ENVISAT ASAR WSM data courtesy ESA(based on SRTM3)

30. Coastal BC: GTCASAR WSM GTC

31. Coastal BC: RTCASAR WSM RTC

32. Coastal BC: NORLIMASAR WSM NORLIM

33. Coherency MatrixScattering Matrix: “Double Bounce”: “Single Bounce”: “Volume Scattering”

34. Coherency Matrixterrain corrected Coherency MatrixArea NormalizationRadiometric Terrain Correctionof Coherency MatrixRadiometric Terrain Correction:Scale all matrix elements by Area NormalizationBut Wait…..For a given class, the ratio of Surface, Double Bounce, and Volume scattering components depend on incidence anglePOLARIMETRIC IMPLICATIONS OF INCIDENCE ANGLE VARIABILITY FOR UAVSARGuritz, Atwood, Chapman, and Hensley

35. Radiometric Terrain Correctionof Coherency MatrixSpan: No Normalization Span: Terrain-model Normalization

36. Radiometric Terrain Correctionof Coherency MatrixPauli: No Normalization Pauli: Terrain-model Normalization

37. Integration of PolSARpro and MapReadyIngest PALSAR data Terrain-correct Perform WishartExport to GIS Generate T3 with MapReadydecomposition Cluster-bustingRadiometric correction using areaLee Sigma Speckle FilterPOA compensation

38. Radiometric Terrain Correctionof Coherency MatrixWishart - No Normalization Wishart - Radiometric Correction

39. Radiometric Terrain Correctionof Coherency MatrixUSGS Reference Wishart– Radiometric Correction

40. Classification ResultsNo Normalization USGS Reference RTC

41. Classification ResultsUrban areas missed / Identified as Open Water

42. Classification ResultsInability to distinguish Mixed Forests and Shrub / Scrub

43. Accuracy AssessmentNo Normalization

44. Accuracy AssessmentWith RTC

45. Accuracy AssessmentComparison RTC yields improved accuracy (particularly for Deciduous Forest)Impact of RTC on forest classificationNo Normalization USGS Reference RTC

46. ConclusionsIn general, PolSAR classification is difficult!

47. Data fusion provides greatest hope for accurate classification results

48. Radiometric variability caused by topography dominates PolSAR classification

49. Area-based RTC offers effective way to “flatten” SAR radiometry

50. RTC of Coherency Matrix shown to improve classification accuracy:

51. Impact most pronounced for Deciduous Forests

52. Although not complete, RTC approach is simple and effective

53. Different scattering mechanisms (SB, DB, Volume) have different sensitivities to topography. RTC does not address this

54. However, RTC is very effective first order correction for segmenting polarimetric data by phenology rather than topographyDiscussionDon Atwooddkatwood@alaska.edu(907) 474-7380Photo Credit: Don Atwood