1. Mobile Bay Water Quality Assessment Using NASA Spaceborne Data Products Jenny Q. Du Mississippi State University

2. Outline Project Objectives Current Research Status Proposed Approaches –Images with higher resolution –Classification methods Preliminary Results Ongoing Efforts

3. Project Objectives To use NASA Spaceborne Imagery (i.e., Landsat, ASTER, Hyperion) in the study of water quality and sediment dynamics in Mobile Bay, AL. To compare with the NASA research products in Mobile Bay using MODIS imagery and assess the improvements.

4. This MODIS satellite image shows sediment plumes moving into the Gulf of Mexico from the main branch of the Mississippi River and through the bayous in its Delta region (visibleearth.nasa.gov)

5. Landsat 11/27/1999

6. Landsat 11/27/1999 (Mobile bay)

7. Landsat 10/15/2001

8. Landsat 10/15/2001 (Mobile Bay)

9. Landsat 2/17/2001

10. Landsat 2/17/2001 (Mobile Bay)

11. Current Research Status MODIS (Aqua/Terra) –Pros: wide spatial coverage, high temporal resolution (covers the entire globe almost everyday) –Con: low spatial resolution (250m-1000m) Hard classification –K-means clustering –ISODATA

12. Proposed Approaches Satellite Images with Higher Resolutions –LANDSAT 30 m spatial resolution (can be enhanced to 15m); 4 VNIR bands –ASTER 15m spatial resolution; 3 VNIR bands –Hyperion 30m spatial resolution; 220 bands Fine Classification –Statistical Classifiers Correlation Study with Ground Truth –In situ sampling (Nov. 2007 – Sep. 2008) –Historic data

13. Preliminary Results Satellite Images with Higher Resolutions –LANDSAT Classification –ISODATA Correlation Study with Ground Truth –Historic data (e.g., Water Resources Database)

14. Landsat 9/26/1991 (Mobile Bay)

15. Land Class 1 Class 2 Class 3 Class 4 Class 5 Observation Stations (Hard) Classification Result

16. Land Class 1 Class 2 Class 3 Class 4 Class 5 Observation Stations Turbidity Stations1=5.1 Stations2=4.7

17. Land Class 1 Class 2 Class 3 Class 4 Class 5 Observation Stations Turbidity Stations3=20 Stations4=25

18. Land Class 1 Class 2 Class 3 Class 4 Class 5 Observation Stations Turbidity Stations5=9.8 Stations6= 10.3

19. Land Class 1 Class 2 Class 3 Class 4 Class 5 Observation Stations Turbidity Stations7=27

20. Satellite Image 09/26/1991 Ground Truth(Turbidity) 09/27/1991 Stations1234567 Class2233445 Turbidity5.14.720259.810.327

21. Landsat 11/27/1999 (Mobile Bay)

22. Land Class 1 Class 2 Class 3 Class 4 Class 5 Observation Stations (Hard) Classification Result

23. Land Class 1 Class 2 Class 3 Class 4 Class 5 Observation Stations Turbidity Stations1=5.1 Stations2=5.5 Stations3=3.3 Stations4=4.4 Stations5=4.0 TSS Stations1=24 Stations2=22 Stations3=17 Stations4=19 Stations5=16 CHL-A Stations1=5.5 Stations2=7.0 Stations3=6.8 Stations4=4.4 Stations5=7.9

24. Land Class 1 Class 2 Class 3 Class 4 Class 5 Observation Stations Turbidity Stations6=7.9 TSS Stations6=43 CHL-A Stations6=37

25. Satellite Image 11/27/1999 Ground Truth 11/27/1999 Stations123456 Class444442 Turbidity3.34.45.55.147.9 TSS171922241643 Chl-A6.84.47.05.57.937

26. Landsat 2/17/2001 (Mobile Bay)

27. Land Class 1 Class 2 Class 3 Class 4 Class 5 Observation Stations (Hard) Classification Result

28. Land Class 1 Class 2 Class 3 Class 4 Class 5 Observation Stations Turbidity Stations1=18

29. Satellite Image 02/17/2001 Ground Truth(Turbidity) 02/08/2001 Stations1 Class3 Turbidity18

30. Proposed Approaches (Cont’d) Classification Approaches r = M a –Unsupervised Linear Mixture Analysis Endmember signature extraction Fully constrained linear unmixing –Blind Source Separation Independent Component Analysis

31. Original Image

32. Linear unmixing result (soft Classification)

33. The (soft) endmember classification map that can be used for detailed water quality mapping

34. ICA result (soft classification)

35. The independent component (soft classification map) that can be used for detailed water quality mapping

36. Ongoing Efforts More detailed correlation analysis –Images and ground truth data collected at the same time. –Images collected during Nov. 2007 and Sep. 2008.