spectral requirements for resolving shallow water information products w. paul bissett and david d....

30
Spectral Requirements for Resolving Shallow Water Information Products W. Paul Bissett and David D. R. Kohler

Upload: beryl-hunter

Post on 30-Dec-2015

214 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Spectral Requirements for Resolving Shallow Water Information Products W. Paul Bissett and David D. R. Kohler

Spectral Requirements for Resolving Shallow

Water Information Products

W. Paul Bissett and David D. R. Kohler

Page 2: Spectral Requirements for Resolving Shallow Water Information Products W. Paul Bissett and David D. R. Kohler

Copyright © 2004 Florida Environmental Research Institute

Unpublished work - all rights reserved

NOAA COAST WorkshopSeptember 29,, 2004

UNCLASSIFEDPROPRIETARY INFORMATION

2

Coastal Ocean Imaging Spectroscopy

Phytoplankton

CDOM-Rich Water

SuspendedSediments

Benthic Plants1/Kd1/Kd

Optically-ShallowOptically-ShallowOptically-DeepOptically-Deep

Micro-bubbles

Whitecaps

Shallow Ocean Floor

Page 3: Spectral Requirements for Resolving Shallow Water Information Products W. Paul Bissett and David D. R. Kohler

Copyright © 2004 Florida Environmental Research Institute

Unpublished work - all rights reserved

NOAA COAST WorkshopSeptember 29,, 2004

UNCLASSIFEDPROPRIETARY INFORMATION

3

Spectral Resolution

Depth = 6.5 m IOP = Case 1Bottom = soft coral

Depth = 13 m IOP = Pure H20Bottom = sponge

Page 4: Spectral Requirements for Resolving Shallow Water Information Products W. Paul Bissett and David D. R. Kohler

Copyright © 2004 Florida Environmental Research Institute

Unpublished work - all rights reserved

NOAA COAST WorkshopSeptember 29,, 2004

UNCLASSIFEDPROPRIETARY INFORMATION

4

Page 5: Spectral Requirements for Resolving Shallow Water Information Products W. Paul Bissett and David D. R. Kohler

Copyright © 2004 Florida Environmental Research Institute

Unpublished work - all rights reserved

NOAA COAST WorkshopSeptember 29,, 2004

UNCLASSIFEDPROPRIETARY INFORMATION

5

Why (Hyper) Spectral?

• Discriminate analysis suggests 10-12 independent vectors of information in hyperspectral data stream.

• But which bands are a priori the rights one to measure?

• Additional bands give greater degrees of freedom with which to attempt more complicated inversion algorithms to retrieve depth-dependent IOPs, which include products such as bathymetry, in situ IOPs, bottom classification.

Page 6: Spectral Requirements for Resolving Shallow Water Information Products W. Paul Bissett and David D. R. Kohler

Copyright © 2004 Florida Environmental Research Institute

Unpublished work - all rights reserved

NOAA COAST WorkshopSeptember 29,, 2004

UNCLASSIFEDPROPRIETARY INFORMATION

6

Joint HSI/LIDAR Data Set

Page 7: Spectral Requirements for Resolving Shallow Water Information Products W. Paul Bissett and David D. R. Kohler

Copyright © 2004 Florida Environmental Research Institute

Unpublished work - all rights reserved

NOAA COAST WorkshopSeptember 29,, 2004

UNCLASSIFEDPROPRIETARY INFORMATION

7

1.9 x 106 Total Points

39% of the LUT results are +/- 10%

73% of the LUT results are +/- 25%

A normal distribution is plotted for reference. The LUT results have less error than would be expected for a normally distributed population.

Page 8: Spectral Requirements for Resolving Shallow Water Information Products W. Paul Bissett and David D. R. Kohler

Copyright © 2004 Florida Environmental Research Institute

Unpublished work - all rights reserved

NOAA COAST WorkshopSeptember 29,, 2004

UNCLASSIFEDPROPRIETARY INFORMATION

8

Page 9: Spectral Requirements for Resolving Shallow Water Information Products W. Paul Bissett and David D. R. Kohler

Copyright © 2004 Florida Environmental Research Institute

Unpublished work - all rights reserved

NOAA COAST WorkshopSeptember 29,, 2004

UNCLASSIFEDPROPRIETARY INFORMATION

9

NOAA/Florida Marine Research Institute

Benthic Assessment of the Florida Keys

Page 10: Spectral Requirements for Resolving Shallow Water Information Products W. Paul Bissett and David D. R. Kohler

Copyright © 2004 Florida Environmental Research Institute

Unpublished work - all rights reserved

NOAA COAST WorkshopSeptember 29,, 2004

UNCLASSIFEDPROPRIETARY INFORMATION

10

Comparison of LUT and NOAA Bottom Types

• 16 m resolution• Unknown pixels removed from both data sets.

NOAA 1992FERI LUT 2002 SandSeagrassCoralHardbottom

Page 11: Spectral Requirements for Resolving Shallow Water Information Products W. Paul Bissett and David D. R. Kohler

Copyright © 2004 Florida Environmental Research Institute

Unpublished work - all rights reserved

NOAA COAST WorkshopSeptember 29,, 2004

UNCLASSIFEDPROPRIETARY INFORMATION

11

Error Matrix of LUT/NOAA Comparison

NOAABottom Type Sand Coral Seagras

sHardbottom

LUT Total

LUT % Error

Producer’s Accuracy

LUT

Sand 23991 8689 16814 97 49591 51.62% 48.38%

Coral 150 49 4311 501 5011 99.02% 0.98%

Seagrass 3488 8288 184576 20154 216506 14.75% 85.25%

Hardbottom 11874 8067 55682 12184 87807 86.12% 13.88%

Total Instances

NOAA Total 39503 25093 261383 32936 358915

NOAA Error 39.27% 99.80%

29.38% 63.01%

User’s Accuracy

60.73% 0.20% 70.62% 36.99%

Total Classification Accuracy

61.52%

Page 12: Spectral Requirements for Resolving Shallow Water Information Products W. Paul Bissett and David D. R. Kohler

Copyright © 2004 Florida Environmental Research Institute

Unpublished work - all rights reserved

NOAA COAST WorkshopSeptember 29,, 2004

UNCLASSIFEDPROPRIETARY INFORMATION

12

Evidence of Algal Overgrowth in the Florida Keys

Typical algal overgrowth on

coral rubble in Florida. Photo courtesy of U.S. Department of the Interior, U.S. Geological Survey, Center for Coastal Geology

Caulerpa brachypus is a nonnative macroalgae that has invaded Florida's coral reefs. Photo courtesy of Harbor Branch Oceanographic Institution, Inc.

Page 13: Spectral Requirements for Resolving Shallow Water Information Products W. Paul Bissett and David D. R. Kohler

Copyright © 2004 Florida Environmental Research Institute

Unpublished work - all rights reserved

NOAA COAST WorkshopSeptember 29,, 2004

UNCLASSIFEDPROPRIETARY INFORMATION

13

Simulating Impacts of Sensor Design• Created 216K different

hyperspectral Rrs(λ) spectra with various water depths, IOPs, and bottom reflectance spectra.

•These spectra were used to create reduced resolution spectra of 5, 15, 25, and 35 nm continuous spectra to test impact of sensor spectral resolution and radiometric discretization.

Page 14: Spectral Requirements for Resolving Shallow Water Information Products W. Paul Bissett and David D. R. Kohler

Copyright © 2004 Florida Environmental Research Institute

Unpublished work - all rights reserved

NOAA COAST WorkshopSeptember 29,, 2004

UNCLASSIFEDPROPRIETARY INFORMATION

14

Spectral Resolution in Shallow Water(OC4)

9,754 Hydrolight runs with various depths (0.1 – 20 meters), IOPs (25 with chl = 0.0. to 0.2), bottom reflectances (60 different spectra).

Max[Rrs(443), Rrs(490), Rrs(510)] /Rrs(555) = 1.0 ± 0.005

OC4 Chlorophyll a = 2.32 mg/m3

Page 15: Spectral Requirements for Resolving Shallow Water Information Products W. Paul Bissett and David D. R. Kohler

Copyright © 2004 Florida Environmental Research Institute

Unpublished work - all rights reserved

NOAA COAST WorkshopSeptember 29,, 2004

UNCLASSIFEDPROPRIETARY INFORMATION

15

Radiometric Discretization

Depth = 25 m IOP = Case 1 / chl of 2 mg/m3Bottom = clean sand

Dynamic Range was set so that the sensor would not saturate over dry sand. However, the Rrs does not include atmosphere.

Inclusion of the atmosphere would dramatically worsen this discretization error, since most of the signal is from the atmosphere, and the bit resolution of the dark signals would thereby be lessened.

Page 16: Spectral Requirements for Resolving Shallow Water Information Products W. Paul Bissett and David D. R. Kohler

Copyright © 2004 Florida Environmental Research Institute

Unpublished work - all rights reserved

NOAA COAST WorkshopSeptember 29,, 2004

UNCLASSIFEDPROPRIETARY INFORMATION

16

Requirements Analysis

• 216k simulated shallow water reflectance spectra

• Resampled each to represent different combinations of sensor bit levels (8, 10, 12, and 14) and band widths (5, 15, 25, and 30)

• Found the absolute spectral difference from the original and resampled spectra to determine the amount of information lost due to the configuration– 1) Σ {absolute[ Rrs(λ) – Rrs_resample(λ)]}– 2) Average (absolute{ [ Rrs(λ) – Rrs_resample(λ)] /

Rrs(λ) } )

Page 17: Spectral Requirements for Resolving Shallow Water Information Products W. Paul Bissett and David D. R. Kohler

Copyright © 2004 Florida Environmental Research Institute

Unpublished work - all rights reserved

NOAA COAST WorkshopSeptember 29,, 2004

UNCLASSIFEDPROPRIETARY INFORMATION

17

Histogram of Average Relative Error (Focus on Discretization)

Page 18: Spectral Requirements for Resolving Shallow Water Information Products W. Paul Bissett and David D. R. Kohler

Copyright © 2004 Florida Environmental Research Institute

Unpublished work - all rights reserved

NOAA COAST WorkshopSeptember 29,, 2004

UNCLASSIFEDPROPRIETARY INFORMATION

18

Contours of 3-D Avg Rel Err Surface (Viewed from Top)

Percentage of counts per Bit Level that fell beyond 50% relative error:

14 bit – 0.00%

12 bit – 0.02%

10 bit – 10.92%

8 bit – 41.92%

Page 19: Spectral Requirements for Resolving Shallow Water Information Products W. Paul Bissett and David D. R. Kohler

Copyright © 2004 Florida Environmental Research Institute

Unpublished work - all rights reserved

NOAA COAST WorkshopSeptember 29,, 2004

UNCLASSIFEDPROPRIETARY INFORMATION

19

Contours of 3-D Avg Rel Err Surface (Viewed from Top)

Percentage of counts per Bit Level that fell beyond 50% relative error:

14 bit – 0.01%

12 bit – 0.02%

10 bit – 6.51%

8 bit – 34.97%

Page 20: Spectral Requirements for Resolving Shallow Water Information Products W. Paul Bissett and David D. R. Kohler

Copyright © 2004 Florida Environmental Research Institute

Unpublished work - all rights reserved

NOAA COAST WorkshopSeptember 29,, 2004

UNCLASSIFEDPROPRIETARY INFORMATION

20

Contours of 3-D Avg Rel Err Surface (Viewed from Top)

Percentage of counts per Bit Level that fell beyond 50% relative error:

14 bit – 0.45%

12 bit – 0.56%

10 bit – 7.82%

8 bit – 36.55%

Page 21: Spectral Requirements for Resolving Shallow Water Information Products W. Paul Bissett and David D. R. Kohler

Copyright © 2004 Florida Environmental Research Institute

Unpublished work - all rights reserved

NOAA COAST WorkshopSeptember 29,, 2004

UNCLASSIFEDPROPRIETARY INFORMATION

21

Contours of 3-D Avg Rel Err Surface (Viewed from Top)

Percentage of counts per Bit Level that fell beyond 50% relative error:

14 bit – 0.96%

12 bit – 1.89%

10 bit – 8.23%

8 bit – 35.28%

Page 22: Spectral Requirements for Resolving Shallow Water Information Products W. Paul Bissett and David D. R. Kohler

Copyright © 2004 Florida Environmental Research Institute

Unpublished work - all rights reserved

NOAA COAST WorkshopSeptember 29,, 2004

UNCLASSIFEDPROPRIETARY INFORMATION

22

Histogram of Average Relative Error (Focus on Band Width)

Page 23: Spectral Requirements for Resolving Shallow Water Information Products W. Paul Bissett and David D. R. Kohler

Copyright © 2004 Florida Environmental Research Institute

Unpublished work - all rights reserved

NOAA COAST WorkshopSeptember 29,, 2004

UNCLASSIFEDPROPRIETARY INFORMATION

23

Contours of 3-D Avg Rel Err Surface (Viewed from Top)

Percentage of counts per width that fell beyond 50% relative error:

35 nm – 0.96%

25 nm– 0.45%

15 nm– 0.01%

5 nm– 0.00%

Page 24: Spectral Requirements for Resolving Shallow Water Information Products W. Paul Bissett and David D. R. Kohler

Copyright © 2004 Florida Environmental Research Institute

Unpublished work - all rights reserved

NOAA COAST WorkshopSeptember 29,, 2004

UNCLASSIFEDPROPRIETARY INFORMATION

24

Contours of 3-D Avg Rel Err Surface (Viewed from Top)

Percentage of counts per width that fell beyond 50% relative error:

35 nm – 1.89%

25 nm– 0.56%

15 nm– 0.02%

5 nm– 0.02%

Page 25: Spectral Requirements for Resolving Shallow Water Information Products W. Paul Bissett and David D. R. Kohler

Copyright © 2004 Florida Environmental Research Institute

Unpublished work - all rights reserved

NOAA COAST WorkshopSeptember 29,, 2004

UNCLASSIFEDPROPRIETARY INFORMATION

25

Contours of 3-D Avg Rel Err Surface (Viewed from Top)

Percentage of counts per width that fell beyond 50% relative error:

35 nm – 8.23%

25 nm– 7.82%

15 nm– 6.51%

5 nm– 10.92%

Page 26: Spectral Requirements for Resolving Shallow Water Information Products W. Paul Bissett and David D. R. Kohler

Copyright © 2004 Florida Environmental Research Institute

Unpublished work - all rights reserved

NOAA COAST WorkshopSeptember 29,, 2004

UNCLASSIFEDPROPRIETARY INFORMATION

26

Contours of 3-D Avg Rel Err Surface (Viewed from Top)

Percentage of counts per width that fell beyond 50% relative error:

35 nm – 35.28%

25 nm– 36.55%

15 nm– 34.97%

5 nm– 41.92%

Page 27: Spectral Requirements for Resolving Shallow Water Information Products W. Paul Bissett and David D. R. Kohler

Copyright © 2004 Florida Environmental Research Institute

Unpublished work - all rights reserved

NOAA COAST WorkshopSeptember 29,, 2004

UNCLASSIFEDPROPRIETARY INFORMATION

27

HES-CW Channel Specifications

Nominal Threshold Channel Center

Wavelength (um)

Nominal Threshold

Resolution (um)

Nominal Difference from Wavelength

Center (um)

Nominal Threshold

Signal to Noise

Nominal GOAL Channel Center

Wavelength (um)

Nominal GOAL Resolution (um)

Nominal Goal Signal to

Noise

0.412 0.02 0.407 through 0.987 0.01

0.443 0.02 0.031 0.57 0.010.477 0.02 0.034 1.38 0.030.49 0.02 0.013 1.61 0.060.51 0.02 0.02 2.26 0.050.53 0.02 0.02 11.2 0.8

0.55 0.02 0.02 12.3 1

0.645 0.02 0.095

Nominal Threshold

Horiz. Resolution

Nominal Goal Horiz.

Resolution

0.667 0.01 0.0220.678 0.01 0.0110.75 0.02 0.072

0.763 0.02 0.0130.865 0.02 0.1020.905 0.035 0.04

300-meters all channels

(at Equator)

150-meters all channels

(at Equator)

300 to 1 all channels

900 to 1 all channels

Page 28: Spectral Requirements for Resolving Shallow Water Information Products W. Paul Bissett and David D. R. Kohler

Copyright © 2004 Florida Environmental Research Institute

Unpublished work - all rights reserved

NOAA COAST WorkshopSeptember 29,, 2004

UNCLASSIFEDPROPRIETARY INFORMATION

28

Summary

• Imaging spectroscopy can address the requirement to simultaneous solve for bathymetry, IOPs, and bottom reflectance in the shallow water environment. This will allow for the production of robust image products, including water clarity, extreme phytoplankton concentrations (red and brown tides), resuspended sediments, river discharges.

• This analysis suggests there may be little difference between 12 and 14 bits discretization, and 5 and 10 nm band width.– However, this analysis does not include TOA reflectance,

which will reduce the bit resolution of the Rrs.– It also does not include noise, which will impact the

simulated results.

Page 29: Spectral Requirements for Resolving Shallow Water Information Products W. Paul Bissett and David D. R. Kohler

Copyright © 2004 Florida Environmental Research Institute

Unpublished work - all rights reserved

NOAA COAST WorkshopSeptember 29,, 2004

UNCLASSIFEDPROPRIETARY INFORMATION

29

Summary

• The real issue are how the impacts of the requirements affect image products. These products require new algorithms, which to date have not been completely developed and tested (cart and horse problem).– One of the issues not addressed in this study is that

average percentage errors are not the same as reduce spectral information. • The requirements for resolution,

calibration, and SNR should be driven by product generation.

• Real high spatial and spectral data sets exist, with future collections planned, that may be used to develop these products.

Page 30: Spectral Requirements for Resolving Shallow Water Information Products W. Paul Bissett and David D. R. Kohler

Copyright © 2004 Florida Environmental Research Institute

Unpublished work - all rights reserved

NOAA COAST WorkshopSeptember 29,, 2004

UNCLASSIFEDPROPRIETARY INFORMATION

30

Total Kelp : 3.93 sq km

Total Kelp : 0.55 sq km

Total Kelp : 5.73 sq km

Total Kelp : 0.76 sq km

Total Kelp : 2.99 sq km

Total Kelp : 1.11 sq km

Monterey Bay

Morro Bay

San Luis Bay

Big Sur

CICORE Central California 2002