USF Libraries
USF Digital Collections

Bottom albedo derivations using hyperspectral spectrometry and multispectral video

MISSING IMAGE

Material Information

Title:
Bottom albedo derivations using hyperspectral spectrometry and multispectral video
Physical Description:
Book
Language:
English
Creator:
Farmer, Andrew Scott
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla.
Publication Date:

Subjects

Subjects / Keywords:
Ocean
Optics
Albedo
Sand
Seagrass
Dissertations, Academic -- Marine Science -- Masters -- USF   ( lcsh )
Genre:
government publication (state, provincial, terriorial, dependent)   ( marcgt )
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Summary:
ABSTRACT: Remote sensing reflectance data collected with a remotely operated vehicle (ROV) were used to derive bottom albedo and optical properties for a shallow marine environment near Lee Stocking Island, Bahamas. Optical model inversion techniques were applied to hyperspectral measurements of remote-sensing reflectance to derive water absorption and backscatter coefficients. Using these derived water properties, path attenuation and radiance effects were removed from bottom observations to derive bottom albedos. Histograms from multispectral, hyperspatial video images were used to determine the albedo range of optical end members observed in scenes of sand and seagrass. Variations of spectral signatures for optical end members caused by path-adjacency effects are shown to influence the reflectance measurements.Low-altitude albedo histograms for heterogeneous scenes demonstrate higher contrast between sand and seagrass than is observed at higher altitudes, even after correction for path radiance and attenuation effects. For example, reflected light from bright sand scatters into the field of view of dark seagrass, while less light scatters out from the seagrass into the field of view of sand. This decreases the apparent sand albedo, and increases that for seagrass when viewed from higher altitudes, including aircraft. Evidence provided suggests that simple bottom classifications based upon expected albedo values for scene end members are in error unless the water depth is very shallow.
Thesis:
Thesis (M.S.)--University of South Florida, 2005.
Bibliography:
Includes bibliographical references.
System Details:
System requirements: World Wide Web browser and PDF reader.
System Details:
Mode of access: World Wide Web.
Statement of Responsibility:
by Andrew Scott Farmer.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 71 pages.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001681064
oclc - 62735150
usfldc doi - E14-SFE0001054
usfldc handle - e14.1054
System ID:
SFS0025375:00001


This item is only available as the following downloads:


Full Text

PAGE 1

Bottom Albedo Derivations Using Hyperspectral Spectrometry and Multispectral Video by Andrew Scott Farmer A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science College of Marine Science University of South Florida Major Professor: Kendall L. Carder, Ph. D. Pamela Hallock Muller, Ph.D. Gabriel Vargo, Ph.D. Date of Approval: April 1, 2005 Keywords: ocean, optics, albedo, sand, seagrass Copyright 2005, Andrew Scott Farmer

PAGE 2

Acknowledgements This project was completed as part of the Office of Naval Research’s Coastal Benthic Optical Properties initiative. Thanks goes out to all of those who participated in this project and to those responsible for funding provided. I would like to give special thanks to the entire College of Marine Science Ocean Optics lab for all of their input and scientific support. I especially thank David Costello for providing the extrapolation data used in this study for inversion data verification. Additionally, I would like to thank my wife Michelle and all of my friends and family members for their various roles of support during my academic career.

PAGE 3

i Table of Contents List of Figures ii Abstract v Chapter 1. Introduction 1 1.1 Background 3 1.2 Study Sites 8 1.3 Instrumentation 13 Chapter 2. Methods 16 2.1 Hyperspectral Albedo Derivations 16 2.2 Video Albedo Derivations 20 Chapter 3. Results and Discussion 22 3.1 Rainbow Gardens 22 3.2 Adderly Cut 38 3.3 Horseshoe Reef 47 Chapter 4. BRDF and Path-Adjacency Effects 54 Chapter 5. Summary and Conclusions 59 References 61

PAGE 4

ii List of Figures Figure 1. Landsat TM-5 1991/03/03 showing location of three study areas near Lee Stocking Island, Bahamas 2 Figure 2. Sony XC-999 color video image of the heavily fouled, sparse seagrass located at Rainbow Gardens, near Lee Stocking Island, Bahamas 10 Figure 3. Elbex, Inc. EXC-525 color navigational video image of thick, unfouled seagrass at Adderly Cut, near Lee Stocking Island, Bahamas 11 Figure 4. Grayscale Xybion 550 nm spectral video image of bright, oolitic sand at Horseshoe Reef, near Lee Stocking Island, Bahamas 12 Figure 5. Image of the ROV Rosebud with se nsors used in this study identified 15 Figure 6 Illustration of vehicl e orientation and shading. 19 Figure 7. Above-water Rrs at Rainbow Gardens, which was used to model water properties and validate in-w ater, modeled water properties 26 Figure 8. rrs measured at vehicle altitude in-water, and used for water-property and albedo derivations 27 Figure 9. Comparison of above-water and in-water derived total absorption coefficients. 28 Figure 10. Comparison of above-water and in-water derived total backscatter coefficients 29 Figure 11. Comparison of in-water derived albedos from the inversion method versus the extrapolation method at Rainbow Gardens 30 Figure 12. Comparison of albedos derived from the hyperspectral inversion method versus the multispectral Xybion video inversion method at Rainbow Gardens 31 Figure 13. Comparison of the multispectral video albedo from the second vertical profile versus the first profile’s, and the hyperspectral’s inversion-derived albedos at Rainbow Gardens 32

PAGE 5

iiiFigure 14. Low-altitude (1.92 m) albedo hist ograms from subsampled a) sand b) seagrass and c) sandy-seagrass ROIs at the second location of Rainbow Gardens 33 Figure 15. Low-altitude (1.92m) albedo histograms from the 550 nm channel data shown in Figure 13a,b, and c. 35 Figure 16. High-altitude (4.58m) albedo hist ograms from subsampled a) sand b) seagrass and c) sandy-seagrass ROIs at the second location of Rainbow Gardens 36 Figure 17. Comparison of the total absorption coefficients derived at Adderly Cut and Rainbow Gardens 41 Figure 18. Comparison of the total backscatter coefficients derived at Adderly Cut and Rainbow Gardens 42 Figure 19. Comparison of the three vertical rrs profiles at Adderly Cut 43 Figure 20. Comparison of in-water derived albedos from the inversion method versus the extrapolati on method at Adderly Cut 44 Figure 21. Comparison of albedos derived from the hyperspectral inversion method versus the multispectral Xybion video inversion method at Adderly Cut 45 Figure 22. Albedo histograms from seagrass at Adderly Cut a) low altitude (1.82m) and b) high altitude (2.64m) 46 Figure 23. Comparison of inversion derived total absorption coefficients at all three study locations 49 Figure 24. Comparison of inversion derived total backscatter coefficients at all three study locations 50 Figure 25. Comparison of albedos derived from the hyperspectral inversion method versus the extrapolati on method at Horseshoe Reef 51 Figure 26. Comparison of albedos derived from the hyperspectral inversion method versus the multispectral video inversion method at Horseshoe Reef 52 Figure 27. Albedo histrograms from a sa nd ROI at Horseshoe Reef a) low altitude (1.05m) and b) high altitude (10.61m) 53

PAGE 6

ivFigure 28. Comparison of 550 nm albedos at different altitudes from Rainbow Gardens 57 Figure 29. Illustration showing how vertical structure can create a canopy effect. 58

PAGE 7

v Bottom Albedo Derivations Using Hyperspectral Spectrometry Multispectral Video Andrew S. Farmer ABSTRACT Remote sensing reflectance data collected with a remotely operated vehicle (ROV) were used to derive bottom albedo and optical properties for a shallow marine environment near Lee Stocking Island, Bahamas. Optical model inversion techniques were applied to hyperspectral measurements of remote-sensing reflectance to derive water absorption and backscatter coefficients. Using these derived water properties, path attenuation and radiance effects were removed from bottom observations to derive bottom albedos. Histograms from multispectral, hyperspatial video images were used to determine the albedo range of optical end members observed in scenes of sand and seagrass. Variations of spectral signatures for optical end members caused by pathadjacency effects are shown to influence the reflectance measurements. Low-altitude albedo histograms for heterogeneous scenes demonstrate higher contrast between sand and seagrass than is observed at higher altitudes, even after correction for path radiance and attenuation effects. For example, reflected light from bright sand scatters into the field of view of dark seagrass, while less light scatters out from the seagrass into the field of view of sand. This decreases the apparent sand albedo, and increases that for seagrass

PAGE 8

viwhen viewed from higher altitudes, including aircraft. Evidence provided suggests that simple bottom classifications based upon expected albedo values for scene end members are in error unless the water depth is very shallow. ROV-collected reflectance data allows for analysis of individual end members and their collective influence on the total upwelling light signal at various altitudes, and suggests that remote-sensing retrievals of accurate bottom albedos for heterogeneous and high-contrast components of the bottom setting will not be possible without corrections for path-adjacency effects.

PAGE 9

1 Chapter 1 Introduction The objective of the Coastal Benthic Optical Properties (CoBOP) initiative at Lee Stocking Island (LSI), Bahamas, was to study light propagation in an optically shallow, aquatic environment, where bottom effects on water-leaving radiance could be observed. Comprehensive data sets were collected to develop and test new methodologies for mapping and classifying bottom constituents. The objectives of this study were to derive bottom albedo using vertical profiles of measured remote-sensing reflectance (rrs) and spectral images over a heterogeneous sandy seagrass mixture, as well as over two homogenous end members, sand and seagrass. Figure 1 illustrates a Rectified LandSat Image of LSI during CoBOP 1998. This image shows the optically shallow environment of the barrier island chain known as the Exumas. The color and intensity of light measured from the water surrounding these islands is largely affected by reflectance from the bottom. The three locations from which albedo derivations were made are also shown in Figure 1.

PAGE 10

2Figure 1. Landsat TM-5 1991/03/03 showing location of three study areas near Lee Stocking Island, Bahamas

PAGE 11

3 1.1 Background Seagrass beds are often studied with the use of remote-sensing reflectance (Rrs) data, as they are a critical component in sustaining many shallow-water ecosystems. They are spawning grounds for many important commercially viable fish and help to remove pollutants and sediment from the water column. The health of an estuarine ecosystem is often judged by the extent of its seagrass coverage (Hemminga and Duarte, 2000). Remote-sensing reflectance provides an efficient method for monitoring and mapping seagrass coverage. In coastal regions where bottom reflectance influences the upwelling light field, interpreting above-water Rrs measurements becomes complex due to wave facets and multiple scattering effects causing non-uniform, or no-Lambertian bottom reflectance. The uniformity of this bottom reflectance is characterized by the bottom reflectance distribution function ( BRDF). Non-Lambertian BRDF can occur from scattering effects such as path-adjacency, shading, and canopy effects. Pathadjacency occurs when bright objects scatter light into the field of view of a sensor when it is observing adjacent dark objects (e.g. Reinersman and Carder, 1995). These light scattering effects induce errors in quantitative remote assessments of seagrass health and coverage. Remote monitoring and classification of land types has been ongoing with the use of aircraft imagery such as the Portable Hyperspectral Imager for Low-Light Spectroscopy (PHILLS) and the Airborne Visible and Infrared Imaging Spectrometer (AVIRIS) (Vane et al., 1993; Davis et al., 2002). In air, some light scattering, or path

PAGE 12

4radiance, occurs for terrestrial scenes, thus decreasing contrast and degrading scene clarity. However, no significant path absorp tion is present for visible wavelengths. While most land-classification schemes derived from data acquired by high-altitude aircraft traditionally have not been corrected for atmospheric effects, Reinersman and Carder (1995) corrected a high-contrast, infr ared land-ocean scene near Big Pine Key, Florida, for both path radiance and path attenuation. Although standard atmospheric corrections were applied in the Reinersman and Carder (1995) study, path-adjacency was not initially accounted for. As a consequence, canals and ponds had apparent reflectances or albedos that were too high b ecause light reflected from adjacent bright land targets was scattered into the sensor’s view of the canal. They then theoretically corrected this scene for path-adjacency effects, which increased the contrast between bright land and dark water. Although these authors proposed an infrared correction for path-adjacency effects in air, this correction is not directly applicable in seagrass studies because at infrared wavelengths the ocean appears black and the bottom is not visible. Current Rrs models for the ocean do not account for path-adjacency. Rrs is the ratio of upwelling, water-leaving radiance to downwelling irradiance measured above water. Similarly, rrs is the ratio of upwelling radiance to downwelling irradiance measured below water. A simple model for deep water (Morel and Prieur, 1977), describes Rrs in terms of path reflectance as a function of solar zenith angle, f (below water) and Q (above water), and attenuation, a( ) and bb( ): f bb( ) Rrs = __________________ Equation (1) Q [a( ) + bb( )]

PAGE 13

5Previous studies of bottom imagery, rrs, and Rrs have separated out the water-column component from seafloor components (Pratt, 1997; Lee et al.,. 1998, 1999). These calculations for bottom albedo represented the combined contributions of multiple constituents of the seafloor for a 10o field of view. As a result, they did not allow for a direct analysis of individual components in a heterogeneous scene. Only in scenes with uniform sand or grass, far from contrast transitions (Lee et al., 2001), could individual end members be independently characterized. Components such as corals, sand, seagrass, ooids, and structural parameters (e.g., shading) all impact the upwelling light fiel d. To characterize individual end members within both heterogenous and homogeneous scenes, Xybion multispectral video cameras were used for bottom imaging (Costello et al., 1995). By removing the water-column effects from the Xybion upwelling imagery with hyperspatial resolution, bottom components were analyzed on a much smaller scale. To do this effectively and accurately, the first step was to apply an algorithm that separated the rrs signal into bottom and water-column components as in Equation 2. Total rrs is the summation of rrs signals from the water column (rrs C) and rrs from the sea floor (rrs B): B rs C rs rsr r r Eq (2) In optically shallow, aquatic environments, the derivation of inherent optical properties from rrs measurements is complicated by bottom reflectance. Equation 2 is for level, homogeneous bottom types. As with most radiative transfer models it assumes a Lambertian, i.e., a uniform in all directions, bottom reflectance. It does not include interaction effects between adjacent pixels or fields of view and thus does not account for path-adjacency effects. For this study, the model for rrs by Lee et al. (1999) was applied.

PAGE 14

6This model, shown in Equation 3, is for nadir-viewing sensors for horizontally homogenous scenes. }. ] [ { }) ] [ { (H D exp H D exp 1 r rB u ) cos( 1 1 C u ) cos( 1 dp rs rsw w Eq (3) This algorithm describes water-column path radiance contributions as functions of downward and upward diffuse-attenuation coe fficients, absorption and backscattering coefficients, and water depth (H). Here, rrs dp is the remote sensing reflectance for optically deep water. It is modified by a bracketed truncation term for finite water column depth and path lengths. A similar function using a coefficient for a Lambertian bottom albedo ( ), accounts for bottom contributions in the second term. Albedo, A( ), as defined by Mobley (1994), is simply the irradiance reflectance of a surface (Eq. 4). In Equation 3, bottom albedo, ( ), is the irradiance reflectance of the bottom. Eu( ) / Ed( ) Eq. (4) Lee et al. (1999) used a predictor-corrector inversion approach to derive the absorption, a( ), and backscattering, bb( ) coefficients described in Equations 1, 5, and 6. In this inversion approach, water absorption aw( ) and backscattering bbw( ) coefficients are known functions (Pope and Fry, 1997; and Morel, 1974, respectively) and spectral shapes of gelbstoff absorption ag( ), phytoplankton absorption a( ), and particulate backscatter bbp( ) are unique (Carder and Steward, 1985; Carder et al., 1991; and Lee et al., 1994, respectively). This allows scaling factors for each component, together with one each for bottom albedo and water depth, to be adjusted or optimized to provide the best fit of modeled rrs to measured rrs curves. (Lee et al. 1999, 2001)

PAGE 15

7a( ) = aw( ) + a( ) + ag( ) Eq (5) bb( ) = bbw( ) + bbp( ) Eq (6) A program developed by Lee et al. (1999) was used to derive these inherent optical properties from Rrs( ) and rrs( ) spectra for different sites.

PAGE 16

8 1.2 Study Sites Study Site 1, Rainbow Gardens, is an area of sparse Thalassia testudinum (turtlegrass) at approximately 9 – 10 meters in depth (Fig. 2). Located at 23.794o N, 76.137o W, the bottom composition was previously characterized as having a heavy epiphytic covering by Drake et al. (2003) with an albedo of about .375 at 550 nm (Louchard, et al., 2003). Data were collected on May 26, 2000, at about 10:00 a.m. EST. Winds were calm and skies were overcast. Study Site 2, Adderly Cut, at approximately 4 meters depth (Fig. 3), is an area of dense turtlegrass meadows with little to no epiphytic load with an albedo of about .100 at 550 nm (Louchard et al., 2003). This location is an area of strong tidal exchange between Exuma Sound and the surrounding Bahama Banks (Louchard et al., 2003; Drake et al., 2003). Data were collected at 23.466o N, 76.113o W on May 24, 2000, approximately 10:00 a.m. EST. When these data were collected, winds were calm and there was little to no cloud cover. Study Site 3, Horseshoe Reef, was located on the fringe of Lee Stocking Island adjacent to Exuma Sound. This area had been previously characterized by Boss and Zaneveld (2003) as oolitic sand flats next to a reef composed of macroalgae, sponges and corals. Flat oolitic sands are considered to be essentially Lambertian reflectors (Mobley and Sundman, 2003) with an albedo of about 0.525 at 550 nm (Louchard et al., 2003). The imagery was collected over an area of bright sand at approximately 11 meters in

PAGE 17

9depth (Fig. 4). These data were collected about 10:00 a.m. EST on May 26, 1999, at 23.772o N, 76.089o W on calm day with little to no cloud cover.

PAGE 18

10Figure 2. Sony XC-999 color video image of the heavily fouled, sparse seagrass located at Rainbow Gardens, near Lee Stocking Island, Bahamas

PAGE 19

11Figure 3. Elbex, Inc. EXC-525 color navigational video image of thick, unfouled seagrass at Adderly Cut, near Lee Stocking Island, Bahamas

PAGE 20

12Figure 4. Grayscale Xybion 550 nm spectral video image of bright, oolitic sand at Horseshoe Reef, near Lee Stocking Island, Bahamas

PAGE 21

13 1.3 Instrumentation The spectral data used in this study were collected by the Bottom Classification and Albedo Package (BCAP), which was mounted on the remotely operated vehicle (ROV) Rosebud (Costello and Carder, 1994). The instrument packages consisted of 512channel upwelling radiance, Lu( ), and downwelling irradiance, Ed( ), sensors (Spectrix), a Xybion, Inc. IMC-301 multispectral (6-channel) intensified video-imaging camera, and a Sony, Inc. XC-999 high-resolution color video camera. The Spectrix sensors were built by the Ocean Optics Laboratory at the University of South Florida and calibrated against a Licor 1800 irradiometer (~5%) and a calibrated Spectrix radiance meter (Cattrall et al., 2002) Multispectral video albedos were calibrated with two standard reflectors (~5% Spectralon) placed within the field of view of the Xybion camera. Additional instrumentation onboard the vehicle used in this study consisted of a Falmouth Scientific, Inc. Model MCTD-DBP-5 CTD, a Tritech International Limited Model PA-200.20-P5 acoustic altimeter, a Elbex, Inc. Model EXC-525 color navigational video camera, and a suite of ancillary optical sensors that were not used in this analysis (e.g., Carder et al., 2001). The instrumentation are shown mounted on the ROV in Figure 5. In this study, two different Xybion video cameras were used between the years 1999 and 2000. Both cameras had six spectral filters mounted on a filter wheel, which were timed-synchronized so every 0.0333 seconds a new filter rotated in front of the camera lense and each frame represented one of the six spectral bandwidths. Each

PAGE 22

14camera had set gain functions, but had self-adjusting integration times to avoid signal saturation in any given scene. The primary difference between the two cameras was the filters. The spectral channels 1 – 6 from the camera used in 1999 were 460 nm, 520 nm, 575 nm, 620 nm, 685 nm, and 730 nm respectively. The spectral channels 1 – 6 from the camera used in 2000 were 415 nm, 445 nm, 490 nm, 520 nm, 550 nm, and 670 nm respectively.

PAGE 23

15Figure 5. Image of the ROV Rosebud with sensors used in this study identified

PAGE 24

16 Chapter 2 Methods 2.1 Hyperspectral Albedo Derivations The Lee et al., (1999) inversion approach to derive bottom depths and inherent optical properties (IOPs) of the water column was used to derive bottom albedo. In my study, water depth (H) was replaced by vehicle altitude to determine the optical properties of the water column between the ROV Rosebud and the bottom. It is important to note that when vehicle was collecting data it was oriented so that the tail end of the vehicle was towards the sun to avoid any vehicle shading from occurring within the imagery. With the low solar angles from data collection around 10:00 a. m., the imagery data was the combination of reflectance of shaded or sky light, and the Fresnel light reflectance which is light reflected direc tly off of objects from within the scenes. Ultimately in scenes with less Fresnel light, due to vehicle orientation collecting data on the shady side of BRDF, the albedo values are lowered (Figure 6). Work by Weidemann et al., (1995) suggests that additional errors may occur from this solar angle and other constituents contributing to total backscattering that are not accounted for in the inversion model. The largest variation in the procedure was the source of the 550 nm-normalized, bottom-albedo data. At the lowest altitude measurement of each vertical profile, the 550 nm-normalized measured rrs curve served temporarily as the bottom albedo within the optimization (predictor-corrector) model. Accordingly, when the inversion program was

PAGE 25

17run, the output altitude was set to zero. Albedo was then retrieved by re-writing the algorithm in Equation 3 so } ] [ { }) ] [ { (H D exp H D exp 1 r r rB u ) cos( 1 C u ) cos( 1 dp rs rs bottom rsw w Eq (7) rbottom rs Eq (8) In Equations 7 and 8, the modeled bottom rrs equals that of the measured rrs and the albedo curve derived from the initial inversion. In the model, rrs, altitude was then adjusted from zero to the known altitude of the vehicle. This adjustment changed the modeled albedo, which was then placed back into the original inversion scheme. This iterative process was continued until variances between the known altitude and modeled altitude were minimized (< 0.001). The final modeled albedo curve was normalized at 550 nm and was used to replace the standard albedo curve for deriving water properties from the hyperspectral rrs spectra measured at higher altitudes in the water column. The first data set was collected at Rainbow Gardens. These data were used to develop methodology from the in-water inversions used in this study. The inherent optical properties (IOPs) derived were validated with a more standard inversion approach using above-water-measured Rrs (e.g., Lee et al., 1999). This approach applies an earlier Rrs model by Lee et al. (1998) where: 0.5rrs Rrs _____________ Eq. 9 1-1.5rrs This allows the direct comparison of IOPs derived from above-water data to those derived from in-water data. The above-water data were collected shipboard on the same date and approximate time. The in-water derived albedos were also compared with

PAGE 26

18albedos from the previously established ex trapolation method described by Costello and Carder (2002) using the same rrs data. For the remaining data sets only the albedo derivations were directly validated. The i nherent optical properties derived by in-water rrs inversion were only validated by relative comparison to the Rainbow Gardens inversion-derived IOP data set.

PAGE 27

19Figure 6. Illustration of vehicle orientation and shading.

PAGE 28

20 2.2 Video Albedo Derivations To derive bottom albedos using the multispectral video, the inherent optical properties derived hyperspectrally for the water column were applied to the video data with the same Lee et al. (1999) algorithm to remove effects of the intervening water layers over the bottom. Here rrs was calculated using the video imagery. Using real-time software, histogram functions for video im ages, end member and standard-reflector regions of interest (ROIs) within each scene were selected and analyzed for their 8-bit grayscale intensity values. They were combined to produce remote sensing reflectance spectra for each scene component as described in Equation 9. End-member pixel intensity was divided by the average reflector intensity. This obtained value was multiplied by the calibrated Spectralon reflectance (g). This operation cancels out the exposure time, gain, and any other common calibration coefficients for each channel and yields the remote sensing reflectance for the region of interest. rrs = (Io/Ir)g Eq (9) where Io is the pixel intensity for the end member region of interest, and Ir is the average pixel intensity of the Spectralon reflector. Th e gray standard reflector had a reflectance, g, that was nearly 0.05, but varied slightly spectrally. The reflector is visible in the top left corner of the XC-999 video scene in Figure 2. The use of this calibrated reflector allowed for a direct measurement of the downwelling irradiance within scenes studied. The ratio of these pixel values with the pixel values from objects on the seafloor yielded

PAGE 29

21reflectance measurements. The absorption and backscattering coefficients, as well as vehicle altitude, were treated as known values from the hyperspectral reflectance model retrievals. The video measured reflectance and the previously derived attenuation coefficients were input into the Lee et al., 1999 algorithm and albedo was solved for on pixel-by-pixel basis. Bottom components of in terest were converted to albedo histograms at various scene locations and altitudes. It is important to note that the magnitude of the frequencies within the histograms varies for each of the averaged video channels. This is due to the fact that when the raw video data is captured using the real-time video software each pixel represents a binned intensity value. The range of the intensities and therefore albedo values within that bin varies as a function of exposure time. Channel 1 of both the 1999 and 2000 cameras was omitted from the experiment because the signal was too weak for the camera array under the natural lighting conditions. Channel 6 of the 1999 camera, 730 nm, was also omitted because this infrared wavelength extended beyond the sensitivity range of the hyperspectral spectrometers. Therefore, no inherent optical properties were derived for these channels. Channels 2 – 5 (520 nm, 575 nm, 620 nm, and 685 nm) were used to analyze the sand end members at Horseshoe Reef. Channels 2 – 6 (445 nm, 490 nm, 520 nm, 550 nm, and 670 nm) were used to study individual end member components associated with the sparse and heavily fouled seagrass from Rainbow Gardens and the thick, unfouled seagrass at Adderly Cut.

PAGE 30

22 Chapter 3 Results and Discussion 3.1 Rainbow Gardens In terms of the water column, this specific data set was taken under ideal conditions. The area was known to be a shallow, well-mixed environment where the water properties were both horizontally and vertically homogenous. On the day of sampling, cloud overcast and an approximate water depth of about 10 m minimized the effects of any wave focusing (e.g., Zaneveld et al., 2001). Figures 7 and 8 illustrate the measured above-water Rrs and the measured in-water rrs profile from which inversion IOPs were derived and validated. These absorption and backscattering coefficients, derived from the inversion process, are compared in Figures 9 and 10. The close fit of the above-water data to the below-water data provides a form of validation for the ROVderived water properties that were used to model the hyperspectral spectrometer and multispectral video albedos. IOP data from Ivey et al. (2002) have validated the Rrs method for deep water near Lee Stocking Island and Lee et al. (1999) validated IOPs for shallow waters. Absorption accuracies exceeding 7% were found for numerical data with errors in ag (440 nm), increasing to about 22% for field data Figure 11 validates the inversion-derived albedo by plotting it against the albedo derived using vertical profiles of Lu and Ed extrapolated to the bottom (Costello et al., 1998). The larger differences between the data sets beyond 580 nm appear be the result of incorrect vehicle-altitude caused by an

PAGE 31

23error in data acquisition software used in the extrapolation method. This impacts greater at longer wavelengths due to the higher absorption coefficients. Additionally, slope in bottom reflectance from approximately 550 to 680 nm in the inversion-derived data is more consistent with other reported bottom reflectance analyses (e.g., Louchard et al., 2003). Coincident inversion-derived albedos from the hyperspectral data and multispectral video are illustrated in Figure 12. The spectra from these two sensors produce very similar results for the same bottom imagery. For additional comparison, a second video albedo scene was sampled from the same area (Fig. 13). From this scene, a thicker patch of seagrass was subsampled. The albedo from this thicker patch of grass has a steeper 550 to 680 nm slope, which further substantiates the error in vehicle altitude in extrapolation data shown in Figure 11. Figure 14 shows albedo histograms from subsampled ROIs from this video scene. Data from subsampled ROIs at a vehicle altitude of approximately 1.92 m are shown for areas of sand (Fig. 14a), seagrass (Fig. 14b), and a sand-seagrass mix (Fig. 14c) within the total bottom regime. At 1.92 m, histogram shapes for sand and seagrass are skewed in opposite directions. All channels of sand had a peak pixel frequency at brighter albedos when compared to their respective channels of seagrass, which had peak frequencies at darker albedos. The histogram in Figure 14c, representing the sandseagrass mix, displays modality features expected in a multi-end member scene. Peaks from the individual end members can be directly matched with these modality features of this mixed regime (Fig. 15). The histograms in Figures 14 and 15 illustrate the combination of specific spectral signatures within the upwelling light field. The combined total-albedo signature of a

PAGE 32

24scene as determined using hyperspectral Spectrix rrs data, though considered a uniform Lambertian reflector, can be broken down in to individual components. Analysis by Lochard et al., (2003) suggest that brighte r-albedo pixels from sand indicate sediment with minimal microalgal composition. The broader, brighter patches of sand indicate algal entrappment and are bright oolitic and other clean, sand-sized sediments. The darker pixels that skew the sand histogram to lower albedo values are most likely due to microalgae and organic detritus entrapment where the sand and seagrass end members meet. The XC-999 high-resolution color video suggests that the brighter pixels that skew the seagrass histograms towards brighter albedo values are from epiphytic growth, oolitic entrapment, and seagrass blades that are oriented normal to the water column. These blades scatter greater amounts of light back towards the sensor. Additional brighter pixels may come from sand within these seagrass beds. The darker regions appear to be from cleaner blades of grass, blades oriented obliquely to the water column, which scatter light forward, as well as self-shading and darker microalgae. Figures 16 shows albedo histograms subsampled from ROIs of the same sand, seagrass, and sandy-seagrass end members mentioned above but from a vehicle altitude of approximately 4.58 m. When the video sensor was closer to the bottom, the fields of view appeared more heterogenous with little path radiance effects on individual pixels in homogenous patches of grass or sand. With a vehicle altitude of only 4.58 m, albedo histograms take on a more homogenous appearance. Histograms have lost the skewed features of the individual end member ROIs (Figs. 16a and 16b) and the modality features of the mixed ROIs (Fig. 16c). When the vehicle ascended from the bottom, the

PAGE 33

25larger field-of-view angles began to include reflected and forward scattered light from other bottom components. This path-adjacency effect is discussed further in Chapter 4.

PAGE 34

26Figure 7. Above-water Rrs at Rainbow Gardens, which was used to model water properties and validate in-water, modeled water properties

PAGE 35

27Figure 8. rrs measured at vehicle altitude in-water, and used for water property and albedo derivations

PAGE 36

28Figure 9. Comparison of above-water and in-water derived total absorption coefficients

PAGE 37

29Figure 10. Comparison of above-w ater and in-water derived total backscatter coefficients

PAGE 38

30Figure 11. Comparison of in-water derived al bedos from the inversion method versus the extrapolation method at Rainbow Gardens

PAGE 39

31Figure 12. Comparison of albedos derive d from the hyperspectral inversion method versus the multispectral Xybion video inversion method at Rainbow Gardens

PAGE 40

32Figure 13. Comparison of the multispectral video albedo from the second vertical profile versus the first profile’s, and the hyperspectral’s inversion derived albedos at Rainbow Gardens

PAGE 41

33Figure 14. Low-altitude (1.92 m) albedo histograms from subsampled a) sand b) seagrass and c) sandy-seagrass ROIs at the second location of Rainbow Gardens a) b)

PAGE 42

34c)

PAGE 43

35Figure 15. Low-altitude (1.92m) albedo histograms from the 550 nm channel data shown in Figure 13a, b, and c.

PAGE 44

36Figure 16. High-altitude (4.58m) albedo histograms from subsampled a) sand b) seagrass and c) sandy-seagrass ROIs at the second location of Rainbow Gardens a) b)

PAGE 45

37c)

PAGE 46

38 3.2 Adderly Cut The location chosen at Adderly Cut was vertically well mixed. Although the data were collected during slack tide for vehicle manuverability, the inherent optical properties were affected by the strong tidal exchange in the area. Absorption in the blue spectral region increased compared to the more isolated Rainbow Gardens environment (Fig. 17). Previous work by Otis et al. (2004) suggests this increase may be due to colored dissolved organic matter (CDOM) transport out through the cut into Exuma Sound eminating from the large drainage area of the waters surrounding the cut (see Fig. 1). The suspended sediment transported with the increased flow also increased the amount of particulate backscattering compared to Rainbow Gardens (Fig. 18). However, even with the increase in light attenuation, this was a clear-water environment that allowed for the growth of thick, unfouled seagrass meadows. Albedo data from scenes from Adderly Cut represent the purest seagrass end member in this study. However, in a shallow water column, ~4m, wave focusing confounds albedo derivations on bright sunny da ys (e.g., see Zaneveld and Boss, 2002). Figure 19 illustrates the increased variability due to wave focusing on three vertical rrs profiles. Note that for these profiles, the rrs values at 3m depth are more red, but are the same at 550 nm for all three depths in the second profile due to wave focusing. The extrapolated albedo shown in Figure 20 is based on the average albedo from all three of these vertical profiles. After averaging the three profiles, the slope of 550 nm to 680 nm of the extrapolated albedo appears flatter than the expected values described by Louchard

PAGE 47

39et al. (2003) for a pure, thick, clean seagrass end member scene. Figure 20 also compares the extrapolated albedo with the inversion albedos for each profile. Each inversion albedo has a much larger 550-680 nm slope. The inversion albedo in Figure 21 is from the third profile to coincide with the video data time frames. These values do are slightly lower than other albedos collected from the same area due to vehicle orientation and the collection of data on the shaded side of seagrass BRDF. Although the video data in Figure 21 show a 550 nm value between that of the extrapolation and inversion albedo, the other channels coincide very well with the inversion albedo. The low value at 550 nm is attributable to wave focusing causing a hot spot on the standard reflector, which increases the apparent Ed. Here, the hyperspectral data are believed to be the most accurate as the data at all channels are taken simultaneously, whereas the video data are collected on a frame-by-frame basis as the filter wheel rotates. The wave focusing causes hot spots on a random basis hitting which can hit the reflector on one filtered frame, but not adjacent filter frames. Clearly, with both instruments, wave focusing does not permit consistent albedos to be derived without a significant amount of temporal averaging of many cycles of focusing and defocusing. In Figure 22a, at 1.82m vehicle altitude, the seagrass histograms have the same characteristic shape as the seagrass from Rainbow Gardens (Fig. 14), but have much lower albedo values. The darker, secondary peaks occur due to darker pigmentation, which predominate ROIs of self-shading, a nd entrapment of forward-reflected light, which occurs within the thick bed of seagrass. The brighter pixels are representative of the photobleached and/or sunglint off blades of grass, which were oriented normal to the downwelling irradiance (Fig. 3). These blades directly reflected light back towards the

PAGE 48

40sensor. As the vehicle ascended in the water-column, to 2.64 m altitude, the albedo histograms again lost their skewed shape (Fig. 22b). There is an increase in the 445 nm derived albedo occurs due to wave focusing on the grass and not the reflector, whereas there is a decrease in 490 nm albedo due to direct wave focusing on the reflector.

PAGE 49

41Figure 17. Comparison of the total absorption coefficients derived at Adderly Cut and Rainbow Gardens

PAGE 50

42Figure 18. Comparison of the total backscatter coefficients derived at Adderly Cut and Rainbow Gardens

PAGE 51

43Figure 19. Comparison of the three vertical rrs profiles at Adderly Cut

PAGE 52

44Figure 20. Comparison of in-water derived al bedos from the inversion method versus the extrapolation method at Adderly Cut

PAGE 53

45Figure 21. Comparison of albedos derived from the hyperspectral inversion method for profile three versus the multispectral Xybion video inversion method at Adderly Cut

PAGE 54

46Figure 22. Albedo histograms from seagrass at Adderly Cut a) low altitude (1.82m) and b) high altitude (2.64m) a) b)

PAGE 55

47 3.3 Horseshoe Reef The final end member in this study was from a clean, bright, sand area located near Horseshoe Reef. This flat area is considered to be a uniform Lambertian (diffuse) reflector. This area is located adjacent to the deep, open, and clear waters of Exuma Sound. Therefore, the lower absorption values in the blue region from the inversionderived IOPs were expected (Fig. 23), and they approach those of Exuma Sound (see Ivey et al., 2002). No significant deviations in total backscatter coefficients from those at Rainbow Gardens were observed (Fig. 24). These data were also collected on a bright sunny day. However, there were no influences from wave focusing due to the increased depth at which the albedo derivations were made. The same vehicle altitude error appears to be present with extrapolation albedo curves when compared with the inversion albedo (Fig. 25). Specifically, the extrapolated albedo did not have the same spectral shape observed by Louchard et al. (2003), Werdell and Roesler (2003), and Zimmerman (2003) with other oolitic carbonate sands in the area around Lee Stocking Island area. Both video and hyperspectral inversion albedos share this common characteristic (Fig. 26). With the exception of the fifth channel (685 nm), the video inversion albedo matches fairly well with the hyperspectral inversion albedo. This may be due to a much lower sensitivity in the photocathode response of the camera’s image intensifier. This response is about half of that of the other three channels from which albedos were derived (data not shown). However, the data were incorporated in this study, because the

PAGE 56

48low-altitude (1.05 m) albedo histogram at that channel shows multiple peaks within what was a homogeneous surface. These peaks within a scene characterized as Lambertian reiterates this methodology’s usefullness in studies trying to deconvolve spectral signatures from multiple reflecting components (Fig. 27a). From the video scenes, these particular peaks, also present at 575 nm and 620 nm, appear to be from the illuminated and non-illuminated sides of small sand waves. Carder et al. (2003), observed this same contrast in larger sand wave features in the Adderly Cut region of LSI. As the vehicle increased to 10.61 m altitude, histogram peaks from the small heterogeneous fragments of the bottom reflectance are lost (Fig. 27b) and the histograms take on a more Gaussian shape. This total bottom reflectance signal represents a more uniform Lambertian reflector as initially characterized. Note light traveling almost 11 m down to the bottom and 11 m back up to the sensor has traveled 22 m. Red light, at 685 nm with an absorption coefficient of about 0.5, would have a total absorption of surface light of approximately b e-0.5 22 or 0.3 e-11. The values at 620 nm would be attenuated by only 0.5 e-0.3 22 or 0.5 e-6.6. The Xybion sensor apparently is unable to measure the very low bottom signals at 685 nm from 10.6 m altitude, while it can measure those at 620 nm. This demonstrates the utility of deriving IOPs for attenuation coefficients, i.e., Kds (Figs. 23 and 24), and bottom albedos (Figure 26) for examining the performance limits of the sensors involved.

PAGE 57

49Figure 23. Comparison of inversion derived total absorption coefficients at all three study locations

PAGE 58

50Figure 24. Comparison of inversion derived total backscatter coefficients at all three study locations

PAGE 59

51Figure 25. Comparison of albedos derive d from the hyperspectral inversion method versus the extrapolation method at Horseshoe Reef

PAGE 60

52Figure 26. Comparison of albedos derive d from the hyperspectral inversion method versus the multispectral video inversion method at Horseshoe Reef

PAGE 61

53Figure 27. Albedo histrograms from a sand ROI at Horseshoe Reef a) low altitude (1.05m) and b) high altitude (10.61m) a) b)

PAGE 62

54 Chapter 4 BRDF and Path-Adjacency Effects Histogram data presented earlier from the Rainbow Gardens, Adderly Cut, and Horseshoe Reef regions all have shown evid ence of path-adjacency effects. Because most radiative transfer models assume uniform reflectance, path-adjacency occurs when there is a deviation of the bidirectional re flectance distribution function, BRDF, from a Lambertian, homogenously reflective surface (Mobley and Sundeman, 2003). NonLambertian reflectance can occur from morphologic complexities existing within bottom regimes such as end member variability or heterogeneity, presence of biomatter, and bottom slope and illumination characteristics (Carder et al., 2003). In the Lee et al. (1999) model, the path elongation factors, (Eq. 4), were based on Hydrolight models of water with varying water properties and homogenous bottom types. It did not account for the heterogeneity that was present with the multispectral video data from the sands and seagrasses studied here. In this particul ar study, the non-Lambertian BRDF was observed at the hyperspatial scale and can be used to deconvolve multiple facets within end member histograms. In the case of all three locations, low altitude albedo histograms showed specific features, which were accountable to varying bottom components within pure end member imagery. However, when the vehicle ascended from the bottom, the larger field-of-view angles began to include forward-scattered light from the adjacent facets within the scene. In Figures 14c and 16c, the multimodal features in the sandy-seagrass mix at Rainbow

PAGE 63

55Gardens are lost when the vehicle is at the higher altitude. At both Rainbow Gardens and Adderly Cut, (Figures 14b and 16b; 22a and 22b respectively) the brighter pixel skewing in the seagrass end-member histograms are transformed into a more typical bell or Gaussian shape. Likewise, Rainbow Garden s and Horseshoe Reef (Figures 14a and 16a; 27a and 27b respectively) both lose the darker pixel skewing in sand histograms. The end member frequencies within each channel’s albedo range became more unified. The effects of path-adjacency deform any multimodal or skewed end member histograms to Gaussian histograms. This can make any spectral deconvolution at larger scales more difficult, if not impossible, without properly accounting for BRDF. In a study mapping the albedos of patch coral on a sandy bottom using an autonomous underwater vehicle (AUV), Hou et al. (2001), measured albedos which were enhanced by as much as 50% above expected values. This significant increase was attributable to path-adjacency effects. The effects of BRDF and path-adjacency on the video-derived albedos were further investig ated with Rainbow Garden’s data. By isolating the individual patches of sand and seagrass at multiple altitudes during vehicle ascent, the effects of a brighter object next to a darker object at 550 nm are shown (Figure 28). Although the same patches of sand and seagrass were selected and the pixel area was reduced to compensate for the decreased angular field of view for the selected components, effects from adjacent, high-contrast bottom components became apparent. In this vertical profile, the albedo of the sand ROI’s decreased with altitude, while the albedo of the seagrass remained relatively constant. These changes in the apparent bottom albedos are believed to be only partially due to path-adjacency effects; i.e., light that is reflected from adjacent bottom components scattered into the sensor from larger,

PAGE 64

56oblique angles. If the changes in albedo were only due to path-adjacency, the seagrass albedo would be expected to increase with altit ude. Here, there appears to be a secondary BRDF factor affecting the derived albedos. Previous work by Mobley et al. (2003) and Zimmerman (2003) suggests that the additional non-Lambertian BRDF factor was due to the vertical structure of the seagrass beds. Though the seagrass at Rainbow Gardens appears to be fairly flat, the video data suggests the presence of a canopy effect. The canopy effectively minimizes the impact of the sand ROIs on the upwelling light signal. At larger, oblique angles the vertical structure of the seagrass reduces the visibility of the brighter sand beneath the canopy. At lower altitudes the impact is less, because the pixels representing sand can be isolated easier. As the vehicle rises, the area of sand that can be sampled is reduced in size, and becomes blocked by blades of grass at larger angles within the field of view. An illustration of how the field of view can be impacted by this canopy effect with an increase in sensor altitude is shown in Figure 29. In areas of taller seagrass meadows, currents can also be nd the grass blades to obscure more of the bottom (e.g., Fig.3) The convergence of sand and seagrass albedos with an increase in altitudes indicates that, for a larger image footprint, the bottom reflects as a more homogenous bottom, and the apparent albedo of bottom end members is affected. If one were to apply what was learned from these data to aircraft or other high altitude imagery, it would suggest that retrievals could be in error. Fortunately, work by Mobley et al. (2003) shows that these effects are often within acceptable tolerances for narrow-viewing sensors such as PHILLS. With larger angle sensors, such as SeaWifs, the non-Lambertian effects become more pronounced.

PAGE 65

57Figure 28. Comparison of 550 nm albedos at different altitudes from Rainbow Gardens.

PAGE 66

58Figure 29. Illustration showing how ver tical structure can create a canopy effect, especially at higher vehicle altitudes.

PAGE 67

59 Chapter 5 Summary and Conclusions Ocean color data can yield information about the water column as well as the ocean floor and it constituents. In this study, inversion techniques were applied to hyperspectral remote sensing reflectance data to derived inherent optical properties and extract bottom albedo. The high resolution, hyperspectral data were then applied and compared to hyperspatial, multispectral video imagery. From these data, retrieval and analysis of sand and seagrass end member albedos were performed at the hyperspatial scale. Low altitude video clearly shows heterogeneous contributions to albedo histograms from within individual end members. The variations at specific wavelengths are from end member components deviati ng from Lambertian BDRFs. At higher altitudes the histograms become more homogenous as path-adjacency effects merge the total upwelling signatures. This suggests that higher altitude sensors can largely assume Lambertian reflectance over a uniform bottom regime. Deriving bottom albedo at a hyperspatial scale gives insight as to how a sandyseagrass bottom type behaves spectrally with a mix of end member components. However, in mixed end member scenes, the apparent albedo of individual end members can be influenced by their adjacent counterpa rts. Light reflected off brighter bottom constituents could be scattered into a sensor’s field of view of darker, adjacent substrates. Evidence of scattered light in the histograms of a vertical profile demonstrates the

PAGE 68

60existence of path-adjacency and possible canopy effects. The errors that can induced due to these effects is further enhance when data is collected at during times of a low solar zenith angle due to increased shade and shadows with the imagery scenes. This suggests that, in multi-end member scenes, albedo retrievals will be in error if not corrected for path-adjacency, shading, and canopy effects. To accurately map bottom components on larger footprint aircraft pixels, more studies are needed regarding path-adjacency and bidirectional reflectance effects on upwelling radiance. Further experiments could be performed to analyze additional footprint coverages with a broader range of end-member fractions. Coupling hyperspatial data with the hyperspectral data will allow for the deconvolution of the total albedo signature. Spectral-unmixing algorithms to derive component fractions could be tested by comparing large-field hyperspectral radiometry of the bottom coincident with hyperspatially calculated end-member fractions. This could then provide estimates of the accuracy of end-member fractions derived by spectral unmixing methods for the spatially larger, aircraft pixels after atmospheric correction.

PAGE 69

61 References Boss, E., and J. R. V. Zaneveld, 2003. The effect of bottom substrate on inherent optical properties: Evidence of biogeochemical processes. Limno. Oceanogr. 48: 346-354. Carder, K. L., and R. G. Steward, 1985. A remote-sensing reflectance model of a red tide dinoflagellate of West Florida. Limno. Oceanogr. 30: 286-298. Carder, K. L., C. Liu, Z. Lee, D. C. English, J. Patten, R. F. Chen, J. E. Ivey, and C. O. Davis, 2003. Illumination and turbidity effects on observing faceted bottom elements with uniform Lambertian albedos. Limno. Oceanogr. 48: 355-363. Carder, K. L., D. K. Costello, L. C. Langebrake, W. Hou, J. T. Patten, and E. A. Kaltenbacher, 2001. Real-time AUV data for command, control, and model inputs. IEEE J. Ocean. Eng. 26: 742-751 Carder, K.L., S. K. Hawes, K. A. Baker, R. C. Smith, R. G. Steward, and B. G. Mitchell, 1991. Reflectance model for quantifying chlorophyll-a in the presence of productivity degradation products. J. Geophys. Res. 96: 20,599-20,611. Cattrall, C., K. L. Carder, K. J. Thome, and H. R. Gordon, 2002. Solar-reflectance-based calibration of spectral radiometers. Geophys. Res. Letter. 29: 20 Costello, D.K., and K. L. Carder, 1994. New platforms for subsurface optical measurements. SPIE Ocean Optics XII vol. 2258. Costello, D. K., K.L. Carder, J. Ivey, 2002. Measurement and Interpretation of Diffuse Attenuation and Reflectance in Clear, Deep-Water Environments: the Effects of Transspectral Phenomena. SPIE Ocean Optics XVI, Proc. Costello, D. K., K. L. Carder, R. F. Chen, T. G. Peacock, and N. S. Nettles, 1995. Multispectral imagery, hyperspectral radiometry, and unmanned underwater vehicles: Tools for the assessment of natural resources in coastal waters. SPIE Visual Communications and Image Processing 2501: 407-415. Costello, D.K., K. L. Carder, W. Hou, T.G. Peacock, and J. Ivey, 1998. Hyperspectral measurements of upwelling radiance during CoBOP: The role of bottom albedo and solar stimulated flourescence. SPIE Proceedings Ocean Optics XIV. Davis, C.O., and others, 2002. Ocean PHILLS hyperspectral imager: Design, characterization, and calibration. Opt. Express 10: 210-221.

PAGE 70

62 Drake, L. A., F. C. Dobbs, and R. C. Zimmerman, 2003. Effects of epiphytic load on optical properties and photosynthetic potential of the seagrasses Thalassia testudinum Banks ex Konig and Zostera marina L. Limno. Oceanogr. 48:456-463. Hemminga, M., and C. Duarte, 2000. Seagrass ecology. Cambridge Univ. Press. Hou, W., K.L. Carder, D.E. English, and D.K. Costello, 2001. Large-scale bottom classification using multi-channel imagery from an autonomous underwater vehicle. Submitted, Limnol. Oceanogr. Ivey, J. E., K. L. Carder, F. R. Chen, and Z. P. Lee, 2002. Absorption measurements in optically clear waters. SPIE Proceedings Ocean Optics XIV Lee, Z.P., K. L. Carder, C.D. Mobley, R.G. Steward, and J.S. Patch, 1999. Hyperspectral remote sensing for shallow waters: 2. deriving bottom depths and water properties by optimization. Appl. Opt. 38: 3831-3843. Lee, Z.P., K. L. Carder, C.D. Mobley, R.G. Steward, and J.S. Patch, 1998. hyperspectral remote sensing for shallow waters: 1. A semi-analytical model. Appl. Opt. 37: 63296338. Lee, Z.P., K. L. Carder, R. F. Chen, and T. G. Peacock, 2001. Properties of the water column and bottom derived from airborne visible infrared imaging spectrometer (AVIRIS) data. J. Geophys. Res. 106: 11,639-11,651. Lee, Z.P., K. L. Carder, R. G. Steward, T. G. Peacock, C. O. Davis, and J. L. Mueller, 1996. Estimating primary production at depth fr om remote sensing. Appl. Opt. 35: 463474. Lee, Z. P., K. L. Carder, S. K. Hawes, R. G. Steward, T. G. Peacock, C. O. Davis, 1994. Model for the interpretation of hyperspectral remote-sensing reflectance. Applied Optics 37(27). Louchard, E. M., R. P. Reid, and R. A. Maffione, 2003, Effects of microalgal communities on reflectance spectra of carbonate sediments in subtidal optically shallow marine environments. Limno. Oceanogr. 48:511-521. Mobley, C. D., 1994. Light and water: Radiative transfer in natural waters. (Academic, New York). Mobley, C. D., H. Zhang, and K. J. Voss, 2003. Effects of optically shallow bottoms on upwelling radiances: Bidirectional reflectance distribution effects. Limnol. Oceanogr. 48: 329-336.

PAGE 71

63Mobley, Curtis D., and Lydia K. Sundman, 2003. Effects of Optically shallow bottoms on upwelling radiances: Inhomogenous and sloping bottoms. Limnol. Oceanogr. 48: 323-328. Morel, A., 1974. Optical properties of pure water and pure sea water, in Optical Aspects of Oceanography. N. G. Jerlov and E.S. Nielsen, eds. (Academic, New York), pp. 1-24. Morel, A., and L. Prieur, 1977. Analysis of variations in ocean color. Limnol. Oceangr. 22: 709-722. Otis, D. B., K. L. Carder, D. C. English, and J. E. Ivey, 2004. CDOM transport from the Bahamas Banks. Coral Reefs 23: 152-160. Pope, R. and E. Fry, 1997. Absorption spectrum (380-700 nm) of pure waters: II. integrating cavity measurements. Appl. Opt. 36: 8710-8723. Pratt, P., 1997. Algorithms for path radiance and attenuation to provide color corrections for underwater imagery, characterize optical properties and determine bottom albedo. SPIE 2963:753-758. Reinersman, P. N. and K. L. Carder, 1995. Monte Carlo Simulation of the Atmospheric Point-Spread Function with an Application to Correction for the Adjacency Effect. Appl. Opt. 34: 4453-4471. Vane, G., R. O. Green, T. G. Chrien, H. T. Enmark, E. G. Hansen, and W. M. Porter, 1993. The airborne visible / infrared imag ing spectrometer (AVIRIS). Remote Sens. Environ. 44:127-143. Weidemann, A. D.; Stavn, R. H.; Zaneveld, J. R. V.; Wilcox, M. R. 1995. Error in predicting hydrosol backscattering from remotely sensed reflectance J. Geophys. Res. Vol. 100, No. C7, p. 13,163-13,178. Werdell, P. J., and C. S. Roesler, 2003. Remote assessment of benthic substrate composition in shallow waters using multispectral reflectance. Limno. Oceanogr. 48: 557-567. Zaneveld, J. R. V., E. Boss, and A. Barnard, 2001. Influence of surface waves on measured and modeled irradiance profiles. Appl. Opt. 40: 1442-1449. Zimmerman, Richard C., 2003. A bioptical model of irradiance distribution and photosynthesis in seagrass canopies. Limno. Oceanogr. 48:568-585.


xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001681064
003 fts
005 20060215071204.0
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 051222s2005 flu sbm s000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0001054
035
(OCoLC)62735150
SFE0001054
040
FHM
c FHM
049
FHMM
090
GC11.2 (Online)
1 100
Farmer, Andrew Scott.
0 245
Bottom albedo derivations using hyperspectral spectrometry and multispectral video
h [electronic resource] /
by Andrew Scott Farmer.
260
[Tampa, Fla.] :
b University of South Florida,
2005.
502
Thesis (M.S.)--University of South Florida, 2005.
504
Includes bibliographical references.
516
Text (Electronic thesis) in PDF format.
538
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
500
Title from PDF of title page.
Document formatted into pages; contains 71 pages.
520
ABSTRACT: Remote sensing reflectance data collected with a remotely operated vehicle (ROV) were used to derive bottom albedo and optical properties for a shallow marine environment near Lee Stocking Island, Bahamas. Optical model inversion techniques were applied to hyperspectral measurements of remote-sensing reflectance to derive water absorption and backscatter coefficients. Using these derived water properties, path attenuation and radiance effects were removed from bottom observations to derive bottom albedos. Histograms from multispectral, hyperspatial video images were used to determine the albedo range of optical end members observed in scenes of sand and seagrass. Variations of spectral signatures for optical end members caused by path-adjacency effects are shown to influence the reflectance measurements.Low-altitude albedo histograms for heterogeneous scenes demonstrate higher contrast between sand and seagrass than is observed at higher altitudes, even after correction for path radiance and attenuation effects. For example, reflected light from bright sand scatters into the field of view of dark seagrass, while less light scatters out from the seagrass into the field of view of sand. This decreases the apparent sand albedo, and increases that for seagrass when viewed from higher altitudes, including aircraft. Evidence provided suggests that simple bottom classifications based upon expected albedo values for scene end members are in error unless the water depth is very shallow.
590
Adviser: Kendall L. Carder, Ph.D.
653
Ocean.
Optics.
Albedo.
Sand.
Seagrass.
690
Dissertations, Academic
z USF
x Marine Science
Masters.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.1054