Presentation is loading. Please wait.

Presentation is loading. Please wait.

J. Cho Department of Integrated Environmental Science Bethune-Cookman University Daytona Beach, FL.

Similar presentations


Presentation on theme: "J. Cho Department of Integrated Environmental Science Bethune-Cookman University Daytona Beach, FL."— Presentation transcript:

1 J. Cho Department of Integrated Environmental Science Bethune-Cookman University Daytona Beach, FL

2 Elements of Remote Sensing data acquisition and analysis sun Camera system

3 Electromagnetic (EM) Radiation Consists of An electrical field (E) A magnetic field (M) Both fields travel at the speed of light (c) c = vλ c: 3 x 10 8 m/sec v: frequency λ: wavelength http://www.astronomynotes.com/light/s3.htm

4 EM Spectrum http://en.wikipedia.org/wiki/File:EM_spectrum.svg

5 (Jensen, 2007)

6

7 http://ian.umces.edu/learn/modulepopup/barrier_islands_and_sea_level_rise/a_closer_look_at_seagrasses

8 535 nm760 nm

9 Spectral Reflectance Characteristics of Vegetation and Conventional Vegetation Indices Image source: http://extnasa.usu.edu/on_target/images_independent/nir_vegetation_graph.gif

10 Spectral reflectance of Submerged Aquatic Vegetation (SAV) at Varying Depths

11 I. Water Correction Algorithm Development

12 Rationale and Objective Remote detection of benthic features (i.e. seagrass) has been limited because of numerous factors including the influence of the water column To develop algorithms that reduce water effects to improve remote detection and classification of shallow underwater features and seagrasses

13 Hyperspectral images corrected for conventional atmospheric distortions Pure water effect removal Water color, turbidity effects reduction Depth effect adjustment Images with enhanced benthic features Selection of critical bands and data Reduction & compression Controlled experiments Algorithm development

14 Water Correction Algorithm Development for Benthic Mapping Water LrLsLv Lb La L L = Ls + Lv + La + Lr + Lb

15 * water absorption can be derived from the following way 1 x (1-A w /2) x R r x (1-A w /2) + R w = R m A w and R w are the functions of water depth * For any bottom panel, R r can be directly measured Algorithm to model water effects

16 53cm Pulley System String 1.2 m 82 cm (Washington et al. 2012)

17

18

19

20 Reflectance (corrected) = f (ref (measured), depth) 1 x (1-A w /2) x R r x (1-A w /2) + R w = R m Water Volumetric Reflectance Water Absorption (Cho et al. 2010))

21 Data Extension into Deeper Depths Beer-Lambert Law where I z, λ is the light intensity at a given depth z, I 0 is the light intensity present before any contact is made with the absorbing medium, and K d is the downwelling attenuation coefficient. (Washington et al. 2012)

22 SAV (Seagrass) Pixels (24 pixels) (Cho et al. 2011)

23 (Gaye et al. 2011)

24 Graphical User Interface (Cho et al. 2013)

25

26

27

28 Airborne AISA Hyperspectral data Acquisition and Analyses Mission-Aransas NERR, TX: July 2008 Application of the Algorithm on image data (Cho et al.)

29 553 nm (Green Color Energy) (Cho et al.)

30 694 nm (Red Energy) (Cho et al.)

31 741 nm (NIR) (Cho et al.)

32 Applications of the Technology John Wood, Ph.D. candidate, Harte Research Institute Fellow in the Coastal and Marine Systems Sciences Program at Texas A&M University-Corpus Christi The dissertation title: Geospatial Analysis of Seagrass Remote Sensing Data From Redfish Bay, Texas..

33 Classification Results Classification Results Depth Extrapolation out to 5 m. 535, 600, 620, 638, 656 nm ClassProducer’s Accuracy User’s Accuracy Bare66.7%69.6% Halodule29.0%39.1% Thalassia46.2%40.0% Ruppia0% Mixed23.3%22.6% Overall Accuracy 38% ClassProducer’s Accuracy User’s Accuracy Bare73%66% Halodule64% Thalassia46%55% Ruppia0% Mixed47%39% Overall Accuracy 62%

34 ClassProducer’s Accuracy User’s Accuracy Bare50%58% Halodule50%6% Thalassia14%33% Ruppia10%4% Mixed27%62% Overall Accuracy 27% ClassProducer’ s Accuracy User’s Accuracy Bare81%71% Halodule37%43% Thalassia40%33% Ruppia00% Mixed34%44% Overall Accuracy 45% Classification Results Depth Extrapolation out to 0.6 m. 554, 695, 723, 742, 809 nm

35 Problems Multi-spectral data and the current chlorophyll algorithms cannot distinguish seagrass from algal signals. Airborne hyperspectral data are costly and have temporal/spatial limitations.

36 Seagrass Vascular plants Generally have higher Chl concentrations compared to macroalgae

37 Macroalgae Non vascular Varying levels of Chlorophyll and colors

38 Benthic Remote Sensing Hyperspectral remote sensing has been suggested to be an effective tool in distinguishing spectral patterns of benthic habitats.(Fyfe 2003; Kutser et al. 2005)

39 Goal and Objectives Goal: To develop a novel approach using satellite data that can be efficiently used to distinguish seagrass and macroalgae signals and help facilitate accurate benthic vegetation mapping Objectives: Find spectral characteristics that can distinguish seagrass signals from those of macroalgae. Map seagrass and macroalgae in Indian River Lagoon using satellite data.

40 HICO The Hyperspectral Imager for the Coastal Ocean (HICO) is a hyperspectral sensor onboard the International Space Station (ISS). HICO has a high signal-to-noise ratio that can facilitate benthic habitat mapping.

41 Study Site (Indian River Lagoon)

42 Methods Obtaining and pre-processing HICO data over the Indian River Lagoon (March 2013). Developing spectral models. Benthic classification using four methods supervised – Spectral Angle Mapper (SAM), unsupervised, and two new models – Slope RED, and Slope NIR ) Performance comparison of the four methods using high resolution aerial photos and field survey data.

43 Cho et al. 2014

44 Results Cho et al. 2014

45 Results Cho et al. 2014

46 Results Cho et al. 2014

47 Results Cho et al. 2014

48 Accuracy Assessment Overall Accuracy (%)Kappa Slope RED 64.210.5264 Slope NIR 63.160.5171 Supervised- (SAM) 47.500.3466 Unsupervised25.000.0722 Cho et al. 2014

49 Accuracy Assessment ClassProducer Accuracy (%) User Accuracy (%) Slope RED Seagrass5268.42 Macroalgae9052.04 Slope NIR Seagrass10042.55 Macroalgae7285.71 Supervised – (SAM) Seagrass2555.56 Macroalgae1633.33 Cho et al. 2014

50 Conclusion The study demonstrates that the advantage of selecting key narrow bands to accentuate the subtle differences between seagrass and macroalgae, which conventional classification methods do not perform well. Combining the slope methods, Slope RED and Slope NIR, with a supervised classification method will lead to higher accuracies in distinguishing key vegitation.

51 Acknowledgments National Geospatial-Intelligence Agency (NGA). U.S. Naval Research Laboratory (NRL). The National Aeronautics and Space Administration (NASA). Oregon State University (OSU). Florida Space Grant Consortium. St. Johns River Water Management District (SJRWMD).

52

53 IRL Initial Water Correction Results

54 OriginalWater-Corrected Slope RED Slope NIR Slope RED Slope NIR 1.658-1.6153.234-2.457

55 IRL Initial Water Correction Results

56 OriginalWater-Corrected Slope RED Slope NIR Slope RED Slope NIR 0.873-0.6771.725-0.842


Download ppt "J. Cho Department of Integrated Environmental Science Bethune-Cookman University Daytona Beach, FL."

Similar presentations


Ads by Google