HABs Identified and Quantified Using Hyperspectral Imagery Authors: Chelpon S.M.¹, Becker, R.², Cline M.² 1

CUNY Hunter College Department of Geography NSF REU Program, University of Toledo 2 University of Toledo Department of Environmental Sciences

0.03

R² = 0.7683

20

10

0 0

10

20

Wynne Validation: Ground Reference Data Derived CI

30

Figure 1. Simis’ method shows good correlation (R²=0.7683) when used with ground reference data

R² = 0.7784

0.02 0.01 0 0

-0.01

Measured [Blue-Green Algae] (mcg/L)

50.00* radiance (W/m^2/micrometer/sr)

Simis Validation: Ground Reference Data Derived [PC]

30

Methods

5

10

15

20

25

30

Uncorrected Spectral Profile: Site 8M 3000 2000

1000

0 400

Measured [Blue-Green Algae] (mcg/L)

500

600

700

800

900

Wavelength (nm)

Figure 2. Wynne’s method shows good correlation (R²=0.7784) when used with ground reference data

Atmospherically Corrected Spectral Profile: Site 8M

0.15

Reflectance

• Ground and on water spectral reference data collected using ASD Spectroradiometer • Ground reference data used to atmospherically correct airborne HSI data • Corrected HSI used to correct HICO satellite imagery • Algorithms by Lee, Randolph, and Wynne were used to estimate phycocyanin concentrations [PC] and cyanobacterial activity (CI) • Algorithms utilize spectral signatures of chl-a and PC for identification of HABs • Statistical analysis used to determine validity of various methods in comparison to the site measured concentrations of blue-green algae

Atmospheric Correction Cyanobacterial Index

• Lake Erie’s Western Basin faces water quality concerns due to Harmful Algal Blooms (HABs) • Remote sensing can be cheaper and more efficient than in situ methods of HAB identification • Analysis of different types of Hyperspectral Imagery (HSI) has allowed us to identify and quantify HABs in Lake Erie’s Western Basin

Results Derived [PC] (mcg/L)

Introduction & Objectives

0.1

0.05

0 400 -0.05

500

600

700

800

900

Wavelength (nm)

Conclusion •

Study Area • • Wynne Validation: Remotely Sensed CI

Simis Validation: Remotely Sensed [PC] HICO

0.03

HSI

R² = 0.4956 20

10

0

-10

0

10

20

30

Measured [Blue-Green Algae] (mcg/L)

Figure 2. Simis’ method shows a moderate correlation (R²=0.4956) when used with remotely sensed imagery

Cyanobacterial Index

Derived [PC] (mcg/L)

30

Simis and Randolph’s method is only accurate at the ground level and is adversely affected by errors in the atmospheric correction technique Wynne’s method provides consistent accuracy at all levels of data and is unaffected by errors related to atmospheric correction Wynne’s method proves to be more useful in the detection and quantification of HAB levels overall

HICO

HSI

Acknowledgements

R² = 0.7794

0.02

0.01

0 0 -0.01

5

10

15

20

25

30

Measured [Blue-Green Algae] (mcg/L)

Figure 4. Wynne’s method maintains a good correlation (R²=0.7794) when used with remotely sensed imagery

This project was aided by National Science Foundation Grant #NSF DBI-1461124 to the University of Toledo's Lake Erie Center, “Undergraduate Research and Mentoring- Using the Lake Erie Sensor Network to Study Land-Lake Ecological Linkages”. We thank the principal investigators: Dr. Carol Stepien and Dr. Kevin Czajkowski, the program manager Rachel Lohner, and teaching assistants Lucas Groat and John Dilworth for help and logistic support. HSI data provided by NASA GRC