You are on page 1of 29

Mapping Chlorophyll

Using Landsat ETM+


Data
Luoheng Han
Department of Geography
University of Alabama
EaGLe

Introduction

The hypothesis (question) being


investigated
Can the remote sensing image acquired
from Landsat TM be used to map
chlorophyll concentration in the Gulf of
Mexico?
It addresses the objective of CEER-GOM
for developing ecosystem/watershed
indicators

The main objective of CEER-GOM

To study, develop and validate indicators


of nutrient and dissolved oxygen (DO)
impacts on estuarine condition at four
levels of increasing biological complexity
(Individual, Population, Community and
Ecosystem/Watershed), and to integrate
the suite of indicator responses through
models that can be used to describe the
health of an estuarine ecosystem.

Why measure chlorophyll


concentration?

Chlorophyll is important for


photosynthesis
Direct indicator of phytoplankton
growth, primary productivity, and
community
Indirect indicator of potential hypoxia
condition
Optically active
Remotely measurable

Interpret through chl information to infer phytoplankton and DO levels.

Principle of Remote Sensing

Pensacola and Escambia bay. 30 x 30 m pixels. Use the brightest band. Much
coarser scale than what Larry Harding is using in ACE INC.

Why use Landsat TM/ETM+ data?

Routinely available and economic


Adequate spatial coverage and
spatial resolution for monitoring
estuaries
Archived data available
Challenges
Inadequate spectral resolution for
measuring some of the WQ parameters,
such as Chlorophyll?

Available every 6 days, and is very cheap to obtain. But spectral resolution is low.

Are the Landsat TM/ETM+ data


spectrally adequate?

Landsat TM Bands (6)


Spectral Reflectance
4

Reflectance (%)

Pensacola, FL

0
400

500

600

700

800

900

Wavelength (nm)

ASD Bands (701)

Example from one of the sampling sites. On the top curve, sunlight has washed
everything out to make it very flat. But near infrared is better.

Pensacola Bay System

Study area: freshwater dominated. 144 sq miles for bays.

Watershed of the bay

May 20, 2002 (ID: 020039000214050)

10

Subset

11

GED sampling Sites

23 water quality
variables are
measured monthly
Samples are collected
from surface and
bottom waters

Working with the EPA GED lab.

12

Chlorophyll-a concentrations
25.00

MEAN = 7.41 g l-1


STDEV = 5.89 g l-1

Chl-a (ug/l)

20.00

15.00

Mouth of Blackwater
River

Mouth of Yellow
River

10.00

5.00

0.00
1

Mouth of Escambia
River

10

11

12

13

14

15

16

Sampling Sites

May 2002 chl a measurements.

13

Our approach
Pre-processing
Masking out land area
Spectral reflectance

Modeling

Water sampling

Mapping chlorophyll

14

Geometric correction
Geometric model: Polynomial
Method: Ground Control Points
Projection: UTM
Datum: NAD 83
Zone 16
Unit: meter
Resampling method: nearest neighbor

Processing of the remote sensing data.

15

Atmospheric correction
Chavez (1996):
= * d2 * (Lsat
(Lsat Lhaze
Lhaze) /(TAUv * ESUN
ESUN *
cos2*TAUz)
Where:

D: distance between Earth and Sun (unit: AU)


Lsat
Lsat: apparent atat-satellite radiance
Lhaze
Lhaze: path radiance
ESUN
ESUN: ExoExo-atmospheric solar irradiance
TAUv: atmospheric transmittance along the path from
the ground surface to the sensor
TAUz: atmospheric transmittance along the path from
the sun to the ground surface
: Sun zenith angle

And atmospheric correction is needed. This is quite tedious. Converts all digital
numbers into reflectance.

16

Landsat ETM+ Bands


Band

Color

Resolution

Blue

.45 toto- .515

30m

Green

.525 to .605

30m

Red

.63 to .690

30m

NIR

.75 to .90

30m

SWIR

1.55 to 1.75

30m

TIR

10.40 to 12.5

60m

SWIR

2.09 to 2.35

30m

.52 to .90

15m

PAN

Use all bands except band 6.

17

Principles of Band Ratios

Select two spectral bands. Often for chlorophylla, one band corresponds to high absorption and
one band corresponds to low absorption
(Gitelson et al., 1996; Gin et al., 2002)
Bands most associated with chlorophyllchlorophyll-a:

Blue
Green
Red
Near Infrared

Using band ratio as the primary way to model the chl a.

18

R2 with different regression models


Regression
Linear

TM1/TM2 TM1/TM3 TM1/TM4 TM2/TM3 TM2/TM4 TM3/TM4

0.33

0.42

0.29

0.3

0.03

0.1

0.4

0.47

0.55

0.3

0.07

0.25

Cubic

0.49

0.5

0.61

0.31

0.11

0.35

Exponential

0.41

0.29

0.03

Quadratic

19

T M Bands vs. spectral reflectance


20

TM1

18

TM2

TM3

TM4
P1
P2

16

P3
P4

Reflectance (%)

14

P5
P6

12

P7
P8

10

P9
P10

8
6

P11
P12

P13
P14

P15
P16

0
400

500

600

700

800

900

Wavelength (nm)

To precisely measure chl a, you need to measure the 680 wavelength, but LandSat
averages it in a wide band.

20

Masking out land area

21

Modeling
Linear regression model of predicting chl-a
1.6
1.4

y = -10.476x + 3.7331
2

R = 0.6201

L o g 10 (C h l-a )

1.2
1
0.8
0.6
0.4
0.2
0
0.24

0.26

0.28

0.3

0.32

0.34

Log 10 (T1/T3+T2+T3+T4+T5+T6)

This is our best model using all the band combinations.

22

Chl map created using the band ratio of band 3 to band 1. Model predicted values a
little lower than the ground truthing showed from GED samplings, but it did indicate
areas where hypoxia developed.

23

Model to predict chl, this one was a little high.

24

Conclusions

Remote sensing provides an effective


alternative means to traditional water
sampling for monitoring estuaries
The unique characteristic of remote
sensing data is its ability to provide WQ
information for an area, not just a point
E.g., From 30 by 30 meter pixel to the entire
Pensacola Bay system

The spatial resolution of Landsat TM data


is very suitable for the size of estuaries
similar to Pensacola Bay system

25

Conclusions, contd

Although the band ratio/combination


algorithms seem to be effective,
estimating chlorophyll concentration
using Landsat TM is still a challenge
due to its limited spectral resolution.

26

What is next?

Further tests on the effectiveness of band


ratio
October 22, 2003 Landsat TM

AISA, an airborne hyperspectral imager


will be flying over Pensacola Bay next year
to provide an alternative tool and verify
the effectiveness of Landsat TM data
Preliminary flights were conducted in July and
November 2003, results were failure and
promising respectively.

Also will be comparing ASA to LandSat TM

27

Acknowledgements
Karen Jordan and Yongtao Luo, graduate
research assistants, helped in data
processing
Gulf Ecology Division of US Environmental
Protection Agency provided water
sampling data and facilitated
hyperspectral reflectance data collection
This research has been supported by a
grant from the US Environmental
Protection Agency's Science to Achieve
Results (STAR) Estuarine and Great Lakes
(EaGLe) Coastal Program.
R82945801

28

Disclaimer

Although the research described in this


presentation has been funded wholly or in
part by the United States Environmental
Protection Agency through cooperative
agreement R82945801 to the Consortium
for Estuarine Ecoindicator Research for the
Gulf of Mexico (CEER-GOM) it has not
been subjected to the Agencys required
peer and policy review and therefore does
not necessarily reflect the views of the
Agency and no official endorsement
should be inferred

29

You might also like