eo-1 hyperion time series for vegetative and … hyperion time series for vegetative and calibration...

EO-1 Hyperion Time Series for Vegetative and Calibration Studies

Petya K. E. Campbell*, David Lagomasino☺Elizabeth Middleton**, Kurtis J. Thome**, Bill Cook**, K. Fred

Huemmrich*, Larry Corp***, Yen-Ben Cheng+ Lawrence Ong++, Nathan Pollack++, Stuart Fry and David Landis***

* Joint Center for Earth Systems Technology (JCET), University of Maryland Baltimore County, Catonsville, MD

☺GSFC Summer Student Intern, Florida International University, Miami, FL** NASA Goddard Space Flight Center, Greenbelt, MD

*** Sigma Space, Greenbelt, MD+ERT, ++SSAI, +++SGT

BackgroundIt has become critical to understand the land cover dynamics as ecosystems cycle through seasonal changes and respond to variable environmental conditions such as water, temperature, light and nutrient availability.

Remote sensing offers an unique opportunity to monitor the changes in land cover, through multiple spectral observations and understanding of the land cover spectral responses.

Multi-date monitoring of the terrestrial environment, requires adequate radiometric calibration in order to carry out biophysical and geophysical parameter surveys that are of sufficient sensitivity and are reproducible over time.

• How do vegetation spectral, environmental and functional characteristics change in concert, under the effects of seasonal dynamics and varying environmental factors?

• How well can Hyperion monitor the spectral response and stability/cycles of major natural land covers or vegetation types?

> By assembling an initial dataset of spectroscopy images for FLUX sites, we hope to contribute toward the understanding of how do natural ecosystems respond to impinging environmental changes, particularly to climate and land cover change and the increasing impacts of urbanization?

Science Questions

1. Assess the spectral properties and variability of EO-1 Hyperion spectra at three calibration sites, namely: Frenchman Flat, Ivanpah Playa and Railroad Valley, located in North America along a similar longitude but at a varying latitude and elevation.

2. At three vastly different vegetative sites (the FLUX sites Mongu, Duke and Konza), seek common spectral trends associated with vegetation function, induced by the seasonal effects/variation of temperature, moisture and humidity.

Research Objectives

EO-1 Observations MSO Sites CEOS Sites Volcanoes > 10 Observations

The EO-1 Time Series

Hyperion DataSite Name Free of Clouds Images

[total used]Analyzed Scenes by Acquisition Date

[Julian date/year]

1. Railroad Valley Playa 16 (2001-2009)

12/2006, 41/2006, 120/2007, 135/2007, 165/2001, 174/2004, 176/2008, 181/2001, 197/2001, 209/2007, 234/2006, 272/2005,

284/2005, 288/2007

2. Ivanpah Playa 10 (2002-2005) 68/2003, 81/2002, 148/2003, 169/2005, 180/2003, 244/2003

3. Frenchman Flat 7 (2003-2008) 12/2009, 178/2003, 199/2008, 219/2007, 222/2008, 270/2007, 337/2008

1. Mongu 25 (2008)22, 40, 107, 115, 141, 145, 158, 163, 171, 173, 176, 181, 191, 194, 199, 201, 204, 225, 230,

253, 284, 291, 296, 302, 365

2. Duke Loblolly Pine 7 (2008-2009) 6/2009, 34/2009, 162/2008, 180/2008, 203/2008, 290/2008, 300/2008

3. Konza Prarie 7 (2008-2009) 13/2009, 28/2009, 202/2008, 205/2008, 261/2008, 266/2008, 302/2008

Data Processing

Level 1R Hyperion data was atmospherically corrected using the Atmosphere CORrection Now (ACORN) model and subsequently, the spectral reflectance data was extracted in the vicinity of existing flux towers. Spectral sensitivity analysis was conducted to determine spectral trends and spectral bio-indicators were computed.

-50

0

50

100

150

200

250

300

350

400

450

500

550

600

-50 0 50 100 150 200 250 300 350 400 450 500 550 600

ACO

RN

ATREM

corn (r = 0.95)

forest (r = 0.98)

water (r = 0.92)

bright target (r = 0.97)

lichens (r = 0.98)

ice (r = 0.99)

0

100

200

300

400

500

600

700

450 700 950 1200 1450 1700 1950 2200 2450

Re

fle

cta

nce

(%)

Wavelength (nm)

ice AC ice ATbright target AC bright target ATcorn AC corn ATlichens AC lichens ATforest AC fores ATwater AC water AT

CEOS Calibration Sites

Site NameCenter

LatitudeCenter

Longitude

Elevation(m asl)

Frenchman Flat, NV 36.80928 -115.93479 940

Ivanpah Playa, CA 35.5692 -115.3976 813

Railroad Valley, NV 38.50 -115.69 1435

0

10

20

30

40

50

60

70

80

90

100

450 700 950 1200 1450 1700 1950 2200 2450

Frenchmen Flat

Ivanpah Playa

Railroad Valley

Wavelength (nm)

Refle

ctan

ce (%

)Re

flect

ance

(%)

ρ(%

, mea

ns a

nd st

anda

rt de

viat

ions

)

CEOS Sites

Temporal profile of Hyperion at Railroad Valley Playa

0

10

20

30

40

50

60

70

80

90

100

12 41 120 135 165 174 176 181 197 209 229 234 272 284 288

2006 2006 2007 2007 2001 2004 2008 2001 2001 2007 2001 2007 2005 2005 2007

Blue447 Green549 Red651

NIR854 FR1003 FR1679

FR2204

Acquisition (Julian day)

ρ(m

eans

at s

elec

t ban

ds,%

)

EO-1 Hyperion Spectral Bands

Year

Jul. day

-25-20-15-10

-505

10152025

450 700 950 1200 1450 1700 1950 2200 2450

Wavelength (nm)

ρ-m

ean(ρ)

Julian days 41-28812/2006

4 km-1 -0.5 0 0.5 +1

N

N

+ +

Left: Natural color composite (RGB: 651, 549, 447)

Right: Getis Gi* statistics (band 549) calculated for 197 Jul day image. The Gi statistics calculated for G, R, NIR, SWIR bands produced similar results.

-1 -0.5 0 0.5 +14 km

N

++


Right: Getis Gi* statistics (band 549) calculated for 197 Jul day image. The Gi statistics calculated for G, R, NIR, SWIR bands produced similar results.

0

10

20

30

40

50

60

70

80

90

100

68 81 148 169 180 244

2003 2002 2003 2005 2003 2003

Blue447 Green549Red651 NIR854FR1003 FR1679FR2204

Acquisition (DOY)

Spectral Bands

ρ(%

, mea

ns a

nd st

anda

rd

devi

atio

ns)

-25-20-15-10

-505

10152025

450 700 950 1200 1450 1700 1950 2200 2450

68 81 148 169180 244 average

Wavelength

DOYρ-m

ean(ρ)

Temporal Profile at Ivanpah Playa

-1 -0.5 0 0.5 +1

N

4 km

+ +


Right: Getis Gi* statistics (band 549) calculated for 197 DOY day image. The Gi statistics calculated for G, R, NIR, SWIR bands produced similar results.

This is the site of the LED Spectrometer (LSpec) autonomous calibration facility (http://lspec.jpl.nasa.gov)

Temporal Profile at Frenchman Flat (FMF)

Vegetative Study Sites

FLUX Site Name Location Climate Vegetation

1. Mongu Zambia Temperate/ warm summer

Kalahari/ MiomboWoodland

2. Duke Loblolly Pine, Hardwoods

NorhtCarolina, US

Temperate/ no dry season/ hot summer

Mixed forest/ Evergreen

3. Konza Prarie Kanzas, US Cold/ no dry season/ hot summer

Grassland

1.

2.3.

MSO Sites

Vegetative Study Sites•The seasonal dynamics in spectral properties are assessed at three vegetative FLUX sites: Mongu (Zambia, Africa), Konza Prairie (Kansas, USA), and Duke Forest (North Carolina, USA). •Each vegetative site represents a distinct vegetative ecosystem type; hardwood forest, grassland, evergreen forest, respectively. •Flux tower data (e.g., CO2 flux, soil heat flux) from each site is compared to spectral information for the same locations.

Mongu, Zambia

-5

0

5

10

15

20

25

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5

0 50 100 150 200 250 300 350

Pre

cipi

tati

on, C

O2

Flu

x, A

lbed

o

Hyp

erio

n S

pect

ral B

io-In

dica

tor

(Dm

ax)

Temporal profile of Hyperion’s Spectral Bio-indicators at Mongu

The temporal profile in the derivative spectral bio-indicator Dmaxcaptures the dynamics in vegetation phenology at Mongu, Zambia

Dmax (x10) Precipitation (cm) CO2 flux (umol m -2 s-2 -1) Albedo (x10)

Day of the Year (2008)

Day of the Year (2008)

-5

0

5

10

15

20

25

0

1

2

3

4

5

6

0 50 100 150 200 250 300 350

Prec

ipita

tion,

CO

2Fl

ux, A

lbed

o

Hype

rion

Spe

ctra

l Bio

-Indi

cato

r Ch

lorop

hyll,

G32

= (R

750-

R445

)/((R

700-

R445

)

Spectral Bio-indicator CO2 flux (umol m -2 s-2 -1) Precipitation (cm) Albedo (x10)

The temporal profile a spectral bio-indicator associated with chlorophyll content (G32, green line) captured best the dynamics in vegetation phenology

at the flux site Mongu, Zambia.

Prec

ipit

atio

n, C

O2

Flux

, Alb

edo

Spec

tral

Bio

-indi

cato

rCh

lorop

hyll,

G32 =

(R75

0-R4

45)/(

R700

-R44

5)Hyperion’s Spectral Bio-indicators at Mongu

Duke Pine Site, NC

-5

0

5

10

15

20

25

0

5

10

15

20

25

30

35

40

45

50

0 50 100 150 200 250 300 350

Ref

lect

ance

(%)

DOY

518 549 651 854 1054 1649 2204 precip (cm) CO2 flux

Prec

ipita

tion,

CO

2flu

x , A

lbed

o

Seasonal Spectral Dynamics at Duke Pine Site

-50510152025

-0.2-0.18-0.16-0.14-0.12

-0.1-0.08-0.06-0.04-0.02

0

0 50 100 150 200 250 300 350Spec

tral

Bio

-indi

cato

r PR

I =

(570

-531

)/(57

0+53

1)

DOY

PRI1 precip (cm) CO2 flux

Prec

ipita

tion,

CO

2flu

x , A

lbed

o-50510152025

-80-70-60-50-40-30-20-10

0

0 50 100 150 200 250 300 350

Spec

tral

Bio

-indi

cato

r

DOY

RFG precip (cm) CO2 flux albedo

Prec

ipita

tion,

CO

2flu

x , A

lbed

o

Seasonal Spectral Dynamics at Konza

-15

-10

-5

0

5

10

15

20

0

5

10

15

20

25

30

35

40

45

50

0 50 100 150 200 250 300 350

Prec

ipita

tion,

CO

2Fl

ux, A

lbed

o

Ref

lect

ance

(%)

DOY

518 549 651 854 1054 1649 2204 precip (cm) CO2_flux_2008

Temporal Spectral Dynamics at Konza

0102030405060708090

100

0 50 100 150 200 250 300 350-5

0

5

10

15

20

25

Spec

tral

Bio

-indi

cato

rP

hys.

Fun

ctio

n,

RFf

r= (R

730-

R65

0)/(R

685+

R65

0 )

DOY

Prec

ipita

tion,

CO

2Fl

ux,

Albe

do

precip (cm) CO2_flux_2008 albedo RFfr

0

1

2

3

4

5

6

0 50 100 150 200 250 300 350-5

0

5

10

15

20

25

Prec

ipita

tion,

CO

2Fl

ux,

Albe

do

DOY

Spec

tral

Bio

-Indi

cato

r D

max

, Veg

. Phy

siol

ogy,

Can

opy

Inde

pend

ent

precip (cm) CO2_flux_2008 albedo Dmax

0

10

20

30

0 50 100 150 200 250 300 350-5

0

5

10

15

20

25

Spec

tral

bio

-indi

cato

rC

hlor

ophy

ll, G

94 =

(R80

0-R

700)

/(R80

0+R

700)

DOY

Prec

ipita

tion,

CO

2Fl

ux, A

lbed

o

precip (cm) CO2_flux_2008 G94 albedo

Temporal profile of Hyperion’s Spectral Bio-indicators at Konza

Multiple Flux SitesKonza (K) 1yr burn & 4 year burn, Mongu (M), Duke (D) pine & hardwoods

y = -0.0024x2 - 1.4794x + 17.814R² = 0.75

0

10

20

30

40

50

60

70

80

-40 -30 -20 -10 0 10

Spec

tral

Bio

-indi

cato

r (D

max

)

CO2 flux (µmol/m2/s)

K1DK4BMDDuke HWhw

p

At a number of sites with vastly different vegetation type, the derivative spectral bio-indicator (Dmax) reflected vegetation dynamics and was negatively correlated with CO2 flux.

Correlation coefficients for the top 5 (based on R values) biophysical indices

Spectral Bio-indicator

FormulaCO2 flux

CO2Respiration

ResSoil Heat Flux

G

Dmax Max. derivative in the 690-730 - 0.82 0.69 0.90

G32, G94 R(750-445)/(700-445), R(800-700)/(800-700) - 0.79 0.67 0.87

RFfr R(650-730)/(730-650) - 0.78 0.62 0.85

Transformed Chl Abs R index

R700, R600, R550 - 0.78 0.25 0.94

NDWI R(870-1240)/R(870-1240) - 0.77 0.29 0.83

• Biophysical spectral indices correlate well with G and CO2 flux measurements from 3 vastly different vegetation types

• Unique seasonal R dynamics corresponded with each site specific phenology

• Highly correlated spectral bio-indicators required continuous spectra, and were using a number of wavelengths near the “red edge”, and

• High spectral resolution imaging is useful for detecting parameters related to CO2 flux and G

Preliminary Findings

Future Directions• expand and refine the spectral datasets and the spectral bio-indicators (add

other locations and vegetation types, and consider BRDF effects)• test the established Canopy Chemistry equations for the sites with time series• use the datasets in pre-launch product simulations of sensors such as EnMAP

and HyspIRI• characterize basic ecological patterns in anthropogenic ecosystems related to

differences in ecosystem function.

! Overlapping satellite observations and land cover records, with 1 to 2 year overlap periods are required for the generation of long term data records for environmental monitoring and climate change assessments.

Project Collaborators• Hank Margolis, Laval University , Quebec, Canada• Jose F. Moreno, Department of Earth Physics and

Thermodynamics, University of Valencia, Spain• Nicholas Coops, University of British Columbia, Vancouver, CA• Caroline J Nichol, School of GeoSciences, University of

Edinburgh, Edinburgh, UK• Valérie Demarez, CESBIO, Toulouse, France• Frederic Baret, CNES, Avignon, France

AcknowledgmentsDr. Brunsell and the KNZ-LTER supplied flux data for the Konza Prairie. The School of the Environment at Duke University supplied flux data for the Duke Forest sites.

The project provided research experience for graduate and undergraduate students, which participate through established programs, assisting with data processing, analysis and preparation of reports.

Thank You

Temporal variations in the spectral characteristics of the Frenchman Flat site (reflectance means and standard deviations for selected Hyperion bands). The spectral characteristics are significantly by the presence of various proportion of clouds (cloud cover expressed in percent of imaged area covered by clouds) in the atmosphere adjacent to the site, as is evident for Julian days 54, 82, 107, 166, 214, 217 and 278.

EO-1 Hyperion:

RGB 752, 609, 508 nm

30 m pixels Reflectance

1 km 1 km

EO-1 Hyperion:

RGB 752, 609, 508 nm


July 13, 2008 August 18, 2008 October 2, 2008

forest

forestcorn

water

forest

forestcorn

water water

forest

forestcorn

1 km

EO-1 Hyperion:

RGB 752, 609, 508 nm


Seasonal Spectral Changes at OPE3

Cover Type Hyperion, 2008 V1 PRI REIP Dmax WBI Albedo Corn 13-Jun 1.03 -0.04 712 0.36 0.96 0.461

18-Aug 1.81 -0.06 722 0.75 1.09 0.197 3-Oct 1.15 0.04 721 0.51 0.98 0.155

Forest 13-Jun 1.12 -0.06 712 0.89 1.00 0.257 18-Aug 1.56 -0.03 722 0.51 1.01 0.140 3-Oct 1.61 -0.10 712 0.42 0.94 0.127

Water 13-Jun 0.15 0.01 712 0.16 1.23 0.058 18-Aug 0.52 0.02 712 0.10 1.46 0.031 3-Oct 0.62 -0.07 712 0.08 0.93 0.036

Seasonal Spectral Dynamics in Land Cover at OPE3

Spring Summer Fall17-Apr 13-Jun 20-Jul 18-Aug 4-Oct

EO-1 Hyperion True color

RGB Color Composite Bands 752, 609, 508 nm

1 km


forest

corn

water


1 km


forest

corn


1 kmRGB Color Composite

Bands 752, 609, 508 nm

1 km1 km

60 mpixelsReflectance

forestforest

corncorn

waterwater


1 km


forest

corn



Bands 752, 609, 508 nm

1 km1 km


forestforest

corncorn

waterwater



Bands 752, 609, 508 nm

1 km1 km


forestforest

corncorn


1 km1 kmRGB Color Composite

Bands 752, 609, 508 nm

1 km1 km

60 mpixelsReflectance

forestforest

corncorn

waterwater

USDA/BARC Beltsville, MD

18-August, 2008

Water Stress Index (ρ1600/ ρ 820)

c

Bio-physical parameter Spectral bio-indicators (examples) References

VSWIR indicators Chlorophyll in crops (RNIR/ R-1

720-730) - 1 Gitelson et al., 2005

Canopy greenness LRCRCRRR

GEVIblueredNIR

redNIR

+−+−

=21

* Huete et al. 2002

Chlorophyll concentration estimation

)16.0/()(*)16.01()]/(*)(*2.0)[(*3/

670800700800

670700550700670700

++−+−−−

=RRRR

RRRRRROSAVITCARI Haboudane et al. 2002, Zarco-Tejada

et al. 2004 Efficiency of photosynthesis PRI=(R570-R531)/(R570+R531) Gamon et al., 1992,

Suarez et al. 2008 Canopy foliar water content 1240

1240

RRRR

NDWINIR

NIR

+−

= Gao, 1996

Canopy foliar water content 1640

1640

RRRR

SIWINIR

NIR

+−

= Fensholt and Sandholt, 2003

Vegetation stress Average (R675…. R705) Vogelman, 1993 Veg. structure and foliar biomass NDVI = (R800-R670)/(R800+R670) Deering, 1978

Stress/Chlorophyll D714/D705, where D is product of first derivative transformation of R Entcheva, 2004

Stress/ Chlorophyll Dmax/D705, where Dmax is the maximum of D in the 670-730nm region Entcheva, 2004

eo-1 hyperion time series for vegetative and … hyperion time series for vegetative and calibration...

Documents