MODIS Vegetation Index ProductMODIS Vegetation Index ProductMODIS Vegetation Index Productfrom C4 to C5from C4 to C5from C4 to C5

Kamel Didan & Alfredo Huete, TBRS Lab.The University of Arizona

MODIS Land Collection 5/LTDR WorkshopUniversity of Maryland University College, January 17-18, 2007

2

OutlineOutline

VI product & AlgorithmC5 changesC4 vs. C5Validation statement Future plans (C6?)

3

Algorithm and Product DependenciesAlgorithm and Product Dependencies

Daily observations (Pixel 60-1860, Tile h20v11)

-0.2

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1 6 11 16 21 26 31 36 41 46 51 56 61

Observation number

Surf

. Ref

l. an

d V

Is.

Red

NIR

Blue

MIR

NDVI

EVI

Cloud LegendClearMixedCloudy

Aerosol LegendHighAverageLow

Observations after quality evaluation and weighted average procedure. (Pixel 60-1860, Tile h20v11)

00.10.20.30.40.50.60.70.80.9

1

1 3 5 7 9 11 13 15 17

Observation number

Surf

. Ref

l. an

d V

Is RedNIRBlueMIRNDVIEVI

4

CVCV--MVC compositingMVC compositingAVHRRAVHRR--MVC vs. MODISMVC vs. MODIS--MVC vs. MODIS CVMVC vs. MODIS CV--MVCMVC

0102030405060708090

Ideal d

ata

Aeroso

l/Sha

dow/View is

...

Snow/Ic

e

Cloud

MVC_avhrrMVC_modiscv_mvc

0

20

40

60

80

100

120

Ideal d

ataAero

sol/S

hadow/View

i...

Snow/Ic

e

Cloud

MVC_avhrrMVC_modiscv_mvc

5

SUMMARY of C5 changesSUMMARY of C5 changes

1. Added per pixel reliability (Second most important change)

2. Added the composite day of the year to the output

3. Replaced the NDVI_QA and EVI_QA with one VI_QA layer

4. Restructured the VI_QA data layer to eliminate redundant information

6

1. Improved input data filtering ( & added adjacent cloud filter)

2. Modified the compositing approach:1. To identify mislabeled cloud2. Modified CV-MVC to favor smaller view angle

3. Adopted a phased production for Terra and Aqua to keep the streams separate while increasing the temporal frequency (Most important change)

SUMMARY of C5 changesSUMMARY of C5 changes

7

One VI QA layer (restructured) One VI QA layer (restructured)

NDVI QA and EVI QA are identicalEliminated this redundancy

8

Composite day of the yearComposite day of the year

Useful for temporal comparisons and validation work

Southeast US (h10v05)

9

Composite day of the yearComposite day of the year

Composite period 2000-161

10

Per pixel data reliabilityPer pixel data reliabilityEliminate the need for QA fields/binary manipulation. Ordinal number & few (4) categoriesPowerful tool for time series/Phenology analysis– Helps in extracting ‘true’ growing season profile

Snow/Ice Ideal Marginal CloudInterpolatedOnly for CMG

11

Data reliability applicationsData reliability applications

C 5 ND V I/E V I ( B ra z i/C e r ra d o )

0

0.2

0.4

0.6

0.8

1

19-Jan-00 23-M ar-00 26-M ay -00 29-Ju l-00 01-O c t-00 04-D ec -00 06-F eb-01 11-A pr-01

D a te

NDV

I/EVI

ND VI_Good E VI_Good ND VI_Marg ina l E VI_Margina lND VI_S now E VI_S now ND VI_C loudy E VI_C loudy

ND VI Pro file (T apajos)

00.2

0.40.60.8

11.2

Jan-00 Jul-00 Jan-01 Jul-01 Jan-02 Jul-02 Jan-03 Jul-03 Jan-04 Jul-04 Jan-05 Jul-05 Jan-06D ate

ND

VI

NDVI_Good ND VI_Marginal NDVI_Snow ND VI_C loudy

12

E V I E x a m p le A n n u a l P ro file s

0

0 .1

0 .2

0 .3

0 .4

0 .5

0 .6

0 .7

0 .8

0 .9

1

1 4 - De c 2 - Fe b 2 4 - Ma r 1 3 - Ma y 2 - Ju l 2 1 - A u g 1 0 - O c t 2 9 - No v 1 8 - Ja n

D a te

EVI

E V I_ No r th _ C A _ C o a s t_ 1

E V I_ No r th _ C A _ C o a s t_ 2

E V I_ E l_ C e n tr o

E V I_ S ib e r ia _ 1

E V I_ S ib e r ia _ 2

E V I_ D e s e r t_ 1

E V I_ Hi_ E V I_ S ite _ 1

E V I_ S E _ A m a z_ B a s in

Data reliability applicationData reliability applicationNDVI Example Annual Profles

00.10.20.30.40.50.60.70.80.9

1

14-Dec 24-Mar 2-Jul 10-Oct 18-JanDate

ND

VI

NDVI_North_CA_Coast_1

NDVI_North_CA_Coast_2

NDVI_El_Centro

NDVI_Siberia_1

NDVI_Siberia_2

NDVI_Desert_1

NDVI_Hi_EVI_Site_1

NDVI_SE_Amaz_Basin

13

Global performanceGlobal performance

14

Phased ProductionPhased Production

Terra & Aqua data are highly identical, but slight differences & other issues makes combining them potentially problematic– Produced on the same day, so in most cases there is no

point in using both (C4 and earlier)

Improve the use efficiency without combining the two– Phased production was an ideal solution

• Increased temporal frequency• Now there is a reason to use both

Only Aqua will be phased

15

Phased ProductionPhased Production

In order for this to work there should be no bias in the selected composite day of the year.

Selected Composite DOY Distribution

0

2

4

6

8

10

12

14

16

0 11 22 33 44 55 66 77 88 99 110

121

132

143

154

165

176

187

198

209

220

231

242

253

264

275

286

297

308

319

330

341

352

363

DOY

Perc

ent (

%)

nadir aperture door was closed

MODIS-L1B data was not available

16

Phased productionPhased production

Jasp er_R id g e_B io lo g ica l_P reserve 3x3 W in d o w @ 312x348

0

0.1

0.2

0.3

0.4

0.5

2/23/2004 3/10/2004 3/26/2004 4/11/2004 4/27/2004 5/13/2004 5/29/2004D ate

EVI

T_8_Da ys A_8_Da ys T_16_Da ys A_16_Da ys

T_A_8_Da ys T_A_16_Da ys T_P _8_Da ys T_P _16_Da ys

17

C4 and C5 ComparisonsC4 and C5 Comparisons

18

Reduced BRDF

Improved data quality screening

C4 and C5 ComparisonsC4 and C5 Comparisons

19

Cloud reductionCloud reduction

C4

C5

20

NDVI comparisonsNDVI comparisons

C4- C5 NDVI Difference Distribution (23 Composite cycles)

0

2

4

6

8

10

12

14

16

18

-0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2NDVI Difference

Perc

ent (

%)

Consistent & slight decrease in C5

21

EVI ComparisonsEVI Comparisons

C4- C5 EVI Difference Distribution (23 Composite cycles)

0

2

4

6

8

10

12

14

16

18

20

-0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2

EVI Difference

Perc

ent (

%)

Consistent & slight decrease in C5

22

Red LSR ComparisonsRed LSR Comparisons

C4 - C5 Red Difference Distribution

0

5

10

15

20

25

30

-0.1 -0.08 -0.06 -0.04 -0.02 0 0.02 0.04 0.06 0.08 0.1

Red Difference

Perc

ent (

%)

23

NIR LSR ComparisonsNIR LSR Comparisons

C4 - C5 NIR Difference Distribution

0

5

10

15

20

-0.1 -0.08 -0.06 -0.04 -0.02 0 0.02 0.04 0.06 0.08 0.1

NIR Difference

Perc

ent (

%)

24

Rain Forest NDVI Error Range

00.10.20.30.40.50.60.70.80.9

1

Dec-99 Apr-01 Sep-02 Jan-04 May-05

Date

ND

VI/E

rror

AvgMinMaxSTD_DevMax_Error

Noise related Error budgetNoise related Error budget

MODIS EVI Range/ErrorBoreal Forest

-0.1

0

0.1

0.2

0.3

0.4

0.5

0.6

12/6/1999 4/19/2001 9/1/2002 1/14/2004 5/28/2005

Date

EVI/E

rror

AvgMinMaxSTD_DevMedianMax_Error

25

Validation statementValidation statement

Maturity= "Validated Stage 2“

The MODIS Vegetation Index product is retrieved accurately using a state of the art MODIS specific, quality driven, constrained view, maximum value composite method. For most cloud and snow free, low to no aerosol load, the VI values are very reliable. Over inland water bodies (rivers, lakes, etc...) surface reflectance inputs and VI values are unstable and should be usedwith caution. Strong correlations are also found with the nadir-adjusted (NBAR) NDVI and EVI products. Various field validationcampaigns indicate good agreement of VI values for most biomes, and cross sensor correlations (AVHRR, Landsat, SPOT, etc…) are very strong. The VI product is particularly reliant on coherentinter-band (red and NIR) atmosphere corrections. Unstable values may be encountered over more extreme bright or dark surfaces.

26

C6 plans (towards CDR)C6 plans (towards CDR)

MODIS specific– Completely cloud free & gap filled

• It is pointless to produce/use cloudy observations• Gaps should be filled with obs. from the historical record

(similar to the CMG approach)– Dynamic composite (First good observation)

• No fixed composite cycle, but VI produced when a good pixel becomes available

General CDR– EVI2 for continuity– EVI Backward compatibility (paper to be submitted) and no-

Blue band sensors– BRDF???