27
Scaling Up Above Ground Live Biomass From Plot Data to Amazon Landscape Sassan S. Saatchi NASA/Jet Propulsion Laboratory California Institute of Technology 4800 Oak Grove Drive Pasadena, California 91109 Tel: 818-354-1051 Fax: 818-393-5285 Email: [email protected]

Scaling Up Above Ground Live Biomass From Plot Data to Amazon Landscape Sassan S. Saatchi NASA/Jet Propulsion Laboratory California Institute of Technology

Embed Size (px)

Citation preview

Page 1: Scaling Up Above Ground Live Biomass From Plot Data to Amazon Landscape Sassan S. Saatchi NASA/Jet Propulsion Laboratory California Institute of Technology

Scaling Up Above Ground Live Biomass From Plot Data to

Amazon Landscape

Sassan S. SaatchiNASA/Jet Propulsion LaboratoryCalifornia Institute of Technology

4800 Oak Grove DrivePasadena, California 91109

Tel: 818-354-1051Fax: 818-393-5285

Email: [email protected]

Page 2: Scaling Up Above Ground Live Biomass From Plot Data to Amazon Landscape Sassan S. Saatchi NASA/Jet Propulsion Laboratory California Institute of Technology

Houghton et al. 2000

Page 3: Scaling Up Above Ground Live Biomass From Plot Data to Amazon Landscape Sassan S. Saatchi NASA/Jet Propulsion Laboratory California Institute of Technology

Houghton et al. 2001

Method SpatialResolution

SecondaryForests

Forest Cover Total Carbon inBiomass (PgC)

Mean Biomass(MgC/ha)

1. Independent Measurement 5 km Avoided Potential 76.5 1922. RADAMBRAZIL (Brown et al. 1992) 1 km Some Potential 62.5 1563. RADAMBRAZIL (Fearnside, 1997) 1 km Some Potential 93.1 2324. Brown (calibrated with 39 points) Brown (calibrated with forest surveys) Brown (calibrated with areas >0.5 ha)

5km Avoided Potential 73.078.078.8

183196197

5. Olson 1o x 1o Included Actual 38.9 1006. Potter (CASA model driven by NDVI) 1o x 1o Some Actual 78.2 1967. DeFries (% woody cover map) 1 km Some Actual 69.2 178Mean 70 177Std. Error 8 20

Conclusion:1. Estimates of biomass for Brazilian Amazon vary by more than a factor of 2.2. Disagreements on regions of high & low biomass3. Methods disagree on spatial patterns of distribution of biomass4. Unknown spatial variations in biomass makes accurate calculations of

flux impossible

Question: What is the spatial distribution of vegetation biomass in the Amazon basin?

Page 4: Scaling Up Above Ground Live Biomass From Plot Data to Amazon Landscape Sassan S. Saatchi NASA/Jet Propulsion Laboratory California Institute of Technology

What is the Regional Distribution of Biomass?

Methods:• Assign average biomass values to vegetation map

2. Interpolate biomass plots to the region

3. Model structural distribution and biomass using environmental variables such as climate, topography, vegetation, soil, etc.4. Use of ecosystem models for carbon allocations and/or assimilation

5. Extrapolate plot data to the region using remote sensing metrics.

Questions:

1. Which environmental variables or RS metrics?

2. Which models or extrapolation approaches?

3. How errors and uncertainties are propagated?

Page 5: Scaling Up Above Ground Live Biomass From Plot Data to Amazon Landscape Sassan S. Saatchi NASA/Jet Propulsion Laboratory California Institute of Technology

Investigator Geographical Location Vegetation TypeE. Moran & E. Brondizio Altamira, Maraju Isl., Brangantina, Tome-Acu, Brazil, Yapu columbia Secondary, Primary

M. Stienneger Manaus Brazil, Santa Cruz Bolivia Secondary Forest, FloodplainW. Laurance Fragmentation site, Manaus, Brazil Secondary, Primary, Fragmented

T. Killeen, S. Brown Noel Kempt National Park, Bolivia Primary, Liana ForestR. Lucas Manaus, Brazil Secondary, Primary

A. Luckman Tapajos, Manaus, Brazil Seondary, PrimaryB.Nelson Acre, Brazil Primary, Bamboo Forest

D. Hoekman Guaviare, Columbia Primary, Secondary, Pasture, BrunedF. Brown Acre, Brazil Secondary, PrimaryWageni Manaus, Brazil Palm ForestD. Alves Rondonia, Brazil Secndary Forest

CIAD Pucalpa, Peru Secondary ForestR. Vasquez Peru Primary ForestM. Silman Manu, Peru Primary, Floodplain, BambooB. Turque Ecuador Primary, Secondary ForestJ. Sanden Mabura Hill, Guyana Primary, Secondary, Sw amp, PastureR. Quiroz Peru Lowland Primary

L. Araujo & J. Santos Roraima, Rondonia, Mato Grosso Brazil Primary, Secondary, Woodland SavannaE. Sano Brazilia National Park Cerrado Vegetation

M. Delaney Venezuela Wet & Dry Montane ForestJB. Kauffman Brazil Submontane Forest

N. Higuchi Brazil Dense Moist Forest

544 plots used for this study

Page 6: Scaling Up Above Ground Live Biomass From Plot Data to Amazon Landscape Sassan S. Saatchi NASA/Jet Propulsion Laboratory California Institute of Technology

Methodology• Collect most recent ground measurements of above ground forest biomass in the basin.

• Acquire and develop a series of remote sensing data and products:Radar backscatter and textureMODIS continuous field product (% forest cover)Canopy roughness derived from RS data fusionDigital Elevation, Slope, and Ruggedness factor (SRTM)NDVI metrics Scatterometer metrics (surface moisture)Vegetation map from data fusion

• Develop a cover specific semi-empirical algorithm between remote sensing data and above ground biomass

• Use ground data in a bootstrapping technique to iteratively improve algorithm and verify results

• Produce a 1km resolution above ground vegetation biomass over the Amazon basin

Page 7: Scaling Up Above Ground Live Biomass From Plot Data to Amazon Landscape Sassan S. Saatchi NASA/Jet Propulsion Laboratory California Institute of Technology

Ground Biomass Measurements

• Measurements are performed on small plots• Allometric equations are species specific• Geographical locations are not accurate• Detailed structural parameters are not always available

0

50

100

150

200

250

300

0 5 10 15 20 25 30 35 40

Biomass and Basal Area Relation of Tropical Forest Regeneration

Basal Area (m 2/ha)

Bio

ma

ss

(to

ns

/ha

)

0

100

200

300

400

500

600

0 5 10 15 20 25

Bio

ma

ss

(to

ns

/ha

)

Height (m)

y = 1.4538 + 0.12147x + 0.79269x 2 R=0.94979

Page 8: Scaling Up Above Ground Live Biomass From Plot Data to Amazon Landscape Sassan S. Saatchi NASA/Jet Propulsion Laboratory California Institute of Technology

Surface Geomorphology & Drainage Systems

TM Amazon Basin

SRTM 90 m

Detailed View

Page 9: Scaling Up Above Ground Live Biomass From Plot Data to Amazon Landscape Sassan S. Saatchi NASA/Jet Propulsion Laboratory California Institute of Technology

SRTM Derived Drainage System and Watersheds

Page 10: Scaling Up Above Ground Live Biomass From Plot Data to Amazon Landscape Sassan S. Saatchi NASA/Jet Propulsion Laboratory California Institute of Technology

DEM < 100 m

Page 11: Scaling Up Above Ground Live Biomass From Plot Data to Amazon Landscape Sassan S. Saatchi NASA/Jet Propulsion Laboratory California Institute of Technology

Dense forestLiana forestBamboo forestSeasonal forestMontane forest

Closed flood forest

Open flood forestHerbaceous floodplain

Mangrove

Savanna

Open woodland

Park savanna

Closed woodland

Mixed forest-woodlandSecondary forest

Nonforest

Vegetation Map of Amazon Basin Vegetation Map of Amazon Basin

Page 12: Scaling Up Above Ground Live Biomass From Plot Data to Amazon Landscape Sassan S. Saatchi NASA/Jet Propulsion Laboratory California Institute of Technology

5.0<Z0<8.03.5<Z0<5.0

Page 13: Scaling Up Above Ground Live Biomass From Plot Data to Amazon Landscape Sassan S. Saatchi NASA/Jet Propulsion Laboratory California Institute of Technology

NDVI SPOT-VEGETATION

Page 14: Scaling Up Above Ground Live Biomass From Plot Data to Amazon Landscape Sassan S. Saatchi NASA/Jet Propulsion Laboratory California Institute of Technology

11 km Radar Mosaic & Texture Maps

Segmentation of radar texture

Page 15: Scaling Up Above Ground Live Biomass From Plot Data to Amazon Landscape Sassan S. Saatchi NASA/Jet Propulsion Laboratory California Institute of Technology

Segmentation of Surface Ruggedness

Page 16: Scaling Up Above Ground Live Biomass From Plot Data to Amazon Landscape Sassan S. Saatchi NASA/Jet Propulsion Laboratory California Institute of Technology

a=6.1298c=6861.9n=2079.5

Kriging Plot Data

Undisturbed Forest: 273-491Disturbed Forest: 127-198Bamboo Forest: 177-340Semi-Dec: 167-252Flooded Forest: 198-334Open Flooded Forest: 53-91Woodlands: 88-144

14

15

1317

3

12

45

6

7/8

910

11

12

16

18

19

20

21

22

23

Page 17: Scaling Up Above Ground Live Biomass From Plot Data to Amazon Landscape Sassan S. Saatchi NASA/Jet Propulsion Laboratory California Institute of Technology

40

42

44

46

48

50

52

54

0 5 10 15 20 25

Ba

ck

sc

att

er

Inte

ns

ity

(m

2/m

2)

Biomass (tons/ha)

Savanna Vegetation

Sigma=37.876 x b 0.093038 R2=0.763

Relationship Between Radar Backscatter & Biomass

20

40

60

80

100

120

140

0 100 200 300 400 500 600

y = 40.684 * x^(0.16572) R= 0.9292

Ba

ck

sc

att

er

Inte

ns

ity

(m

2/m

2 )

Biomass (tons/ha)

Secondary Regrowth & Primary Forest Plots

Secondary & Primary Forests Savanna Vegetation

Page 18: Scaling Up Above Ground Live Biomass From Plot Data to Amazon Landscape Sassan S. Saatchi NASA/Jet Propulsion Laboratory California Institute of Technology

90

100

110

120

130

140

100 150 200 250 300 350 400

Old regrowth & Primary Forest

y = 36.937 * x^(0.21599) R= 0.75105 Co

-oc

cu

ran

ce

En

erg

y (

10

0 p

ixle

s)

Biomass (tons/ha)

20

25

30

35

40

45

50

55

60

0 10 20 30 40 50 60 70 80

Young Regeneration

Co

-oc

cu

ran

ce

En

erg

y (

10

0 p

ixe

ls)

Biomass (tons/ha)

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

0 50 100 150 200 250 300 350

y = 1.8844 * x^(-0.29041) R= 0.94135

No

rma

lize

d C

oe

f. V

ari

ati

on

(1

00

pix

els

)

Biomass (tons/ha)

Regrowth & Primary Forest

Relationship between biomass &1. Co-occurance Energy2. SNR adjusted Coef. Variation

Page 19: Scaling Up Above Ground Live Biomass From Plot Data to Amazon Landscape Sassan S. Saatchi NASA/Jet Propulsion Laboratory California Institute of Technology

R2 = 0,28220

20

40

60

80

100

120

140

0 100 200 300 400 500 600

Ground Biomass

SP

OT

metr

ic 6

amazon_ndvi_metric6_reg1.byt

Expon.(amazon_ndvi_metric6_reg1.byt)

R2 = 0,0811

150

200

250

300

350

400

0 50 100 150 200 250 300 350 400 450

ground biomass

nd

vi

sat_bio

Linear (sat_bio)

Page 20: Scaling Up Above Ground Live Biomass From Plot Data to Amazon Landscape Sassan S. Saatchi NASA/Jet Propulsion Laboratory California Institute of Technology

82 111 140

31 71 110

Metric 6: Max. NDVI Dry season

Dry Season Radar Backscatter

Baysian Approach : Maximum Likelihood Estimationgi (x)lnP(i )

1

2ln

i

1

2(x i)

T i 1(x i)

Biomass Estimation Approach for Undisturbed Forests

Page 21: Scaling Up Above Ground Live Biomass From Plot Data to Amazon Landscape Sassan S. Saatchi NASA/Jet Propulsion Laboratory California Institute of Technology

Biomass Estimation Algorithm

1. Use land cover map for algorithm implementation2. Use the bootstrapping method and nonlinear estimation techniques to estimate the coefficients of the following function:

bi = a1i T + a2i Z + a3i A where: bi = biomass of class i

T = SNR corrected coefficient of variation Z = Canopy Roughness A = Backscatter Amplitude power coefficients

3. The optimum coefficients from the bootstrapping technique is used in final algorithm

Page 22: Scaling Up Above Ground Live Biomass From Plot Data to Amazon Landscape Sassan S. Saatchi NASA/Jet Propulsion Laboratory California Institute of Technology

0 50 100 150 300Mg/ht

Vegetation Above Ground Live Biomass

Page 23: Scaling Up Above Ground Live Biomass From Plot Data to Amazon Landscape Sassan S. Saatchi NASA/Jet Propulsion Laboratory California Institute of Technology

1. Biomass of deforested areas & Secondary Regrowth along roads are delineated.

2. Vegetation biomass along river channels are separated from terre firme forest.

3. Mico topography can cause errors in biomass estimation.

Analysis of Results

Page 24: Scaling Up Above Ground Live Biomass From Plot Data to Amazon Landscape Sassan S. Saatchi NASA/Jet Propulsion Laboratory California Institute of Technology

Analysis of Results

Biomass of transitional vegetation such as bamboo and liana forests are accurately estimated.

Page 25: Scaling Up Above Ground Live Biomass From Plot Data to Amazon Landscape Sassan S. Saatchi NASA/Jet Propulsion Laboratory California Institute of Technology

Vegetation Carbon Distribution Over Two Transects Along the Amazon Basin

East-West Transect

North-South Transect

Page 26: Scaling Up Above Ground Live Biomass From Plot Data to Amazon Landscape Sassan S. Saatchi NASA/Jet Propulsion Laboratory California Institute of Technology

Leaf Biomass Wood Biomass

Page 27: Scaling Up Above Ground Live Biomass From Plot Data to Amazon Landscape Sassan S. Saatchi NASA/Jet Propulsion Laboratory California Institute of Technology

Summary

0

10

20

30

40

50

60

70

80

0 100 200 300 400 500

Es

tim

ati

on

Err

or

(to

ns

/ha

)

Biomass (tons/ha)

Bootstrapping Approach

• No validation has been performed except the internal error analysis of the algorithm using the ground biomass.• Estimation Error increases with biomass because: 1. Lack of sensitivity of existing radar and optical instruments to high biomass values. 2. Quality of biomass data over dense forest • Spatial distribution of Amazon biomass/Carbon can be improved as more data becomes available.• Similar to RADAMBRAZIL & Brown interpolation map high biomass runs east-west through central Amazonia with some high values in Peruvian Amazon