Investigation of biophysical forest biomass and density from radar image texture

(1)

HAL Id: hal-02806314

https://hal.inrae.fr/hal-02806314

Submitted on 6 Jun 2020

HAL is a multi-disciplinary open access archive for the deposit and dissemination of sci- entific research documents, whether they are pub- lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

Investigation of biophysical forest biomass and density from radar image texture

Isabelle Champion, Pascale Dubois-Fernandez, Xavier Dupuis

To cite this version:

Isabelle Champion, Pascale Dubois-Fernandez, Xavier Dupuis. Investigation of biophysical forest biomass and density from radar image texture. EARSEL ”1st Forestry Workshop: Operational Remote Sensing in Forest Management”, Jun 2011, Prague, Czech Republic. 21 pl. �hal-02806314�

(2)

‘Investigation of biophysical forest biomass and density from radar image texture’

and density from radar image texture

Isabelle Champion p

INRA, UR1263 EPHYSE, F-33140 Villenave d’Ornon

Pascale Dubois Fernandez Xavier Dupuis Pascale Dubois-Fernandez, Xavier Dupuis

ONERA, Centre de Salon de Provence, France

Investigation of biophysical forest biomass and density from radar image texture. Champion I., Dubois-Fernandez P., Dupuis X.

(3)

SAR polarimetric HH, HV, VH, VV images (ONERA airborne system)

Experimental site in Les Landes

(Pi i t d lti t d t d )

(Pinus pinaster, even-aged cultivated stands)

SAR flight direction

SAR polarimetric image

Investigation of biophysical forest biomass and density from radar image texture. Champion I., Dubois-Fernandez P., Dupuis X. EARSEL "1st Forestry Workshop: Operational Remote Sensing in Forest Management", Prague, Czech Republic, 2-3 June 2011.

Bande P, HV polarisation

(4)

Equally sized homogeneous zones are sampled Stand ages are known

Stand ages are known

0

15 -10 -5

ma0

-25 -20 -15

sigm

HV VH

0 10 20 30 40 50 60

-30

stand age

HH VV

Whatever the polarization, meanp ,

sigma0 dynamics with forest growth is low for mature stands

→ Image texture ?

→ Image texture ?

(5)

Texture indicators are calculated (Haralick, 1973)

1) f  di ib i 2) f l l i

1) from  distribution 2) from grey-level co-occurrence matrix

50 60 70

0 10 20 30 40

âge = f(Variance)

10 0 10 20 30 40

âge = f(Energie)

300 350 400 450 500

P band, HV, 2004 N = 32 niveaux parcelle 23 ans

20 25 30

2.7 2.8 2.9 3 3.1 3.2 3.3 3.4

x 10^-3 -20

-10

Variance: R²=0.79228t student =7.3075 ErrorEst=5.5822

60 70

0.028 0.029 0.03 0.031 0.032 0.033 0.034 0.035 0.036 -20

-10

Energie: R²=0.77148t student =6.8749 ErrorEst=6.2565

50 60 70

-50 -40 -30 -20 -10 0 10

0 50 100 150 200 250

P band, HV, 2004 parcelle---23ans 32 Niveaux

5 10 15 20 25 30

5 10 15

0 10 20 30 40 50

âge = f(Entropie 1)

0 10 20 30 40

âge = f(Entropie 2)

5 10 15 20 25 30

5.6 5.65 5.7 5.75 5.8 5.85

-20 -10 0

Entropie 1: R²=0.79904t student =7.461 ErrorEst=5.0965

-0.184 -0.182 -0.18 -0.178 -0.176 -0.174 -0.172 -0.17 -0.168 -0.166 -0.164 -10

0

Entropie 2: R²=0.78931t student =7.242 ErrorEst=5.9519

image variance kewness kurtosis entropy

HV 0.79 0.45 0.60 0.80

image energy contrast IDM homog correlation entropy

HV 0.79 0.39 0.31 0.33 0.64 0.82

Texture indicators are highly correlated with forest growth

HV 0.79 0.45 0.60 0.80

VH 0.74 0.00 0.11 0.85

HH 0.77 0.00 0.21 0.80

VV 0.57 0.03 0.20 0.66

VH 0.76 0.58 0.49 0.50 0.47 0.78

HH 0.81 0.59 0.50 0.52 0.44 0.80

VV 0.68 0.17 0.57 0.57 0.54 0.66

Texture indicators are highly correlated with forest growth

Investigation of biophysical forest biomass and density from radar image texture. Champion I., Dubois-Fernandez P., Dupuis X. EARSEL "1st Forestry Workshop: Operational Remote Sensing in Forest Management", Prague, Czech Republic, 2-3 June 2011.

(6)

60 70

120

Texture indicator: 140

intensity variance

40 50

riance)

80 100 120

= f(Variance)

Â

10 20 30

âge = f(Va

40 60

stem biomass =

Âge R²=0.77

3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9

x 10^-3 -10

0

Variance: R²=0.77105t student =5.1906 ErrorEst=5.3913 0.7

3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9

x 10^-3 0

20

Variance: R²=0.7226t student =4.5649 ErrorEst=8.218 180

Trunk dbh R²=0.81

0.5 0.6

e)

140 160

nce)

Stem biomass R²=0.72

0.3 0.4

unk dbh = f(Varianc

80 100 120

l biomass = f(Varian

Total biomass R²=0.67

0 0.1

tru0.2

20 40 total 60

P band, VH

3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9

x 10^-3 0

3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9

x 10^-3 20

(7)

60 70

120

40 50

riance)

80 100 120

= f(Variance)

Â

10 20 30

âge = f(Va

40 60

stem biomass =

Âge R²=0.77

3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9

x 10^-3 -10

0

3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9

x 10^-3 0

20

0.5 0.6

e)

140 160

nce)

Stem biomass R²=0.72

0.3 0.4

unk dbh = f(Varianc

80 100 120

Total biomass R²=0.67

0 0.1

tru0.2

20 40 total 60

P band, VH

EARSEL "1st Forestry Workshop: Operational Remote Sensing in Forest Management", Prague, Czech Republic, 2-3 June 2011.

3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9

x 10^-3 0

3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9

x 10^-3 20

,

(8)

60 70

120

40 50

riance)

80 100 120

= f(Variance)

Â

10 20 30

âge = f(Va

40 60

stem biomass =

Âge R²=0.77

3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9

x 10^-3 -10

0

3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9

x 10^-3 0

20

0.5 0.6

e)

140 160

nce)

Stem biomass R²=0.72

0.3 0.4

unk dbh = f(Varianc

80 100 120

Total biomass R²=0.67

0 0.1

tru0.2

20 40 total 60

P band, VH

3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9

x 10^-3 0

3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9

x 10^-3 20

,

(9)

160

Retrieving biomass directly through 180

a texture indicator:

120 140

f(Variance)

80 100

total biomass = f

Total biomass

20 40

t 60

3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9

x 10^-3 20

T t l bi 1 186 +05* i 321 08 Total biomass =1.186e+05*variance-321.08 Mean error : 11.4 t.ha^-1

R²=0.67

(10)

120

Retrieving biomass directly through 140

80 100 120

f(Variance)

40 60

stem biomass = f

Stem biomass

3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9

0 20

x 10^-3 Variance: R²=0.7226t student =4.5649 ErrorEst=8.218

St bi 1 01 +05 * i 286 51

Stem biomass = 1.01e+05 *variance =-286.51 Mean error : 8.2 t.ha^-1

R²=0.72

(11)

0.7

Retrieving biomass with using allometric equation

0.5 0.6

e)

biomass _{per tree}= a*dbh^b * age^c

0.3 0.4

nk dbh = f(Variance

a) Age is known

0.1 trun 0.2

b) dbh is derived from texture :

3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9

x 10^-3 0

dbh=489.18*variance-1.38, Error=0.0354 m

1500

Biomass stand = biomass _{per tree} *stand density c) stand densit f(stand age)

1000

nsity (t.ha-1 )

calculated density = 2.45*10⁶*age^-3+185 observed values

c) stand density = f(stand age)

0 500

stand den

10 15 20 25 30 35 40 45 50 55

0

stand age

(12)

Retrieving biomass from texture

160 180 200

ss

1:1

direct: biomass/texture

indirect biomass with dbh/texture 160 180 200

ss

60 80 100 120 140

ulated stem bioma

60 80 100 120 140

ulated total biomas

I di tl bi f(t t )

0 20 40 60 80 100 120 140 160 180 200

0 20 40 60

observed stem biomass

calcu

0 20 40 60 80 100 120 140 160 180 200

0 20 40 60

observed total biomass

calcu

1:1

direct: biomass/texture indirect: biomass with dbh/texture

I_ directly biomass=f(texture)

II_ with allometric equation with dbh=f(texture) mean error < 20%

bi l bi

Mean error Stem biomass Total biomass Direct method 8.1 t.ha^-1 11.5 t.ha^-1 Indirect method 18 9 t ha^-1 23 7 t ha^-1

Indirect method 18.9 t.ha 23.7 t.ha