Remote sensing –Beyond images
Mexico 14-15 December 2013
The workshop was organized by CIMMYT Global Conservation Agriculture Program (GCAP) and funded by the Bill & Melinda Gates Foundation (BMGF), the Mexican Secretariat of Agriculture, Livestock, Rural Development, Fisheries and Food (SAGARPA), the International Maize and Wheat Improvement Center (CIMMYT), CGIAR Research Program on Maize, the Cereal System Initiative for South Asia (CSISA) and the Sustainable Modernization of the Traditional Agriculture (MasAgro)
Disease monitoring in wheat through remotely sensed data
1. Disease monitoring in wheat
through remotely sensed data
Perla Chávez-Dulanto1
Pawan K. Singh1
Christian Yarlequé2
Matthew P. Reynolds1
1CIMMYT,
Mexico
2CIP, Peru
p.chavez@cgiar.org
2. Disease monitoring through remotely sensed data
Justification
Conventional visual field
monitoring: stress is detected
after significant damage has
occurred and yield reduced.
Pests and diseases pressure
increased due to climate
change.
3. Early disease monitoring through remotely sensed data
Approach and advantage
3) Space-borne assessment
2) Air-borne assessment
Reducing environmental risks and
footprint of farming by reducing use
of agrichemicals.
Target application (local and extend)
of pesticides.
Breeding purposes: identification of
resilient genotypes.
Non-destructive and large-scale
applicable approach.
1) Ground-based assessment
Stresses can be detected before symptoms
development
Boosting competitiveness through
more efficient practices (e.g.
improved management of inputs).
4. Early disease monitoring through hyperspectral remotely sensed data
SVI
NDVI
SRa
NPQI
PRI
WI1
WI2
Name
Normalized difference vegetation index
Simple ratio
Normalized pheophytinization index
Photochemical reflectance index
Water index 1
Water index 2
Calculation
(RNIR-Rred)/(RNIR+Rred)
R800/R680
(R415-R435)/(R415+R435)
(R531-R570)/(R531+R570)
R900/R970
R900/R950
Spectral vegetation indices (SVIs) calculated from the hyperspectral reflectance data of the
wheat genotypes under study did not show reliable results
5. Method 1 : Physiology and Statistics (Chavez et al., 2009)
100
∫
b
60
f(x) dx
a
40
20
Wavelenght (nm)
Blue
Green
Red
NIR
988
952
916
879
843
806
769
732
695
658
620
582
544
506
468
429
0
390
Reflectance (%)
Reflectance (%)
80
6. Method 2 : Physics, Physiology and Statistics (Chavez et al., 2010)
Fractal dimensions and formalism of the time series hyperspectral data
7. Decision tree for classification of fungal disease severity of wheat with hyperspectral time
series data
8. Three main diseases evaluated:
• Septoria tritici blotch → Toluca
• Tan spot → El Batán
• Spot blotch → Agua Fria
10. SEPTORIA TRITICI BLOTCH - Toluca
Genotype
AUDPC
Index3m
PhysioStat
Met1+Met2
PhyStat
AUDPC
Index3m
Met1
Met2
Rank
Rank
CROC_1/AE.SQUARROSA (205)//BORL95/3/2*MILAN I
131.07
I
0.47
25588
A
0.955
G
1
1
MURGA
139.71
H
0.49
H
30022
A
0.708
A
2
2
FINSI/METSO
218.93
G
0.54
G
23896
A
0.919
F
3
3
6B662
237.66
F
0.55
F
28227
A
0.858
D
4
4
GLENLEA
250.62
E
0.56
E
31360
A
0.748
B
5
5
CATBIRD
259.26
D
0.58
D
20345
A
1.024
H
6
6
ERIK
381.69
C
0.66
C
29257
A
0.87
E
7
8
ND-495
393.21
B
0.66
C
30350
A
0.83
C
8
7
HUIRIVIS #1
445.06
A
0.71
B
19691
A
1.048
I
9
9
KACHU #1
445.06
A
0.83
A
20036
A
1.187
J
10
10
2
r wi th AUDPC
SPOT BLOTCH - AguaFria
Genotype
1.00
0.95
AUDPC
0.35
Index3m
Met1+Met2
0.99
0.52
Phys i oSta t
Met1
PhySta t
Met2
AUDPC
Ra nk
Index3m
Ra nk
MURGA
58.23
J
0.62
CHIRYA.3
60.08
I
0.62
CATBIRD
124.69 F
0.65
CROC_1/AE.SQUARROSA (205)//BORL95/3/2*MILAN
61.73
H
0.66
KACHU #1
113.99 G
0.68
I
I
H
G
F
23817
21145
27001
25151
27988
DE
F
AB
CD
A
2.59
3.01
3.00
3.44
3.19
C
G
F
J
I
1
2
5
3
4
1
2
3
4
5
FINSI/METSO
133.54
E
0.69
E
22261
EF
2.67
D
6
6
HUIRIVIS #1
158.64
D
0.72
D
23880
DE
2.58
B
7
7
FRANCOLIN #1
231.69
C
0.73
C
26642
ABC
3.16
H
8
8
CIANO T 79
391.77
B
0.75
B
25325
BCD
2.91
E
9
9
SONALIKA
643.83
A
0.78
A
28389
A
2.42
A
10
r 2 wi th AUDPC
TAN SPOT - El Batan
Genotype
1.00
AUDPC
0.85
Index3m
Met1+Met2
CROC_1/AE.SQUARROSA (205)//BORL95/3/2*MILAN
115.23
J
0.66
MURGA
122.43
I
0.67
6B662
270.78
H
0.68
CATBIRD
292.39
G
0.69
HUIRIVIS #1
325.51
F
0.70
FINSI/METSO
328.40
E
0.71
KACHU #1
335.60
D
0.70
ERIK
371.60
C
0.71
GLENLEA
427.78
B
0.76
ND-495
488.27
A
0.81
r
2
wi th AUDPC
1.00
0.53
0.89
Phys i oSta t
Met1
H
G
F
E
D
C
D
C
B
A
34824
35934
28178
37926
30195
34093
31912
27127
23147
20738
0.77
PhySta t
Met2
C
B
G
A
F
D
E
H
I
J
2.75
2.57
3.29
2.27
2.67
3.17
2.73
2.73
3.03
3.09
0.37
10
0.96
0.46
AUDPC
Ra nk
E
H
A
I
G
B
F
F
D
C
Index3m
Ra nk
1
2
3
4
5
6
7
8
9
10
1
2
3
4
5
7
6
8
9
10
0.99
Same level of
sensitivity like
AUDPC to
discriminate
susceptible/
resistant
genotypes
11. Early disease detection through hyperspectral remotely sensed data :
Yellow rust pilot trial (isolines) – Chavez P., Yahyaoui A., Singh P.K. et al.
Pictures from CIMMYT Toluca 06/09/2012
Healthy
Diseased
Yellow Rust Detection:
Merging both methods: 94% matching with conventional visual monitoring.
Discrimination between susceptible and resistant cultivars: Resilience level among
genotypes.