Vous êtes sur la page 1sur 44

California Agriculture

$36.6 Billion Industry (2007)


400 different commercial crops.

California produces about half of U.S.-grown fruits, nuts and


vegetables.

Almonds: 5th most


valuable crop
$2,127,375 x $1000

Cotton lint: 14th most


valuable crop
$599,352 x $1000
500,000 acres
in almond
production
(99% of
almonds U.S.),
It is the most
widely planted
tree crop in
California
CALIFORNIA:

• California is 2nd largest Cotton Producing State


• Cotton is One of Top 3 Water Demanding Crops
“ARID”
Eto = 9.30 in.
Picture 4
Water Shortage Drought impacts from this
year’s water shortages are
most severe in the west side of
the San Joaquin Valley.

CVP deliveries for that area


are at 10 percent of
contractors’ allocations this
year, following deliveries of 40
percent in 2008 and 50 percent
in 2007.
Picture 9

Can RS contribute to more


Data Source: California Department of Water Resources efficient irrigation methods?
Revised June 18, 2009
California’s
unemployment rate is at
its highest level since
World War II. The loss
of 68,900 jobs in May
pushed the seasonally
adjusted rate to 11.5%.
Paramount Farms Almond Orchards, Lost Hills,
Southern San Joaquin Valley, CA

6 replicates of 4 nitrogen and water treatments,


studied for the past 8 years
What biophysical properties do we want
to measure about plants, soil and water?
What leaf and canopy properties are
measured in optical spectrum?
Can we estimate how much plant
material is present and can we
estimate its “health”?

9
How to Retrieve
Information from Spectral
Spectral Measurements
Indexes
As the amount of the absorbing molecule increases, the
absorption feature to deepens and broadens.

Indexes are correlated to the amount of the absorbing material


present.

Narrow-band spectral indexes are simple calculations (e.g.:


ratios, differences) that accentuate and summarize spectral
information from contrasting bands where absorption occurs and
bands that don’t absorb

BioPhysical indexes use spectral features that capture


information that are used to interpret the condition of vegetation,
soil, water
Vegetation Indexes: ratios of bands sensitive
to a biochemical of interest to bands
insensitive to that biochemical

Natural
color
Color IR
Physiological
Indexes Simple Index (SR)
SR = Rred / RNIR

and Normalized
NDVI = Difference
0.813 Vegetation Index
(NDVI)
NDVI = RNIR − Rred
0.547 NDVI =
RNIR + Rred
NDVI =
0.139
Detects canopy
“greenness”
Separates green
plants and
soil/plant litter
Normalized Difference
Vegetation Index

The normalized difference 
vegetation index

NDVI = (RNIR ­ RR) / (RNIR + RR) 

Developed to reduce albedo 
differences due to topography 
and to normalize reflectance to 
adjust for calibration differences 
Chlorophyll
Concentration:
relationships to
reflectance
Red or blue
reflectance

and Carbon
Assimilation
chlorophyll concentration
0.35
Veget

Photosynthesis 0.3 IcePlant

(Assimilation) 0.25

Reflectance
Potential 0.2
Potential canopy

0.15
assimilation

0.1

0.05
carbon

0
400 600 800 1000 1200 1400

Amount of chlorophyll in Wavelength (nm )

canopy
Reflectance changes
during drying

ydration of canopy varies with soil water availability Jensen, 2007


Physiological Water
Indexes
Water Band Index
(WBI) R900
WBI =
R970

WBI = WBI = Ratio of reflectance


1.52
1.02
at 900 nm(yellow
band) and 970
nm liquid water
absorption
feature (cyan
band).
Infrared Water Index

NDMI
NDMI =
NDWI
(NIR 800 – MIR1700)/
(NIR 800+ MIR1700)
NDWI =
(NIR800 – MIR2300)
/ (NIR800 +
MIR2300)
NDMI NDWI
Low Low

High High
Physiological Index formula Details citation
Pigment indexes

R NIR Index of green vegetation cover. Various Tucker, 1979


SR, Simple Ratio RR wavelengths used, depending on sensor.(eg:
NIR=845nm, R=665nm)
Index of green vegetation cover. Various Tucker, 1979
NDVI, Normalized Difference R NIR − RR
wavelengths used, depending on sensor.(eg:
Vegetation Index R NIR + RR
NIR=845nm, R=665nm)
R750 − R705
mNDVI, modified NDVI leaf chlorophyll content Fuentes et al. (2001)
R750 + R705
599
Summed green reflectance ∑R
n =500
n
Index of green vegetation cover. Fuentes et al. (2001)
Xanthophyll response to photosynthetic
PRI, Photochemical R531 − R570
efficiency. Also sensitive to Rahman et al. (2001)
Reflectance Index R531 + R570
carotenoid/chlorophyll ratio
R600−699
Red/Green ratio R500−599 anthocyanins/chlorophyll Fuentes et al. (2001)

NPCI, Normalized Pigments R680 − R430 Peñuelas et al.


total pigments/chlorophyll
Chlorophyll Ratio Index R680 + R430 (1995)
SRPI, Simple Ratio Pigment R430
R680 carotenoid/chlorophyll a content Zarco-Tejada (1998)
Index
NPQI, Normalized R415 − R435 chlorophyll degradation, detects stress at
R415 + R435
Zarco-Tejada (1998)
Phaeophytinization Index early states
SIPI, Structure Intensive R800 − R445
carotenoid/chlorophyll a concentrations Zarco-Tejada (1998)
Pigment Index R800 − R680
R695
PI1, Pigment Index 1 plant stress status Zarco-Tejada (1998)
R420
R695 Zarco-Tejada
PI2, Pigment Index 2 plant stress status
R760 (1998)
R440 vegetation health, based on Lichtenthaler et al.
PI3, Pigment Index 3 R690 chlorophyll fluorescence ratios (1996)
R440 vegetation health, based on Lichtenthaler et al.
PI4, Pigment Index 4 R740 chlorophyll fluorescence ratios (1996)

Physiological Index formula Details citation


Water indexes
NDWI, Normalized R860 − R1240
R860 + R1240 leaf water content Gao (1996)
Difference Water Index
R900 Peñuelas et al.
WBI, Water Band Index leaf water content
R970 (1997)
Foliar chemistry indexes
log 1680 
R
NDNI, Normalized  R1510  Serrano et al.
foliar nitrogen concentration
Difference Nitrogen Index log 1
 R1680 R1510 

 (2002)
log 1680 
R
NDLI, Normalized  R1754  Serrano et al.
log 1  foliar lignin concentration
Difference Lignin Index  R1680 R1754 
 (2002)
based upon cellulose & lignin
CAI, Cellulose Absorption 0.5 * (R2020+R2220)- Nagler et al.
R2100 absorption features, used to
Index (2000)
discriminate plant litter from soils
Burn Ratio and Normalized Burn Ratio
(NDBR)
Natural
color

NDBR = NBRpre – NBRpost


Larger Numbers = burns

NDBR
Reflectance of Different Soils in Red and NIR wavelengths
Picture 3

From review, J.L. Hatfield, 2004


Red and Near-Infrared “Soil Line”

Example of Soil line extracted from


NIR vs RED data
Continuum Removal

The continuum removal method provides


a normalized way to estimate absorption
depth that can provide a more quantified
estimate of the concentration of the
material.

Both indexes and continuum removal


standardize reflectance spectra to allow
for comparison of absorption features.
Continuum
Shrink wraps a curve over
the measured spectrum Removal
(creates a “convex hull”). continuum
970 nm
Absorption
Can be approximated by feature

Reflectance (%)
linear segments spanning 1200 nm
Absorption
absorptions. 470 nm
feature 2300 nm
Absorption
Absorption feature
feature

The line connecting the 670 nm Absorption feature 2100 nm Absorption feature
500 1000 1500 2400 2400
spectrum without the Wavelength (nm)
absorption features is the
continuum.
Continuum
R Removal
ENVI Algorithm: r = R
'

where R is the reflectance


and RC is the continuum
reflectance for each λ across
the feature.
Rescales values to 1.0 where
the continuum and measured
reflectance are equal;
values <1 means reflectance
is less than continuum (i.e.,
absorption ).
Useful products include
•Feature depth
•Feature area
Continuum
Removal
Common absorption features:
Liquid water absorptions in vegetation: 970
nm, 1240 nm
Lignin/Cellulose absorptions in dry veg.: 2100
nm
Carbonate absorption in soil: 2300 nm
Clay absorptions in soil: 2200 nm
Mineral absorptions for different chemical
groups like sulphates, hydroxides, carbonates,
Continuum Removal of Soil Reflectance Features

Relative +
λi +
band ++++
spectrum with depth
++
continuum removed +
++
σ

water overtones

spectrum with continuum λ


0

clay carbonate
band band Fitting
Soil Reflectance Continuum for
Spectrum and Water Absorption
Continuum Removal to Soil Spectrum
Spectrum Whiting, 2004
Full Spectrum Methods
% Reflectance * 10000

0.5 1.00 1.50 2.00 2.50


Wavelength, µm
Spectral Angle Mapper

Each vector represents a a


t
spectrum. The smaller

Band j
the angle between two ta
b
spectra, the more likely
they are belong to the
same class. tb

If the classifier is trained


with spectrum- t; Band i
spectrum-a is more
likely to be assigned to  n 
class-t than spectrum-b −1  ∑ ti ri 
is because ta < tb θ = cos  ni=1 n 
 ∑t 2 ∑ r2 
 i=1 i i=1 i 
The Delta The Species The Method Step 1 Step 2 Step 3 Step 4 Step 5 Step 6

Delineate Emergent and


3 Compute Spectral
Levee Vegetation with Angle EMERGENT
SWIR Reflectance
Spectral Vector

4 Delineate Tule/Cattail
Delineate low reflectance using NIR band
1 vegetation using NIR average – 25% TULE/CATTAIL
band average – 40%

UNHEALTHY
Delineate unhealthy WATER HYACINTH
5 Profusely
Spectra Reference l Vector flowering Water
Hyacinth from riparian
Using water absorption RIPARIAN

Delineate healthy Water


Partial Spectral
2 Hyacinth using Spectral Endmember VEGETATIVE
WATER HYACINTH
Mixture Analysis
Library
HEALTHY
Delineate Water Hyacinth WATER HYACINTH
3 similar in spectral response
to pennywort using water
absorption and green index PENNYWORT
%

Wavelength, µm
The Delta The Species The Method Step 1 Step 2 Step 3 Step 4 Step 5 Step 6

Emergent Aquatic Species

Rhode Island

Pennywort Pennywort

Water Hyacinth Water Hyacinth


MNF Transform
What is Principal Component
U2
x2 Analysis?
A linear transformation

1
that identifies orthogonal

U
axes by computing the
X1
eigen decomposition of
the sample covariance
of the data
∑ = E [ xx T
] = U T
∑U = Λ
MNF?
Noise-adjusted PCA
Two-stage linear transformation
Noise is first decorrelated and rescaled
33
The Delta The Species The Method Step 1 Step 2 Step 3 Step 4 Step 5 Step 6

SAV Species Discrimination

Sago
Brazilian waterweed
Curly leaf pondweed
Cabomba
% Reflectance

Milfoil
Coon’s tail
Broad leaf Sago
Common waterweed
Water

Wavelength, nm

10 samples per plant per species


The Delta The Species The Method Step 1 Step 2 Step 3 Step 4 Step 5 Step 6

Brazilian waterweed
z

Cabomba

Common Broad leaf Sago


waterweed Milfoil
y x

water
Coon’s tail Curly leaf
pondweed

Sago

SAV Species Discrimination


San Joaquin Valley, CA
Near Lemoore, Hanford

MNF 3,2,1 (RGB)


July 8, 2002
When Pixels are Mixtures
Linear unmixing of surface materials within pixels
Spectral Mixture Analysis

Partial Spectral
Endmember Decomposes a
Library
mixed pixel into a
collection of
constituent
spectra, or
endmembers, and
35 % red
a set of fractional
17% green abundances
13% dark blue
35% light blue indicating the
proportions of
each endmember
Linear Spectral Unmixing
Two parts to the algorithm:
N

∑F =F +F
i 1 2 +...+FN =1
i=1

Fi = sum fractions of endmembers in


pixel i = 1
DN =
Rλ λ=FF
1DN
1Rλλ1 F2 DN
,1 +
+ F 2R
λ +
,2λ2 + FFNNR
...+ DNλNλ+E
,N Eλλ
+

Rλ (or DN) = the pixel reflectance for band λ


Rλ,I (or DN) = the reflectance for band λ of
endmember 1
Near Lemoore, Hanford

SMA GV
July 8, 2002
SMA NPV 07082002
SMA Soil Fraction 07082002
San Joaquin Valley, CA
Near Lemoore, Hanford

Inner Coast Range

SMA soil, GV, NPV, 07 08 20002


Pistachio

Garlic Tomato

Wheat

Vous aimerez peut-être aussi