Analysis of water and chlorophyll features in cotton agriculture

1
Analysis of water and chlorophyll features in
cotton agriculture

• Jonathan Greenberg, George Scheer,
• Michael Whiting, and Susan Ustin

2
Cotton Production in California

• Cotton is a 1 billion a year industry in CA.
• 87 of water in CA is used for irrigation.
• Almost all water in CA is developed.
• Cotton is one of the top three largest water
  consuming crops in California.
• Over-watering can cause decreased bole
  production, increased fungal and insect damage,
  and increased production costs.
• Under-watering can cause decreased bole
  production and increased insect damage.

3
Estimating Water Stress Through Pressure Bomb
Measurements
Growers base irrigation decisions on a
field-by-field basis using point sampled leaf
water potential measurements.
4
Crop Condition Using Multispectral Imaging (NDVI)

• Images provide greater spatial discrimination.
• NDVI can predict vegetation status at LAIs lt 3.0.
• NDVI is not a direct measure of water content.

5
Hyperspectral Signature of Cotton
Water absorption features indicates canopy water
content
Water stressed cotton
Well irrigated cotton
Red edge position indicates canopy structure and
chlorophyll content
Cotton Spectra
6
Questions

• Is there significant within-field and
  between-date variation of canopy condition that
  would improve irrigation management?
• How well can we estimate leaf water potential (an
  indicator of leaf water concentration) from
  canopy reflectance data?
• Multispectral NDVI
• Hyperspectral Continuum Removal Ratio (899-1060
  and 1061-1265), Red Edge Position, NDWI, WI.

7
Methods

• For each point, we recorded
• Geographical position using GPS.
• Spectrum using an ASD FR spectrometer (3-9
  spectra per sample).
• Leaf Water Potential using a pressure bomb.

8
Methods

• For each reflectance spectrum, the following
  analyses were performed
• NDVI
• CR 899-1060 and 1061-1265
• Red Edge Position
• NDWI (Gao, 1996)
• WI (Peñuelas et al., 1993)
• We used a correlation matrix and Fishers r to Z
  to determine correlation and significance of leaf
  water potential and the indices.

9
Methods Continuum Removal Ratio
The CR Ratio is given by the formula Area under
absorption feature Area under continuum
10
Methods Red Edge Position
11
Results Within-Field Variation
12
Results Continuum Removal vs. Time
13
Results Continuum Removal vs. Time
14
Results Red Edge Position vs. Time
15
Results Pressure Bomb Correlations
16
Results Summary of LWP vs. Indices
17
Conclusions

• There is significant within field variation in
  crop condition.
• Crop condition indices change in a predictable
  manner over time.
• Narrow-band hyperspectral based indices (CR,
  NDWI, WI) are superior to traditional
  multispectral indices for predicting leaf water
  potential.

• The CR at 950 and 1150 nm. may be sensitive to
  cotton development stage.
• Noise in correlations may be due to pressure bomb
  error more than errors in the estimations.

18
Upcoming Research2001 Field Season

• Several AVIRIS flights over the Sheely Farms (we
  hope!)
• Weekly flights of the OKSI VNIR hyperspectral
  sensor (pending).
• Weekly leaf water potential measurements with GPS
  by farm management.

• Biweekly/monthly liquid water content
  measurements.
• Biweekly/monthly chlorophyll content.
• Biweekly/monthly ASD FR spectrometer to support
  LWT and chlorophyll measurements.

19
Upcoming Research2001 Field Season

• Geocorrect 2000 AVIRIS scenes with supporting
  IKONOS imagery for ground truthing of pressure
  bomb data.
• Application of radiative transfer modeling
  (PROSPECT and SAIL) in the estimation of water
  content and canopy chemistry.

• Application of improved techniques to determine
  red edge position using the inverse Gaussian.
• Development of software to determine canopy water
  content for immediate (lt 48 hrs.) use by the
  grower.

20
(No Transcript)