A New Method for Spectroscopically Estimating Leaf Nitrogen Concentration

1
A New Method for Spectroscopically Estimating
Leaf Nitrogen Concentration
Zachary J. Bortolot
CEARS
2
Acknowledgements

• Dr. Randolph H. Wynne
• Virginia Tech Conservation Management Institute
• U. S. National Guard
• NCASI
• McIntire-Stennis Research Program
• NASA
• International Society of Arboriculture Research
  Trust

3
Why Foliar Nitrogen?

• Fertilization
• Nitrogen is a macronutrient. In order for a
  plant to grow and be productive, it must be able
  to extract large quantities of nitrogen from the
  soil.
• However, too much nitrogen can be detrimental to
  plant growth and productivity, and the excess may
  leach into streams and lakes, causing
  environmental degradation.
• Managers use foliar nitrogen measurements to
  determine whether nitrogen needs to be added, and
  if so how much.

4
Why Foliar Nitrogen? (contd)

• Forest Ecology
• The carbon to nitrogen ratio in leaf litter
  determines how quickly the litter will decompose,
  and the quantify and chemical form of the
  nutrients after release.
• This is important for monitoring the nitrogen and
  carbon cycles within a forest.
• Foliar nitrogen is also very useful in
  determining a trees photosynthetic rate.

5
Chemical Determination of Foliar Nitrogen

• In the past, most managers have determined foliar
  nitrogen by collecting leaf samples from a plant
  and then sending the sample to a laboratory for
  analysis.
• Although the results are very accurate, there are
  limitations
• Cost
• Timeliness
• Sample size
• Sample collection

6
Alternatives

• To solve these problems, several alternatives
  have been found.
• The Cardy ion meter
• Involves inserting electrodes into the petiole
  (stalk) of a leaf to measure the nitrate
  concentration.
• Inexpensive and quick, but the results are poor.
  Sample size and collection may still be a
  problem.

7
Alternatives (contd)

• Chlorophyll (a.k.a. SPAD) meters
• Transmit light through a leaf to measure the leaf
  chlorophyll concentration in terms of SPAD
  units.
• Nitrogen concentration is correlated to SPAD
  units because nitrogen is a major building block
  of chlorophyll.
• This technique is quick, cheap and works well for
  a single species on a single site at a single
  point in time.
• However, sample collection may be difficult and
  samples must be sent to a lab to obtain
  calibration coefficients.

8
The Use of Reflectance Spectroscopy

• Reflectance spectroscopy involves using the
  fraction of light reflected by an object to find
  out about the object.

9
Advantages of Reflectance Spectroscopy

• You do not need to physically be in contact with
  the leaves.
• It is possible to sample many leaves at the same
  time (may or may not be an advantage).
• If you use an airborne sensor, you may be able
  obtain a complete sample.
• If a ground based collection method is used, it
  is also easy to obtain many samples since the
  procedure is quite fast.
• The results are usually very accurate.

10
Disadvantages of Reflectance Spectroscopy

• Cost
• The current cost of collecting the type of
  airborne data needed for these measurements make
  it infeasible for most applications.
• Handheld units involve a high one-time cost
  (20,000) but future costs are quite small.
• In the past, the techniques used to determine
  foliar nitrogen yielded equations that were site
  and species dependent, or required that samples
  be dried and ground prior to measurement.

11
A Bit of Spectroscopic Theory

• Objects reflect different amounts of light at
  different wavelengths.
• A plot showing the amount of light reflected at
  different wavelengths is known as a spectral
  reflectance curve.

12
Spectral Reflectance Curves
13
Spectral Features

• The patterns seen in spectral reflectance curves
  are a result the chemistry and structure of the
  object being measured.
• For measuring foliar nitrogen levels, we are most
  interested in a type of feature known as
  vibrational absorptions.

14
Vibrational Absorptions

• Vibrational absorptions occur when light causes
  molecules or parts of molecules to vibrate.
• This happens at certain critical wavelengths.
  Which wavelengths these absorptions occur at is
  dependent on the structure on the molecule.

15
Example Water
16
Vibrational Absorptions of Organic Materials
A Modern materials B Ancient materials
1 OH 2 NH 3 CH 4 Organic pigments (not
vibrational)
17
Absorption vs. Reflectance

• An important concept in spectroscopy is that of
  reflectance versus absorption.
• Reflectance is the fraction of light at a given
  wavelength that strikes the surface and then
  returns from the surface at the same wavelength.

18
Absorption vs. Reflectance (contd)

• Reflectance is not an ideal means of measurement
  in spectroscopy.
• This is because the decrease in reflectance due
  to an absorption is not proportional to the
  amount of the absorbing material present.

19
Absorption vs. Reflectance (contd)

• Example
• You add an amount of a material that absorbs 50
  of the incident light to a bright surface
  (reflectance 0.80) and to a dull surface
  (reflectance 0.20).
• Case one
• The reflectance decreases from 0.80 to 0.40, a
  difference of 0.40.
• Case two
• The reflectance decreases from 0.20 to 0.10, a
  decrease of 0.10.
• Even though the same amount of material is
  present, the reflectance decreases different
  amounts.

20
Absorption vs. Reflectance (contd)

• To fix this problem, we convert from reflectance
  to absorption using Beers law
• A log10(1 / R)
• A Absorption
• R Reflectance
• Going back to the example
• Case 1 R 0.80 ? A 0.097 R 0.40 ? A
  0.40 Change in A 0.30
• Case 2 R 0.20 ? A 0.70 R 0.10 ? A 1
  Change in A 0.30

21
Foliar Nitrogen

• For measuring foliar nitrogen, there are a number
  of relevant absorption features. Of particular
  interest are two vibrational absorptions caused
  by proteins that are closely related to nitrogen
  concentration.
• Unfortunately, there are absorptions due to other
  molecules at the same wavelengths as these two
  absorptions.

22
The Absorption of Interest
The Absorptions Together with its Neighbors
The Absorptions on their Own
23
Leaf Water

• 40 80 of a green leafs mass is water.
• Unfortunately, the leaf water has an dramatic
  affect on the reflectance at the wavelengths
  where the absorptions we are looking at occur.

24
The Affect of Leaf Water
25
Methods for Determining Leaf Nitrogen The
Derivative Method

• In the past, several approaches have been used to
  predict leaf nitrogen concentration from
  spectroscopic data.
• The original and most common method is known as
  the derivative method.
• The first step is to convert leaf spectrum from
  reflectance to absorption. Generally the entire
  leaf spectrum is used, not just spectral regions
  known to contain useful features.
• Next you subtract the absorption at each
  wavelength at from the absorption at the previous
  wavelength. The result is known as a first
  difference spectrum.

26
Methods for Determining Leaf Nitrogen The
Derivative Method (contd)

• Stepwise multiple linear regression is then used
  to develop a model relating the values in the
  first difference spectrum to foliar nitrogen
  concentration.
• This method works because it is able to pick up
  the subtle changes in curvature caused by
  vibrational absorptions related to nitrogen.

27
Methods for Determining Leaf Nitrogen The
Derivative Method (contd)
Note the changes in curvature at different N
concentrations.
28
Methods for Determining Leaf Nitrogen The
Derivative Method (contd)

• However, this method has limitations
• A large number of wavelengths are typically used,
  necessitating a large number of training samples.
• The models often include absorptions that are
  indirectly related to nitrogen content. These
  absorptions may be related to nitrogen
  differently in different species or at different
  sites.
• The curvature changes dramatically with leaf
  water content differences.

29
Methods for Determining Leaf Nitrogen The
Derivative Method (contd)
30
Methods for Determining Leaf Nitrogen The
Derivative Method (contd)

• As a result of these limitations, the models
  produced using this technique tend to be site and
  species dependent.
• This is problematic, since each time you want to
  use this technique in a new area, ground samples
  must be collected and chemically analyzed.

31
Methods for Determining Leaf Nitrogen Continuum
Removal

• In 1999, Ray Kokaly and Roger Clark presented a
  new technique that they had developed that solved
  the problems of species and site dependence.
• Their new approach was based on a technique that
  is widely used in geological spectroscopy, known
  as continuum removal.

32
Methods for Determining Leaf Nitrogen Continuum
Removal (contd)

• Continuum removal is implemented by drawing a
  straight line on a spectral reflectance curve
  between the beginning and end of an absorption
  you are interested in.
• Next, the reflectance at each wavelength of
  interest in the spectral reflectance curve is
  divided by the value of the line at that
  wavelength.

33
Methods for Determining Leaf Nitrogen Continuum
Removal (contd)
34
Methods for Determining Leaf Nitrogen Continuum
Removal (contd)

• The values are normalized by either the area
  under the line or the maximum distance between
  the line and the curve.
• The main purpose of performing continuum removal
  is that it helps reduce the affects of
  absorptions and other spectral features that
  cause broad trends in the data.
• In this case, the broad trends that are removed
  are primarily due to leaf structure and leaf
  water.

35
Methods for Determining Leaf Nitrogen Continuum
Removal (contd)

• Finally, stepwise multiple linear regression is
  used to relate the resulting values to foliar
  nitrogen concentration.
• The results of this analysis showed that this
  technique worked very well, with one major
  exception.

36
Methods for Determining Leaf Nitrogen Continuum
Removal (contd)

• The Achilles heel of this approach is that it is
  unable to cope with green leaves, since the
  algorithm is unable to handle large amounts of
  leaf water.
• This is because the effect of leaf water on a
  leaf spectrum is not a linear function of
  wavelength.

37
Methods for Determining Leaf NitrogenThe Hybrid
Method

• After Kokaly and Clarks approach was made
  public, several researchers attempted to modify
  the approach to work with green leaves.
• The obvious solution (and the one suggested by
  Kokaly and Clark) was to
• Estimate the leaf water concentration
• Use this information to create the absorption
  spectrum of the leaf water present in the leaves
  being measured.
• Subtracted out the leaf water absorption
  spectrum.
• This sounds easy, but in reality it is very hard
  to do well. To the best of my knowledge, no one
  has gotten this approach to work.

38
Methods for Determining Leaf NitrogenThe Hybrid
Method (contd)

• In the summer of 2001 I came up with a different
  approach to solving the leaf water problem.
• It is based on combining the traditional
  derivative method with Kokaly and Clarks method.

39
Methods for Determining Leaf NitrogenThe Hybrid
Method (contd)

• The first step in implementing my approach is to
  reduce the spectral resolution to 10nm by
  averaging channels. This was done to reduce
  noise.
• Next, the values are converted from reflectance
  to absorption using Beers law.

40
Methods for Determining Leaf NitrogenThe Hybrid
Method (contd)

• Once the absorption values have been calculated,
  a transform is applied to all the channels in the
  area of the protein absorption features discussed
  earlier in this presentation.
• The transform simply involves subtracting the
  absorption in the middle channel from the average
  of its two immediate neighbors.

41
Methods for Determining Leaf NitrogenThe Hybrid
Method (contd)

• This is similar to applying a series of continuum
  removals to the data

42
Methods for Determining Leaf NitrogenThe Hybrid
Method (contd)

• Finally, stepwise multiple linear regression is
  used to relate the transformed values to foliar
  nitrogen concentration.
• The idea behind this method is that, like
  continuum removal, the transformation removes the
  effects of broad absorption features. However,
  like the derivative method, it is very localized
  and emphasizes curvature.
• My hope was that this method would allow changes
  in curvature directly related to nitrogen
  concentration to isolated.

43
Testing the Hybrid Method

• I tested the hybrid method by using data
  collected for the Accelerated Canopy Chemistry
  Program (ACCP).

44
Testing the Hybrid Method (contd)

• The dried samples collected from eastern forests
  were used as training data.
• The fresh samples of western tree species were
  used for testing.

45
Testing the Hybrid Method (contd)
46
Testing the Hybrid Method (contd)
Fresh
Dried
47
Testing the Hybrid Method (contd)
48
Noise

• The one major drawback with my method is that the
  technique is very sensitive to noise.
• This not an intractable problem at all, but some
  research will be required to find the best means
  of collecting low-noise spectra.

49
Conclusions and Future Research

• Reflectance spectroscopy presents a means of
  accurately measuring foliar nitrogen
  concentrations in the field.
• Unlike other methodologies, my hybrid technique
  is site and species independent and works with
  green leaves.

50
Conclusions and Future Research (contd)

• In the future I am planning to use this technique
  together with spatial statistics to create a
  nitrogen concentration map for a commercial pine
  plantation.
• I also plan to use this technique for additional
  chemical measurements, including leaf lignin and
  leaf energy content.
• Additional research could focus on algorithm
  improvement and data collection issues.