1. Characterizing non-pigment canopy biochemistry from imaging spectrometer data for studying ecosystem processes
Gregory P. Asner,
Mary E. Martin,
Scott V. Ollinger, &
Carol A. Wessman
Raymond F. Kokaly
U.S. Geological Survey
Ohio State Univ.

2. General Research Goals
To quantify non-pigment biochemistry at the leaf and canopy levels using reflectance spectroscopy
Apply knowledge of leaf spectroscopy for detailed characterization of vegetation
Quantify canopy biochemistry to increase understanding of fine-scale variations in ecosystem processes over large areas

4. Types of Remote Sensing
Electromagnetic
Wavelength (mm)
Passive
Systems
Active
Ultraviolet (UV)
0.4 – 0.7
Visible
0.7 – 1.4
1.4 – 3.0
Near-Infrared
Shortwave Infrared
3.0 – 1,000
Infrared
1,000-
300,000
Microwave
LIDAR
Airborne Cameras
Radiometers
Spectrometers
Radar (SAR)

5. Imaging Spectrometers
EO-1
Spacecraft
Satellite-based
Imaging Spectroscopy Data:
EO-1 Hyperion
Airborne
HyMap
Probe
CASI
AVIRIS
NASA ER-2

6. Trends in quantifying non-pigment leaf biochemistry
Greater understanding of lab, field and remote sensing level spectra in relation to biochemical composition
Increased use of full spectrum or specific biochemical absorption features to characterize vegetation
Greater application of imaging spectroscopy to understand ecosystem processes

7. Understanding vegetation spectra in relation to biochemical composition
Water
Nitrogen (in chlorophylls and proteins)
Lignin and Cellulose
Non-Photosynthetic Vegetation

8. Kokaly, REMOTE SENS. ENVIRON. 84:437-456 (2003)

9. Water
Water content/potential is a limiting factor for plant transpiration (photosynthesis and carbon gain) and thus growth and development

11. Kokaly and Clark, REMOTE SENS. ENVIRON. 67:267–287 (1999)

12. Spectroscopic Estimates of Leaf/Canopy Water
Leaf/Canopy Liquid Water Thickness
Gao & Goetz, 1994
Canopy Equivalent Water Thickness
Green et al., 1991 & Roberts et al. 1997
Canopy Relative Water Content
Serrano et al. 2000
Foliar Water Potential
Stimson et al. 2005

13. Serrano et al., REMOTE SENS. ENVIRON. 74:570–581 (2000)

14. NDWI = (R860 - R1240)
(R860 + R1240)
NDII = (R819 - R1649)
(R819 + R1649)

15. Foliar Water Potential (MPa)
Stimson et al., Remote Sensing of Environment 96 (2005) 108– 118

16. Leaf (Lab) Canopy (AVIRIS) 980 1200 nm nm 850 1350 1850
Clark et al., Remote Sensing of Environment 96 (2005) 375 – 398

17. Nitrogen By dry weight, very low abundance in leaves, only 0.5 to 4%
Present in important biochemical constituents of plants
Chlorophyll
Protein
Linked by field studies to rates of ecosystem functioning (carbon fixation) and widely used in ecosystem models

18. Nitrogen in Proteins
Proteins composed of amino acids ranging from a 100 to 100,000 in number

19. RuBisCO
Ribulose-1,5-bisphosphate carboxylase/oxygenase, catalyzes the first major step of carbon fixation
RuBisCO is the most abundant protein in leaves (maybe the most abundant on Earth).

20. Kokaly, REMOTE SENS. ENVIRON. 75:153–161 (2001)

21. Kokaly, REMOTE SENS. ENVIRON. 75:153–161 (2001)

22. Kokaly and Clark, REMOTE SENS. ENVIRON. 67:267–287 (1999)

23. Nitrogen in Chlorophyll

24. Mutanga et al., Remote Sensing of Environment 96 (2005) 108– 118

25. Biochemical Components of Plants
Cellulose
Cellulose monomers (β-glucose) are linked together through 1→4 glycosidic bonds
Cellulose is a common material in plant cell walls
Cellulose is the most abundant form of living terrestrial biomass (R.L. Crawford 1981).

26. Biochemical Components of Plants
Lignin
10% to 40% by dry weight
Lignin is a large macromolecule with molecular mass in excess of 10,000 amu.
Structural component of plant cell walls.
It is hydrophobic and aromatic in nature.
Lignin concentration in litter has strong influence on ecosystem processes

27. Lignin
Cellulose
Since their chemical structures are different, we expect different spectra and that
Is indeed what we see. Though their spectral features overlap which complicates
Things when looking at the spectrum of an intact leaf or plant. What we see is
That the lignin absorption in the 2.1 micron regoin is at slightly longer wavleengths
And the 2.3 micron feature in cellulose is a doublet (two dips) compared to lignin’s
Single absorption feature.

28. Kokaly et al., Remote Sensing of Environment, in press
= 3.8
Grass
Cellulose:Lignin
Pine
Cellulose:Lignin
= 1.5
What we found is that pine needles have relatively higher lignin-cellose ratios compared
To other dry vegetation in the area – mostly dry grass. And we see this shift in the pine
Spectra to loner wavelength for the pine sample and the doublet nature of the 2.3 micron
Absorption for the grass spectrum. This is from lab data.
Kokaly et al., Remote Sensing of Environment, in press

29. Apply knowledge of leaf spectroscopy for detailed characterization of vegetation

30. Cerro Grande Fire Los Alamos, New Mexico
The case study we will look at for these materials is a conifer forest that burned
In New Mexico, near Los Alamos, which is the home of atomic research. Obviously there
Was a lot of concern about the fire’s effect on the area (it’s effect on materials that might
Have been say nuclear material that may have been misplaced but mostly concern about
The post-fire impacts from erosion and landslides on the lab’s facilities.

31. Here is what a burned conifer forest looks like
Here is what a burned conifer forest looks like. We can see a lot of scorched trees that
Have been heated by the fire. We were interested in detecting these trees and quantifying
What percentage of the surface was affected in that way. The heat pulse actually penetrates
Further than this orange ring, the trees just a ways inward suffer from water loss but not
Chlorophyll degredation. So we needed to examine dry pine needles in relation to other
Dry vegetation to see if we could identify it.

32. Kokaly et al., Remote Sensing of Environment, in press
=3.8
Grass
Cellulose:lignin
Pine
Cellulose:lignin
=1.5
What we found is that pine needles have relatively higher lignin-cellose ratios compared
To other dry vegetation in the area – mostly dry grass. And we see this shift in the pine
Spectra to loner wavelength for the pine sample and the doublet nature of the 2.3 micron
Absorption for the grass spectrum. This is from lab data.
Kokaly et al., Remote Sensing of Environment, in press

33. Kokaly et al., Remote Sensing of Environment, in press
AVIRIS
Dry Grass
AVIRIS
Dry Pine
In the AVIRIS data we see these same spectral differences. The spectra from pixels containing dry conifers have a broadened 2.1 micron feature and a stronger central absorption in the 2.3 micron feature when compared to pixels from areas where dry straw matting was placed on the surface. Thus, the effects of chemistry on leaf level spectra are also present in AVIRIS pixels.
This highlights the real power of imaging spectroscopy, that is the ability to link spectral differences to chemistry and thus discriminate materials or identify substances. Having a highly capable and reliable sensor is crucial to such research.
Kokaly et al., Remote Sensing of Environment, in press

34. Results: AVIRIS Maps
Green
Vegetation
Mineral/Ash
Mineral-1mm
Mineral-2mm
Dry Conifer
Dry & Green
Conifer
Straw matting
Ash/Charcoal
& Green grass
Again by comparing the pixels to the spectra in the library we can look for these materials. We see
Quite a diversity in materials in the burned area.
Kokaly et al., Remote Sensing of Environment, in press

35. Kokaly et al., Remote Sensing of Environment, in press
Green
Vegetation
Mineral/Ash
Mineral-1mm
Mineral-2mm
Dry Conifer
Dry & Green
Conifer
Straw matting
Ash/Charcoal
& Green grass
Looking in more detail at the severely burned area in the center of the image. We can see
The dry trees
Kokaly et al., Remote Sensing of Environment, in press

36. Kokaly et al., Remote Sensing of Environment, in press
This image shows the zoning that occurs at the boundary of severely burned areas. Why is this
Important, well the dry needles will fall soon and contribute to nutrient cycling and these
Zones are different from the completely burned areas here and the relatively unaffected
Trees on the other side. A recent study by a USGS colleague has also found that these
Moderately affected trees are important habitat for avian species following fire.
30m
High Burn Severity
Kokaly et al., Remote Sensing of Environment, in press

37. Analysis of spectral shape (full spectrum and/or absorption features) for vegetation characterization
Multiple Endmember Spectral Mixture Models:
Roberts, et al. (1998) Remote Sens. Environ. 65:267–279
Monte Carlo spectral unmixing model:
Asner and Heidebrecht (2002) Int. J. Remote sensing 23: 3939–3958
Tetracorder, absorption feature analysis:
Clark, et al. (2003) . J. Geophys. Research 108 (E12): 5-1 to 5-44

38. Understanding ecosystem processes by quantifying canopy biochemistry
Wessman, et al. (1988) Nature 333: 154–156
Asner and Heidebrecht (2003) IEEE Trans. on Geoscience & Remote Sensing 41:
Smith, et al. (2002) Ecological Applications 12: 1286–1302
Ollinger and Smith, (2005) Ecosystems 8: 760–778
Asner et al. (2005), Remote Sensing of Environment 96: 497–508

39. Asner and Heidebrecht, Int. J. remote sensing, 2002, vol. 23, no

40. Non-Photosynthetic Vegetation and Shrubland Ecosystems
Asner and Heidebrecht, IEEE Trans. on Geoscience & Remote Sensing (2003), 41:

41. Kokaly, REMOTE SENS. ENVIRON. 75:153–161 (2001)
In lab spectra we can see that high N leaves have stronger absorptions at the
Protein wavelengths
Kokaly, REMOTE SENS. ENVIRON. 75:153–161 (2001)

42. Mutanga et al., Remote Sensing of Environment 96 (2005) 108– 118

43. V
Asner and Vitousek,
Proc. Nat. Acad. Sci.,
Vol 102 , 2005.

44. Smith et al., Ecological Applications, 12(5), 2002, 1286–1302
r2-=0.71
See=0.19
Smith et al., Ecological Applications, 12(5), 2002, 1286–1302

45. Smith et al., Ecological Applications, 12(5), 2002, 1286–1302

46. Ollinger and Smith, Ecosystems (2005) 8: 760–778
Alsoworked withwood production
Ollinger and Smith,
Ecosystems (2005) 8: 760–778

47. Ollinger and Smith, Ecosystems (2005) 8: 760–778

48. Ollinger and Smith,
Ecosystems (2005) 8: 760–778

49. Remaining Challenges/ Future Directions
Overlapping of biochemical absorption features
Leaf water masking effect
Understanding protein/lignin/cellulose changes which all affect the 2.1 and 2.3 mm features
Quantification of other biochemicals (starch, tannin, hemi-cellulose, phosphorous)
Independent-site/Multi-site validation
Atmospheric correction needs improvement to lessen impact of residual features