Environmental Engineering Reference
In-Depth Information
20.4 Spectroscopic Retrieval of Leaf and Canopy Liquid
Water Contents
Olsen ( 1967 ) published the first spectral measurements of leaves drying in the
laboratory, and the data look similar to Fig. 20.2 . Typically, dehydration initially
increases near-infrared reflectance, indicating changes in the cellular structure of a
leaf (Aldakheel and Danson 1997 ). Relative to the near-infrared wavelengths
(Fig. 20.1a ), leaf dehydration causes increased reflectance at longer near-infrared
and shortwave-infrared wavelengths (Fig. 20.2 ). The largest percent changes in
reflectance occur at wavelengths around 1495 and 1950 nm; however, water vapor
strongly absorbs near these wavelengths, making the atmosphere opaque to solar
radiation (Green et al. 2006 ). Thus, the 1495- and 1950-nm liquid water absorption
features can be used only to estimate LWC under an artificial light source.
At wavelengths greater than 2000 nm, there are strong biochemical absorption
features from lignin and cellulose (Daughtry and Hunt 2008 ), so the reflectance
spectrum of dehydrated leaves becomes more complex. Therefore, the three regions
that have potential for routinely monitoring changes in LWC and CWC are the local
absorption maxima around 970 and 1240 nm (Fig. 20.1a ) and the local absorption
minimum around 1650 nm (Fig. 20.1b ). Below, spectroscopic retrieval of LWC and
CWC are discussed. In the next section, remote-sensing indices using these three
regions are described.
Allen et al. ( 1969 ) showed that LWC could be estimated by comparing the leaf
reflectance spectrum with the expected spectrum determined from different equiva-
lent thicknesses of pure liquid water; however, calculators or personal computers were
not available so the comparisons were qualitative. If the reflectances from leaves and
canopies are assumed to follow the Beer-Lambert law (Downing et al. 1993 ; Gao and
Goetz 1990 , 1995 ; Roberts et al. 1998 , and Sims and Gamon 2003 ), then:
lnðR λ Þ¼α λ l
(20.1)
where R λ is the reflectance at wavelength
λ
, ln is the natural logarithm operation,
α λ
is the absorption coefficient at wavelength
, and l is the optical path length of liquid
water. The path length l , that is, the LWC or CWC, is determined from the slope of a
linear regression between ln( R λ ) and
λ
α λ over some wavelength region. The regres-
sion intercept is supposed to account for fixed effects from errors in the assumptions
(Roberts et al. 2004 ; Sims and Gamon 2003 ). Gao and Goetz ( 1990 , 1995 ) use
predicted CWC to estimate the amount of water vapor for atmospheric correction of
imaging spectrometer data (Gao et al. 2009 ).
Gao and Goetz ( 1995 ) showed that the retrieved CWC at 1,000 nm was greater
than the retrieved CWC at 1600 nm. Using the two spectra in Fig. 20.2 , regressions
of ln( R λ ) versus
α λ around 970, 1240, and 1650 nm were made according to
Eq. 20.1 . The measured LWC for the fresh leaf and dehydrated leaf were 0.14
and 0.04 kg m 2 , respectively. The retrieved LWC for the fresh leaf were 0.70,
0.59, and 0.45 kg m 2 for 970, 1240, and 1650 nm, respectively, whereas the
retrieved LWC for the dehydrated leaf were 0.24, 0.19, and 0.16 kg m 2 ,
Search WWH ::




Custom Search