The following three sections summarize the spectral basis for retrieval of the
three most important water quality characteristics—clarity, chlorophyll, and
CDOM—from remote sensing imagery.
2.2.4 Water Clarity Variables
Three water clarity variables discussed here include Secchi depth (SD), turbidity,
and TSS.
Water clarity, whether measured as light scattering in laboratory turbidimeters,
as the in situ depth of disappearance of a white disk (SD), or as the slope of the
logarithm of light attenuation with depth (K d ) in a water body, provides critically
important information to both users of water bodies and to water resource managers. Fortunately, because of their close relationship to both IOPs and AOPs of
water, clarity parameters are well suited to measurement by ORS.
In part because of the widespread availability of calibration data from citizen and
agency monitoring programs, SD has been the subject of many ORS studies (see
Sect. 4 for details). Retrieval of SD from satellite imagery also is facilitated by the
fact that the broad Landsat bands are suitable for SD retrieval. Numerous studies
have yielded good relationships for SD that involve bands 1 and 3 in two-term
equations like ln(SD) ¼ a(TM1/TM3) + bTM1 + c [31], where ln(SD) is the natural
logarithm of Secchi depth; a, b, and c are regression coefficients; and TM1 and
TM3 are reflectance values for thematic mapper bands 1 and 3. As SD decreases,
reflectance in the red band (TM3) increases. The blue band (TM1) tends to
normalize brightness in the red and improves algorithm performance. R
2 values
for such equations are in the range 0.71–0.96 for lakes in Minnesota [32]; others
[33] reported similar ranges of fit. Olmanson et al. [1] found that MERIS and
Landsat imagery worked equally well for SD, but the coarser spatial resolution of
MERIS allowed assessment of only about 8 % of the Minnesota lakes accessed by
Landsat.
Models for turbidity and TSS in optically complex waters, where phytoplankton,
CDOM, and SS min all may affect IOP features, should avoid the absorption
characteristics of chlorophyll in the red and CDOM in the blue region and use the
scattering peak at ~705 nm or band combinations in the NIR or green regions
(where plant pigments have minimal absorption). For example, Gitelson et al. [34]
found that a difference ratio algorithm (R560 À R520)/(R560 + R520) was highly
correlated with TSS in lakes and rivers with TSS values < 66 mg/L. Phytoplankton
absorption is at a minimum near 560 nm, but reflectance at this wavelength is
sensitive to TSS; in contrast, reflectance at 520 nm is relatively insensitive to
changes in TSS [35].
Numerous studies have shown the usefulness of NIR bands for turbidity and TSS
(for reviews, see [10, 35]). The scattering peak at ~700 nm was found to be strongly
correlated with TSS by many studies (e.g., [36–38]), and Senay et al. [39] reported a
good relationship for turbidity. The difference in reflectance at 710 and 740 nm was
Remote Sensing for Regional Lake Water Quality Assessment: Capabilities and. . .
119
three most important water quality characteristics—clarity, chlorophyll, and
CDOM—from remote sensing imagery.
2.2.4 Water Clarity Variables
Three water clarity variables discussed here include Secchi depth (SD), turbidity,
and TSS.
Water clarity, whether measured as light scattering in laboratory turbidimeters,
as the in situ depth of disappearance of a white disk (SD), or as the slope of the
logarithm of light attenuation with depth (K d ) in a water body, provides critically
important information to both users of water bodies and to water resource managers. Fortunately, because of their close relationship to both IOPs and AOPs of
water, clarity parameters are well suited to measurement by ORS.
In part because of the widespread availability of calibration data from citizen and
agency monitoring programs, SD has been the subject of many ORS studies (see
Sect. 4 for details). Retrieval of SD from satellite imagery also is facilitated by the
fact that the broad Landsat bands are suitable for SD retrieval. Numerous studies
have yielded good relationships for SD that involve bands 1 and 3 in two-term
equations like ln(SD) ¼ a(TM1/TM3) + bTM1 + c [31], where ln(SD) is the natural
logarithm of Secchi depth; a, b, and c are regression coefficients; and TM1 and
TM3 are reflectance values for thematic mapper bands 1 and 3. As SD decreases,
reflectance in the red band (TM3) increases. The blue band (TM1) tends to
normalize brightness in the red and improves algorithm performance. R
2 values
for such equations are in the range 0.71–0.96 for lakes in Minnesota [32]; others
[33] reported similar ranges of fit. Olmanson et al. [1] found that MERIS and
Landsat imagery worked equally well for SD, but the coarser spatial resolution of
MERIS allowed assessment of only about 8 % of the Minnesota lakes accessed by
Landsat.
Models for turbidity and TSS in optically complex waters, where phytoplankton,
CDOM, and SS min all may affect IOP features, should avoid the absorption
characteristics of chlorophyll in the red and CDOM in the blue region and use the
scattering peak at ~705 nm or band combinations in the NIR or green regions
(where plant pigments have minimal absorption). For example, Gitelson et al. [34]
found that a difference ratio algorithm (R560 À R520)/(R560 + R520) was highly
correlated with TSS in lakes and rivers with TSS values < 66 mg/L. Phytoplankton
absorption is at a minimum near 560 nm, but reflectance at this wavelength is
sensitive to TSS; in contrast, reflectance at 520 nm is relatively insensitive to
changes in TSS [35].
Numerous studies have shown the usefulness of NIR bands for turbidity and TSS
(for reviews, see [10, 35]). The scattering peak at ~700 nm was found to be strongly
correlated with TSS by many studies (e.g., [36–38]), and Senay et al. [39] reported a
good relationship for turbidity. The difference in reflectance at 710 and 740 nm was
Remote Sensing for Regional Lake Water Quality Assessment: Capabilities and. . .
119