Olmanson et al. [42] using airborne hyperspectral imagery on waters of the Minnesota, Mississippi, and St. Croix Rivers. For chlorophyll a, the ratio of reflectance at
705 to 670 nm yielded r
2 values of 0.75–0.93. Of the recently or currently available
and forthcoming satellite sensors, only MERIS and the Sentinel-2 and Sentinel-3
satellites have appropriately narrow red-edge bands.
Despite the above comments, it must be noted that many studies have reported
strong empirical relationships between the broad bands of Landsat sensors and
chlorophyll a (e.g., [1]); typical predictive equations involve the ratio of TM or
ETM+ bands 1 and 3. Absorption by chlorophyll a is strong in bands 1 (450–
520 nm) and 3 (630–690 nm). Nonetheless, increased scattering by phytoplankton
cells counteracts some of the absorption effects and leads to increased reflectance
with increasing chlorophyll a levels in band 1 and even larger increases in band 3. If
it is known that the optical properties of the water bodies being studied are
dominated by phytoplankton, the use of these empirical relationships may be
considered acceptable. However, for regional assessments where specific water
quality characteristics are not known, lakes with high CDOM and/or SS min may
be misclassified.
As an example, when we used Landsat 8 imagery to estimate chlorophyll a and
SD in lakes of northeastern Minnesota for August 31, 2013, we found strong
relationships for both variables (r
2
¼ 0.70 and 0.77; RMSE 0.758 and 0.406,
respectively). Calibration data (Æ3 days) for the images are from the Minnesota
Pollution Control Agency (for chlorophyll a, n ¼ 99; for SD, n ¼ 258) [51]. For
most Minnesota lakes, the results are believed accurate because phytoplankton
dominates their optical properties. When we used the same models for the
St. Louis River Estuary (SLRE), where optical properties of the waters are dominated by CDOM and SS min , however, the resulting maps misrepresented SS min as
chlorophyll (Fig. 3 zoomed into Duluth, MN & Superior, WI area: SLRE at the
western edge of Lake Superior). Consequently, we believe it is best to limit
regional-scale assessments using Landsat to water clarity or turbidity, which is
appropriate for the spectral characteristics of the Landsat sensors, unless independent data are available to verify that SS min and CDOM are not important factors in
the lakes being assessed.
The characteristics of Landsat, MERIS, and MODIS sensors for regional water
quality measurements were analyzed by Olmanson et al. [1]. Imagery from the three
sensors was compared for spatial and spectral characteristics, and empirical models
were developed for chlorophyll a using various bands and band ratio combinations
as dependent variables. MERIS provided a better fit for chlorophyll a (R
2
¼ 0.85,
n ¼ 90) than Landsat and MODIS (R
2
¼ 0.79 for both, n ¼ 177 and 42, respectively).
The red-edge band at 708 nm improved the fit and allowed discrimination between
phytoplankton and SS min , but the Landsat and MODIS results misclassified high
SS min levels as chlorophyll a, similar to Fig. 3.
Phycocyanin, a pigment occurring in cyanobacteria (formerly known as bluegreen algae), can serve as a marker for the presence of these microorganisms in
surface waters and is amenable to measurement by ORS. Cyanobacteria are common in eutrophic water bodies, and some species produce substances that are toxic
122
L.G. Olmanson et al.
705 to 670 nm yielded r
2 values of 0.75–0.93. Of the recently or currently available
and forthcoming satellite sensors, only MERIS and the Sentinel-2 and Sentinel-3
satellites have appropriately narrow red-edge bands.
Despite the above comments, it must be noted that many studies have reported
strong empirical relationships between the broad bands of Landsat sensors and
chlorophyll a (e.g., [1]); typical predictive equations involve the ratio of TM or
ETM+ bands 1 and 3. Absorption by chlorophyll a is strong in bands 1 (450–
520 nm) and 3 (630–690 nm). Nonetheless, increased scattering by phytoplankton
cells counteracts some of the absorption effects and leads to increased reflectance
with increasing chlorophyll a levels in band 1 and even larger increases in band 3. If
it is known that the optical properties of the water bodies being studied are
dominated by phytoplankton, the use of these empirical relationships may be
considered acceptable. However, for regional assessments where specific water
quality characteristics are not known, lakes with high CDOM and/or SS min may
be misclassified.
As an example, when we used Landsat 8 imagery to estimate chlorophyll a and
SD in lakes of northeastern Minnesota for August 31, 2013, we found strong
relationships for both variables (r
2
¼ 0.70 and 0.77; RMSE 0.758 and 0.406,
respectively). Calibration data (Æ3 days) for the images are from the Minnesota
Pollution Control Agency (for chlorophyll a, n ¼ 99; for SD, n ¼ 258) [51]. For
most Minnesota lakes, the results are believed accurate because phytoplankton
dominates their optical properties. When we used the same models for the
St. Louis River Estuary (SLRE), where optical properties of the waters are dominated by CDOM and SS min , however, the resulting maps misrepresented SS min as
chlorophyll (Fig. 3 zoomed into Duluth, MN & Superior, WI area: SLRE at the
western edge of Lake Superior). Consequently, we believe it is best to limit
regional-scale assessments using Landsat to water clarity or turbidity, which is
appropriate for the spectral characteristics of the Landsat sensors, unless independent data are available to verify that SS min and CDOM are not important factors in
the lakes being assessed.
The characteristics of Landsat, MERIS, and MODIS sensors for regional water
quality measurements were analyzed by Olmanson et al. [1]. Imagery from the three
sensors was compared for spatial and spectral characteristics, and empirical models
were developed for chlorophyll a using various bands and band ratio combinations
as dependent variables. MERIS provided a better fit for chlorophyll a (R
2
¼ 0.85,
n ¼ 90) than Landsat and MODIS (R
2
¼ 0.79 for both, n ¼ 177 and 42, respectively).
The red-edge band at 708 nm improved the fit and allowed discrimination between
phytoplankton and SS min , but the Landsat and MODIS results misclassified high
SS min levels as chlorophyll a, similar to Fig. 3.
Phycocyanin, a pigment occurring in cyanobacteria (formerly known as bluegreen algae), can serve as a marker for the presence of these microorganisms in
surface waters and is amenable to measurement by ORS. Cyanobacteria are common in eutrophic water bodies, and some species produce substances that are toxic
122
L.G. Olmanson et al.