waters are CDOM dominated and cyanobacterial blooms take
place every summer. Lake Mälaren in Sweden is the ninth
largest lake in Europe (1140 km”). Mälaren has sophisticated
morphology and due to that optical water properties vary greatly
in different basins. Lake Vörtsjärv is the second largest lake in
Estonia (270 km?). It is a hypetrophic lake with relative high
concentrations of optically active substances (phytoplankton,
CDOM and suspended matter). Lake Tamnaren in Sweden
(32.62 km?) is a very shallow (max depth 1.8 m) brown-
watered lake. We happened to sample there in a windy day and
therefore the suspended matter concentration was the highest
from all sampled lakes (64.95 mg/l). Lake Harku is the smallest
from studied lakes with it's 1.64 km? area. The concentrations
of chlorophyll-a there were the highest from all studied lakes
(203.31 mg/m?) while the CDOM absorption at 400 nm was in
the same range with Lake Peipsi (up to 11.7 m™).
The Estonian lakes were sampled three times during spring-
summer season 2011 (in May, July, August). The Swedish
lakes were sampled in September 2011.
2.2 Remote sensing data
The original idea was to collect MERIS imagery from all
studied lakes, test performance of the standard products and
develop new methods/algorithms if the standard products do not
perform well. The first two field campaigns in Estonia took
place in practically cloud free conditions. Unfortunately, there
were small clouds or processing errors in MERIS Level 2
products in the places where our sampling took place. In
Sweden we sampled during a week with variable weather. Some
parts of Lake Málaren were sampled during cloudy days. In one
day our sampling stations happened to be in an area between
two MERIS swaths. As a consequence, the number of matchup
points is too low to validate MERIS Level 2 products.
We carried our reflectance measurements with two Ramses
radiometers in each sampling station. The reflectance
measurements were carried out in two different ways. Firstly we
measured reflectance in a “normal” way where both the
downwelling irradiance and upwelling radiance sensors were
above the water surface. These measurements were followed by
measurements where the irradiance sensor was above the water
surface but the upwelling radiance sensor was just below the
water surface. This allowed us to collect reflectance spectra
without any sun or sky glint contamination.
2.3 Optical measurements
We studied optical water properties with a Wetlabs instrument
package, besides the reflectance measurements, in each
sampling station. The package consists of a hyperspectral
absorption and attenuation meter ac-s, a backscattering senor
eco-bb3 measuring backscattering coefficient at three
wavelengths (412, 595, 715 nm), and a volume scattering sensor
eco-vsf3 that measures scattering at three wavelengths (460,
532, 660 nm) and under three angles (100, 125 and 150
degrees). It is possible to calculate the backscattering coefficient
also from the ec-vsf3 data. This way we get backscattering
coefficient at six different wavelengths.
2.4 Laboratory measurements
For chlorophyll concentrations 0.5—1.0 liters of water was
filtered through Whatman GF/F-filters (pore size 0.7 um),
extracting pigments with hot ethanol (90%, 75°C) and
measuring absorption coefficients at the wavelengths of 665
and 750 nm (ISO 10260, 1992 (E)).
The concentration of total suspended matter, Crss, was
measured gravimetrically after filtration of the same amount of
water through pre-weighed and pre-combusted (103—105 ?C for
1 h) filters; the inorganic fraction, Csp;, was measured after
combustion at 550 °C for 30 min. The organic fraction Csrom
was determined by subtraction of Cspyy from C7ss (ESS method
340.2, 1993).
Absorption by coloured dissolved organic matter (CDOM,
called also yellow substance), ay, was measured with a
spectrometer (Hitachi U-3010 UV/VIS, in the range 350—750
nm) in water filtered through Millipore 0.2 um filter, measured
in a 10 cm cuvette against distilled water and corrected for
residual scattering according to Davis-Colley and Vant (1987),
who estimated the accuracy of the method to be better than
0.017 m”.
The total particulate absorption, a„(A), was measured with a
spectrometer (Hitachi U-3010 UV/VIS, in the range 400-800
nm) with the Whatman GF/F filter pad technique (Tassan &
Ferrari, 1995, 2002) and using later depigmentation with
sodium hypochloride (Ferrari & Tassan, 1999), which separates
phytoplankton pigment absorption a,,(4) and the absorption by
the rest of particles, a,(4) (tripton).
2.5 Radiative transfer modelling
Hydrolight 5.0 radiative transfer model (Mobley and Sundman
2001) was used to simulate remote sensing reflectance spectra.
A Case 2 model was parameterised in Hydrolight with some
input from our local knowledge about the optical properties of
the lakes under investigation. For example specific absorption
and scattering coefficients of Anabaena circinalis (Metsamaa et
al. 2006) were used as we know that Anabaena species
dominated in some of the lakes during certain periods.
We are in the early stages of creating the modelled spectral
library that covers the whole possible range of chlorophyll,
CDOM and suspended matter variations in our lake waters. The
spectral library used in this study was created using chlorophyll-
a concentrations 1, 5, 10, 30, 60, and 200 mg/m’; CDOM
absorption coefficients (at 420 nm) 1, 5, and 10 m™; and
mineral particle concentrations 1, 5, 10, and 20 mg/l.
In this study we used reflectance spectra where the water
leaving radiance (L,) was divided by the downwelling
irradiance (Ey). These spectra were compared with in situ
reflectance spectra measured with upwelling radiance sensor
just below the water surface and the downwelling irradiance
sensor above the water surface i.e. both the modelled and
measured reflectance spectra did not contain any glint.
Spectral Angle Mapper (SAM) was used to find the modelled
spectrum most similar to any measured reflectance spectrum.
Different angles from 0.2 to 0.5 were used in SAM. All the
angles produced identical classification results but the number
of unclassified reflectance spectra was the lowest for the angle
of 0.5. Therefore, the results presented in this study are for this
angle.
Results c
prelimina
The num
model w:
concentra
whole ra
Therefore
concentra
concentre
estimatec
in Figur
correlatic
mg/m’.
250
no
©
e
150
£o
a
o
Estimated chlorophyll-a, mg/m?
un
e
Figure
measure
estimat:
We are
spectral
different
previous
bio-opti
compari
of varie
reduce
concent
when tt
with ir
differen
become
There i
Hydroli
able to
to cyan
absorpt
the Ana
exactly
Howev:
backscz
cyanob
backsc:
Spectra
wavelei
signific
creatin,
during