APPLICATION OF REMOTE SENSING FOR ADJUSTMENT AND CALIBRATION OF WATER QUALITY MODEL - Integrating GIS, Remote Sensing, and Mathematical Modelling for Surface Water Quality Management - page 179

Although the quantitative calibration of TSM did not show a big variation, the same trend is well
recognized between the model and the image values. The important result from the calibration using
the image data is that it showed close relation to the field measurements and it showed that the
mathematical model could be insensitive to the TSM re-suspension. This explains the underestimated
concentrations from the model. More detailed analysis could be done using analytical remote sensing
approach or bio-optical modelling to develop more accurately calibrated algorithms for the TSM maps
because the remote sensing of TSM involves a big level of uncertainty that may occur due to the
presence of intensive submerged and floating vegetation in the lake and this may reflect higher
measured signals of TSM. Further investigation is needed to develop an accurate algorithm to estimate
suspended sediment concentration over Lake Edko using MODIS images. In order to achieve accurate
estimates it is necessary to develop a site-specific algorithm that fits the conditions of the Lake.
8.6. SPATIAL AND TEMPORAL QUANTITATIVE CALIBRATION OF THE SCREENING
EUTROPHICATION MODEL
8.6.1. Chlorophyll-a Retrieval from Remotely Sensed Imagery
CHL-a is not uniformly distributed in inland waters. This light-absorbing molecule in the chloroplasts
of algal cells is the universal algal pigment, and CHL-a is a simple measure of phytoplankton biomass
in surface waters. Thus, as mentioned above, we use CHL-a as an indicator of eutrophication in this
research.
Remote sensing of CHL-a has had limited success in turbid productive coastal waters. Far offshore, in
case I waters , ocean color (upwelling radiance) is largely determined by water and CHL-a, and
extraction of CHL-a from ocean color measurements has been done successfully (e.g., Gordon and
Morel 1983). However, in turbid productive coastal waters known as case II waters , color is
determined by water, CHL-a, colored dissolved organic matter (CDOM), and non-phytoplanktonic
particulates (seston or tripton). Although inland waters (e.g., lakes, rivers, reservoirs) usually have a
higher range of CHL-a and thus a stronger signal, the independent variations in CDOM and
particulates have impeded the routine extraction of CHL-a from ocean color measurements inshore.
We refer to turbid waters with high CHL-a concentrations as “turbid, productive waters ” whether
inland or at the land margins. Most inland waters are classified as Case 2 waters (Morel and Prieur
1977; Kirk 1994). Much of the research on Case 2 waters has been focused on methods for analyzing
the reflectance as a function of the concentrations of phytoplankton groups, inorganic matter, and
dissolved organic matter. Most of the algorithms developed for estimating CHL-aorophyll
concentration from inland aquatic systems are based on reflectance spectra derived from ground
measurements and airborne sensors (Lathrop and Lillesand 1986; Dekker 1993; Novo et al. 1991) or
from satellite sensors not tuned to the radiometric and spectral resolution requirements of Case 2
waters. The algorithms are generally site specific and empirically derived through statistical
relationships between reflectance and CHL-a concentration.
Although the MODIS images used in this research have (250m) resolution they have only two bands;
one red and one near -infrared. The lack of green band makes this MODIS data set more suitable for
the TSM analysis, but it is not applicable for the CHL-a extraction. Therefore, the SPOT-5 imagery is
adopted to derive the calibration data, because of its high spatial resolution (10 m) and useful spectral
resolution (including the green band). The SPOT scene was used as guidance for concentrations of
CHL-a at that time of the year since there was no available scene at the same time of the field
measurements. The image was taken twelve days after the field work measurements, on the 10 th of
July 2006, but it was still comparable to the modelling results and the data available from literature.
To have a practically applicable SPOT scene for water quality variables extraction the image has to
pass several image processing steps. The following section explains the detailed SPOT image