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 165

8.3. LINKING REMOTE SENSING WITH WATER QUALITY AND EUTROPHICATION MODELS :
CALIBRATION PROCEDURES
Water quality modelling is associated with a recognised level of uncertainty, which arises from
different sources including the measurements, model inputs, model parameters and the extensive
complicated interactive processes that involve different water quality parameters. Therefore these
uncertainties are also expected to strongly influence the calibration process of the water quality
models. The problem becomes more complicated when the model being calibrated is a data scarce
model, that is, when the data sets required for model calibration are insufficient for the spatial and
temporal requirements needed to perform a standard model calibration, as shown in Chapter 7 . In the
present chapter, remote sensing is introduced as a tool for calibrating water quality and eutrophication
models for the parameters TSM and CHL-a. Different levels of model calibration could be done using
different levels of water quality data sets generated from remote sensing data. Remote sensing can be
seen as a complementary tool in the water quality modelling process for similar situations to Lake
Edko, as this study proves, where the lack of in situ measurements is a common feature due to
economic and technical constraints.
In the following sections, a detailed explanation of the advantages of using remote sensing data for
calibration is highlighted, and the different types of used satellite images and remotely sensed data are
presented. The different methodologies for calibrating the water quality and eutrophication models are
explained and the advantages and limitations of each used methodology are discussed.
8.3.1. Advantages of Calibration Using Remote Sensing Data
Most water quality models have a complex structure and include a number of processes such that the
simulation results are most of the time linked with high uncertainty. However, most of the model
parameters that greatly affect the results should be assigned in advance by referring to typical values
found in the literature. This is because field data work usually misses these modeling parameters.
Consequently, the error caused by the uncertainty in the coefficients greatly decreases the model's
reliability (Canale and Seo, 1996).
Therefore, developing a suitable calibration procedure for these model parameters is important before
the model can be implemented for a real aquatic system. Model calibration is an essential process to
test and tune a model by comparing the simulated results and field data. The simplest method is the
parameter tuning and trial-and-error, which was introduced in Chapter 7 due to lack of temporal field
measurements. By minimizing the difference between simulated results and field data, the modeller
can set up a water quality model with one set of rational parameters for a specific water body.
However, this guesswork is time-consuming and relies considerably on the user's experience. The
reasons why statistical methods are not commonly used in model calibration are the complexity of the
models and the limitation of a large field data set (Henderson-Sellers and Davies, 1991). Moreover,
field data are not always enough in both spatially and temporally. Where there are not enough data sets
for calibration the statistical measures of goodness of fit will not be possibly applicable, therefore
remote sensing technique could provide missing spatial and temporal data sets, which therefore could
be used for calibration.
Early studies that took on board the use of remote sensing data in integration with mathematical
models, showed promising results. Several research projects were done on rivers, lakes and estuaries
using one dimensional model and integrating it with remote sensing data, such as (Yang et al., 2000),
where algal growth rate was modelled with QUAL2E, and 2D spatial data set derived from a SPOT
image was used for calibration. In the present research one of the main objectives is to define the most
appropriate way of calibrating water quality models with remote sensing.