Digital Signal Processing Reference
In-Depth Information
Model Calibration
A successful model application requires a model calibration that compares simulated results with
measured lake conditions. This section discusses the model calibration procedures and details. In
general the model results are tested or calibrated against field measurements. This field data must be
different from the data set used during the construction of the original model (Thomann, 1982;
Thomann and Mueller 1987). The objective of model calibration is to adjust the input parameters so
that there will be closer agreement between the simulated values and observed data (Ambrose, 1992;
Bierman, 1986). There are several methodologies and techniques applied for water quality models
calibration. Dilk's
et al.,
(1990)
have calibrated a DO model by modifying the nitrogen rate, COD and
BOD deoxygenation rate, and reaereation rate. Ambrose (1992) has calibrated
an estuary water quality
model by adjusting dispersion values in transport processes and reaction rate coefficients in water
quality interactions. Additionally lung and Larson (1995) have used low flow conditions to calibrate a
water quality model. Another different methodology was applied by (Masato
et al.,
2002), where the
water quality model calibration was regarded in their study as an optimisation problem to minimize
the discrepancy between the observed and calculated results; global optimisation was used.
In this study the water quality model calibration is done on different levels and by applying different
techniques. The first level is presented in this chapter. Here the conventional water quality parameters
or oxygen group (DO, COD, BOD and Nutrients group NH4 and NO3) and their associated model
process parameters are selected for calibration. In particular, the process parameters are adjusted for
this level of calibration. In the next chapter, further calibration procedures based on the application of
remote sensing techniques are used for enhancing the model performance for specific parameters such
as TSM and CHL-a. The first level calibration was based on a comparison between simulations and
measurements and the calculation of the Mean Relative Error (MRE) and the Root Mean Square Error
(RMSE) to examine the performance of the model. Due to the scarcity in data inside the lake the
comparison and calibration was done by comparing concentrations on a spatial basis between
predicted and measured values. The following section presents the different calibrated parameters and
the calibration results.
Calibration of the Oxygen Group Parameters
The first level of model calibration was carried out by visual comparison of simulations and
measurements in graphs together with the calculation of the statistical error values such as Mean
Relative Error (MRE); the overall performance of the model was examined as well. The output
variables of the model such as DO, COD, BOD with respect to the observations in Lake Edko during
field survey were plotted in graphs to make comparisons, which were used to check how the
simulations fit the observations. Besides, MRE was used to quantify the agreement of the model, by
dividing the residuals by the observed values. In this study the calculation of RE and MRE was based
on Equations (6-5
)
and (6-6):
Csim
Cobs
RE
100
(7-5)
Cobs
RE
n
MRE = Sum
(7-6)
Here Csim and Cobs are the simulated and observed values respectively, and n is the number of cases.
The MRE denotes the mean relative difference between simulations and observations. Other statistical
methods for error calculations were used such as Mean Error (ME), Mean Absolute Error (MAE) and
the Root Mean Square Error (RMSE). The oxygen group calibration parameters are shown in the
following
Table (7-2)
:
Search WWH ::
Custom Search