Page 292 - Soil and water contamination, 2nd edition

Model calibration and validation 279

the model output always depends on the quality of the input. Therefore, one should beware
of relying too heavily on sophisticated model output and be open to a critical assessment of
both model structure and model input.

EXERCISES
1. Describe in brief the similarities and differences between model calibration, verification,
and validation.

2. Explain in your own words the need for model verification, calibration, and validation.
3. What are the disadvantages of ‘trial and error’ model calibration?

4. A model has four model parameters to be calibrated. The brute force method is used and
the predefined ranges of model parameter values are subdivided into six discrete steps.
How many model runs are needed to calibrate the model? How many model runs would
be needed if the model had six parameters to calibrate?

5. The following table presents the observed and simulated concentration data from a
biodegradation experiment.
Time Observed Simulated Time Observed Simulated
(d) (mg/l) (mg/l) (d) (mg/l) (mg/l)
1 2.61 2.28 13 0.76 0.48
2 2.54 1.98 14 0.76 0.43
3 1.48 1.72 15 0.52 0.39
4 1.64 1.50 16 0.35 0.35
5 1.48 1.31 17 0.24 0.32
6 1.48 1.14 18 0.47 0.29
7 0.85 1.00 19 0.56 0.27
8 1.29 0.88 20 0.19 0.25
9 0.90 0.77 21 0.52 0.23
10 0.95 0.68 22 0.07 0.22
11 0.74 0.61 23 0.04 0.20
12 0.82 0.54 24 0.15 0.19
2
a. Calculate the squared Pearson’s correlation coefficient (R ).
b. Calculate the Nash efficiency coefficient.
c. Perform a Student’s t test.
d. Discuss the performance of the model, using the outcomes of questions a–c.

6. Explain why a complex model does not necessarily perform better than a simple model.

10/1/2013 6:45:23 PM
Soil and Water.indd 291 10/1/2013 6:45:23 PM
Soil and Water.indd 291