Geology Reference
In-Depth Information
training data and associated predictions. Weijs et al. [
80
] proposed that hydrological
system should be based on information-theoretical scores.
2.5.1 Modeling Uncertainty
In hydrology and water resources research, there are two major bases of uncertainty
attitudes; one is based on stochasticity as a necessary factor and the other on the
deterministic nature of the system. The de
nition of the uncertainty is much more
uncertain about the modeled numerical values; it relates much deeper processes and
pertains to the governing mechanisms of the model. Distinguishable uncertainties in
hydrology are data uncertainties (mainly associated with measurements), sample
uncertainties (e.g., number of data for calibration), and model uncertainty [
58
]. Klir
[
41
] made an attempt to consider uncertainty in terms of the complexity of the
model. He found both categories have a con
ictive nature, i.e., if complexity
decreases, uncertainty grows. Halfon [
28
] also addressed the issue of modeling in
the context of Lake Ecosystem models; he evaluated the performance of several
models of varying complexity. There are many ways to assess roughly the modeling
uncertainty and these methods range from the use of statistical parameters such as
standard deviation to analytical calculations to find the propagation of error.
Recently, very powerful tools such as Monte Carlo analysis have been used for
sensitivity estimation of more complicated methods. Other criteria such as fractals,
Bayesian fuzzy-sets, and random
fields have been applied successfully to solve
uncertainty problems in hydrology and other
fields such as applied mathematics,
physics, systems theory, etc. Wagener et al. [
77
] applied a Monte Carlo analysis
toolbox, combining a number of analysis tools to investigate parameter identifi-
-
ability, model behavior, and prediction uncertainty to establish a sensible rela-
tionship between model parameters and catchment characteristics. Beck [
12
] points
out valid reasons to concentrate more on the uncertainty of model structure as an
important area of study. Mizumura [
53
] combined a conceptual tank model and a
fuzzy logic model to yield satisfactory results with minimum uncertainty issues.
Kindler and Tyszewski [
39
] acknowledged the applicability of fuzzy theory to a
diagnostic approach of problem solving and uncertainty assessment. Feluch [
26
]
applied non-parametric estimation methods to two classes of hydrological prob-
lems. Various studies have been carried out to ascertain the
model
in environmental chemistry using modern statistical approaches [
43
,
76
], consid-
ering uncertainty. Beven [
15
,
16
] introduced the concept of equi
'
best and right
'
nality which is
related to the uncertainty associated with parameters. Equi
nality arises when, in a
hydrological model, many different parameter sets are equally good at reproducing
an output signal.
