Geology Reference
In-Depth Information
Chapter 2
Hydroinformatics and Data-Based
Modelling Issues in Hydrology
Abstract This chapter highlights and addresses some basic issues associated with
data-based modeling. The chapter starts with a brief description of emergence and
development of hydroinformatics as a potent segment of mainstream hydrology and
proceeds to the ignored or least considered modeling queries existing in hydrology,
e.g., how much bene
t could be gained by increased complexity in data-based
models or whether increased complexity adversely affects model performance. The
chapter reminds one of the need to evaluate existing hypothetic assumptions on
various modeling properties.
Data-based modeling is subject to different types of uncertainties and ambiguities
because of the presence of different unsolved queries and deliberately over-sim-
pli
ed assumptions. Several studies in hydrology have pointed out the contradictory
fact that, under certain circumstances, a poor model may give acceptable results,
while, under other circumstances, a good, re
ned model may fail to give better and
more reliable answers. The main reason for this is that developers view modeling as
a rigorous mathematical exercise rather than as a subjective activity [
56
]. Previously
successful model in one phase of the hydrological cycle might not give relevant
results in a new situation. Proper vision or insight into the working of the actual
hydrological system is necessary, even in data-based modeling; else modeling
results from even a sophisticated mathematical model would be irrelevant or mis-
guiding with regard to the behavior of the actual system. In data-based modeling,
the model trusts the quality of the data which should be inherent in the actual
behavior of the system. Savenije [
60
] emphasized the need to change the modeling
process to a
approach, i.e., learning from the data to the physical theory
rather than giving a lower preference to the strength of the data. Barnes [
11
]
suggests that an
“
top-down
”
model is a model which represents all the information
contained in the data (so that there is effectively no residual information). As per
this de
“
adequate
”
cial intelligence models fall into that cate-
gory. Another required aspect of a model is how ef
nition, all data-based and arti
cient the model is in tackling a
particular phenomenon or situation in hydrology. The problems of overparame-
terization and scaling issues have gained much attention and invited detailed studies
in the context of processes modeling on a wide catchment and regional scale.
