| System Identification or Data-based Modeling is an important tool in the field of data science
which deals with modeling a dynamic system using available data. For some natural systems,
different modeling techniques have been proposed and applied but the most successful
techniques are process-based modeling and data-based modeling. Process-based models provide
a detailed description about the process involved in any mechanical system whereas data-based
models mainly focus on the behavior of the system itself. In this thesis, data-based modeling is
proposed for the Portneuf River in Idaho, United states. The aim of this thesis is to find a suitable
model for the river ecosystem which depends on various factors like the ambient temperature of
the surroundings, the flow of the river, and the number of organisms present in the river itself.
The System Identification models used in this thesis mainly deal with linear models, but nonlinear grey box models were also proposed. Starting with a choice of three different black-box
linear MIMO models, experiments were carried out to find a model which can best describe the
Portneuf River ecosystem. Also, a major contribution of this research was to develop a means of
making use of irregularly spaced temporal data, to produce regularly sampled temporal data. Our
best linear time-invariant models were found to be those produced using transfer function
estimates on weekly sampled data. Upon successful completion, this research might be very
beneficial in protecting the river ecosystem and the animals living in and around it. |