Supervisors info:
Κυριακόπουλος Κ. Καθηγητής (Επιβλέπων), Κίλιας Σ. Καθηγητής, Μερτζάνης Α. Αναπλ. Καθηγητής ΤΕΙ Στερεάς Ελλάδος
Summary:
Floods usually occur during periods of excessive precipitation or thawing in
the winter period (ice jam) and have a huge impact. The effects of natural
disasters are both tangible –financial (such as loss of property and damage to
public property) and intangible (such as injuries and loss of life). Their
prediction is therefore needed and constitutes a challenge for conducting
research. However, there are indications. Floods are usually accompanied by an
increase in river discharge. The model devised by Tsakiri et al., 2014, which
is used in the present work, is able to predict river’s water discharge to a
great extent. There are many data available to be processed by various bodies,
using this model. However, because the volume of data is daunting, most
scientists use case studies. Data mining techniques are used that aid to solve
this problem. Data mining software can handle this volume of data (big data)
efficiently and can even operate in real time. The advantage of resulting
conclusions in real time makes it possible to envisage their use in early
warning systems for predicting floods. Knime, which is used in this project, is
a good example of such software. In this study it is presented a workflow that
uses a statistical model for the prediction and explanation of the timeseries
of water discharge in real time, using an example from New York State. This
model for predicting water discharge does decomposition of the timeseries of
hydrologic and climatic variables in seasonal, long-term and short-term
components. It also analyzes runoff using a summer and a winter model. This
improves the explanation of water discharge up to 81%. We also observe that
with increasing water discharge in the long term, the water table is
replenished, while in the seasonal term it depletes. In the short term, the
groundwater level falls during the winter season and increases during the
summer season.
Keywords:
Real time data mining, Flood warning system, Water discharge prediction, Time series decomposition, Prediction model
