Discussion
This study shows the effectiveness of streamflow analysis using the
information and complexity theory to detect different discharge patterns
that may describe some hydrological processes during low and high
frequency scales as well as during flood assessment. Indeed, the
selection of suitable metrics to quantify streamflow patterns at
different temporal scales is of paramount importance. Interesting issues
were emerged from this research and discussed below.
The role of word length and word
number in characterizing streamflow
patterns
The importance of word length and word number, to the best of our
knowledge, was not highlighted sufficiently in the literature. Pachepsky
et al., (2016), indicated that investigating the role of word length
efficiency to improve the information and complexity metrics would open
new avenue for further explorations. Thus, the important question here
that we can ask ourselves is what is the recommended word length that
should be selected to process the symbolic strings method
professionally. In fact, there is no direct rule available to advise an
ideal word length that generates the finest results in information and
complexity theory. However, since this work investigates the
classification and characterization of various streamflow patterns
during different scales, we suggested two types of word length (two and
four characters). In the case of low and high frequency scales,
streamflow patterns are mainly affected by the intensity of various
rainfall events, hence, it is vital to examine the influence of
different intense events on the discharge data. In this regard, we
considered the hydrograph separation method proposed by (Raghunath,
2006) depicted in Fig. 8 to separate the hydrograph into runoff flow and
baseflow. As a result, in the case of the Gōno River watershed, the
number of N days after peak for the streamflow to get rid of a
rainfall inputs are 4 days approximately and hence, we used 4 characters
as a word length. Thus, in the case of AL=24 hours, it means that each
24 hours of streamflow records (i.e. 1-day discharge data) were grouped
together, and then averaged to form a character, and correspondingly the
word pattern is formed from 4 days to describe the state of a system for
each 4 successive days.