Streamflow Forecasting Using Wavelet Coupled Soft Computing Techniques and Fuzzy Logic-Based Approach for Stream Water Quality-Quantity Assessment
Date
2023
Authors
S, Shruti Kambalimath
Journal Title
Journal ISSN
Volume Title
Publisher
National Institute Of Technology Karnataka Surathkal
Abstract
Hydrologic time series is a collection of timely recorded variables such as streamflow,
temperature, evaporation, etc. over a period of time. Forecasting of such time series
necessarily aid future predictions based on past records as well as filling of missing
data or extension of available data. Accurate and timely forecasting of hydrologic
time series can be a great aid for various applications in water resources planning and
management.
During the last few decades, several types of stochastic models have been proposed as
well as developed for modeling hydrological time series and generating synthetic
stream flows. Some of such stochastic models are autoregressive (AR), Moving
Average (MA), Autoregressive Moving Average (ARMA), and Autoregressive
Integrated Moving Average (ARIMA).
In contrast to the analytical models, soft computing methods learn from past records
and require limited input parameters. These techniques are very useful in cases where
there are limitations in terms of data availability. The collection of techniques under
Soft Computing renders low-cost solutions to imprecisely formulated problems and
attempt to mimic the behavior and learning ability of human beings into computers.
One such soft computing technique is “Fuzzy Logic”. We have developed three soft
computing models to forecast daily streamflow time series for different lead times for
Malaprabha sub-basin in Karnataka state of India. The performance of Support Vector
Machine (SVM), Adaptive Neuro-fuzzy Inference System (ANFIS), and Fuzzy
models to forecast daily streamflow is tested for 1-day, 3-days and 5-days ahead
forecasts. The results indicate that the performance of the models significantly
decreases with an increase in lead times. The models show high R2values for 1-day
ahead streamflow forecasts, whereas it is low for 3 and 5-days lead time.
It is necessary to provide a powerful tool to reduce the noise in the data so that
accuracy of the model is increased. Wavelet transformer is one such powerful tool
used to decompose the data set into different scales. The wavelet method effectively
decomposes the original time series in to sub-series at different resolution levels there
by facilitating denoising of the data. In this research work, discrete wavelet transform
is coupled with the fuzzy logic method to improve the accuracy of the forecast. The
iperformance of all the three models significantly increased when the wavelet is
coupled, especially for longer lead times such as 5 days. The Wavelet coupled fuzzy
(WT-fuzzy) model outperformed Wavelet coupled ANFIS (WT-ANFIS) and Wavelet
coupled SVM (WT-SVM) models. However, WT-ANFIS performed better than WT-
SVM.Longer lead time forecasts find applications in flood forecasting and evacuation
programs. This research aims at improving the efficiency of forecasting models
especially for longer lead times such as 3 days and 5 days which are crucial times for
undertaking quick flood evacuation measures.
The second phase of this research is stream water quality-quantity modeling. Water
quality and quantity are the two aspects that are interrelated and hence should be
studied together within an integrated framework. In today's world, demand for water
essentially takes into account both quality and quantity aspects for various uses of
water. Having a sufficient accessible quantity of water becomes meaningful only if
this quantity of water is acceptable in terms of its quality. This study aims at studying
the role of the quantity of water in determining its quality along with the other quality
parameters. The Water Quality Index (WQI) is an efficient tool which can describe
the status of water by translating a large amount of data in to a single value. The
results in this study indicate that streamflow can be considered as one of the inputs to
determine the WQI.
Description
Keywords
Soft computing, Fuzzy Logic, Support Vector Machine, Adaptive Neuro- fuzzy Inference System