Advances in Streamflow Forecasting: From Traditional to Modern Approaches

By Deepesh Machiwal

Advances in Streamflow Forecasting: From Traditional to Modern Approaches covers the three major data-driven approaches of streamflow forecasting including traditional approach of statistical and stochastic time-series modelling with their recent developments, stand-alone data-driven approach such as artificial intelligence techniques, and modern hybridized approach where data-driven models are combined with preprocessing methods to improve the forecast accuracy of streamflows and to reduce the forecast uncertainties.

This book starts by providing the background information, overview, and advances made in streamflow forecasting. The overview portrays the progress made in the field of streamflow forecasting over the decades. Thereafter, chapters describe theoretical methodology of the different data-driven tools and techniques used for streamflow forecasting along with case studies from different parts of the world. Each chapter provides a flowchart explaining step-by-step methodology followed in applying the data-driven approach in streamflow forecasting.

This book addresses challenges in forecasting streamflows by abridging the gaps between theory and practice through amalgamation of theoretical descriptions of the data-driven techniques and systematic demonstration of procedures used in applying the techniques. Language of this book is kept simple to make the readers understand easily about different techniques and make them capable enough to straightforward replicate the approach in other areas of their interest.

This book will be vital for hydrologists when optimizing the water resources system, and to mitigate the impact of destructive natural disasters such as floods and droughts by implementing long-term planning (structural and nonstructural measures), and short-term emergency warning. Moreover, this book will guide the readers in choosing an appropriate technique for streamflow forecasting depending upon the given set of conditions.

• Contributions from renowned researchers/experts of the subject from all over the world to provide the most authoritative outlook on streamflow forecasting
• Provides an excellent overview and advances made in streamflow forecasting over the past more than five decades and covers both traditional and modern data-driven approaches in streamflow forecasting
• Includes case studies along with detailed flowcharts demonstrating a systematic application of different data-driven models in streamflow forecasting, which helps understand the step-by-step procedures

• Language: English
• Publisher: Elsevier Science
• Release date: Jun 20, 2021
• ISBN: 9780128209240

Book preview

Advances in Streamflow Forecasting

From Traditional to Modern Approaches

Editor

Priyanka Sharma

Editor

Deepesh Machiwal

Table of Contents

Cover image

Title page

Copyright

Dedication

Contributors

About the editors

Foreword

Preface

Acknowledgment

Chapter 1. Streamflow forecasting: overview of advances in data-driven techniques

1.1. Introduction

1.2. Measurement of streamflow and its forecasting

1.3. Classification of techniques/models used for streamflow forecasting

1.4. Growth of data-driven methods and their applications in streamflow forecasting

1.5. Comparison of different data-driven techniques

1.6. Current trends in streamflow forecasting

1.7. Key challenges in forecasting of streamflows

1.8. Concluding remarks

Chapter 2. Streamflow forecasting at large time scales using statistical models

2.1. Introduction

2.2. Overview of statistical models used in forecasting

2.3. Theory

2.4. Large-scale applications at two time scales

2.5. Conclusions

Conflicts of interest

Chapter 3. Introduction of multiple/multivariate linear and nonlinear time series models in forecasting streamflow process

3.1. Introduction

3.2. Methodology

3.3. Application of VAR/VARX approach

3.4. Application of MGARCH approach

3.5. Comparative evaluation of models’ performances

3.6. Conclusions

Chapter 4. Concepts, procedures, and applications of artificial neural network models in streamflow forecasting

4.1. Introduction

4.2. Procedure for development of artificial neural network models

4.3. Types of artificial neural networks

4.4. An overview of application of artificial neural network modeling in streamflow forecasting

Chapter 5. Application of different artificial neural network for streamflow forecasting

5.1. Introduction

5.2. Development of neural network technique

5.3. Artificial neural network in streamflow forecasting

5.4. Application of ANN: a case study of the Ganges River

5.5. ANN application software and programming language

5.6. Conclusions

5.7. Supplementary information

Chapter 6. Application of artificial neural network and adaptive neuro-fuzzy inference system in streamflow forecasting

6.1. Introduction

6.2. Theoretical description of models

6.3. Application of ANN and ANFIS for prediction of peak discharge and runoff: a case study

6.4. Results and discussion

6.5. Conclusions

Chapter 7. Genetic programming for streamflow forecasting: a concise review of univariate models with a case study

7.1. Introduction

7.2. Overview of genetic programming and its variants

7.3. A brief review of the recent studies

7.4. A case study

7.5. Results and discussion

7.6. Conclusions

Chapter 8. Model tree technique for streamflow forecasting: a case study in sub-catchment of Tapi River Basin, India

8.1. Introduction

8.2. Model tree

8.3. Model tree applications in streamflow forecasting

8.4. Application of model tree in streamflow forecasting: a case study

8.5. Results and analysis

8.6. Summary and conclusions

Chapter 9. Averaging multiclimate model prediction of streamflow in the machine learning paradigm

9.1. Introduction

9.2. Salient review on ANN and SVR modeling for streamflow forecasting

9.3. Averaging streamflow predicted from multiclimate models in the neural network framework

9.4. Averaging streamflow predicted by multiclimate models in the framework of support vector regression

9.5. Machine learning–averaged streamflow from multiple climate models: two case studies

9.6. Conclusions

Chapter 10. Short-term flood forecasting using artificial neural networks, extreme learning machines, and M5 model tree

10.1. Introduction

10.2. Theoretical background

10.3. Application of ANN, ELM, and M5 model tree techniques in hourly flood forecasting: a case study

10.4. Results and discussion

10.5. Conclusions

Chapter 11. A new heuristic model for monthly streamflow forecasting: outlier-robust extreme learning machine

11.1. Introduction

11.2. Overview of extreme learning machine and multiple linear regression

11.3. A case study of forecasting streamflows using extreme machine learning models

11.4. Applications and results

11.5. Conclusions

Chapter 12. Hybrid artificial intelligence models for predicting daily runoff

12.1. Introduction

12.2. Theoretical background of MLP and SVR models

12.3. Application of hybrid MLP and SVR models in runoff prediction: a case study

12.4. Results and discussion

12.5. Conclusions

Chapter 13. Flood forecasting and error simulation using copula entropy method

13.1. Introduction

13.2. Background

13.3. Determination of ANN model inputs based on copula entropy

13.4. Flood forecast uncertainties

13.5. Flood forecast uncertainty simulation

13.6. Conclusions

Appendix 1. Books and book chapters on data-driven approaches

Appendix 2. List of peer-reviewed journals on data-driven approaches

Appendix 3 Data and software

Index

Copyright

Elsevier

Radarweg 29, PO Box 211, 1000 AE Amsterdam, Netherlands

The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, United Kingdom

50 Hampshire Street, 5th Floor, Cambridge, MA 02139, United States

Copyright © 2021 Elsevier Inc. All rights reserved.

No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Details on how to seek permission, further information about the Publisher’s permissions policies and our arrangements with organizations such as the Copyright Clearance Center and the Copyright Licensing Agency, can be found at our website: www.elsevier.com/permissions.

This book and the individual contributions contained in it are protected under copyright by the Publisher (other than as may be noted herein).

Notices

Knowledge and best practice in this field are constantly changing. As new research and experience broaden our understanding, changes in research methods, professional practices, or medical treatment may become necessary.

Practitioners and researchers must always rely on their own experience and knowledge in evaluating and using any information, methods, compounds, or experiments described herein. In using such information or methods they should be mindful of their own safety and the safety of others, including parties for whom they have a professional responsibility.

To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors, assume any liability for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions, or ideas contained in the material herein.

Library of Congress Cataloging-in-Publication Data

A catalog record for this book is available from the Library of Congress

British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library

ISBN: 978-0-12-820673-7

For information on all Elsevier publications visit our website at https://www.elsevier.com/books-and-journals

Publisher: Candice Janco

Acquisitions Editor: Louisa Munro

Editorial Project Manager: Aleksandra Packowska

Production Project Manager: Kiruthika Govindaraju

Cover Designer: Mark Rogers

Typeset by TNQ Technologies

Dedication

This book is dedicated to my parents Sudha Sharma and Pramod Sharma, my husband Basant Mishra, and son Advit Mishra.

—Priyanka Sharma

This book is dedicated to my loving family, sisters, brother, parents Devki Machiwal and Durga Prasad Machiwal, my wife Savita, and daughter Mahi.

—Deepesh Machiwal

Contributors

Kevin O. Achieng

Department of Civil and Architectural Engineering, University of Wyoming, Laramie, WY, United States

Department of Crop & Soil Sciences, University of Georgia, Athens, GA, United States

Jan F. Adamowski,     Department of Bioresource Engineering, Faculty of Agricultural and Environmental Science, McGill University, Montreal, QC, Canada

Sheikh Hefzul Bari,     Department of Civil Engineering, Leading University, Sylhet, Bangladesh

Lu Chen,     School of Civil and Hydraulic Engineering, Huazhong University of Science and Technology, Wuhan, Hubei, China

Nastaran Chitsaz,     National Centre for Groundwater Research and Training, College of Science and Engineering, Flinders University, Bedford Park, South Australia, Australia

Ali Danandeh Mehr,     Department of Civil Engineering, Antalya Bilim University, Antalya, Turkey

Ravinesh C. Deo,     School of Agricultural Computational and Environmental Sciences, International Centre of Applied Climate Sciences (ICACS), University of Southern Queensland, Springfield, QLD, Australia

Farshad Fathian,     Department of Water Science and Engineering, Faculty of Agriculture, Vali-e-Asr University of Rafsanjan, Rafsanjan, Kerman Province, Iran

Salim Heddam,     Faculty of Science, Agronomy Department, Hydraulics Division, Laboratory of Research in Biodiversity Interaction Ecosystem and Biotechnology, University 20 Août 1955, Skikda State, Algeria

Md Manjurul Hussain,     Institute of Water and Flood Management, Bangladesh University of Engineering and Technology, Dhaka, Bangladesh

Saeid Janizadeh,     Department of Watershed Management, Faculty of Natural Resources and Marine Sciences, Tarbiat Modares University, Noor, Mazandaran Province, Iran

V. Jothiprakash,     Department of Civil Engineering, Indian Institute of Technology Bombay, Mumbai, Maharashtra, India

Özgur Kişi,     Department of Civil Engineering, School of Technology, Ilia State University, Tbilisi, Georgia

Anil Kumar,     Department of Soil and Water Conservation Engineering, College of Technology, G.B. Pant University of Agriculture and Technology, Pantnagar, Uttarakhand, India

Andreas Langousis,     Department of Civil Engineering, School of Engineering, University of Patras, University Campus, Rio, Patras, Greece

Deepesh Machiwal,     Division of Natural Resources, ICAR-Central Arid Zone Research Institute, Jodhpur, Rajasthan, India

Ishtiak Mahmud,     Civil and Environmental Engineering, Shahjalal University of Science and Technology, Sylhet, Bangladesh

Arash Malekian,     Department of Reclamation of Arid and Mountainous Regions, Faculty of Natural Resources, University of Tehran, Tehran, Iran

Anurag Malik

Punjab Agricultural University, Regional Research Station, Bathinda, Punjab, India

Department of Soil and Water Conservation Engineering, College of Technology, G.B. Pant University of Agriculture and Technology, Pantnagar, Uttarakhand, India

Georgia Papacharalampous,     Department of Water Resources and Environmental Engineering, School of Civil Engineering, National Technical University of Athens, Iroon Polytechniou 5, Zografou, Greece

P.L. Patel,     Department of Civil Engineering, Sardar Vallabhbhai National Institute of Technology, Surat, Gujarat, India

Mir Jafar Sadegh Safari,     Department of Civil Engineering, Yaşar University, Izmir, Turkey

Priyank J. Sharma,     Department of Civil Engineering, Sardar Vallabhbhai National Institute of Technology, Surat, Gujarat, India

Priyanka Sharma,     Groundwater Hydrology Division, National Institute of Hydrology, Roorkee, Uttarakhand, India

Mohammad Istiyak Hossain Siddiquee,     Data and Knowledge Engineering, Otto-von-Guericke University of Magdeburg, Magdeburg, Saxony-Anhalt, Germany

Vijay P. Singh,     Department of Biological & Agricultural Engineering and Zachry Department of Civil & Environmental Engineering, Texas A&M University, College Station, TX, United States

Doudja Souag-Gamane,     Leghyd Laboratory, University of Sciences and Technology Houari Boumediene, Bab Ezzouar, Algiers, Algeria

Yazid Tikhamarine

Southern Public Works Laboratory (LTPS), Tamanrasset Antenna, Tamanrasset, Algeria

Department of Science and Technology, University of Tamanrasset, Sersouf, Tamanrasset, Algeria

Mukesh K. Tiwari,     Department of Irrigation and Drainage Engineering, College of Agricultural Engineering and Technology, Anand Agricultural University, Godhra, Gujarat, India

Hristos Tyralis,     Air Force Support Command, Hellenic Air Force, Elefsina Air Base, Elefsina, Greece

Mehdi Vafakhah,     Department of Watershed Management, Faculty of Natural Resources and Marine Sciences, Tarbiat Modares University, Noor, Mazandaran Province, Iran

About the editors

Dr. Priyanka Sharma is currently working as a Research Associate under National Hydrology Project in Groundwater Hydrology Division at National Institute of Hydrology (NIH), Roorkee, India. She completed her B.Tech (Agricultural Engineering) from Chandra Shekhar Azad University of Agriculture and Technology, Kanpur, India, in 2012. She obtained her M.Tech in 2014 and PhD in 2018 from Maharana Pratap University of Agriculture and Technology (MPUAT), Udaipur. Between March and June 2016, she worked as a Senior Research Fellow in Department of Soil and Water Engineering, College of Technology and Engineering (CTAE), MPUAT, Udaipur, India. From January to June 2018, Priyanka worked as an Assistant Professor at the School of Agriculture, Lovely Professional University, Punjab, India. She also worked as an Assistant Professor in the Faculty of Agriculture Science, Maharishi Arvind University, Jaipur, Rajasthan, India. Her research interests include application of statistical and stochastic time series modeling techniques and modern data-driven techniques such as artificial intelligence in solving problems related to hydrology and water resources. She has published seven research papers in reputed peer-reviewed journals and conferences. She has also contributed three book chapters. She has been conferred with JAE Best Paper Award and Distinguished Scientist Associate Award for her outstanding research works in the field of hydrology. She is a life member of two national professional societies.

Dr. Deepesh Machiwal is a Principal Scientist (Soil and Water Conservation Engineering) in Division of Natural Resources at ICAR-Central Arid Zone Research Institute (CAZRI), Jodhpur, India. He obtained his PhD from Indian Institute of Technology, Kharagpur, in 2009. He has more than 20 years of experience in soil and water conservation engineering and groundwater hydrology. His current research area is modeling groundwater levels in Indian arid region under the changing climate and groundwater demands. Deepesh served from 2005 to 2011 as an Assistant Professor in the all India coordinated research project on groundwater utilization at College of Technology and Engineering, Udaipur, India. He has worked as co-principal investigator in three externally funded research projects funded by ICARDA, ICAR, and Government of Rajasthan, India. He has authored 1 book, edited 2 books, and contributed 19 book chapters. Deepesh has to his credit 39 papers in international and 19 papers in national journals, 2 technical reports, 4 extension bulletins, 16 popular articles, and 33 papers in conference proceedings. His authored book entitled, Hydrologic Time Series Analysis: Theory and Practice, has been awarded by Outstanding Book Award for 2012-13 from ISAE, New Delhi, India. He has been awarded Commendation Medal Award in 2019 by ISAE, Best Paper Award 2018 by CAZRI, Jodhpur, Achiever Award 2015 by SADHNA, Himachal Pradesh, Distinguished Service Certificate Award for 2012-2013 by ISAE, and IEI Young Engineer Award in 2012 by the Institution of Engineers (India), West Bengal. He is a recipient of Foundation Day Award of CAZRI for 2012, 2013, and 2014 and Appreciation Certificate from IEI, Udaipur, in 2012. Earlier, he was awarded Junior Research Professional Fellowship by IWMI, Sri Lanka, to participate in International Training and Research Program on Groundwater Governance in Asia: Theory and Practice. He has been conferred with Second Best Comprehensive Group Paper Award by IWMI, Sri Lanka, in 2007. He was also sponsored by FAO, Rome, and UN-Water for participating in two international workshops at China and Indonesia. He is a life member of eight professional societies and associations. Currently, Deepesh is serving as an Advisory Board Member of Ecological Indicators (Elsevier) and has served as an Associate Editor for Journal of Agricultural Engineering (ISAE) during 2018-20. He is a reviewer of several national and international journals related to soil and water engineering and hydrology.

Foreword

Water-related disasters, also called hydro-hazards, are among the most frequently occurring natural hazards that threaten people as well as socioeconomic development. The Emergency Events Database of 2019 shows that globally water-related disasters accounted for about 74% of all natural disasters between 2001 and 2018, and floods and droughts alone caused more than 166,000 deaths during the past 20 years, affected over 3 billion people, and caused an economic damage of almost US$700 billion worldwide. Unfortunately the number of people to be impacted and the economic losses to be caused are projected to rise in the future due to growing population in flood-prone areas, climate change, global warming, deforestation, loss of wetlands, increasing hurricanes, unplanned development, and rising sea level. Flood hazards may be mitigated through flood preparedness involving the development of flood risk management systems which depend on streamflow modeling and forecasting.

Reliable and accurate streamflow forecasting is essential for the optimal planning and management of water resources systems. A variety of models, ranging from knowledge-based physical models to data-driven empirical models, have been proposed over the years to derive streamflow forecasts. The past few decades have witnessed a proliferation in data-driven models for streamflow forecasting. This past decade, in particular, has seen a precipitous shift from traditional models employing stand-alone data-driven techniques to advanced hybrid models integrating more than one data-driven technique with some sort of a data preprocessing technique to improve forecast accuracy. Even the classical autoregressive moving average (ARMA) models have been advanced into nonlinear autoregressive with exogenous input (NARX) models, self-exciting threshold autoregressive (SETAR) models, generalized autoregressive models with conditional heteroscedasticity (GARCH), and hybridized SETAR-GARCH models to improve the accuracy of streamflow forecasting. Details about such advances are normally not found in a single source. Furthermore, several advances have been made in artificial intelligence techniques, such as artificial neural networks, support vector machines, support vector regression, and genetic programming with both stand-alone and hybrid approaches that have improved model performance and have provided better streamflow forecasts. Therefore, this book is a valuable contribution to the field of hydrology where traditional as well as modern models of streamflow forecasting involving statistical, stochastic, and artificial intelligence techniques are described along with case studies selected from different parts of the world that illustrate their applications to real-world data. The book also provides an overview of major approaches employed in advancing streamflow forecasting over half a century. Each book chapter maintains a proper balance of theory and practical demonstration of the underlying data-driven technique employed for streamflow forecasting. The book provides information on data-driven tools for streamflow forecasting, which will guide the reader in selecting a suitable technique to forecast streamflow under a given set of conditions. Chapter contributors are active researchers who have been involved in advancing streamflow forecasting. The book editors deserve applause for bringing out this valuable book.

Vijay P. Singh, P.D., D.S., P.E., P.H., Hon. D. WRE

University Distinguished Professor

Regents Professor

Caroline and William N. Lehrer Distinguished Chair in Water Engineering

Department of Biological and Agricultural Engineering and

Zachry Department of Civil and Environmental Engineering,

Texas A&M University, College Station, TX, United States

Preface

Streamflow forecasting plays a key role in water resources planning and management including irrigated agriculture, hydropower production, and destructive natural disasters. Accurate forecasting of streamflow is a challenging task and at the same time a very complex process. Therefore, researchers from all over the world have proposed many models to forecast streamflow with improved performance. Until now, a lot of research has been conducted to address this complex process by using many data-driven models. Accurate streamflow forecasting has received much attention over the last two decades. Since 2010, efforts of researchers in increasing forecasting accuracy of streamflow have gained a considerable momentum for long-term water resources planning and management. Before 2000s, academicians and researchers mainly focused on streamflow forecasting by using traditional methods such as statistical and stochastic time series modeling methods. However, in 2010s, focus of the researchers has been shifted toward the advances in streamflow forecasting methods by adopting new stand-alone data-driven approaches such as artificial intelligence techniques and hybrid data-driven approaches where more than one technique, preferably artificial intelligence technique, is amalgamated with other methods. Furthermore, it is learned that the recent advances made in the subject of streamflow forecasting are mostly confined up to research articles where detailed procedures of applying the modern methods for streamflow forecasting are not adequately dealt. It is revealed from the literature that a book describing the step-by-step procedures of the advanced data-driven models along with their case studies in streamflow forecasting is not available. Therefore, this book is an attempt to abridge this gap by providing theoretical descriptions, systematic methodologies, and practical demonstration of the traditional and modern tools and techniques adopted for streamflow forecasting over the years. Thus, the book is very useful for the readers to gain an insight about the developments made in the field of the streamflow forecasting. Furthermore, this book deals with theory-to-practice approach where procedures for applying the different methods are explained in concrete and ordered steps that may be easily followed by the readers if they wish to apply any of the methods in their studies. It may be difficult for the readers to find all such details about the advanced methods at a single source in the literature. Moreover, the methods detailed in this book may be useful for the readers to make forecast of a variable in any subject including hydrology and water resources engineering.

This book comprises a total of 13 chapters that cover most-promising hydrological data-driven prediction approaches, which have been used to develop accurate yet efficient forecasting models and forecast streamflows during the earlier and recent times. Chapter 1 provides a detailed overview of the streamflow forecasting models and advances made in traditional as well as modern data-driven techniques. A comprehensive review of the literature is provided based on the studies related to streamflow forecasting at different timescales based on data-driven techniques. Chapters 2 and 3 describe traditional methods and their recent advances including statistical linear and nonlinear time series models such as exponential smoothing and autoregressive fractionally integrated moving average (ARFIMA) models, vector autoregressive without/with exogenous variables (VAR/VARX), and multiple/multivariate generalized autoregressive conditional heteroscedasticity (MGARCH). Chapters 4–10 contain the advance stage of development and verification of highly complex or nonlinear streamflow forecasting models involving artificial intelligence approaches. These chapters explain the concept, procedure, and application of stand-alone artificial intelligence models such as artificial neural network, adaptive neuro-fuzzy inference system, genetic programming, gene expression programming, model tree technique, support vector regression, and extreme learning machines to forecast streamflows. These chapters also involve the comparison among the salient artificial intelligence methods. When the streamflow data are highly nonstationary, artificial intelligence methods may not be able to simulate and forecast the streamflow without pre/postprocessing of the input/output data. Recently, in such situations, hybrid approaches that combine data preprocessing and artificial intelligence techniques are being increasingly adopted as an important tool to improve the forecast accuracy of streamflow. Hence, Chapters 12 and 13 of the book include the recent hybrid approaches that are progressively being used to improve the forecast accuracy and to reduce the uncertainties in streamflow forecasting. Editors believe that the book may be very much useful to students, researchers, and academicians as well as planners, managers, and policy-makers involved in sustainable development and management of water resources.

Priyanka Sharma

Deepesh Machiwal

Acknowledgment

This book would have not been possible without support of many people. The first editor (Priyanka Sharma) would like to express her sincere gratitude and due respect to Dr. Surjeet Singh, Scientist F, and Dr. J.V. Tyagi Director, National Institute of Hydrology (NIH), Roorkee, India. She is immensely grateful for their valuable support and continuous encouragement. The second editor (Deepesh Machiwal) gratefully acknowledges the support and motivation provided by Dr. O.P. Yadav, Director, ICAR-Central Arid Zone Research Institute (CAZRI), Jodhpur, India. Deepesh would like to express his feelings to his mentor, Dr. Madan Kumar Jha, Professor, Indian Institute of Technology (IIT) Kharagpur, India, who has been a constant source of inspiration to him. He further feels a sense of indebtedness to Dr. Adlul Islam, Principal Scientist, NRM Division, Indian Council of Agricultural Research (ICAR), New Delhi, India, for his stimulation to do something creative. The editors acknowledge the generous support and inspiration of their friends and colleagues received during the entire course of this book project.

The editors thank all the chapter contributors for their selfless and quality contribution to this book. They sincerely appreciate a lot of valuable insights arising from experience and dedication of the contributors that enriched the material presented in the book. They would like to thank all the reviewers for their intuitive suggestions and comments, which have led to numerous improvements in content of the chapters. Each chapter of the book has been revised at least twice. The editors are glad to have a foreword from Professor Vijay P. Singh, Department of Biological & Agricultural Engineering and Zachry Department of Civil & Environmental Engineering, Texas A&M University, College Station, TX, United States.

The editors are thankful to the Elsevier book project team that assisted them from time to time at various steps of publishing process since the beginning of this book project. The editors are delighted to specially acknowledge Louisa Munro (Acquisitions Editor: Aquatic Sciences), Fisher Michelle (Acquisitions Editor for Molecular Biology), Hannah Makonnen (Editorial Project Manager), Aleksandra Packowska (Editorial Project Manager), and Kiruthika Govindaraju (Senior Project Manager) for their professional assistance throughout the publishing process of the book. A special word of appreciation is also extended to the designer team of the publisher for their prompt actions in revising and improving the cover page of the book.

Priyanka Sharma

Deepesh Machiwal

Chapter 1: Streamflow forecasting

overview of advances in data-driven techniques

Priyanka Sharma ¹ , and Deepesh Machiwal ²       ¹ Groundwater Hydrology Division, National Institute of Hydrology, Roorkee, Uttarakhand, India      ² Division of Natural Resources, ICAR-Central Arid Zone Research Institute, Jodhpur, Rajasthan, India

Abstract

Reliable and realistic streamflow forecasting plays a crucial role in hydrology and water resources engineering as it can directly affect the dams operation and performance, groundwater recharge/exploitation, sediment conveyance capability of river, watershed management, etc. However, an accurate streamflow forecasting is not an easy task due to the high uncertainty associated with climate conditions and complexity of collecting and handling both spatial and nonspatial data. Therefore, hydrologists from all over the world have developed and adopted several types of data-driven techniques ranging from traditional stochastic time series modeling to modern hybrid artificial intelligence (AI) models for future prediction of streamflow. In literature, studies dealing with streamflow forecasting used a variety of techniques having dissimilar concepts and characteristics and streamflow datasets at different timescales such as daily, monthly, seasonal, and yearly, etc. This chapter first describes and classifies available data-driven techniques used in streamflow forecasting into suitable groups depending upon their characteristics. Then, growth of the salient data-driven models both single and hybrid such as time series models, artificial neural network models, and other AI models is discussed with their applications and comparisons as reported in studies on streamflow forecasting over time. Thereafter, current approaches used in the recent 5-year streamflow forecasting studies are briefly summarized. Also, challenges experienced by the researchers in applying data-driven techniques for streamflow forecasting are addressed. It is concluded that a vast scope exists for improving streamflow forecasts using emerging and modern tools and combining them with location-specific and in-depth knowledge of the physical processes occurring in the hydrologic system.

Keywords

Artificial intelligence technique; Data-driven models; Hybrid models; Streamflow forecasting; Time series modeling

1.1. Introduction

Runoff water generated from the precipitation may reach a stream by overland flow, subsurface flow, or both and move toward the oceans in a channelized form and is called streamflow or river flow. Streamflow is generated by a combination of baseflow (return from groundwater), interflow (rapid subsurface flow through macropores and seepage zones), and saturated overland flow (Mosley and McKerchar, 1993). A schematic diagram of three components of streamflow, i.e., interflow, saturated overland flow, and baseflow, is shown in Fig. 1.1 under absence and presence of rainfall event. Streamflow, expressed as discharge in units of cubic feet per second (ft³/s) or cubic meters per second (m³/s), is the only phase of the hydrological cycle in which the water is confined in well-defined channels which permit accurate measurements to be made of the quantities involved. At the same time, streamflow is one of the most complex quantitative parameters that takes place in a stream or channel and varies in time and space (Wiche and Holmes, 2016). Analysis of the streamflow data provides us description of river flow regime, enables us to compare between rivers, and helps in prediction of future river flows (Davie, 2008). Streamflow is experiencing long-term changes as the freshwater demands are increasing worldwide due to increase in population and changing climate (Oki and Kanae, 2006). However, it is difficult to predict the changes in future streamflows as it involves a physical process that depends upon more than one random variable such as precipitation, evapotranspiration, topography, and human activities. Hence, it is a very complex and nonlinear process of hydrologic cycle that is not well understood. This necessitates prediction or forecasting of future streamflows for efficient water resources planning and management. Streamflow forecasting may be performed for short-term for defending from floods, medium-term for reservoir operation, and long-term for water resources planning and management.

Figure 1.1 Schematic diagram showing interflow, saturated overland flow, and baseflow components of streamflow during (A) dry period and (B) rainy period. 

Modified from Mosley, M.P. and McKerchar, A.I. 1993. Streamflow, chapter 8. In: D.R. Maidment (Editor In Chief), Handbook of Hydrology, McGraw-Hill, Inc., New York, pp. 8.1–8.35.

Streamflows have been forecasted using both physically-based and data-driven models. In literature, a wide range of both types of models has been proposed and applied in streamflow forecasting. The subject of streamflow forecasting has been flourishing well with a large number of studies being continuously reported over the past five decades where numerous forecasting approaches are used. Hydrologists are dynamic in proposing and adopting new tools and techniques and/or refining the existing methods in streamflow forecasting to overcome the drawbacks of the older approaches. It has been revealed that the streamflow forecasts are mainly affected by the great uncertainty especially in ungauged or poorly gauged basins where good-quality streamflow data of adequate time period are not available. In addition, chaotic behavior of streamflow process with nonlinear and nonstationary time series is another major reason of unsatisfactory forecasts. Therefore, researchers have made advances in understanding the physical process as well as in analyzing the process empirically using modern computing techniques over the years. This chapter aims at presenting advances made in streamflow forecasting using data-driven techniques/models. First, data-driven models ranging from traditional to modern techniques are classified into suitable groups depending on the nature of the models. Then historical development and application of different data-driven models in streamflow forecasting is detailed. Furthermore, a section of comparative evaluation of data-driven models is presented and current trends in the recent studies are highlighted. Moreover, key challenges experienced in making accurate streamflow forecasts are discussed and concluding remarks are provided.

1.2. Measurement of streamflow and its forecasting

Measuring, estimating, and/or predicting streamflow is an important task in surface water hydrology. There exist a variety of methods for monitoring streamflow and each method remains specific to a particular type of stream. The methods to quantify and monitor the streamflow are grouped into four categories (Dobriyal et al., 2017): (i) direct measurement methods, (ii) velocity-area methods, (iii) constricted flow methods, and (iv) noncontact measurement methods (Fig. 1.2). An overview of the methods used for streamflow monitoring is provided by Mosley and McKerchar (1993), and their advantages and disadvantages are presented by Dobriyal et al. (2017). Suitability of a method for streamflow monitoring depends on the factors such as water quantities to be measured, degree of accuracy, permanent or temporary installation, and cost incurred (Dobriyal et al., 2017).

Many activities associated with the planning, operation, management, and control of a water resource system require forecasts of future streamflow, which is a challenging task for water resources engineers, planners, and managers. Accurate streamflow forecasts are needed for the efficient operation of water resources systems within technical, economical, legal, and political priorities (Salas et al., 2000). Streamflow forecasts should take into account spatial and temporal variability of entire streamflow field for a sound control and management of the water resources system. It is worth-mentioning that streamflow forecasting can be of two types depending upon the temporal scale at which forecasts are made (Yaseen et al., 2015): (i) short-term or real-time forecasting with lead times of hours and days, which is crucial for real-time reservoir operation and reliable operation of flood warning and mitigation systems, and (ii) long-term forecasting with lead times varying from weeks to months, which is important for the planning and operation of reservoirs, hydropower generation, sediment transport, drought analysis and mitigation, irrigation management decisions, scheduling releases and many other applications.

Figure 1.2 Classification of methods used for streamflow monitoring. 

Modified from Dobriyal, P., Badola, R., Tuboi, C. and Hussain, S.A. 2017. A review of methods for monitoring streamflow for sustainable water resource management. Applied Water Science, 7: 2617–2628.

1.3. Classification of techniques/models used for streamflow forecasting

In literature, the main objective of employing streamflow forecasting models is studying the operation of a hydrologic system in order to predict its behavior. Thus, streamflow prediction and forecasting has become one of the most important issues in hydrology over the past few decades and is currently attracting hydrologists to advancing research on accurate hydrologic predictions. In this section, a brief overview of the state-of-the-art scientific approaches applied in hydrology especially for streamflow modeling and forecasting is provided. Presently, there exist a variety of techniques used for modeling a hydrologic process, and it is very imperative to first distinguish among several types of techniques/models used to enhance the reliability and accuracy of forecasting hydrological variables and to account for the scale- and time dependency of their errors in evaluation. Over the last three to four decades, a large number of models have been proposed for hydrologic time series prediction and forecasting and to improve the hydrologic forecasting accuracy. Broadly, these models are classified into three categories (Bourdin et al., 2012; Devia et al., 2015; Liu et al., 2018): (i) physical-based models, (ii) conceptual models, and (iii) black-box models (Fig. 1.3). The physical-based models, also called white-box or process-based models, describe hydrological characteristics in detail by solving differential equations describing the physical laws of mass, energy, and momentum conservations. The conceptual models, also known as gray-box models, are a descriptive representation of hydrologic system that incorporates the modeler’s understanding of the relevant physical, chemical, and hydrologic conditions. On the other hand, black-box models, sometimes called empirical or statistical models, are based on input–output relationships from a statistical point of view rather than physical principles. The latter two categories of the models do not describe the underlying hydrologic processes. Now-a-days, the hydrologic models are classified into only two categories