EGU23-10952, updated on 26 Feb 2023
https://doi.org/10.5194/egusphere-egu23-10952
EGU General Assembly 2023
© Author(s) 2023. This work is distributed under
the Creative Commons Attribution 4.0 License.

Trend Analysis and Forecasting of Streamflow in the Upper Narmada Basin using Random Forest (RF) and Long Short-Term Memory (LSTM) Models 

Siddik Barbhuiya1, Meenu Ramadas2, Suraj Jena3, and Shanti Biswal4
Siddik Barbhuiya et al.
  • 1PhD Scholar, Indian Institute of Technology Bhubaneswar, School of Infrastructure, Odisha, India (sab12@iitbbs.ac.in)
  • 2Assistant Professor, Indian Institute of Technology Bhubaneswar, School of Infrastructure, Odisha, India (meenu@iitbbs.ac.in)
  • 3Postdoctoral Scholar, Department of Biological and Ecological Engineering, College of Agricultural Sciences, Oregon State University, USA(jenas@oregonstate.edu)
  • 4Research Associate, , School of Infrastructure, Indian Institute of Technology Bhubaneswar, Odisha, India(ssb14@iitbbs.ac.in)

In order to effectively plan, design, and manage water resources, it is necessary to understand the trends present in hydro-climatic variables such as streamflow and rainfall. In this study we used the Pettitt’s test as well as the standard normal homogeneity test (SNHT) to discover the trends in streamflow in the Upper Narmada Basin during the 1990 to 2018 period. The Upper Narmada basin extends over an area of 45, 580 square kilometers lies between latitudes 21°20’ N and 23°45' N and longitudes 72°32' E and 81°45’ E in India. From the flow records from gauges in this study basin, change points in the flow regime are thus identified.

Additionally, we performed Mann–Kendall (MK) test, modified Mann–Kendall (MMK) test, Sen's slope (SS) analysis to quantify the trends in streamflow time series. While the MK and MMK tests determine whether a trend is monotonically increasing or decreasing over time, SS suggests the rate of temporal change of streamflow variable. Further, we used advanced machine learning algorithms such as random forest (RF) and long short-term memory (LSTM) to develop flow forecasting models for few gauging sites in the study basin. In this way it is possible to address gaps in the flow records and perform long term analysis of gauge data.

Keywords: Trend analysis, Change point detection, Machine Learning Algorithm, LSTM, Upper Narmada Basin

 

 

 

How to cite: Barbhuiya, S., Ramadas, M., Jena, S., and Biswal, S.: Trend Analysis and Forecasting of Streamflow in the Upper Narmada Basin using Random Forest (RF) and Long Short-Term Memory (LSTM) Models , EGU General Assembly 2023, Vienna, Austria, 24–28 Apr 2023, EGU23-10952, https://doi.org/10.5194/egusphere-egu23-10952, 2023.