Project Info
Enhancing Streamflow Forecasting through Machine Learning Techniques
Adrienne Marshall
adriennemarshall@mines.edu
Project Goals and Description:
Estimating streamflow plays a pivotal role in the management of water resources, serving various critical functions such as mitigating floods, issuing drought warnings, and optimizing reservoir operations. However, predicting streamflow presents significant challenges due to the inherent complexity and variability of the hydrological processes involved. Traditionally, streamflow prediction has relied on physically based models that simulate the processes governing the movement of water through a watershed. These models are grounded in the fundamentals of hydrology, requiring detailed information about the watershed’s physical properties and the climatic conditions. While they can be informative, these models are often computationally intensive and require extensive calibration and validation. In contrast, data-driven methods have emerged as a powerful alternative for streamflow forecasting, leveraging the growing availability of hydrological and meteorological data. These methods use machine learning algorithms to identify patterns and relationships in the data without requiring detailed physical descriptions of the watershed. In recent decades, a variety of machine learning techniques have been applied to streamflow forecasting, including neural networks (NNs), support vector machines (SVMs), fuzzy logic, wavelet transforms, and long short-term memory (LSTM) networks. This project's primary goal is to develop a robust machine learning algorithm for forecasting streamflow at the basin scale. A data-rich basin will be chosen for the initial analysis to ensure comprehensive, high-quality data for model training and validation. The model's predictions will be compared with observed data and outputs from process-based models. This project aims to enhance both the accuracy and efficiency of streamflow forecasting, making it a more adaptable tool for water resource management.
More Information:
Grand Challenge: Provide access to clean water.
The MURF student will work with a graduate student funded on <a href="https://www.nsf.gov/awardsearch/showAward?AWD_ID=2241892">NSF Award #2241892</a>.
<em>Additional References</em>:
Cheng, M., Fang, F., Kinouchi, T., Navon, I. M., & Pain, C. C. (2020). Long lead-time daily and monthly streamflow forecasting using machine learning methods. <em>Journal of Hydrology</em>, <em>590</em>, 125376. <a href="https://doi.org/10.1016/j.jhydrol.2020.125376">https://doi.org/10.1016/j.jhydrol.2020.125376</a>
Khand, K., & Senay, G. B. (2024). Evaluation of streamflow predictions from LSTM models in water-and energy-limited regions in the United States. <em>Machine Learning with Applications</em>, <em>16</em>, 100551.
Konapala, G., Kao, S. C., Painter, S. L., & Lu, D. (2020). Machine learning assisted hybrid models can improve streamflow simulation in diverse catchments across the conterminous US. <em>Environmental Research Letters</em>, <em>15</em>(10), 104022. 10.1088/1748-9326/aba927
Parisouj, P., Mohebzadeh, H., & Lee, T. (2020). Employing machine learning algorithms for streamflow prediction: a case study of four river basins with different climatic zones in the United States. <em>Water Resources Management</em>, <em>34</em>(13), 4113-4131. <a href="https://doi.org/10.1007/s11269-020-02659-5">https://doi.org/10.1007/s11269-020-02659-5</a>
Primary Contacts:
Adrienne Marshall, <a href="mailto:adriennemarshall@mines.edu">adriennemarshall@mines.edu</a> | Surabhi Upadhyay, <a href="mailto:surabhi_upadhyay@mines.edu">surabhi_upadhyay@mines.edu</a>
Student Preparation
Qualifications
<ul>
<li>Enthusiastic in learning about streamflow and hydrology.</li>
<li>Ability to handle large datasets and perform intensive computations in Python or R or willingness to learn.</li>
<li>Basic knowledge of or willingness to learn machine learning algorithms.</li>
</ul>
TIME COMMITMENT (HRS/WK)
4-5
SKILLS/TECHNIQUES GAINED
Throughout this project, the student will acquire a comprehensive understanding of streamflow and hydrological processes. They will gain practical experience in handling large datasets, data preprocessing and visualization techniques, enhancing their ability to derive meaningful insights from hydrological data. Moreover, the student will delve into machine learning concepts tailored to the water domain, enabling them to construct and validate predictive models for streamflow forecasting.
MENTORING PLAN
Initially, an introductory session will orient the student(s) to the project's goals and methods. Subsequently, regular meetings (weekly or bi-weekly depending on the load and schedule) will be scheduled to monitor progress, address challenges, and provide feedback for next steps. Moreover, educational resources like research papers and tutorials will be provided to enrich the student's comprehension and aid their learning process. Furthermore, the student will be motivated to showcase their work at Mines seminars, conferences, or similar events, and to pursue publication opportunities as a primary or co-author.
Preferred Student Status
Sophomore
Junior
Senior