Registration is open for the next AIH Webinar! Register here.

AIH members = $0.00 (contact admin@aihydrology.org or (540) 500-1933 for the promotion code)
Lapsed/non-members = $35.00

WEBINAR SUMMARY:
Recently published literature has confirmed time and time again that machine learning (ML) algorithms (including LSTMs, GRUs, and Transformers) and classical lumped hydrological models (such as SAC-SMA and HBV) perform more reliably in hindcast and forecast flood prediction intercomparison experiments than more sophisticated high-resolution hydrological models. These provocative results have challenged decades of development of physics-based hydrological models for streamflow prediction, which seem to be more sensitive to the errors present in forcing precipitation data, and the spatial description of landscape attributes. Thus, the long-standing promise that a better and more detailed understanding and description of hydrological processes would yield better predictions of streamflow fluctuations (including floods, droughts, etc.) is yet to be fulfilled. I will show results from recent numerical experiments using high-resolution physics-based distributed hydrological models that shed some light into the applicability and limits of data-driven tools to predict the hydrological cycle under conditions of non-stationarity and noisy input information. The analysis is extended to investigate the class of statistical tools used for Regional Flood Frequency Analysis (RFFA) to determine the physical conditions required for classical techniques such as Index Flood Method. The results of this investigation highlight the need to establish causal physical links to the patterns in data reveled by data-driven analysis tools.

LEARNING OBJECTIVES:

1. Evaluate the minimum requirements of data-driven tools to provide acceptable predictions of the hydrological cycle.
2. Recognize the limited information provided by classical performance metrics of data-driven models for streamflow prediction.
3. Design numerical experiments to test the capabilities of data-driven tools that are used in practical engineering applications.

SPEAKER:

Ricardo Mantilla
Associate Professor
University of Manitoba

Ricardo Mantilla is an Associate Professor in the Civil Engineering Department at the University of Manitoba. Ricardo holds a BSc and an MSc degree in Civil Engineering from Universidad Nacional de Colombia and PhD degree from the University of Colorado at Boulder in Civil and Environmental Engineering. Ricardo led the development of a real-time fully autonomous operational flood forecasting system for the State of Iowa and a prototype forecasting system for the State of Nebraska as part of the Iowa Flood Center activities. His current research is focused on developing and testing state-of-the-art technologies and techniques for flood forecasting. His core research is devoted to developing theories that use the observed self-similarity in river networks and landscape features to create accurate and efficient models for multiscale predictions of the hydrologic cycle.

Click here to register for the webinar!
Participating in this webinar qualifies as a continuing education credit for professional hydrologists. 1 Contact hour = 1 PDH/PDC. Learn more about AIH’s continuing education guidance online here.

Interested in the AIH join/certification process? Learn more here.