The geometry of flow: Advancing predictions of river geometry with multi-model machine learning

Shuyu Y Chang,Zahra Ghahremani,Laura Manuel,Mohammad Erfani,Chaopeng Shen,Sagy Cohen,Kimberly Van Meter,Jennifer L Pierce,Ehab A Meselhe,Erfan Goharian

2023-11-28

Abstract:Hydraulic geometry parameters describing river hydrogeomorphic is important for flood forecasting. Although well-established, power-law hydraulic geometry curves have been widely used to understand riverine systems and mapping flooding inundation worldwide for the past 70 years, we have become increasingly aware of the limitations of these approaches. In the present study, we have moved beyond these traditional power-law relationships for river geometry, testing the ability of machine-learning models to provide improved predictions of river width and depth. For this work, we have used an unprecedentedly large river measurement dataset (HYDRoSWOT) as well as a suite of watershed predictor data to develop novel data-driven approaches to better estimate river geometries over the contiguous United States (CONUS). Our Random Forest, XGBoost, and neural network models out-performed the traditional, regionalized power law-based hydraulic geometry equations for both width and depth, providing R-squared values of as high as 0.75 for width and as high as 0.67 for depth, compared with R-squared values of 0.57 for width and 0.18 for depth from the regional hydraulic geometry equations. Our results also show diverse performance outcomes across stream orders and geographical regions for the different machine-learning models, demonstrating the value of using multi-model approaches to maximize the predictability of river geometry. The developed models have been used to create the newly publicly available STREAM-geo dataset, which provides river width, depth, width/depth ratio, and river and stream surface area (%RSSA) for nearly 2.7 million NHDPlus stream reaches across the rivers and streams across the contiguous US.

Geophysics,Machine Learning

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is to improve the accuracy of predicting river width and depth. Traditionally, methods based on power - law relationships (such as those proposed by Leopold and Maddock) have been widely used to understand and predict river systems and flood inundation situations. However, these methods have limitations, especially when dealing with rivers in different geographical regions and stream orders, and their predictive ability is limited. Therefore, the authors of this paper propose to use machine - learning models to go beyond the traditional power - law relationships in order to provide more accurate predictions of river width and depth.

Specifically, the authors used machine - learning models such as Random Forest, XGBoost, and neural networks, combined with an unprecedented large - scale river measurement data set (HYDRoSWOT) and a series of watershed prediction data, to develop new data - driven methods to better estimate the river geometric characteristics in the contiguous United States (CONUS). The research results show that these machine - learning models are significantly better than the traditional area - based power - law equations in predicting river width and depth. In particular, the performance differences in different stream orders and geographical regions also demonstrate the value of the multi - model machine - learning approach.

Through this research, the authors not only evaluated the performance of data - driven methods in estimating river geometric characteristics, but also developed a comprehensive, high - quality, and fine - grained river geometric database for the contiguous United States - STREAM - geo. This database provides information on river width, depth, width - to - depth ratio, and river and stream surface area (%RSSA) for nearly 2.7 million NHDPlus river segments. This provides a basis for improving the theoretical knowledge of continental - scale hydrological models and customizing end - to - end models, and helps to improve the understanding of the river systems in the United States.