Date of Award

May 2021

Degree Type

Thesis

Degree Name

Master of Science

Department

Computer Science

First Advisor

Zeyun Yu

Second Advisor

Qian Liao

Abstract

In the field of water resources management, one vital instrument utilized is the stream gage. Stream gages monitor and record flow and water height within some water body. The United States Geological Survey maintains a network of stream gages at many locations across the country. Many of these sites are also equipped with webcams monitoring the state of the water body at the moment of measurement. Previous studies have outlined methods to approximate stream gage data remotely with limitations such as the requirement of detailed depth information for each site. This study seeks to create a process for training a deep neural network that will utilize the webcam and stream gage data to generate a water height prediction based on the visual state of the water body. The goal of this study is to outline a training process and model that can be utilized on a variety of different sites while only requiring that location’s existing webcam and stream gage data. This paper outlines the experiments on stream gages located at the Clear Creek in Iowa, Auglaize River in Ohio, and Milwaukee River in Wisconsin. The process outlined utilizes transfer learning and well-known image classification models as a basis for a generalized river height regression model. Across the training, validation, and deployment experiments, the developed process shows great success in creating an accurate model for various sites of different conditions and river clarity. The results of this study show confidence in future studies utilizing remote stream gaging with image regression.

Download

Share

COinS