Modeling the temperature response of small rivers to land cover changes using satellite-based spatial data

The influence of urbanization and land cover alteration on water quality, including river temperatures, has important ecological implications. However, there is a dearth of information on temperature of rivers smaller than about 60 m wide (approximately fifth order and below), which constitute roughly 97% of total global stream length: for such rivers, field collection of temperature data is labor intensive and often covers short time scales, while satellite-based temperature predictions are often inaccurate for smaller rivers. This lack of high-resolution spatial temperature data has hindered large scale assessment of river temperature patterns both spatially and temporally. This work aims to model the temperature of small, especially urban, rivers across the contiguous United States without requiring field data. To generate sufficiently high-resolution temperature data, a machine learning model was developed for high-accuracy, satellite-based stream temperature estimation, trained with United States Geological Survey stream temperature gage data. The first model developed (called “TempEst”, for Temperature Estimation) had a median validation Root Mean Square Error across gages of about 1.7 K. Building on the outputs from this model as well as validation against paired stream temperature gages, a second machine learning model was developed to predict the relationship between longitudinal changes in river temperatures and nearby land cover conditions, both along the riverbanks and in the general vicinity (without the specific use of satellite-based temperature data). Validated against upstream-downstream pairs of stream temperature gages, the second model (called “LOST”, for Longitudinal Stream Temperatures) had a median downstream Root Mean Square Error across rivers of 1.2 °C. LOST was developed to predict changes in temperature of small rivers anywhere in the contiguous United States based on publicly available land cover, climate, and regional data.

Name thesis:degree_name

Master of Science (M.S.)

Level thesis:degree_level

Masters

Discipline thesis:degree_discipline

Civil and Environmental Engineering

Grantor dc:publisher

Colorado School of Mines. Arthur Lakes Library

Year dc:date.issued

2022

Author dc:creator

• Philippus, Daniel

Advisor dc:contributor.advisor

• Hogue, Terri S.

Committee members dc:contributor.committeemember

• Anderson, Eric J.
• Sytsma, Anneliese
• Rust, Ashley J.

• land cover
• landsat
• machine learning
• remote sensing
• stream temperature

Statement dc:rights

• Copyright of the original work is retained by the author.

Language dc:language.iso

eng, English

Identifier

T 9472

OAI identifier oai:identifier

oai:repository.mines.edu:11124/176630