Machine learning and remote sensing in the physical sciences : applications in Earth surface and atmospheric modeling.

It is imperative that the observable metrics of land and atmospheric systems (i. e., soil moisture and cloud formation) be accurately quantified prior to simulating water and energy transfer, where the components of flux are often difficult or impossible to observe. In this work, I employ strategies to minimize prediction uncertainty from local to global scales by using remote sensing, physics-based modeling, and artificial neural networks (ANNs) to: (1) develop data-driven models with enhanced prediction capabilities for land-surface water balance and (2) determine signatures in observational atmospheric data that can be used to understand underlying physical processes of cloud formation that are not directly observable. Part 1 developed an ensemble of ANNs to estimate continental-scale water balance. Soil moisture (SM) is a key component in water budgets and there is a well-documented need for accurate SM estimates for use is risk analyses, as well as the simulation of long-term trends. ANNs were trained using the USDA's Soil & Water Assessment Tool inputs and weather time series data mapped onto NASA Soil Moisture Active Passive (SMAP) SM retrievals from January 1, 2016 – January 1, 2019. The best-performing model predicted SM with a mean error of 0.02 cm3 cm–3 with respect to SMAP SM retrievals. Part 2 analyzed aerosol-cloud interactions (ACIs) using NASA satellite retrievals and general circulation model simulations using the Global Earth Observing System version 5. Aerosols influence Earth's radiative balance both directly and indirectly by intercepting and scattering/absorbing incoming solar radiation and modifying cloud macro- and microphysical properties, respectively. This work analyzed remotely sensed and modeled ACIs following two volcanic events. These may be considered "natural experiments" where aerosol effects on clouds and climate can be partitioned from inherent effects from anthropogenic aerosols. Our results determined that ACIs for liquid clouds are strongly affected by aerosol concentrations while processes related to ice cloud formation are dominated by aerosol plume heights and ash concentrations.

Name thesis:degree_name

Ph.D.

Level thesis:degree_level

Doctoral

Grantor

Baylor University.

Year dc:date.issued

2020

Author dc:creator

• Breen, Katherine Heather, 1987-

Advisor dc:contributor.advisor

• James, Scott C.

• Remote sensing.
• Machine learning.
• Artificial neural network (ANN)
• Soil moisture.
• Aerosol-cloud interactions.

Statement dc:rights

• Baylor University works are protected by copyright. They may be viewed from this source for any purpose, but reproduction or distribution in any format is prohibited without written permission. Contact libraryquestions@baylor.edu for inquiries about permission.

Language dc:language.iso

en

Handle dc:identifier.uri

https://hdl.handle.net/2104/11116

OAI identifier oai:identifier

oai:baylor-ir.tdl.org:2104/11116