Predictive Analysis of Arizona Dust Storms (1996–2023): GIS, Artificial Intelligence (MaxEnt), and Spatiotemporal Data

<p>This dissertation looks at how well Maximum Entropy (MaxEnt) model can predict duststorms in Arizona, United States. MaxEnt is a machine learning tool that was first developed to predict where different species might live. MaxEnt uses presence-only data which makes it especially useful in environments where reliable absence data is hard to find or unavailable. This study analyzes if the combination of MaxEnt within Geographic Information Systems (ArgGIS Pro) and spatiotemporal data can provide data-driven and reliable predictions of dust storms susceptibility in Arizona. The framework that was used for this study is Design Science Research because of its problem-solving tendency and its emphasis on designing and evaluating artifacts that tackle real-world problems. Ecological Niche Theory (Hutchinsonian Niche Theory) was used as the theoretical basis for this research, as the concept of ‘niche’ as a set of environmental factors under which a species can exist can also apply to dust storms. The analysis utilizes historical dust storms instances from 1996 to 2023 and integrated land cover and land use (LCLU), effective specific humidity, specific humidity, wind speed, maximum wind speed, elevation, and temperature. These variables were selected as, based on the literature review, they represent drivers that directly influence dust storms creation. The resultsshow that temperature, elevation and wind speed are all positively linked with dust storms, while effective specific humidity has a negative relationship with dust storm occurrences. Furthermore, LCLU categories such as developed open space, developed low, medium, and high-intensity areas correspond to a higher susceptibility while shrubland and evergreen forest areas demonstrate a negative association. As part of external evaluation, an independent dust storms data sets from Sistan watershed in Iran was run by MaxEnt, in which MaxEnt achieved an AUC of 73%, to further underscore its utility in dust storm prediction.</p>

Name thesis:degree_name

Information Systems and Technology, PhD

Level thesis:degree_level

Open Access Dissertation

Discipline thesis:degree_discipline

Center for Information Systems and Technology

Year dc:date.available

2025

Author dc:creator

• Bouzari, Soheil

Contributors dc:contributor

• June Hilton
• Chinazunwa Uwaoma

• Dust storms prediction
• Geographic Information Systems
• Maximum Entropy
• Spatiotemporal data
• Land cover and land use
• Geographic Information Sciences

Repository record dc:identifier

https://scholarship.claremont.edu/cgu_etd/1041

OAI identifier oai:identifier

oai:scholarship.claremont.edu:cgu_etd-2063