Date of Award

12-2022

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Civil Engineering

Committee Chair/Advisor

Dr. Abdul Khan

Committee Member

Dr. Ashok Mishra

Committee Member

Dr. Nigel Kaye

Committee Member

Dr. Ravi Ravichandran

Abstract

The spatial and temporal distribution of temperature and precipitation determines the availability of water resources for ecosystems. The hydrological cycle can be disrupted by differences in meteorological variables and the random nature of water demand. Horizontal visibility degradation is an indicator of the changes in local meteorological variables, rainfall deficits, and drought conditions. Long-term analysis of meteorological variables and visibility records helps understand the causes of visibility degradation. The first objective of this dissertation is to investigate the long-term variability of horizontal visibility along with metrological variables based on annual and seasonal time scales over Saudi Arabia. The second objective is to develop classification models to predict low visibility events and the levels of PM10­ based on an hourly time scale in Riyadh, Saudi Arabia.

This dissertation examines trend direction/magnitudes and abrupt changes for the seasonal and annual time series of horizontal viability and maximum temperature over Saudi Arabia. The spatial and temporal variability of horizontal visibility, along with the Standardized Precipitation Evapotranspiration Index (SPEI) and meteorological variables, are investigated for the study period and two subperiods to comprehensively understand the interrelationships between the variables. In addition, a binary machine learning classifier is developed using five meteorological variables and four pollutants, based on different feature selection methods and sampling approaches, to select the best model for predicting hourly low-visibility events. Furthermore, a multi-classification task is developed to predict hourly PM10 levels for four-time windows based on only meteorological variables. The Synthetic Minority Oversampling Technique (SMOTE) is used for oversampling PM10 minority classes to achieve a balanced dataset. The dataset oversampled using SMOTE is compared to the imbalanced dataset for all time windows.

Available for download on Sunday, December 31, 2023

Share

COinS