Electronic Theses and Dissertations

Ehsan Momeni

Date

2022

Document Type

Dissertation

Degree Name

Doctor of Philosophy

Department

Earth Sciences

Committee Chair

Angela Antipova

Committee Member

Reza Banai

Committee Member

Hsiang-te Kung

Committee Member

Esra Ozdenerol

Committee Member

Dorian Burnette

Abstract

Land use and land cover (LULC) data analyses disclosed large land conversion in Shelby County, TN, between 1990 and 2010. LULC mixtures have significant associations with socio-demographics and travel behavior. Mathematical modeling is used to forecast urban expansion in order to promote sustainable development. Cellular automata (CA) is a popular approach for the simulation of urban growth. Nevertheless, precise calibration of CA is challenging due to uncertainties and its knowledge-intensive process. Shannon relative indices (SRIs) have been used in this study as an indicator of land patterns to calibrate a CA model of urban growth in Shelby County. The results of using a Genetic Algorithm (GA) indicate that including patterns in the calibration improves the simulations’ accuracy (from 93.21% to 94.84%). Furthermore, an Imperialistic Competitive Algorithm (ICA) was implemented for the first time in the field of urban planning to calibrate the CA model of urban growth. Two alternatives including total disagreement and the Kappa coefficient have been separately implemented in the cost function of ICA. The findings indicate that the Kappa coefficient achieves higher overall accuracy in ICA, compared with the total disagreement (93.86% vs 92.37%). Moreover, adding patterns to the Kappa coefficient improves overall accuracy and increases the maximum (from 93.86% to 94.65%), the mean (from 79.76% to 81.11%), and the median (from 79.97% to 82.54%) of simulations’ accuracy. The pattern-based calibration resulted in a more realistic simulation of urban growth of over 9.49 sq. km of land in Shelby County (in comparison with the simulation without adding patterns). The results also demonstrated that ICA surpasses logistic regression (LR). While LR's overall accuracy was 92.98%, ICA's was 94.88% with patterns added, 94.40% using the Kappa coefficient alone, and 94.03% using the total disagreement. Using ICA and including patterns resulted in a correct simulation of urban growth of over 37.56 sq. km of land in Shelby County (in comparison with when an LR is used). Urban planners can utilize these findings to more accurately forecast urban growth while construction companies, transportation engineers, tax assessors, and utility providers will all benefit from accurately modeled urban land.

Comments

Data is provided by the student.

Library Comment

Dissertation or thesis originally submitted to ProQuest.

Notes

Open access

Download

Share

COinS