Quantifying urban expansion using Landsat images and landscape metrics: a case study of the Halton Region, Ontario

Liyuan Qing, Hasti Andon Petrosian, Sarah Narges Fatholahi, Michael A. Chapman, Jonathan LiGEOMATICA 2020 · DOI

Abstract

Urbanization is a primary driver of global environmental change, necessitating precise spatial analysis for effective regional planning. This study examines thirty years of urban expansion in Ontario’s Halton Region (1989–2019) using multi-temporal Landsat imagery and machine learning. By applying landscape metrics and the Vegetation-Impervious Surface-Soil (V-I-S) model, the research quantifies the transition from natural to built environments. The findings reveal a dominant "leapfrog" development pattern and a high correlation (96.2%) between population growth and urban land expansion, particularly along major transportation corridors.

Introduction

The Halton Region, situated within the Greater Toronto Area (GTA), represents one of Canada’s most rapidly developing sectors. While regional growth is evident, urban planners require high-resolution data to distinguish between different modes of expansion: infilling, edge expansion, and leapfrog development. Understanding these spatial dynamics is critical for sustainable infrastructure development and resource management. This research addresses the gap in long-term, localized spatial analysis for the Halton Region, providing a framework for monitoring urban "texture" and growth trajectories.

Methodology

The study utilized a comprehensive geospatial workflow integrating satellite remote sensing and statistical modeling.

Data Sources

Analysis was based on Landsat 4-5 (Thematic Mapper) and Landsat 8 (OLI/TIRS) imagery spanning 1989 to 2019. Validation was performed using 12 cm resolution digital aerial photography and Google Earth imagery to ensure classification accuracy. Socioeconomic data from Statistics Canada provided the basis for correlating urban expansion with population and GDP trends.

Classification and Metrics

A Support Vector Machine (SVM) algorithm classified the landscape into four categories: built-up, barren land, water, and vegetation. To enhance feature extraction, spectral indices such as the Normalized Difference Built-up Index (NDBI) and Soil Adjusted Vegetation Index (SAVI) were employed.

Landscape metrics were calculated to quantify the spatial configuration of urban patches. These metrics, detailed in the table below, provide insights into the fragmentation and complexity of the urban fabric.

Table of landscape metrics used to quantify urban spatial patterns.

Results and Discussion

The SVM classification achieved exceptional accuracy, with Overall Accuracy (OA) ranging from 94.25% to 99.68%. Over the 30-year study period, built-up areas in the Halton Region increased by approximately 39%, with a mean annual growth rate of 1.21%.

Spatiotemporal Patterns of Growth

The analysis identified a distinct "leapfrog" dominance. An increase in the Number of Patches (NP) and Patch Density (PD) suggests that new, isolated urban clusters are emerging faster than they are merging into existing urban footprints. Furthermore, rising values in the Landscape Shape Index (LSI) and Area-Weighted Mean Shape Index (AWMSI) indicate that urban boundaries are becoming increasingly irregular and complex.

Map showing urban expansion in the Halton Region from 1989 to 2019, highlighting growth along major highways.

Growth was notably concentrated along major transportation corridors (e.g., Highway 401, QEW) and lake shores, underscoring the influence of infrastructure on sprawl direction. The V-I-S model further illustrated a shift from low-density residential areas (high vegetation/soil) toward medium-to-high density impervious surfaces, reflecting a transition toward commercial and industrial land use.

Socioeconomic Correlation

A near-perfect 96.2% correlation was found between population surges and urban land expansion. This confirms that the Halton Region's fringes are absorbing significant population pressure from the broader GTA, necessitating proactive regional planning.

Conclusion

This study demonstrates the efficacy of combining satellite imagery with landscape metrics to monitor urban sprawl. While the patterns of growth are now well-quantified, future research should investigate the specific policy and economic drivers behind these trends. Integrating predictive modeling could further assist regional agencies in balancing development with environmental sustainability.

Summary updated

BibTeX
@inproceedings{qing2020quantifying,
  author = {Liyuan Qing and Hasti Andon Petrosian and Sarah Narges Fatholahi and Michael A. Chapman and Jonathan Li},
  title = {{Quantifying urban expansion using Landsat images and landscape metrics: a case study of the Halton Region, Ontario}},
  booktitle = {GEOMATICA},
  year = {2020},
  doi = {10.1139/geomat-2020-0017},
}
Figure