Ecological Multi-Decadal Landsat Satellite Imagery Pixel-Level Analysis on Vancouver Dataset
Description
This dataset captures multi-decadal patterns of vegetation dynamics, urban expansion, and ecological risk across the City of Vancouver using consistent Landsat satellite observations from 1990 to 2025. Vancouver provides a uniquely relevant case study due to its rapid population growth, strong environmental planning reputation, and location within the highly biodiverse and climate-sensitive Pacific Northwest. This Dataset is primarily analysis results of Normalized Difference Vegetation Index (NDVI) and Normalized Difference Built-up Index (NDBI) were calculated from Landsat spectral bands for each year at the pixel level. The dataset offers a comprehensive view of how ecological conditions in a model green city have evolved under sustained urban pressure by comparing urban sprawl and vegetation loss during the peak vegetation summer months.
Files
Steps to reproduce
This study employed an integrated remote sensing, Geographic Information Systems (GIS), and statistical geospatial analysis framework to assess long-term ecological change and risk across the City of Vancouver. Multi-temporal Landsat satellite imagery was acquired for five representative years spanning 1990 to 2025, selected to capture long-term urban and ecological dynamics while minimizing cloud contamination. All imagery was radiometrically and atmospherically corrected using surface reflectance products to ensure temporal comparability. Normalized Difference Vegetation Index (NDVI) and Normalized Difference Built-up Index (NDBI) were calculated from Landsat spectral bands for each year. Raster datasets were clipped to the Vancouver municipal boundary and resampled to a consistent spatial resolution. GIS techniques were used to generate spatial layers representing vegetation cover, built-up intensity, and a composite ecological risk index integrating heat exposure, smoke, ozone, flood susceptibility, vegetation loss, and urban sprawl. Descriptive statistics were computed to characterize temporal trends in NDVI and NDBI. Change detection analysis quantified annual and cumulative changes between time periods. Spatial autocorrelation was assessed using Global Moran’s I to evaluate the presence and strength of clustering, with statistical significance determined using z-scores and p-values. Hotspot and cold spot analyses were conducted using the Getis–Ord Gi* statistic to identify statistically significant clusters of high and low ecological risk at multiple confidence levels. All analyses were performed using GIS software and reproducible Python-based workflows, ensuring methodological transparency and replicability. This integrated approach allows consistent detection of spatial and temporal ecological patterns and can be applied to other urban regions with comparable remote sensing data.
Institutions
- Boston University