TL;DR
This study investigates how urban amenities influence the transformation of urban green spaces (UGS) in Guangzhou, China, over a decade. By integrating remote sensing and Point of Interest (POI) data, the research reveals that certain amenities negatively impact UGS, while others, like scenic spots, support their preservation, offering crucial insights for sustainable urban planning.
What Problem Does This Paper Address?
Rapid urbanization presents a significant challenge for maintaining and expanding urban green spaces (UGS), especially in densely populated cities. UGS are vital for urban ecosystems, providing ecological, environmental, and economic benefits, and enhancing residents' quality of life. However, the development of urban amenities—such as schools, hospitals, and commercial centers—often competes with land available for green spaces. Understanding the complex interplay between the distribution and type of urban amenities and UGS dynamics is crucial for effective urban planning and management. While previous studies have examined UGS changes or amenity distribution separately, there's a research gap in effectively integrating these two data sources to analyze their spatial and temporal correlations and understand their nuanced interactions.
Methods
The study employs a multi-faceted approach combining remote sensing and Point of Interest (POI) data analysis.
- Remote Sensing for UGS Dynamics: High-resolution remote sensing data (Landsat scenes from Google Earth Engine) were used to classify land cover into six types: cropland, forest, grassland, water bodies, built-up land, and unused land. The grassland and forest layers were merged to define the "green space" component. This allowed for the analysis of annual changes in UGS distribution from 2013 to 2022.
- POI Data for Urban Amenities: Point of Interest (POI) data from the Gaode Map Open Platform were utilized to represent urban amenities. This data was preprocessed, deduplicated, and aggregated into 23 categories. The study calculated the number and density of these POIs within hexagonal grids to represent amenity distribution.
- Geospatial Analysis and Modeling: A hexagonal grid system was overlaid on the study area to analyze UGS area and proportion within each cell annually.
- Ordinary Least Squares (OLS) Regression: Used to assess linear correlations between annual UGS area and the 23 POI categories.
- Random Forest (RF) Regression: Employed to capture more complex, non-linear relationships and interactions between amenities and UGS, and to identify feature importance. This model is known for its higher predictive accuracy and ability to handle complex datasets.
- Spatial-Temporal Analysis: Both OLS and RF models were applied across the 10-year period to track changes in the relationship and identify temporal trends.
Datasets
- Remote Sensing Data: 30 m annual land cover dataset of China from 1990-2019, derived from Landsat scenes available on Google Earth Engine. This dataset was used for the period 2013-2022.
- Point of Interest (POI) Data: 2024 POI data (1,015,533 records) from the Gaode Map Open Platform, covering the period 2013-2022. This data was processed to yield 274,308 valid records categorized into 23 types.
Key Insights and Ideas
- Dual Role of Amenities: Urban amenities are not monolithic; some actively hinder green space development while others actively support it. This nuanced understanding is critical for targeted planning.
- Negative Impact of Infrastructure-Heavy Amenities: Amenities like auto repair shops, shopping services, and transit facilities show a negative correlation with UGS. This suggests that the expansion of these types of infrastructure, especially in dense areas, often comes at the expense of green spaces.
- Positive Influence of Recreational Amenities: Scenic spots (parks, plazas) and life services exhibit a positive correlation with UGS. This indicates that areas designed for recreation and public enjoyment are more likely to be preserved or even expanded, highlighting their value in urban planning.
- Temporal Shift in UGS Dynamics: The study observed a decline in UGS from 2013-2018 due to rapid urbanization, followed by a stabilization and even reversal of this trend from 2018 onwards. This shift is attributed to increased public awareness and government initiatives.
- Impact of External Events: The COVID-19 pandemic appears to have influenced UGS dynamics, potentially slowing down development and highlighting the importance of accessible green spaces for well-being.
- Model Sophistication Matters: The Random Forest model provided a more detailed understanding of the complex, non-linear relationships between amenities and UGS compared to the OLS model, particularly in identifying feature importance.
Key Results
- Negative Correlations: Amenities such as auto repair, traffic facilities, and shopping services were found to have a negative correlation with UGS.
- Positive Correlations: Scenic spots and life services showed a positive correlation with UGS.
- Temporal Trend: UGS declined from 2013 to 2018, then stabilized and began to recover.
- Model Performance: The explanatory power of both OLS and RF models (measured by R²) showed a decreasing trend over the 10-year period, with OLS R² declining from 0.166 to 0.140 and RF R² from 0.2929 to 0.2475. This suggests that while the models capture some relationships, other factors might be at play, or the relationship itself is weakening or becoming more complex.
- Specific Findings: The sharpest decrease in OLS R² occurred between 2016 and 2017. The RF model showed a temporary increase in R² between 2018 and 2019.
Limitations and Future Work
- POI Data Limitations: The reliance on POI data may not fully capture informal or unregistered amenities that also influence UGS dynamics. Future research could incorporate crowd-sourced or field survey data.
- Low R² Values: The relatively low R² values across both models suggest that while UGS and amenities are related, the relationship might not be strong enough to explain a large portion of the variance, or more explanatory variables are needed.
- Spatial Analysis Refinement: Further refinement of spatial analysis techniques is needed to account for the heterogeneity of urban environments.
- Data Granularity: Integrating more detailed land use data (e.g., building heights, impervious surface cover) could improve model accuracy and policy recommendations.
- Future Research Directions: The authors suggest incorporating high-precision remote sensing imagery and temporal changes in POI data to enhance future research. Exploring more explanatory variables is also recommended to improve model interpretability.
Summary updated