- The paper introduces an automated method that uses OSM road networks and POI data to identify and characterize over 80,000 urban parcels in China.
- It employs a vector-based cellular automata model and logistic regression, achieving 74.2% accuracy and a 71% overlap with traditional datasets in Beijing.
- The framework enables efficient, cost-effective urban planning by integrating crowdsourced data, offering a replicable approach for regions with limited traditional data.
Automated Identification and Characterization of Parcels with OpenStreetMap and Points of Interest
The paper introduces a method for Automated Identification and Characterization of Parcels (AICP) leveraging OpenStreetMap (OSM) and Points of Interest (POI) data, tackling the challenge of insufficient parcel data in China. Land parcels are foundational for urban modeling, planning, and policy implementation. Traditional parcel identification is cumbersome, reliant on remote sensing and field surveys, which are often costly and labor-intensive, yielding limited accessibility in developing countries. This research posits OSM road networks to identify parcel shapes, while POI data infer parcel characteristics, aiming to optimize efficiency and access to accurate parcel data.
The proposed methodology overcomes challenges associated with conventional parcel data acquisition by automating the process of delineating parcels and attributing them with urban functions and densities. The choice to integrate OSM and POIs stems from OSM's growing data coverage and density in urban regions, alongside the high-resolution and continuously updated nature of POI data. The utility of this approach is demonstrated across 297 Chinese cities, identifying 82,645 urban parcels.
Key Results and Methodological Insights
The approach delineates parcels using OSM by first trimming, merging, and buffering OSM road data, effectively forming polygon areas identified as parcels. The model uses a vector-based constrained cellular automata (CA) mechanism to select urban parcels, utilizing administrative boundaries for accuracy. Parcel characteristics—function, density, and land use mix—are induced from POI data.
A critical numerical outcome is the identification of over 80,000 parcels with urban characteristics out of hundreds of thousands across various cities, suggesting a significant efficiency improvement over traditional manual methods. The framework achieves an accuracy of 74.2% in identifying urban parcels, with verification via logistic regression on parcel attributes demonstrating its robustness.
Despite using open crowdsourced datasets, the method's parcel outputs closely align with conventional datasets, evidenced by a parcel overlap of over 71% in Beijing-based empirical validation. Moreover, urban parcels generated by OSM data maintain a Pearson correlation coefficient of 0.858 with manual estimates of development density, reinforcing the reliability of using POI data as proxies for urban density indicators.
Implications and Future Research Directions
The implications are multi-faceted. Practically, the provision of timely, accurate parcel data enables fine-grained urban planning and can accommodate vector-based simulation models, adaptable to analyze urban expansion and dynamic spatial changes efficiently. Theoretically, this approach pushes the boundaries of volunteered geographic information (VGI) application, setting a precedent for its integration into more comprehensive urban datasets.
Future research avenues include refining the vector-based CA model by incorporating additional constraints, leveraging evolving sources of urban mobility data for improved land-use intensity inference, and expanding validation beyond Beijing to a wider array of cities. The method's adaptability suggests potential applicability in other regions lacking traditional data infrastructure, provided the continuous improvement of OSM and POI data quality.
In conclusion, AICP exemplifies how cutting-edge data sources and automatic techniques can substantially complement traditional urban parcel identification, offering a replicable framework that promises to enhance urban studies and planning methodologies.