Assessing Individual-Building Vertical Light Exposure in Urban Environments with a Residual Cascade Framework

Research Challenge

Conventional two-dimensional nighttime-light imagery primarily characterizes upward radiance from horizontal surfaces and cannot directly represent illumination on building facades. Detailed three-dimensional simulations can address this vertical blind spot, but their data and computational requirements make city-scale repeated assessment difficult.

Proposed Framework

The study develops the Physics-Informed Residual Cascade Framework (PIRCF) to estimate individual-building vertical light exposure from two-dimensional multisource geospatial data. Here, “physics-informed” refers to using exposure-related geometry, distance attenuation, spatial topology, and environmental occlusion as inductive biases rather than directly enforcing physical governing equations in the loss function.

PIRCF combines graph-based neighborhood inference with XGBoost residual correction. This design captures broad relationships among nearby buildings while correcting localized variation that the graph model does not fully explain.

Key Results

• In Guangzhou, PIRCF achieved test-set R² values of 0.78 for panchromatic exposure and 0.85 for blue-light exposure, outperforming the selected statistical baselines.
• Applied directly to Shanghai without retraining or parameter adjustment, the Guangzhou-trained model achieved R² values of 0.70 and 0.73, respectively.
• Panchromatic exposure showed broader, more continuous gradients associated with road networks, while blue-light exposure formed more fragmented clusters near commercial and vertically developed urban areas.

Significance

The framework offers a scalable alternative to repeated three-dimensional simulation for building-level urban light-exposure screening. It can help identify priority locations for detailed field investigation and support more spatially refined urban-lighting assessment and governance.

Read the open-access article