Technology Category
Machine Learning: Bayesian LearningReasoning under Uncertainty: Relational Probabilistic ModelsData Mining & Knowledge Management: Mining of Spatial, Temporal or Spatio-Temporal Data
A nonstationary spatial model of PM2.5 with localized transfer learning from numerical model output
🤖 AI Summary
Sparse and unevenly distributed air quality monitoring stations limit the accuracy of high-resolution PM₂.₅ modeling. To address this, we propose a nonstationary spatial statistical model based on local transfer learning. Our method transfers spatially varying features from numerical model outputs into a Bayesian framework, constructing a location-adaptive nonstationary covariance function. By integrating localized parameter estimation with an efficient computational algorithm, the model jointly assimilates sparse monitoring data and numerical model outputs. Compared to conventional stationary approaches, our model achieves significantly improved predictive accuracy and effectively captures pollution heterogeneity across complex terrain and urban areas. It enables scalable, high-resolution nationwide PM₂.₅ mapping, supporting robust environmental health assessments and fine-grained air quality management.
Technology Category
Application Category
📝 Abstract
Ambient air pollution measurements from regulatory monitoring networks are routinely used to support epidemiologic studies and environmental policy decision making. However, regulatory monitors are spatially sparse and preferentially located in areas with large populations. Numerical air pollution model output can be leveraged into the inference and prediction of air pollution data combining with measurements from monitors. Nonstationary covariance functions allow the model to adapt to spatial surfaces whose variability changes with location like air pollution data. In the paper, we employ localized covariance parameters learned from the numerical output model to knit together into a global nonstationary covariance, to incorporate in a fully Bayesian model. We model the nonstationary structure in a computationally efficient way to make the Bayesian model scalable.
Problem
Research questions and friction points this paper is trying to address.
Modeling nonstationary PM2.5 spatial distribution with localized covariance
Integrating numerical model output with sparse monitoring network data
Developing computationally efficient Bayesian methods for air pollution inference
Innovation
Methods, ideas, or system contributions that make the work stand out.
Nonstationary spatial model with localized transfer learning
Bayesian model integrating numerical output and monitor data
Computationally efficient scalable covariance parameterization
🔎 Similar Papers
No similar papers found.
💼 Related Jobs
No related jobs found.
