What question did this study set out to answer?

This research aims to enhance the prediction of NO2 concentrations in data-sparse urban areas using transfer learning and satellite data.

Cross-City Transfer Learning for Sentinel-5P-Driven NO 2 Prediction in Data-Sparse Urban Environments

View Full Paper

Ask AI

Bookmark

Authors

Randomized trial demonstrates improved NO2 prediction in urban environments, indicating AI's potential in air quality management.

This research aims to enhance the prediction of NO2 concentrations in data-sparse urban areas using transfer learning and satellite data.
Implemented a transfer learning model utilizing satellite-derived NO2 columns and various environmental data across five cities.
Applied a CatBoost regression model within a Leave-One-City-Out framework for effective model transfer.
Conducted a Leave-One-Block-Out ablation study to assess the contribution of different feature groups.
Achieved a 7% reduction in Root Mean Square Error (RMSE) and a 2.7% increase in Coefficient of Determination (R2) compared to baseline models.
In Toronto, R2 improved from 0.58 to 0.66 and RMSE decreased from 6.44 μg/m3 to 5.84 μg/m3 under the transfer learning model.
Identified spatial coordinates and meteorological features as key predictors of NO2 concentration, enhancing model generalization.

Janku et al. (2026) studied this question.

View Full Paper

Ask AI

Bookmark