Maintained by the Centre for Research on Energy and Clean Air (CREA) Data Team.
High-resolution surface PM2.5 concentration for India, built from satellite products, surface monitors, and machine learning pipelines.
This is a reimplementation of the Improved daily PM2.5 estimates in India reveal inequalities in recent enhancement of air quality (Kawano, et al., 2025), developed and published by researchers at Stanford University, and built by the Centre for Research on Energy and Clean Air (CREA).
Our aim with this project is to regularly produce daily, 10 km resolution gridded PM2.5 data for India, and to make this openly available to researchers, journalists, policymakers, and the public. The latest dataset covers the period from September 2018 to September 2025.
This map displays the monthly mean PM2.5 concentration (µg/m³) for each 10 km grid cell across India. Use the selectors to choose a year and month, then explore the data directly on the map.
Evaluating against air quality monitoring station data, our implementation, covering an extended time period (September 2018 to March 2025), achieves a similar level of spatial accuracy to the original study: we achieved R2=0.60 and RMSE=27.14 µg/m³ compared to the findings R2=0.67 and RMSE=27.7 µg/m³ reported by Kawano et al. (2025). With every version of the dataset we release, we will include the cross-validation results for that dataset for comparison.
The PM2.5 data is produced using data-driven methods. In regions with limited availability of data (lack of air quality stations or low coverage in satellite data), the results are more uncertain.
We publish our results to South Asia PM2.5 map: India | Data (Zenodo), where you can access NetCDF files for the dataset, available for the full analysis period. There, you can also find more details for the file format and versioning.
You can download daily PM2.5 data from the original authors for 1 January 2005 to 30 September 2023 from High-Quality Daily PM2.5 Datasets for India at 10 km Resolution (Zenodo).
When using this data, please reference:
IndiaMap PM2.5 provides gridded estimates of PM2.5 across India using a two-stage ML workflow: first imputing gaps in satellite products (TROPOMI NO₂/CO, MODIS AOD) and then training a full model with satellite reanalysis data and static predictors.
For details of the implementation, see the pm25ml GitHub repository.
Kawano, A., Kelp, M., Qiu, M., Singh, K., Chaturvedi, E., Dahiya, S., … Burke, M. (2025). Improved daily PM2.5 estimates in India reveal inequalities in recent enhancement of air quality. Science Advances, 11(4), eadq1071. https://doi.org/10.1126/sciadv.adq1071
Kawano, A., Kelp, M., Qiu, M., Singh, K., Chaturvedi, E., DAHIYA, S., Azevedo, I., & Burke, M. (2024). High-Quality Daily PM2.5 Datasets for India at 10 km Resolution (Version 2) [Data set]. Science Advances. https://doi.org/10.5281/zenodo.13694585
Centre for Research on Energy and Clean Air (CREA) (n/d). pm25ml: PM2.5 Estimation for India. [Computer software] https://github.com/energyandcleanair/pm25ml (MIT license).
The input datasets for the published datasets are detailed on the data repository.
For questions or feedback, please contact the CREA Data Team: [email protected]