Increasingly frequent wildfires have significant impact on solar energy production as the atmospheric aerosols generated by wildfires diminish the incoming solar radiation. Atmospheric aerosols can be measured by aerosol optical depth (AOD), and multiple spatiotemporal AOD data streams are available from geostationary satellites. Although these multisource remote-sensing data streams are measurements of the same underlying AOD process, they often present heterogeneous characteristics, such as different measurement biases and errors, data missing rates, etc. It is usually not known which data source provides more accurate measurements, and simply averaging multiple data streams is not always the best practice. Hence, this paper proposes a statistical model that is capable of inferring the underlying true AOD process by simultaneously integrating heterogeneous multisource remote-sensing data streams and the advection-diffusion equation that governs the dynamics of the AOD process. A bias correction process is included in the model to account for the bias of the physics model and the truncation error of the Fourier series. The proposed approach is applied to California wildfires AOD data obtained from the GOES East (GOES-16) and GOES West (GOES-17) satellites operated by the National Oceanic and Atmospheric Administration. Comprehensive numerical examples are provided to demonstrate the predictive capabilities and model the interpretability of the proposed approach. History: Kwok-Leung Tsui served as the senior editor for this article. Funding: This work was funded by the U.S. National Science Foundation [Grant 2143695]. Data Ethics & Reproducibility Note: Our code requires packages that are not available on CodeOcean (for example, the Rfast package in R). Hence, code and sample data are provided on GitHub: https://github.com/gz-wei/Physics-Informed-Statistical-Modeling-for-Wildfire-Aerosols-Propagation . The online appendix is available at https://doi.org/10.1287/ijds.2021.0058 .
Read full abstract