Spatial-Temporal Autoregressive Dynamic Model

Although a myriad of methods have been advanced to tackle spatial and temporal structures in data separately, it becomes difficult to analyze these data using classical linear regression models when spatial-temporal structures coexist, especially when the data size is relatively large. In this article, we demonstrate a simple to implement method to handle spatial-temporal structures simultaneously. Building on the classical linear regression model, we propose to account for spatial-temporal autocorrelation in the residuals from the linear regression model iteratively until the main assumptions of the model are satisfied. Neighbors and their weights are defined using the strength of correlation at different spatial-temporal lags. Using the proposed methodology aerosol optical depth (AOD) from satellite data will be corrected for meteorological conditions and spatial-temporal structure to predict air quality for Delhi and its neighboring areas. This methodology will have greater applications for estimating air quality at unprecedented spatial-temporal resolutions for air quality surveillance and management, epidemiological and environmental justice research, because the existing network of air pollution monitoring stations has limited spatial-temporal coverage.