Extended Data Fig. 10: Monitor locations, weighting, and influence on impact estimates.
From: Disparate air pollution reductions during California’s COVID-19 economic shutdown
The public (CARB) and private (PurpleAir) PM2.5 sensor networks used in this study are not evenly distributed across the state, which affects how different census block groups contribute to estimated impacts. On the left we show post-shutdown concentration changes across various census block group gradients (as in Fig. 3), but estimated using different samples – the public CARB network only (green), the private PurpleAir network only (purple), both together but unweighted (brown), and both together and weighted (red). (These weighted estimates correspond to the estimates presented in Fig. 3.) The panels on the right show the representation of demographic and geographic features due to sensor placement within the different sensor networks. Compared to the distribution of these features by all census block groups in California (black lines), the distribution of census block groups with CARB or PurpleAir monitors can be quite different. The distribution of CARB and PurpleAir combined after weighting (red) matches the all-group state-wide distribution much more closely (see Supplementary Information for details).
