Fig. 5: The effect of sampling frequency on the accuracy of the hotspot scores.
From: Detecting urban PM2.5 hotspots with mobile sensing and Gaussian process regression
The Spearman rank correlation between hotspot scores and spatial PM2.5 averages in Beijing is computed while filtering to grid tiles with n_measurements > t. In Beijing LP, filtering to the most-sampled tiles improves this correlation, while in Beijing CLT the correlation remains similar.
