Figure 1.
Study area and locations of PM2.5 monitoring stations and meteorological sites with daily wind roses for May 18th.
Figure 2.
Observed wind-field and continuous wind-field generated from the monitoring data of May 28th.
Figure 3.
Steps for calculating SWPDs using a wind-field.
(A) Grid-based representation of a wind-field. (B) Computing the cost associated with the edge between adjacent cells. (C) Reforming a grid-based wind-field into a graph. (D) SWPD calculated between starting point A and ending point F.
Table 1.
Comparison of the interpolation accuracies achieved using MSE, MAE and MRE based on a cross-validation analysis.
Table 2.
Distributions of the relative errors of the two methods for the daily PM2.5 estimation on 21st May.
Figure 4.
Comparison between the daily PM2.5 (May 21st) estimations obtained using IDW based on the SWPD and the Euclidean distance.
(A)Interpolation results obtained using IDWS. (B) Interpolation results obtained using IDWE.
Figure 5.
Temporal variation in the hourly PM2.5 concentration and the improvement ratio of IDWS with respect to IDWE on May 26th.
Figure 6.
Interpolation results of the hourly PM2.5 concentrations on May 26th at various times: (A) 8 am, (B) 12 pm, (C) 4 pm and (D) 8 pm.