Table 4. Model Accuracy Validated by the Fixed-site Routine Monitors for NO2 (n = 82).
| external
long-term validation |
||||
|---|---|---|---|---|
| model category | model name | R2 | MAE (μg/m3) | RMSE (μg/m3) |
| feature-based transfer learning LUR models | CPH2AMS_Coral | 0.39 | 6.47 | 7.61 |
| RTM2AMS_Coral | 0.30 | 4.57 | 5.47 | |
| IDW_Coral | 0.35 | 4.47 | 5.36 | |
| directly applied SLR | CPH2AMS_SLR | 0.17 | 7.86 | 10.77 |
| RTM2AMS_SLR | 0.19 | 5.19 | 6.67 | |
| IDW_SLR | 0.21 | 4.94 | 6.28 | |
| local reference model | AMS_SLR_160Da | 0.52 | 3.72 | 4.70 |
| AMS_SLR_30Da | 0.39 | 4.83 | 5.47 | |
a
AMS_SLR_160D is the SLR model trained using Amsterdam mobile measurements collected for 160 days. AMS_SLR_30D is trained with 30 days of mobile measurements.