Table 2.

Model performance with/without Twitter data. Table shows $R^2$ (coefficient of determination) and root mean squared error (RMSE) for Models 1, 2, 3, 1T and 2T. All scores are averaged from leave-one-out validation tests across all 11 pollen monitoring stations.

	${\mathit{R}}^2$	RMSE
Model 1	0.810	0.115
Model 2	0.847	0.099
Model 3	0.745	0.188
Model 1T	0.858	0.100
Model 2T	0.858	0.096