Fig. 1: Evaluation results of the models.

The dotted line is the 1:1 line, and N is the number of samples. “Val” and “Test” represent model validation and test, respectively. Each subplot is labeled with a combination of letters and numbers, where a, b, and c respectively refer to RF/GTW-RF, LGBM/GTW-LGBM, XGBoost/GTW-XGBoost, and the numbers 1–4 refer to the validation and test results of the global models and local models, respectively. The equations stand for the regression relationships between the estimated and observed PM2.5.