Find in Library

In response to the nonlinear fitting difficulty of the traditional weighted average temperature (<i>T_m</i>) modeling, this paper proposed four machine learning (ML)-based <i>T_m</i> models. Based on the seven radiosondes in the Yangtze River Delta region from 2014 to 2019, four forecasting ML-based <i>T_m</i> models were constructed using Light Gradient Boosting Machine (LightGBM), Support Vector Machine (SVM), Random Forest (RF), and Classification and Regression Tree (CART) algorithms. The surface temperature (<i>T_s</i>), water vapor pressure (<i>E_s</i>), and atmospheric pressure (<i>P_s</i>) were identified as crucial influencing factors after analyzing their correlations to the <i>T_m</i>. The ML-based <i>T_m</i> models were trained using seven radiosondes from 2014 to 2018. Then, the mean bias and root mean square error (<i>RMSE</i>) of the 2019 dataset were used to evaluate the accuracy of the ML-based <i>T_m</i> models. Experimental results show that the overall accuracy of the LightGBM-based <i>T_m</i> model is superior to the SVM, CART, and RF-based <i>T_m</i> models under different temporal variations. The mean <i>RMSE</i> of the daily LightGBM-based <i>T_m</i> model is reduced by 0.07 K, 0.04 K, and 0.13 K compared to the other three ML-based models, respectively. The mean <i>RMSE</i> of the monthly LightGBM-based <i>T_m</i> model is reduced by 0.09 K, 0.04 K, and 0.11 K, respectively. The mean <i>RMSE</i> of the quarterly LightGBM-based <i>T_m</i> model is reduced by 0.09 K, 0.04 K, and 0.11 K, respectively. The mean bias of the LightGBM-based <i>T_m</i> model is also smaller than that of the other ML-based <i>T_m</i> models. Therefore, the LightGBM-based <i>T_m</i> model can provide more accurate <i>T_m</i> and is more suitable for obtaining GNSS precipitable water vapor in the Yangtze River Delta region.