Estimating Daily Dew Point Temperature Based on Local and Cross-Station Meteorological Data Using CatBoost Algorithm

Accurate estimation of dew point temperature (Tdew) plays a very important role in the fields water resource management, agricultural engineering, climatology and energy utilization. However, there are few studies on applicability local Tdew algorithms at regional scales. This study evaluated performance new machine learning algorithm, i.e., gradient boosting decision trees with categorical features support (CatBoost) to estimate daily using limited cross-station meteorological data. The random forests (RF) algorithm was also assessed for comparison. Daily data from 2016 2019, including maximum, minimum average (Tmax, Tmin Tmean), relative humidity (RHmax, RHmin RHmean), global solar radiation (Rsmax, Rsmin Rsmean) three weather stations Hunan China were used evaluate CatBoost RF algorithms. results showed that both achieved satisfactory accuracy target (on RMSE = 1.020°C, R2 = 0.969, MAE 0.718°C NRMSE 0.087) absence complete parameters (with only as input). 1.900°C 0.835) better than 2.214°C 0.828). stability positively correlated number input parameters, three-parameter higher two-parameter developed methodology is helpful predict scale.