Estimation of daily dew point temperature using genetic programming and neural networks approaches

The present study investigates the ability of two different artificial neural network (ANN) models and gene expression programming (GEP) technique for estimating daily dew point temperature by using recorded weather data. The weather data used consist of 8 years of daily records of air temperature, wind speed, relative humidity, atmospheric pressure, incoming solar radiation and dew point temperature from two weather stations (Seoul and Incheon, in the Republic of Korea). Two different data management scenarios are applied in this paper. In the first scenario, weather data obtained from each station are used to estimate T_dew at the same station (at-station approach). In the second scenario, the ANN and GEP models are used for estimating dew point temperature of each station by using the data of the other station (cross-station application), through the optimal input combinations of the first scenario. Comparison of the results reveals that the GEP model surpasses ANN in estimating daily dew point temperature values.