ARTIFICIAL INTELLIGENCE APPLICATION FOR ASSESSMENT OF SEASONAL RELATIONSHIP BETWEEN AIR HUMIDITY AND OTHER CLIMATE FACTORS IN NAMBO PLAIN

Abstract

This study evaluated the relationship between air humidity and other climatic factors at Nha Be, Can Tho and Rach Gia meteorological stations in Nam Bo plain during both dry - rainy seasons. The use of Artificial Neural Networks (ANN) to simulate air humidity for daily climate data in 2013 - 2020 period provided quite good results with 0.84 < R < 0.94 (or 0.71 < R² < 0.88). In particular, the analysis also showed that all obtained R_max series did not follow the normal distribution for 1000 simulations, although some series demonstrated a tendency to follow the Johnson Transformation distribution (most common), 3 - Parameter Weibull and Smallest Extreme Value. Elbow method of graphical evaluation has demonstrated the suitability of ANN structure with 1 hidden layer of 8 neurons using Tansig transfer function for simulating air humidity for all 3 stations. The calculation results proved that humidity value in dry season is most strongly related to daily evaporation, followed by the previous day's average humidity. However, for rainy season, the relationship strength in decreasing order is average temperature, followed by daily evaporation. Thus, daily evaporation is the most important factor with stable relationship to air humidity in survey area throughout the year.