Determination of dew point temperature based on simultaneous multivariate models and vector time series considering heterogeneity in meteorological stations in eastern Iran

Abstract

n this research, meteorological data from eleven stations were monitored to build a suitable model for predicting dew point values. Given the importance of dew point temperature in forecasting frosts, rainfall, and other meteorological applications, accurate prediction of this parameter is crucial. The stations included in the study are Bam, Birjand, Chabahar, Iranshahr, Kerman, Mashhad, Sabzevar, Tabas, Torbat Heydarieh, Zabul, and Zahedan, all located in dry climates. Initially, the correlation between various weather parameters and dew point was analyzed. Based on the highest correlation, three parameters—average, maximum, and minimum temperature—were selected as input variables for the model. CARMA and VAR models were used for analysis, and the stability of the residuals from both models was calculated. The series were then developed using the GARCH model. As a result, dew point modeling for the eleven meteorological stations was achieved with the CARMA-GARCH and VAR-GARCH models. Our findings show that the VAR-GARCH model outperformed the CARMA-GARCH model in both training and testing phases, making it the best model for this research. One key factor in the VAR-GARCH model’s superior performance is its enhanced memory for processing time series data. The definitive result indicates that developing the residuals using the GARCH model improves the accuracy of both primary models by 6% to 30% in the testing phase