Precipitation is among the main variables in weather and climate studies. The length of the statistical period plays a pivotal role in the accurate analysis of precipitation. One of the limitations of meteorological stations is having missing data. Analysis based on incomplete data leads to biased analysis. The historical monthly precipitation of the five stations in Iran is available since 1880 with missing data. The name of these synoptic stations are Mashhad, Isfahan, Tehran, Bushehr, and Jask. The data in the period of 1941–1949 have a gap that was during and following World War II (1939–1945). The present study aimed to use several classic and meta-heuristic methods to estimate these missing data. The Root Means Square Error (RMSE) criteria were used for comparison. The neighboring stations of Iran were selected as independent variable to estimate missing rainfall data. First, missing data were restored with the fitting of several new regression models for monthly precipitation (with RMSEs: 9.79, 7.89, 13.43, 6.65, and 20.96 millimeter(mm)). Then, the parameters of regression models were optimized by methods of genetic algorithm (GA) and Ant Colony (ACO). It was observed that RMSEs reduced to 2.56, 2.51, 3.49, 2.48, and 4.02 mm. Besides, Artificial Neural Network (ANN) and Support Vector Regression (SVR) methods were used to model the data. ANN and SVR could not increase the accuracy of the estimated data. The missing data were imputed using evolutionary methods (GA and ACO). As a result, the length of the statistical period of the stations reached over 125 years, and the data could be considered a valuable basis for water resources, drought analyses, evaluation trends, climate changes, and global warming.

Keywords