Performance evaluation of the penalized maximal T and F algorithms in the quality control of monthly and daily climatic time series on the southwest coast of the Caspian Sea

Given the broad application of long-term meteorological data in various sciences and the need to predict their possible changes at local and global scales, it is very important to ensure the accuracy and homogeneity of such data. The penalized maximal T and F tests in the Rhtests software package were used to control the climatic parameters of the western half of the Caspian region on a daily and monthly scale. The data included precipitation, maximum and minimum temperature, wind speed, relative humidity, and sunshine hours, during 1979-2017 in Bandar Anzali, Rasht, and Ramsar stations. The missing data were estimated using the standards of the World Meteorological Organization and using the k-nearest neighbors algorithm. For the monthly time series, 23 change points were identified, which were homogenized in two stages, before and after reconstruction. The standard normal method was less sensitive than the penalized maximal F method by identifying 9 change points. Then, the daily data of the mentioned parameters were homogenized, which was homogenized by identifying a total of 32 points of change. However, it was not possible to thoroughly homogenize the sunshine hours due to the consecutive missing data. Data homogenization reversed the trend in 33% of cases. The studied method had acceptable results in homogenizing meteorological data in the study area based on the obtained results