The Identification of Climatic Patterns of Iran Based on Spectral Analysis and Clustering of Precipitation and Temperature Extreme Values

Authors

Abstract

Introduction
Extreme weather events are rare events from intensity and frequency perspective and at the time of their occurrence, ecosystem and human societies hardly can adapt themselves to occurred changes. Some examples of these events are Heat waves, Cold waves, Floods, Tropical cyclones, Tornadoes, Wildfires and Dust storms. Since extreme weather events occur when precipitation and temperature have extreme values, studying the extreme climatic data is a matter of vital importance. An important issue in the climatological studies is the identification of regions with the same climatic behavior. The results of regionalization based on extreme climatic events can aid decision-makers and planners in dealing with crises aroused from extreme weather events, especially in the development of management policies.

Methodology
The Regionalization was implemented for 65 synoptic stations in the Islamic Republic of Iran. Three climatic signals were selected comprising maximum 24 hour monthly precipitation, maximum monthly temperature and minimum monthly temperature for a twenty-year period from 1986 to 2005. The climatic data were extracted from the website of Islamic Republic of Iran Meteorological Organization (IRIMO). The regionalization was implemented by means of Spectral analysis and Clustering methods. The main advantage of the Spectral analysis method is the spatio-temporal analysis of climatic data in the frequency domain instead of the time domain which can significantly decrease the calculation volume. Calculating the spectral characteristics of climatic data was a time-consuming process; therefore, software named “Dadisp” was applied for performing the calculation. The climatic data were imported to the software. Then, linear trend of the climatic data was removed and main spectrums were identified using power spectrum density function (PSD). Four key spectrums of precipitation signals and one key spectrum of temperature signals were identified and selected. Clustering is one of the categorization methods. Members of each cluster have the same characteristics. For clustering, a deterministic approach (K- means) was used. Clustering the results of spectral analysis was performed using “MATLAB” software based on three different scenarios. In scenario one, spectral analysis results and geographical characteristics of stations are applied. In scenario two, spectral analysis results and elevation of stations are applied and finally in scenario three, only spectral analysis results are applied. According to the minimum variance criteria, 14 to 17 clusters were selected. Finally, stations with the same climatic behavior and their spatial distribution in Iran were illustrated in six figures.

Results and Discussion
One of the most obvious results was the singularity of Ramsar station in all scenarios. Based on spectral analysis results, it was presumable that this station has a singular climatic behavior in the clustering. Among southern stations, Kenarakchabahar station was another singular station. The same climatic behavior of Urmia, Shiraz, Tehran and Abadeh stations in the first scenario was also remarkable. Also similarities between results of second and third scenario, like the same climatic behavior of Mahabad, Shiraz, Sanandaj, Karaj and Zahedan stations were notable. The results of clustering are presented in figures 5 to 10 and table 3.

Conclusion
The identification of regions with the same climatic behavior is an important issue from different perspectives. Hence, the occurrence of extreme weather events like floods in Iran is so prevalent; the identification of regions with the same climatic behavior based on extreme climatic data can aid planners and decision-makers in the process of disaster management. In this paper, a regionalization is implemented by analyzing the content of climatic data and employing a systematic approach for finding the same climatic patterns. Clustering the stations based on spectral analysis results and geographical situation of stations impose a strong assumption on pattern explorer system. This assumption is the behavior similarities of different stations based on neighborhood. Although this assumption is true in most cases, but implementation of this assumption in a large region like Iran with the significant climatic variation need a more conservative approach. Therefore, another two scenarios (second and third scenario) were developed and the results were compared. The results of second and third scenario had inconsiderable differences but both of these scenarios had totally different results in comparison with first scenario. The final results show that the applied systematic approach in this paper was successful in finding stations with singular climatic behavior like Ramsar and Kenarakchabahar stations and it could specify regional patterns in climatic variation clearly. One of the proposed fields for future studies is the implementation of fuzzy clustering methods for specifying the climatic behavior of different regions.

Keywords