Comparison of correlation and mean methods in selecting Indicator days of sea level pressure patterns combined with heavy and pervasive rainfall in Iran

Spatial diversity and temporal variability of natural phenomena make their study time-consuming and difficult. On the other hand, there are mutual relationship between the phenomena and aspects and the position of their performance, that have spatial and temporal homogeneity. Based on these criteria, space (time) can be divided into independent units (Kavyani and Alijani, 2001: 344). therefore, classification is the systematic grouping of phenomena or events that have common characteristics or relationships within groups (Heydari, 1999: 8). Climatic phenomena are one of the important and most variety things in the nature of the world. There is no environmental-climatic phenomenon that is not caused by a specific pattern of air pressure distribution (Alijani, 2006: 201). Air pressure is an element that can be used to determine the characteristics of climatic elements, including temperature, precipitation, etc... Therefore, pressure systems are the main and independent factors that control characteristics of humidity, temperature and pressure with their temporal and spatial changes. Most of these pressure systems have a medium or synoptic scale (Boucher, 2006: 5). Therefore, in this study, sea level pressure was chosen as the climate index. Heavy rainfalls are one of the climate phenomena that can have many effects on the environment. This kind of rainfalls are most important natural reason of floods that cause irreparable damage to the natural environment and human societies. Therefore, the study of heavy rainfalls, especially in pervasive areas, is essential to understand the behavior of this type of rains. Identifying the behavior of this type of precipitation is an important step in water resources management. According to the amount of available data and for the ease of knowing the behavior of heavy rainfalls, we will need to use classification. Because despite the variety of patterns leading to heavy rainfall, there are similarities in these patterns.

In this research, heavy and pervasive rainfall in Iran has been studied by using the third version of Esfazari database. For this purpose, the values of the 90th, 95th and 99th percentiles of precipitation on rainy days were calculated for all cells of the map. This means that for each cell a specific threshold was obtained. Then, data above the 90th, 95th, and 99th percentiles were identified respectively as heavy, very heavy, and extremely heavy rain days. To determine the threshold for the definition of " pervasive rainfall " in this research, at first, anomaly of heavy rainfall area was calculated, and then the days with a positive anomaly was selected. Then the 75th percentile of the area of the selected days were obtained. The 75th percentile value obtained for the mentioned days is equal to 16% of the country's area and it was determined as the threshold of pervasive area. Result shows that 160 days with pervasive heavy rainfall (with precipitation equal or more than 90th percentile) occurred during the study period. Rainfall data of 160 days classified by using Binary cluster analysis. For this purpose, the Euclidean distance method and "Ward" linkage method were used. After identifying the number of heavy rainfall patterns, Indicator Days for each pattern identified by using correlation method (Lund) and the average area method. After identifying the Indicator days by both methods, by using sea level maps and the upper atmosphere levels and checking moisture convergence flux maps, Omega and statistical methods, the Indicator days obtained through both methods are compared.

As mentioned above, 160 days with pervasive heavy rainfall were identified and then classified by using binary cluster analysis method. Result show that the 160 studied days are placed in 10 groups (patterns). After determining the number of patterns, the indicator days for each pattern were identified by using correlation method and the average area method. Then, the indicator (obtained from both methods) days were compared by using sea level pressure maps and upper atmosphere levels and checking moisture convergence flux maps, Omega and statistical methods. Result show that average area method can choose indicator days better than correlation method. In the other hand indicator days chosen by correlation method most of the patterns are attributed to the Red sea low pressure and the role of other factors is small or ignored. For example, in this method, the role of the Indian Ocean in providing moisture for the heavy rains in the east and southeast of the country has been very small. But on the indicator days selected by the area average method, the pressure centers are arranged in such a way that a relative balance has been established in the role of factors that cause heavy rainfall.

The results of this research show that 160 days with heavy rainfall (with precipitation equal or more than 90th percentile) and in the area of more than 16% of the country (pervasive) occurred during the study period. The classification shows that 8 synoptic patterns cause heavy rainfalls. Also, Result show that average area method can choose indicator days better than correlation method.