Identification of the temperature regimes of Iran using multivariate methods

Identification of temperature regimes is very crucial for a better management of energy resources, recreations and truisms as well as for adequately determining agricultural calendars in different parts of the country. Few attempts have been made to adumbrate the Iranian temperature regimes; thus it is necessary to identify the most realistic temperature regimes of Iran using as many available stations across the country as possible. For this purposes, 155 Iranian synoptic stations with a relatively regular distribution across the country, mostly having full data records for the common period of 1990 to 2014, were used for the identification of the temperature regimes. In all stations, the average of the mean monthly temperature  was computed  in the mentioned period and  further employed for the analysis. A principal Component Analysis (PCA) was applied to the inter-stations correlations matrix (155×12) composed of 155 stations and 12 mean monthly temperature values for each station. The computed Kaiser-Meyer-Olkin (KMO) measure of the sampling adequacy indicated that the matrix with the KMO value of 0.87 is useful for a PCA application. The first three leading PCs were considered for further analysis based on the scree plot and the sampling errors of the PCs (North et al., 1982). The remaining PCs were then rotated using varimax orthogonal criterion. The PC scores of both rotated and un-rotated solutions were separately used as inputs for Cluster Analysis (CA) to partition the considered stations into distinctive clusters. Moreover, all agglomerative CA methods as well as K-means CA were examined so as to find out the most appropriate method for partitioning the data. The cophenetic correlation coefficient was employed to measure how well the hierarchical dendrogram of a given CA represents the relationships within the input data. The results indicated that all the clustering approaches properly represented the inherent structure of the input data, yet the Ward method was selected as the most appropriate method since it resulted in much realistic clusters that quite perfectly matched the topographic and geographical features of the country. The correct number of clusters was also selected based on the Silhouette index (Rousseeuw, 1987) that measures how well objects lie within their cluster, and which ones are merely somewhere in between the clusters. The average silhouette width provides an evaluation of clustering validity, and might be used to select an ‘appropriate’ number of clusters. Computing the index for a set of predefined cluster numbers (2 to 15 clusters), it was observed that six is the most appropriate number for clusters  for a better representation of the inherent structure of the data. As such, all 155 stations were classified into six clusters applying Ward CA method on the three leading un-rotated PC scores.
The maps of varimax rotated PC scores properly represented areas characterized with seasonal temperature variability. The first and second varimax rotated PC sores respectively display the winter and summer temperature variability across the country. Applying Ward clustering on un-rotated PC scores resulted in seven distinct clusters that appropriately specified the Iranian temperature regimes. It was found that the western and northern mountainous areas have a mountainous temperature regime whereas the central-eastern Iran hosting the Iranian deserts has a plain temperature regime which is considerably warmer than the earlier one. The third temperature regime includes stations scattered across the mountainous region, all of which are characterized with a very high elevation, hence having the coldest winters and the coolest summers in the country. The hot temperature regime which is the warmest temperature regime in the country belonged to the southwest and certain parts of the south. The stations located in the coastal areas of the northern and southern Iran respectively  have the Caspian coastal, Persian Gulf coastal and Oman Sea coastal temperature regimes, all with the lowest annual temperature ranges in the country. Four out of the seven Iranian temperature regimes are continental but the ‌two mountainous temperature regimes are the most continental regimes in the country due to their wider temperature ranges. The results conduce to a better management of energy resources, recreations and tourisms as well as an optimal determination of agricultural calendars in different parts of the country.