Selection of the best clustering technique in order to zone the frequency of dust storms in Iran

In this study, the frequency of sand and dust storms in 87 synoptic stations for the period of 1987–2019 has been prepared. The purpose of this study is to compare different hierarchical clustering techniques including single linkage, average linkage, Ward method, and fuzzy c -means clustering and to choose the best technique for determining homogeneous areas in terms of dust storm frequency and evaluated with Dunn validation index and Davies–Bouldin score (DBS). The results showed that among the different techniques used, c -means fuzzy clustering method provided better performance, which is clustered in 5 classes. The highest number of dust storm days was observed in cluster 1, which includes Zabol station with 869 days and then cluster 2 with a frequency of 252–447 days in Abadan, Ahvaz, Bostan, Masjed Soleyman, Bandar Mahshahr, Konarak, and Zahedan. The minimum number of sand and dust storm days is in cluster 5 with a frequency of 0–80 days. The spatial distribution map based on the RBF method shows a decrease in the frequency of sand and dust storms from the southwest and southeast to the north and northwest.