Forecasting of Heat Wave Occurrence and its Return Time in Iran Using Markov Chain Model

In this paper the prediction of heat waves have been done using Markov Chain. In order to identify heat waves the max temperature of 1437 stations have been applied. The data covers the period of 1962/1/1 to 2004/12/31. Then these stations have been interpolated on 15 km × 15 km grids and has consisted a matrix in 15695 × 7187. The rows represent days and columns represent the pixels. This index was applied on 15695 and on all the pixels. So a matrix in 15695 × 7187 was consisted. Because the Z above 2 indicates the positive abnormality, this value was chosen as the heat wave. Then the values of P and Q that represent the occurrence and nonoccurrence of heat waves were utilized to predict heat waves over Iran. In general the mean annul distribution of heat waves show that southern parts of the country experience more heat waves than northern parts. The investigation of heat waves in spring with two days of return showed that in many parts of the country the value is 100 days but in regions around Kerman this value is around 2000 days with a significant gradient. The time for the occurrence of heat waves with two days of return in fall is reduced to 50 days in comparison to spring.