A STUDY ON OCCURRENCE WET/DRY SPELLS AND WEATHER CYCLES USING MARKOV CHAIN MODEL

Abstract

IN THIS STUDY , ANALYSIS OF OCCURRENCE OF WET/DRY SPELLS AND WEATHER CYCLES AT CHANDIGARH HAS BEEN MADE USING DAILY RAINFALL DATA ( FROM 1958-89) WITH THE HELP OF MARKOV CHAIN MODEL X2 - TESTS FOR APPROPRIATENESS OF MARKOV CHAIN MODEL OF ORDER I (MCI) AGAINST THAT OF ORDER II ( MCI II) AND INDEPENDENCE SUGGESTED THAT (I) MCI WAS SIGNIFICANTLY DIFFERENT THAN MCI II AT 1% LEVEL OF SIGNIFICANCE. CONDITIONAL PROBABILITIES WORKED OUT DURING DIFFERENT MONTHS OF A YEAR SHOWED THAT VALUES OF P WERE THE HIGHEST (0.5) DURING JULY AND AUGUST AND WAS THE LOWEST DURING NOVEMBER ( 0.17) AND SUCCESSIVELY FOLLOWED DURING OCTOBER ( 0.19) . EXPECTED FREQUENCIES OF WET/DRY SPELLS AND WEATHER CYCLES WORKED OUT THROUGH APPLICATION OF MC I WERE INTERCOMPARED WITH THE CORRESPONDING OBSERVED FREQUENCIES BY X2 - TEST. THE TEST REVEALED THAT (I) TWO SETS OF FREQUENCIES OF WET SPELLS WERE NON- SIGNIFICANTLY DIFFERENT FROM EACH OTHER AT 5% LEVEL OF SIGNIFICANCE FOR ALL THE MONTHS (II) TWO SETS OF DRY SPELL ( EXCEPT FOR MAY AND DECEMBER) COULD NOT BE REGARDED AS DIFFERENT AT 5% LEVEL OF SIGNIFICANCE AND (III) THE WEATHER CYCLE COULD BE PREDICTED WITH THE HELP OF MC I FOR 8 MONTHS ONLY DURING A YEAR ( EXCEPT JANUARY , MAY , JUNE AND JULY.)