Climatic variability and prediction of annual rainfall using stochastic time series model at Jhansi in central India

Abstract

A study was conducted on rainfall variability/change and to develop a stochastic time series model for annual rainfall prediction using rainfall data for the period of 72 years (1939-2010) at Jhansi (25°27ʹ N latitude, 78°35ʹ E longitude, 271 m above mean sea level).The analysis of long term rainfall data for the period of 77 years i.e., 1939-2015 revealed that annual rainfall varied between 375 to 1510 mm over 77 years with a decreasing trend of 4.2 mm/year. The long term mean annual rainfall is 908.3 ± 248.2 mm with a coefficient of variation of 27.3%. The rainfall of the region had been decreased by 319.5 mm over the period of 77 years from 1068.4 mm to 748.4 mm. Autoregressive (AR) models of order 0, 1 and 2 were tried and developed. The autoregressive model of the order 2 was able to predict the annual rainfall of Jhansi within ±20% in 74% of the years. Correlation (r) between the anomaly of observed and predicted annual rainfall from the climatological mean was 0.76. The goodness of fit and adequacy of models were tested by Box- Pierce Portmanteau test, Akaike information Criterion and by comparison of historical and generated data. The graphical representation between historical and generated rainfall was a very close agreement between them. The comparison between the measured and predicted rainfall by AR (2) model clearly shows that the developed model can be used efficiently for the annual prediction of rainfall at Jhansi.