Long Term Temperature Trend and Change Point: A Statistical Approach

Abstract

The long term temperature trends since 1941 to 2010 for Krishnanagar weather observatory, West Bengal, India are analysed. The significant change points in the trend are identified by applying CUSUM chart and Bootstrapping test to the time series of annual mean, annual mean maximum and annual mean minimum temperature time series. Different multi-length segments of the time series with similar trend have been detected by this method at different levels. For every segment, the independent error were adjusted the Mean Standard Error (MSE) with positive correlation and indicating change points at particular confidence level and confidence interval. This principle has been maintained by following bootstrapping technique. The major change point in the annual mean temperatures occurred around 2001 (confidence level 100%). This change point has been confirmed at level-1 (confidence interval 2000, 2001). Prior to the level1 change, the associated average temperature was 27.58◦C; while after the change, the same became 28.94◦C. On the other hand, the annual mean maximum and annual mean minimum temperatures have distinct change points at different levels. There are abrupt changes in the year of 1951, 1964, 1971, 1978, 1984, 1991 for annual mean maximum and 1971, 1978, 1986, 1991, 2000 for annual mean minimum temperature time series respectively. Moreover, the total number of change points, including level-1 change, is 18 for all the three annual mean, annual mean maximum and annual mean minimum temperature time series.