project paper on analysis of data on rainfall of Dhaka division

whether data on a rainfall data division of Bangladesh is stationary or not is checked through ACF and PACF plots and if this high seasonality then take a seasonal difference and the non seasonal difference to make the data stationary the predictive performance of the seasonal ARIMA model are just against the value of widely used model selection criteria, the AIC Criteria ARIMA (1,1,1) (0,1,1)12 at last rainfall for 6 years of Dhaka Station are predicted by the ARIMA. It is hoped that the forecasts of meteorological events would be great help to the user researcher and policy maker.
Weather forecasting is the application of science and technology to predict the state of the atmosphere for a future time and a given location. Human beings have attempted to predict the weather informally for millennia, and formally since at least the nineteenth century. Weather forecasts are made by collecting quantitative data about the current state of the atmosphere and using scientific understanding of atmospheric processes to project how the atmosphere will evolve.
These global forecasts are adjusted by tools of statistical adaptation to local conditions. These adapted forecasts include energy sector relevant meteorological variables, e.g.:
• Atmospheric pressure
• Temperature
• Wind
• Humidity
• Radiation
• Cloudiness
• Rainfall and indicators of extreme weather events, e.g. storms, lightning or ice.

Weather influences almost every human activity. It dictates the clothes we wear, the houses we build, the routes aircraft fly, and when best to sow seeds, and spray or harvest crops; it influences demand for energy. It is life-threatening; few sailors or mountaineers venture forth without the latest weather forecast. Increased accuracy over recent years has undoubtedly saved many lives.

The growing popularity of the time series data has been the corollary of its frequent and intensive use in empirical research. The interest of a prodigious number of econometricians has expedited its use only to help it to reach its zenith. In this study, endeavors have been made to forecast the meteorological events it is necessary to characterize and elucidate their various components. This study has considered the monthly rainfall of Dhaka station.

Testing for stationary of time series data is an important aspect of time series analysis. Hence concepts such as unit root, random walk and integrated time series are relevant. Assuming that the time series is stationary or can be made so by appropriate transformation(s), it is interesting to show how the Seasonal ARIMA modeling made popular by Box-Jenkins methodology to analyze time series data of monthly rainfall of Dhaka station from 1981-2009.

1.2 Background of the Study
The relationship between man and nature is so tied up that it is almost inviolable. Thanks to it, man has to depend on nature to a large extent. Utilizing the nature has set up the present base of civilization. However, still today, when he is at the apex of development, man is yet to learn more about nature. Moreover, to maintain the dynamic development of the civilization, the unimpeded study of meteorological events is must, throughout the world, especially in developing areas like ours.

Meteorological phenomena, such as rainfall, humidity, temperature, draught etc. influence the economy of Bangladesh to a large extent. The geographical location of Bangladesh is also responsible for its dependence on climatic behavior. According to Bangladesh data sheet 1999 published by BBS, 77.79% of the total population of Bangladesh is still living in pastoral areas. Most of them live by agriculture and their productivity is also influenced by the climatic change. Being a reverine country, Bangladesh has formed the largest delta in the world. Excepting a small portion of hilly areas in the northeast and southeast part, this delta is by and large a flat one. The Himalayan Mountains lie at the north and the Bay of Bengal lies at the south of Bangladesh.

For, making valid forecasting, we have to, make a choice from a number of competing alternative models. The comparison results lead us to a model that provides the best characterization in the viewpoint of the data. The model selection procedure largely depends on the process of searching for a suitable specification. In forecasting point pf view, the best model is the one that fits the observed data well and for which the forecast error, on average, is minimum in some sense. Another thing should also be borne in mind that a simpler model is always preferable to a complicated model.