Variability of Rainfall by Some Others Climatic Phenomena in the Northern Part of Bangladesh

Abstract

The main objectives of this study are to find the seasonal variations of rainfalls and it’s related some others climatic variables in the northern part of Bangladesh. Also a suitable panel regression model of rainfalls on these climatic variables is fitted. In this study, climatic data for different phenomena i.e. total monthly rainfall in millimeters, humidity in percentage, cloud covers in hour per day, average temperature per day in degree Celsius, bright sunshine in hour per day from 1981 to 2017 are collected from Bangladesh Meteorological Department. The selected explanatory variables humidity, temperature, cloud cover and bright sunshine are considered in this analysis. For the availability of data for all of these variables considered in this study, the weather stations of Bogura, Rajshahi and Rangpur are considered which are situated in the northern part of Bangladesh. Among these three cities, the overall average monthly rainfall is the highest in Rangpur with high variability and the lowest is in Rajshahi with low variability. For all these three cities seasonal effect of rainfall is the highest for the month of July, seasonal effect of sunshine is the highest for the month of March, and that of cloud cover is the highest for the month of July. But for humidity variable, the seasonal effects of Bogura and Rajshahi stations are the highest for the month of July but that for Rangpur station is in September. Since all of these time series data have seasonal variations and non-stationary so to run panel data regression model it is needed to be made transformation for getting stationary time series data. For rainfall data it is needed two times transformation to get stationary form. Firstly, using fourth root power transformation and the secondly make a twelve periods lag difference transformation to obtain stationary form. But for all other climatic variables only twelve periods lag difference transformation is sufficient for getting stationary form.  Insignificant Cook-Weisberg test statistic suggests the homoskedasticity and lowest variance inflating factor conferred the absent of multicollinearity among explanatory variables. A panel regression model is employed to check the effect of humidity, temperature, cloud cover and bright sunshine on rainfall which is conferred by Lagrange multipliers test statistic. Since hausman test statistic is insignificant hence the random effect panel regression model is considered. Coefficient of non-determination is 0.66, implies that about 34% variation of rainfall can be explained by these explanatory variables. Cloud cover, humidity and sunshine have significant positive effects on rainfall whereas average temperature has significant negative effect on rainfall. One can try to fit a dynamic panel regression model such as GMM with Arellano Bond correction or two steps analysis of panel data models for further study.

Key Words: Rainfall, Variability, Hausman Test Statistic, Cook-Weisberg Test Statistic, Random Effect Panel Model.