A Project submitted in partial fulfillment of the award of masters of science degree in applied statistics of Meru University of Science and Technology

Meru town is the headquarte of Meru county and its population has been increasing over the years at an average annual rate of 2.36% per annum. As the population increases, the demand for water increases which in turn influences increase in total connections as the population density in town also increases. This growth in population will definitely mean more consumption of water. Therefore it si necessary to model water consumption in Meru Township and use the result to estimate population explosion.This study modeled the determinants of water consumotion in Meru township which included, water consumption, per captia consumption and connections. In doing so, the exploration of data for population, per captia consumption and total connections since the year 2005 to 2015 was done in order to understand the behavior of the data and ultimately to get the best model to predict water consumption in Meru township. After exploration of the data, growth models such as:Thomas R.Malthus's Model which is similar to Exponential Growth Model,Logistic Differential Model which is also similar to Logistic Population Growth Model and Logarithmic Growth Model were found not ideal for analysis of this data that favoured Time Series Models.The multi linear model was found to be appropriate for this study and ANOVA technique was employed to establish the validity of the model. The coefficient of the determinant of the model,R2 , is 99.9%, henec the best model for estimating population explosion using determinants of water consumption in Meru Township.According to the data at hand for the determinants of water consumption, it will take a very long time for population explosion to be experienced, however caution must be taken to conserve the natural resource.The AIC criterion was used to select the best model among AR model of different orders. AR (2) model with of 2516.47 was identified as the most suitable model to describe our time series data.