Probabilistic Weather Forecasting Using Bayesian Model Averaging: A Case of SAGCOT Regions

dc.contributor.author	Joel, Emmanuel
dc.date.accessioned	2020-09-21T10:27:56Z
dc.date.available	2020-09-21T10:27:56Z
dc.date.issued	2018-11
dc.identifier.citation	APA	en_US
dc.identifier.uri	http://hdl.handle.net/11192/4209
dc.description	A Dissertation Submitted in Partial Fulfillment of Requirements for Award of the Degree of Master of Science in Applied Statistics of Mzumbe University.	en_US
dc.description.abstract	Over the past decade, Tanzania has experienced spontaneous population increase (1.556 mil annual). But the number is estimated to further increase by 2050 to 2.982 mil annual, thus Tanzania is estimated to have a population of 137 million people in 2050 (UN, 2015). The fast-growing population is mainly depending on rainfed agriculture, which contributes 29 percent of the country GDP and providing employment to 65.5 percent of Tanzanians (Deloitte, 2016). The diversity in climatic and weather activities has posed a challenge in rainfed agriculture especially on when to plant seeds. Therefore in order to promote agricultural activities, stable and reliable weather information are crucial in order for production to match with population increase. This study explores the challenge facing the Numerical Weather Prediction (NWP) namely WRF-ARW, by creating the system of equation (ensembles) from WRF-ARW resulting from the use of different initial conditions. Ensemble allows for the probabilistic forecast to take the form of predictive probability function (PDF). But, the raw ensemble forecast system is finite hence they only capture some of the uncertainty of the NWP. Thus, this study used the Bayesian Model Averaging (BMA) methods of postprocessing ensemble forecast to maximize the sharpness of the parameter and calibration. The findings show BMA method successively removes most of under dispersion showed by raw ensembles. Thus, calibrated and sharp results of BMA approach resolves a number of the weaknesses of the ensemble forecasts including their under dispersion and the discrepancy between forecasts and observations. Therefore BMA can be used to attain higher consistency in the probabilistic forecasts of an operational model.	en_US
dc.description.sponsorship	Private	en_US
dc.language.iso	en	en_US
dc.publisher	Mzumbe University - Faculty of Science & Technology	en_US
dc.subject	Bayesian Model	en_US
dc.subject	Weather forecasting	en_US
dc.subject	Probability forecasts	en_US
dc.subject	WRF-ARW model	en_US
dc.subject	Numerical Weather Prediction	en_US
dc.subject	SAGCOT regions- Tanzania	en_US
dc.title	Probabilistic Weather Forecasting Using Bayesian Model Averaging: A Case of SAGCOT Regions	en_US
dc.type	Thesis	en_US