Ethiopia is the second most populous country in sub-Saharan Africa. Although fertility rates in Ethiopia gradually decreased from 5.5 in 2000 to 4.6 in 2016, regional variations in fertility became wider, as evidenced by a TFR of 5.2 in the Somali region and 2.2 in Addis Ababa in 2016. From the perspective of demographic theory, geographical variations in fertility are often seen as either a reaction to different socioeconomic conditions (the adaptationist approach) or the diffusion of social acceptability of fertility control through geographical distance or linguistic similarity (the diffusionist approach). Recent fertility studies in high- and middle-income countries used spatial models to assess how the adaptation and spatial diffusion effects can jointly account for district-level fertility variations. However, such studies are rare in sub-Saharan Africa due to the shortage of district-level data. This DrPH uses a spatial approach to explore geographical variations in district-level fertility in relation to key selected determinants of fertility for 981 districts using the four Ethiopia Demographic and Health Surveys (2000, 2005, 2011, and 2016). I began by applying a Bayesian geostatistical approach to estimate the total fertility rate (TFR) and two proximate factors (modern contraceptive prevalence (mCP) and median age at first marriage) and two socioeconomic factors (proportions of women living in urban areas and with secondary education) and one ethnolinguistic factor for 981 districts in 2000, 2005, 2011, and 2016. I found that district-level TFRs within the same region were similar in 2000 and 2005, but they substantially varied in 2011 and 2016. In particular, spatial spreads of lower fertility were observed from the capital city to the northern and western parts of the country in 2011 and 2016. I then used spatial models to explore spatial autocorrelation of district-TFR and the spatially heterogeneous relationship between TFR and key selected factors affecting TFRs. I found that spatial autocorrelation of TFR became stronger in recent years. Results show that urban-rural differences in fertility were more associated with different socioeconomic conditions, and the recent spatial spread of lower fertility from Addis Ababa to the Amhara region was more associated with spatially heterogeneous effects of mCP, age at marriage, and ethnolinguistic diversity. This DrPH thesis demonstrates that the geographical location of and distance between districts are important aspects of the recent geographical variations in fertility in Ethiopia. Socioeconomic and cultural characteristics of districts substantially differ even within the same region in Ethiopia, and fertility in a district is affected by where a district is spatially located and the characteristics of nearby districts (diffusion effects), as well as by its own characteristics (adaptation effects). This DrPH thesis provides additional insights into how spatial aspects, as well as socioeconomic, cultural characteristics, and reproductive behaviours in districts, can jointly shape geographical variations in fertility in Ethiopia.