Malaria is a leading cause of morbidity and one of the most targeted communicable diseases. Malaria transmission varies in space and time. Understanding and extracting trends and noticeable patterns from district-level malaria surveillance data is crucial for achieving malaria reduction targets. The interplay of climate and environmental variability with malaria epidemiology is not explored. This study aimed to examine spatio-temporal patterns and trends of malaria epidemic by accounting for climate variabilities. Moreover the study strives to detect space-time clustering and develop a spatio-temporal predictive model for malaria transmission in the Amhara region, Ethiopia. The monthly malaria case incidence and environmental data were collected from Amhara Public Health Institute, NASA, CHIRPS and World Global Climate Databases. We have employed advanced statistical models such as parametric and nonparametric spatio-temporal trend models, Bayesian generalized Poisson model, Kulldorff’s retrospective space-time scan statistic, spatio-temporal generalized additive models, classification and regression training for spatio-temporal data (CAST), and Bayesian spatio-temporal predictive models. The linear parametric spatio-temporal trend estimate revealed a decreasing trend, and differential trends indicate that districts have substantial variations in their malaria transmission scenarios; meanwhile, most of the town districts have a lower reducing rate of malaria incidence. However, the thin-plate regression estimate shows a decreased trend between 2012 and 2016, and a drastic change has occurred since 2017. Malaria transmission is highly seasonal and modeled using cyclic cubic regression spline and shows that higher malaria cases occurred between September and December, with a higher seasonal index in October and December. Climate and environmental variability had a complex interplay with malaria transmission. The average altitude of districts, monthly average minimum temperature, monthly total rainfall, and coverage of at least one long-lasting insecticide net (LLINs) had significant effects on malaria transmission. However, the non-linear effects of climate variables revealed that the effects of climate variabilities were not homogenous across all the ranges of its values. Clustering of districts with a higher malaria risk was detected using Kulldorff’s retrospective space-time scan statistics, and the most likely space time cluster was seen in the western and northwestern parts of the region and occurred between July 2012 and December 2013. The spatio-temporal predictive model was performed using nonparametric and parametric supervised machine learning predictive models. Key words: Bayesian approach; Climate variability; Generalized additive models; Malaria surveillance; Predictive model; Spatial risk; Spatio-temporal; Spatio-temporal clustering.