The advancement of technology has enabled us to collect increasing quantities of spatial and spatiotemporal data at rapidly increasing rate through sensor systems, automated geocoding abilities and social media platforms, such as Facebook or Twitter. Processing, analyzing and making sense of big data, which is characterized by high volume, velocity and variety, is challenging and hence, calls for increased computing performance. Exploratory spatial data analysis approaches, such as kernel density estimation, allow us to detect patterns that facilitate the formation of hypotheses about their driving processes. However, it is important to recognize that patterns of disease and other social phenomena emerge from an underlying population, which has to be accounted for in order to extract actual trends from the data. My dissertation research challenges a key assumption of many prominent methods of estimating disease risk, which is that population is static through time. I put forward the method of adaptive kernel density estimation by accounting for spatially and temporally inhomogeneous background populations. In addition, I develop a flexible spatiotemporal domain decomposition approach, which allows for tackling the big data challenge of developing scalable approaches to compute spatiotemporal statistics, using high-performance parallel computing. Last, I propose a framework for sensitivity analysis of spatiotemporal computing, which allows for quantifying the effect of model parameter values on computing performance and scalability. The results of my dissertation contribute to scalable applications for analyzing social geographic phenomena and elucidate the computational requirements of spatiotemporal statistics.