Understanding Spatiotemporal Big Data

Spatiotemporal big data can be understood, to a large extent, as location and trajectory information collected through location-based service (LBS) applications. Because of its fine spatial and temporal resolution and large-scale population coverage, it has significant commercial as well as scientific value.

The essence of big data does not lie simply in its volume, but in the change of perspective it brings. Big data analysis provides a new way of discovering patterns that were previously hidden. It can be regarded as a microscope for observing human society and a dashboard for monitoring nature. It represents the added value that emerges when countless individual records are aggregated, forming a bottom-up approach to empirical research driven by massive data.

Broadly speaking, big data analysis serves two roles. Looking backward, it enables descriptive analysis, identifying historical patterns and hidden regularities beneath the surface of data. Looking forward, it enables predictive analysis, offering forecasts of future trends. This transition from the “known” to the “unknown,” from the past to the future, is the true vitality and essence of big data.

The strengths of spatiotemporal big data are particularly relevant to geography. As the problems facing both society and science become increasingly complex, traditional approaches often fall short. Big data makes interdisciplinary, integrative research possible—the very spirit of GIS. Compared to remote sensing, spatiotemporal big data brings stronger attention to human-centered socio-economic domains. Compared to traditional human geography, it offers speed, authenticity, and convenience. It is, in itself, an added value arising from data aggregation.

However, spatiotemporal big data is not without its limitations. Data acquisition remains difficult, and most analyses focus on correlations rather than causal mechanisms. This is reminiscent of many remote sensing models: correlations without underlying explanation. A classic example is the Google Flu Trends (GFT) project. In 2013, GFT predictions of influenza outbreaks turned out to be inaccurate. The reasons were twofold: big data hubris—the assumption that big data could replace rather than complement traditional data collection—and algorithmic change, since Google’s own updates altered the way data was generated, artificially inflating search counts and leading to overestimation.

These issues highlight that while big data holds great potential, it cannot yet replace established methods or theory. Without sufficient contextual understanding, relying solely on large numbers can be misleading. Instead, the future of big data should lie in combining big data with small data—traditional controlled datasets—to create deeper, more reliable representations of human behavior. Rather than a “big data revolution,” what is needed is an all data revolution, where new technologies and methods are applied comprehensively for richer and more accurate analysis.

In this sense, spatiotemporal big data is not just a trend but a critical resource that, when thoughtfully integrated, can transform the way we observe society, monitor the environment, and support decision-making across domains.