Article Preview
TopIntroduction
The frontiers of environmental science studies have been driven by hypotheses which are posed by researchers observing the living things and their surrounding settings (Botkin & Keller, 2010). These hypotheses are then formalized into research questions and tested through rigorous scientific investigations. Since environmental problems are related to real-world issues, research questions in environment science can all be decomposed into three basic dimensions: space, time and statistics. The spatial dimension defines the geographical objects and range of the study, such as gauging stations along rivers and the drifting objects in the ocean. The temporal dimension represents the span of the study, and the intervals of observations collected range from seconds to years. The statistical dimension exhibits the arrangement of attributes in the statistical space. These three dimensions are actively intertwined with each other and are inseparable in the analytic process (Ye & Rey, 2013). Over the last few decades, an increasing abundance of available space-time data in various domains is seen (Batty, 2005; Mennis & Guo, 2009; Ye & Shi, 2012). These big data are efficiently collected thanks to the growing capacity and ubiquity of information and communication technologies (Wang, Wilkins-Diehr, & Nyerges, 2012; Yang, Raskin, Goodchild, & Gahegan, 2010). In the meantime, numerous mathematical and statistical models have emerged with the technological and methodological advances. These models are dispatched through both commercial and open-source software packages, which greatly lowered the barriers for sophisticated and large-scale scientific computation (Anselin, 2012; Bivand, 2011; Rey & Anselin, 2010; Steiniger & Hunter, 2013). Benefiting from these advancements, researchers are able to look at data from multiple dimensions, new research questions can be identified this way.
Environmental investigation involves the descriptions and predictions of complex and high dimensional processes (Cressie & Holan, 2011). The diversity of perspectives is essentially the combinations of spatial, temporal, and statistical dimensions of different observations at different scales. Researchers of environmental science deal with heterogeneous data collected from numerous sources, thus it’s critical to be aware of difference of the raw data (unit of observation) and the data of interest (unit of analysis) (Ye & Rey, 2013). While the unit of observation is defined by the data collection process, the process itself is not always determined by researchers. The unit of analysis, on the other hand, is related to how the data are analyzed, interpreted, and framed by the research question. This conceptual difference indicates that the unit of analysis is represented by multiple underlying units of observations through processes such as aggregation, partition, or sophisticated statistical methods. Different types of unit of analysis thus lead to different perspectives of looking at the same data. This diversity of perspectives essentially reflect the scales of the analysis in the three basic dimensions.