Are current spatial data bases useful for meaningful analysis?

abstract:

With increasing amounts and variety of and access to open data collected in our environment, the distance between those who understand the data observation processes and those who analyse the data increases. This enlarges the risk of carrying out operations that are not meaningful, such as summing temperature values or interpolating coal power plant emissions. To avoid the need for tables with permitted analysis procedure for every phenomenon, we attempt to categorize phenomena first. Combining the ideas of Stevens' (1946) one-dimensional measurement scale types with those of spatial and temporal reference systems, we construct compound reference systems that address how space, time, quality, and entity cohere. What we find formalizes ideas established in spatial statistics over three decades ago, but has not been reflected well in spatial data standardization efforts, or spatial data base design. The talk will discuss the potential of these findings, implementation challenges, and their value for relational and array data bases for spatio-temporal data.

Edzer Pebesma leads the spatio-temporal modelling laboratory at the institute for geoinformatics, and is director of this institute. He holds a PhD in geosciences, and is interested in spatial statistics, environmental modelling, GI Science, geoinformatics, and optimizing monitoring strategies. He is one of the authors of Applied Spatial Data Analysis with R (second edition), and serves as associate editor for Spatial Statistics, Computers & Geosciences, the Journal of Statistical Software and Environments. He believes that research is useful in particular when it helps solving real-world problems.