The paradigm of a „Spatial Data Science“, its methods and models for ...

Today we are living in a starting digital age. There are many new techniques and automated work flows supporting our daily life, decision making, industrial production etc. Data is produced anywhere and anytime and used for solving many problems. Geographical data is also used for solving geographical problems like orientation, location, regionalisation, assessment, spatial prediction etc. How a new geographical paradigm “Spatial Data Science” is supporting the solution of such problems is discussed in this paper.

During the last decades different terms like Locational Analysis, Quantitative Geography, Computational Geography or Spatial Analytics were used to name the scientific work based on data in geography. The background for these approaches always were current “methodical paradigms” like Multivariate Methods, System Dynamics, Fractals, Cellular Automata, Fuzzy Sets, Agent-based Modelling etc. Though these paradigms consisted of very different methods, the basic abstract problems, which could be solved by geographical data processing, always were the same. These ageless basic problem types in geography are Data Description, Data Reduction, Classification, Location/Allocation, Assessment, Interaction Modelling, Process Modelling, Prediction and Prescription. Because of the different methods being developed in the changing paradigms the solving methods for the various basic problem types in geography were adopted to them and changed too. In the first part of this paper three basic problem types, namely classification, location/allocation and process modelling are described and the different solving methods are listed and discussed.

The newest methodical paradigm, which has been used for data processing and modelling in geography since half a decade is “Spatial Data Science”. Using this paradigm most of the basic problem types can be solved in a proper way. Usually the problem solving or the modelling process follows the “Cross Industry Standard Process for Data Mining (CRISP-DM)”, which is the basic procedure in most modern data science software. For solving problems using geographical data some special geographical problems, like Spatial Autocorrelation, the Modifiable Areal Unit Problem or Ecological Fallacy have to be considered. In combining geographical data, software for solving the special geographical problems and a data science procedure very good applicable problem solving procedures of spatial data science can be generated. In the second part of this paper spatial data science approaches were described which can be used for solving the three addressed basic geographical problem types.

As a final best practice example a model from the context of energy potential spatial analysis, prediction and prescription using spatial data scientific methods are presented.

There are still many methodical, software and interpretational problems in applying spatial data science methods. Finally they are shortly discussed. The paper intends to show that nevertheless a spatial data science approach can provide useful solutions for the most data based geographical problems already.