The impact of varying semantics in spatial statistics

This paper extends consideration of semantics to identify the sources of semantic variation in modelling, using a case study on soil conductivity in the Loess Plateau, China through spatial regression as the model. Given a simple case of y = f(x), noting that y will only ever be a predicted, y*, the paper examines the impacts on semantics of Measurement of x (the ‘x’ issue), the Choice of x (the ‘m’ issue) and the Support of x (the ‘v’ issue). These affect the semantics of y* through model inputs, specification and granularity, and although some of these factors are known, they have not been considered in this context together before.