Multidimensional Exploratory Spatial Data Analysis (original) (raw)
The assessment of spatial autocorrelation is one of the primary tasks in geographical data analysis. Identifying and examining deviations from the expected autocorrelation is key to gaining a thorough understanding of the phenomenon under investigation. Traditional measures of geospatial sciences focus on the detection of spatial clusters or spatial heteroscedasticity, often in low-dimensional data. However, phenomena are often multidimensional and interdependent-both with and without their spatial dependencyand the toolbox of geospatial sciences is not yet well developed in this regard. The present study aims to contribute to this toolbox for scientists and practitioners. The proposed approach focuses on the detection of spatial discontinuity, considering heteroscedasticity by spatially contrasting residuals from a fitted spatial error model (SEM). This contrastenhancing technique identifies locations whose attributes differ significantly from those of the surrounding features, and with that the technique indicate spatial breaks. The approach is evaluated using agro-ecological field data to identify anomalies and was originally motivated for application in the context of precision farming. Our results enhance understanding of the underlying spatial processes of agricultural fields. The findings contribute to advanced, multidimensional, exploratory, spatial data analysis and present an alternative approach to conventional methods.
Loading Preview
Sorry, preview is currently unavailable. You can download the paper by clicking the button above.