|
Abstract:
|
Geostatistical models have been widely used to provide an accurate understanding of geospatial data at different formats. These models revealed their value from their flexibility in modeling geospatial data and predicting missing data in unsampled locations. Therefore, many methods and tools have been proposed in the literature to tackle the modeling challenges associated with geospatial data. Although many studies have explored geospatial modeling/kriging methods and tools, no existing framework can effectively assess them. In this work, we propose a benchmarking suite over ExaGeoStatR, an R package for large-scale geospatial data modeling on manycore systems, to assess existing modeling/kriging tools. The benchmarking suite relies on synthetic datasets generated by the ExaGeoStatR package and predefined assessment metrics. We evaluate five existing R packages in literature to explore the internal structure of the proposed benchmarking suite using small-size datasets. Medium-size and large-size modeling/kriging tools are also evaluated to show the capabilities of the proposed benchmarking suite.
|
