Rapid Spatial Aggregation

Data visualization is an important component of spatial data analysis. We demonstrate the visualization of spatial/spatio-temporal data on map tiles as implemented in the R package RgoogleMaps. We argue that extremely large spatial or location data sets can lead to clutter and information overload necessitating aggregation to higher geographical identities. Such aggregation requires associating each coordinate point from the set to a particular spatial polygon in the search space. Examples for such polygon-based spatial partitions would be zip codes, census blocks, or school districts. Unless efficient data structures are used, this can be a computationally expensive task involving an exhaustive search across all prospective polygons. In this paper, we propose a methodology that exploits kd-trees as an efficient nearest neighbour search algorithm to significantly reduce the effective number of polygons being searched and expedite the lookup process. The kd-tree is built from either the polygon centroids and/or carefully chosen other points within the polygons. We further demonstrate a successful hybrid strategy by combining a range search with the tree based ranking. Our code has been made publicly available as the R package RapidPolygonLookup.