Spatial variation of feature density in multiscale topographic data

Digital topographic maps are created in a series of scales from large to small, and the underlying spatial data is commonly organized as a multiscale database consisting of several levels of detail (LoDs). Spatial density of features (or spatial objects) in such database varies both between LoDs (coarser levels are less densely populated with features) and within each LoD (feature density changes over the area). While the former type of density variation is caused by generalization, the latter one is mainly conditioned by geographic location and its properties, such as landscape complexity or fraction of urban areas. Since topographic database LoDs are derived using different data sources and generalization techniques, there is a need for a method that can help with automated evaluation of resulting feature density in terms of its appropriateness for the specified location and level of detail. This paper provides such method by uncovering dependencies between the location properties and the density of spatial data in multiscale topographic database. Changes in feature density are modeled as a function of spatial (landscape complexity and terrain ruggedness) and non-spatial (land cover types ratio) measures estimated via independent data sources. Resulting model predicts how much higher or lower is the expected spatial density of features over the area in comparison to the average density for the LoD. This information can be used further to assess the fitness of the data to the desired level of detail of the topographic map.