Complexity reduction in choropleth map animations by autocorrelation weighted generalization of time-series data

Abstract

Choropleth map animation is widely used to show the development of spatial processes over time. Although animation congruently depicts change, the rapid succession of complex map scenes easily exceeds the human cognitive capacity, causing map users to miss important information. Hence, a reduction of the visual complexity of map animations is desirable. This article builds on research related to complexity reduction of static choropleth maps. It proposes value generalization of choropleth time-series data in space and time, by using a method that adapts to the degree of global spatiotemporal autocorrelation within the dataset. A combination with upstream algorithms for local outlier detection results in less complex map animations focusing on large-scale patterns while still preserving significant local deviations in space and time. An according software application allows for in-depth exploration of the spatial and temporal autocorrelation structures in time-series data and provides control over the whole process of generalization.

Empirical Insights from a Study on Outlier Preserving Value Generalization in Animated Choropleth Maps

“…For both tasks, he could not find benefits from temporal averaging of data, but rather found a non-significant tendency that participants who used temporally smoothed data even performed worse in picking the "maximum" map frame, which is understandable due to the smoothing of instantaneous peaks. Given the notion that animated choropleth maps are most useful to gain an overview of the development of spatial process [22,24,25] and significant local outliers (polygons with values greatly differing from their neighbors in space and time) thereof [26], it is questionable whether the tasks in McCabe's experiment were the best choice to tackle potential benefits of animated choropleth maps generalization. To fill this gap, we conducted an experiment to evaluate the effect of local outlier preserving value generalization in space, in time, and in a combination of both dimensions on the detection of overall trends and local outliers in animated choropleth maps.…”

Design of an Experiment to Evaluate Modes of Value Generalization in Animated Choropleth Maps

“…To construct the basemap, we clipped a roughly circular subset of Counties from the US State of Kentucky and merged, split or freely changed several County geometries to approximate their areas. According to Traun and Mayrhofer (2018), choropleth map animation of timeseries data seems useful only for data that exhibits a high level of autocorrelation in space and time, as this will lead to a continuous development of patterns throughout the animation. In turn, weakly autocorrelated data result in uncoordinated flicker, hampering perception, and thus calling into question the use of map animation generally.…”

“…Local outliers were determined by a high value-difference in relation to their first-order neighbourhood in space and time. To define thresholds, we used the heuristic of Traun and Mayrhofer (2018) that evaluates local differences in the context of global autocorrelation. It is based on the rationale that a certain local difference might qualify as a local outlier in a smoothly changing, highly autocorrelated dataset, but not in a less autocorrelated, 'rougher surface' (see Traun & Mayrhofer, 2018 for further detail).…”

“…To define thresholds, we used the heuristic of Traun and Mayrhofer (2018) that evaluates local differences in the context of global autocorrelation. It is based on the rationale that a certain local difference might qualify as a local outlier in a smoothly changing, highly autocorrelated dataset, but not in a less autocorrelated, 'rougher surface' (see Traun & Mayrhofer, 2018 for further detail). The explicit generation of a few (the probability of a polygon being a local outlier was set to = 0.002) but highly deviating local values within the otherwise relatively 'smooth' data from the GAMA model allowed a fairly clear separation of local outliers from 'regular' polygons.…”

“…Because of our interest in different types of generalization, we derived three differently generalized versions for each of the test stimuli by smoothing the polygon values that we generated by their  spatial neighbours  spatiotemporal neighbours  temporal neighbours (Figure 2). Using the methods and software provided by Traun and Mayrhofer (2018), the amount of smoothing adapts to the degree of autocorrelation in the data. Since all stimuli result from the same model parameters, spatiotemporal autocorrelation is similarly high, with Moran's Is (Moran, 1950) being between 0.84 and 0.93.…”