Visualising Data From Dolphin Observations Through Adaptively Ordered Space-time Matrices

2019-09-16T01:19:09Z (GMT) by Judy Rodda Antoni Moore
Spatio-temporal visualisations (e.g. space-time cube) offer a method to observe interactions and
patterns that may not be apparent in a traditional two-dimensional view. Yet the mapping of such time
series data can still easily lead to visual clutter, leading to the development of abstracting REMO
(Relative Motion) and ARM (Adaptive Relative Motion) techniques. ARM uses greedy or simulated
annealing algorithms to optimally reorder object-time matrices from an arbitrary object order at each
time interval to one that keeps geographically proximal objects close to each other in columns and
across rows in the matrix (effectively applying a travelling salesperson algorithm to traverse object
locations for a given time). Testing each algorithm on eight seasons worth of space-time data for an
endemic species of New Zealand dolphin in a large southern bay revealed better results for the greedy
algorithm. However, given the pronounced orientation of the data (close to the SW-facing shoreline of
the bay), testing on more area filling data is required. Nevertheless, these results demonstrate
encouraging advances into algorithm-derived perceptive rationales of dolphin movements across
space and time.