Multivariate Hybrid Visualisation of Ornithological Sensor Data

Abstract

To overcome the challenges of displaying multivariate sensor data, we propose a novel work-in-progress, hybrid, polar method of visualisation. Sensor data is collected by marine biologists in high volumes and using multiple sensors. Challenges arise when trying to unlock the marine wildlife behaviour from the vast amount of time series data collected. The proposed method filters uninteresting behaviour and isolates the features of interest within the set. Multi-layer polar plots are used to visualise local pressure, temperature, temporal behaviour and energy expenditure, all of which are given upper and lower plotting ranges to ensure no overlap. This results in a feature centred visualisation that focuses on the most important behaviour. The value in this method is that the visualisation can show many instances of the chosen activity. Each animal can be examined individually, or multiple animals and behaviours can be compared side-by-side for the first time. An interactive software system enables the user to navigate such that individual instances of the marine wildlife behaviour can be studied at high resolution or the user may choose an overview of every animal. This paper uses ornithological sensor data to demonstrate the proposed visualisation. Although it can be applied to other multivariate data sets that are linked with a temporal dimension.