Elsevier

Computers & Graphics

Volume 31, Issue 3, June 2007, Pages 370-379
Computers & Graphics

Visual Analytics
Designing a visual environment for exploration of time series of remote sensing data: In search for convective clouds

https://doi.org/10.1016/j.cag.2007.01.028Get rights and content

Abstract

Interactive animated images are often the only means to explore large time series of meteorological data sets. However, despite being interactive, animations still lead to information overload. We firstly look at the factors limiting the exploratory use of animations for studies of precipitating cloud and argue that two main factors are responsible for that: data complexity and animation design based on images that mimic reality. Then we present an example of how the current approach to visualize time series of meteorological images can be improved by computational methods, particularly by feature tracking. Next, we describe the visualization environment and discuss the representational, data mining and interactive functionality resulting from such a combination in an environment that is specifically dedicated to visual exploration and analysis of precipitating clouds.

Introduction

Earth system science is a multidisciplinary science strongly linked to remote sensing data. Satellite sensors are often the only way to obtain the required information of large spatial coverage and of high temporal frequency. As a result, satellite image repositories are becoming the fastest growing archives of spatio-temporal information and the users are frequently confronted with this continuous flow of data that need to be explored. For example, METEOSAT second generation is continuously observing the earth's full disk and monitoring the earth–atmosphere system. On a regular day, it collects 96 images with the frequency of one image in 15 min (where each image is a composite of 12 channels). These images are mainly used for meteorological applications. Exploring, analyzing and nowcasting precipitating clouds, is one of such applications, since one type of precipitating clouds, namely convective clouds, is often associated with storms and severe weather conditions.

The purpose of exploration is to find patterns, trends and relationships in these data sets in order to detect and predict the behavior of the precipitating cloud. One way of supporting the exploratory process is by developing visualization methods. The satellite images are presently explored mainly by animating 2D or 3D image sequences with user controlled video-type interactions (play forward/backward, pause/stop, etc.) (see, for example, [1], [2]) . However, evaluation studies show that despite having these interactions, animations still lead to information overload. In addition, the users are dissatisfied, because this kind of exploration remains a subjective and time-consuming process. One way of supporting the exploratory process is by developing alternative visualization methods and enabling more interaction than the usual video type of controls. In this paper, we look at the visualization aspects of precipitating cloud exploration and the way this process can be improved.

We firstly look at the factors limiting the exploratory use of animated image sequences for meteorological studies and argue that two main factors are responsible for that: data complexity and animation design, containing images that are mimicking reality. Then we present an example of how the current approach to visualize time series of satellite data can be improved by computational methods, particularly by feature tracking. Next, we describe the visualization environment.

Section snippets

Factors limiting visual search in animated image sequences

Why does an animated sequence of images overwhelm the users, and why is its exploration a subjective, tedious and time-consuming process? There are two main reasons linked to the characteristics of the data displayed and animation design.

If the images are played in an animated sequence, the complexity of the real world is mimicked. But users who are interested in evolution of precipitating systems have to identify that type of clouds and visually concentrate on these highly dynamic objects.

Detecting and tracking clouds

Generally, tracking can be performed in either a preprocessing or postprocessing mode [15]. In the preprocessing approach, the objects are extracted simultaneously using temporal (motion) and spatial (intensity) segmentation. If the postprocessing mode is used for tracking, individual images are first preprocessed to extract features,1

Visualizing the evolution of precipitating clouds

The object paths are input into the visual environment designed for exploration of cloud objects and identification of precipitating type of clouds. In the next sections, we will first describe the different views on the data that are provided, and then we describe the interactions that are included to support exploration and classification of precipitating clouds.

Interactivity for exploration: visual analysis

Interactivity is added to the visualizations, to increase their exploratory role. Besides linking the different views mentioned in the previous section, we implemented some controls on the visualization, helping the user to search for specific type of precipitating clouds and to focus attention on the selected objects.

The main exploration tasks in the example presented in this paper is following the precipitating clouds and identifying the convective types based on their lifetime properties.

Discussion and conclusion

We have indicated how the current approach to visually explore large time series of satellite images can be improved by computational methods. The main exploration tasks in the case presented in this paper is tracking the precipitating clouds and classification of convective clouds that are often associated with storms and severe weather conditions.

The tedious and lower level visual process of detecting and tracking each cloud is in our environment no longer a visual exploration task, since it

References (20)

  • H. Feidas

    A software tool for monitoring the features of convective cloud systems with the use of meteosat images

    Environmental modelling and Software

    (2003)
  • M. Molenaar et al.

    Fuzzy spatial objects and their dynamics

    ISPRS Journal of Photogrammetry and Remote Sensing

    (2000)
  • K.A. Robbins et al.

    Visualization of splitting and merging processes

    Journal of Visual Languages and Computing

    (2000)
  • W. Hibbard et al.

    Java distributed objects for numerical visualization in VisAD

    Communications of the ACM

    (2002)
  • Treinish LA. Task-specific visualization design: a case study in operational weather forecasting. In: IEEE computer...
  • M. Monmonier

    Summary graphics for integrated visualization in dynamic cartography

    Cartography and Geographic Information Systems

    (1992)
  • R.E. Lloyd

    Attention on maps

    Cartographic Perspectives

    (2005)
  • C. Ware

    Information visualization. Perception for design

    (2004)
  • R.A. Rensink

    Change detection

    Annual Review of Psychology

    (2002)
  • C.L. Folk et al.

    Made you blink! Contingent attentional capture produces a spatial blink

    Perception and Psychophysics

    (2002)
There are more references available in the full text version of this article.

Cited by (29)

  • Scientific visualization of landscapes and landforms

    2012, Geomorphology
    Citation Excerpt :

    Closed surfaces (spheroids) represent short term gain at the elevation level examined. This visualization concept is similar to the space–time cube approach proposed for epidemiologic studies (Kraak and Madzudzo, 2007) or remote sensing meteorological data (Turdukulov et al., 2007). The previous sections focused on visualization of terrain state and dynamics as captured by surveys using virtual terrain models representing real-world topography.

  • A conceptual framework and taxonomy of techniques for analyzing movement

    2011, Journal of Visual Languages and Computing
    Citation Excerpt :

    Relations between situations in different time units may be analyzed by means of change detection techniques, which are developed in the areas of computer vision, image processing, and remote sensing. In particular, these techniques are used for the reconstruction of events, movers, and trajectories from images (e.g. [26,27]). We are not aware of any specific computational techniques extracting relations of moving objects or events to elements of the spatial and temporal context; however, many kinds of such relations can be extracted by means of database queries (assuming that spatial and temporal predicates are supported).

View all citing articles on Scopus
View full text