Constructing an Image of the Universe
Data visualization has driven discovery in astronomy and helped communicate new findings to the general public
What is the history of our universe? What are the origins of galaxies like our Milky Way? Where does dark matter, which apparently determines the basic structure of the universe, come from?
These are some of the big questions that astronomy deals with. As one of the oldest natural sciences, astronomy explores the basic connections of our universe and it has always been a data-driven discipline. Theory and experiments are in a close and constant interplay. However, experiments about the universe cannot usually be carried out under laboratory conditions or repeated as often as required because the phenomena in question are dynamic processes in distant stars, galaxies, or even the big bang itself.
To overcome the limits of experimentation, astronomers take advantage of the unimaginable size of the universe and the infinite number of objects within. While the death of a single star cannot be observed multiple times, statistical surveys allow for the detection and comparison of many different dying stars. In this way, the similarities and differences of the objects can be examined.
Very large datasets are therefore common in astronomy and the size of the datasets grow with the development of increasingly efficient observation methods [8]. More and more data are being collected during a single observation, which creates new challenges for data processing and its interpretation [10][11]. On the one hand, data test our concept of nature (physics). On the other hand, it constructs its own abstract and incomplete image of nature, which leads us to think of new theories and discoveries. It’s like watching the reflection in a very small mirror and trying to find the big picture that fits the reflection. So the question is, how can data be turned into information and ultimately into knowledge?
Information designers are also concerned with this question. Alberto Cairo describes the translation of data into information and knowledge through various processes of coding, decryption, and comparison with experience:
Reality is understood as unstructured information. Data represents a first encoding of reality. By means of a suitable arrangement and linkage, this data becomes structured information in a second encoding phase which can finally be transformed into knowledge by decoding it and by comparing it with one’s own experiences and memories. In terms of science, the first coding resembles the design and conception of an experiment by the scientist. It decides which statistical parameters are detected and how. Decoding the data is a crucial step in the chain of knowledge acquisition. This step involves recognizing patterns in the data and building a cognitive model. In order to obtain information and statements from the observation data, astronomers examine the underlying patterns and structures of the datasets. In the natural sciences, these are usually expressed concisely by abstract mathematical relations and correlations. For humans, however, visual representation is above all an effective method to understand complex data structures [2][14].
Data visualization is an important tool for understanding and communicating data and theoretical concepts in astronomy and science in general. Although the role of visualization in scientific analysis has changed over the course of history and with the method of analysis, from an outside perspective, information design plays a key role in knowledge acquisition and transfer. Hence, modern data analysis from large, complex datasets not only requires new information technology, but also new creative approaches. How can big data sets with many correlated parameters be visualized? The human brain has an outstanding ability in pattern recognition [5]. Due to the complexity of today’s datasets, however, intelligent algorithms, such as artificial neural networks, are increasingly being used to extract the knowledge from datasets and their internal correlations. Can data visualization itself become a tool to understand data more intuitively?
For the reasons outlined above, astronomy is naturally dependent on visualization. Humans are simply not able to perceive most astronomical objects and phenomena sensuously. Almost every popular scientific publication on an astronomical topic contains at least one artistic graphic representation or schematic drawing of an astronomical phenomenon. Joan Costa summarizes the task of visualization as follows [3]:
“[The goal] is to make certain phenomena and portions of reality visible and understandable; many of these phenomena are not naturally accessible to the bare eye and many of them are not even of visual nature.”
Not only the general public, but researchers as well, want to get an idea of the reality “out there.” For example, consider Fig. 1, which shows “the first image” of (the shadow of) a black hole. Technically speaking, it is a 2D color diagram that shows the brightness distribution of radio radiation in the galaxy M87. The fascination with this data visualization is based on the one hand on the context that the dark spot in the blurry cloud is supposed to be a black hole. On the other hand, it is fascinating to know that this is real data that conveys an image of something that is not visible to the eye. Black holes, stars, or planets are not natural objects that you can encounter on Earth. In order be able to understand them, one uses visualization. The fact that some astronomers are masters of coding reality is demonstrated by the famous Golden Record of the Voyager missions (cf., Fig. 2), which provides those who are able to decrypt it with fundamental information about our location in the universe.
The entire history of astronomy testifies to the ability of humans to form a visual image of invisible phenomena based on information. In this respect, world views themselves are of a visual nature. First of all, they do not correspond to a direct observation of reality. Instead, based on observations of the night sky, a mental space is created and the sky objects are located in it. Figure 3 shows a historical visualization of famous world views which show how different our construction in space can be and how it changes with time and the level of knowledge.
The order of celestial objects also plays an important role in epicyclic theory (see Fig. 4). The epicyclic theory is a geometric method with which, among other things, the Greek scientist Claudius Ptolemy described the planetary movements in the sky. He placed the planets on epicycles (circles whose centers move on another circle), which were on the main circles of the planetary orbits. This resulted in loop tracks with different distances to the center. With the description Ptolemy searched for an explanation for the observed changes in brightness of the planets.
For thousands of years, the human eye has been the primary instrument for observing and detecting astronomical sources. In order to be able to archive, interpret, and communicate what was seen, the observations were drawn by hand. Figure 5 shows one of the first historical drawings of the planet Mars, made by Christiaan Huygens around 1659 and Giacomo Maraldi in 1719. The drawings shows a large dark structure, which today probably corresponds to the Syrtis Major, the largest extensive plateau on the surface of Mars. Huygens made several of these drawings and determined the rotational speed of the planet from the recurring position of the structure.
The drawing as an archiving tool lost its importance with the invention of light-sensitive materials (photo plates). Observations of the sky were archived directly on the photo plates. The analysis was initially based on visual comparisons and the handy transfer to statistical graphics. The blink comparator is an example of a direct visual method as a tool [7]. Due to the rapidly changing illumination of two superimposed photoplates of the same sky region, the stars whose brightness has changed between observations seem to flash. So new variable objects could be detected, such as, in 1930, the dwarf planet Pluto was discovered with this visualization technique.
Visual comparisons were quickly translated into statistical data. As a result, information graphics and diagrams increasingly developed into statistical representations. The development of the butterfly diagram (Fig. 6) is based, for example, on the same principle as Huygen’s observations of Mars: repeated observations of a phenomenon are recorded at different times. While Huygens graphically depicted his observations, the observation by the sun observer Edward Maunder already shows an abstraction and thus a coding of the observation. By simultaneously recording the position of a sunspot and the time of observation in a diagram, he was able to recognize the pattern of a periodic geographic shift of the sunspots. The diagram thus represents an extension of the simple local order of data by visualizing both local and temporal parameters. The consideration of additional dimensions enables the viewer to place the observation in a larger context and makes correlations visible.
Graphic visualization can be understood as a tool for analyzing information and data from astronomical observations. Be it in the representation of concepts, through the direct geographic order and mapping of observations, or in the abstract representation in the form of statistical diagrams. The role of visualization finally changed due to the development of computer-aided analyses. The digital detection of light by semiconductor technology triggered a revolutionary change. Measurements spread across the entire electromagnetic spectrum and due to the direct conversion of light into electronic signals, digital and algorithmic analysis methods became more important. The examples above, in which the visual analysis was used, are datasets with few parameters that could be reasonably plotted and perceived in simple two-dimensional, statistical diagrams. With the development of computer-assisted generation and storage of data, the amount and complexity of data sets grew and computers took over their exploratory analysis. According to a review by Goodman [6], data analysis and its visualizations are now viewed as separate aspects of research. The visualization serves primarily to assess and communicate results and only follow after the computer-based analysis and processing of the data.
In general, demands for scientific data visualizations depend heavily on the analysis goal and the data sets. For example, visualization can continue to serve as a tool for precise mapping planet surfaces. Visual examinations are sometimes also used at the beginning of analysis processes to filter databases according to areas of interest and objects (see Fig. 7). Statistical diagrams are then used to classify and evaluate the results. In this sense, data visualization serves more as a control tool.
“Oddly though, as astronomy’s wavelength coverage increased, the value of the ‚visual‘ to astronomers seems to have declined — not as a wavelength, but as a tool.” (Goodman, 2012)
Contrary to Goodman’s hypothesis, there are still areas in which visual analysis comes into play. In these cases, it is often a matter of examining the structures and distributions of astronomical objects in so-called parameter spaces. For the Cosmic Flows Project [4] (see Fig. 8) an entire series of publications was based on a film. This represented an animated data visualization and showed the discovery of the super galaxy cluster Laniakea, in which our own Milky Way is located. Another class is physical simulations, which convey images of causal relationships and provide clues for new analyzes. A prominent example is the millennium simulation (see Fig. 9) of the structure distribution in the universe based on cosmological models.
The more efficient and prominent statistical analyses were made by computers, the more visual representations moved into the background as analysis tools. With the increasing dimensionality and complexity of the data sets, it became more and more difficult for the astronomer to get an overview of the data situation and its context. Instead, they are often forced to trust the results of the computer.
A problem with focusing on statistical methods is the lack of or difficult access to contextual information, or the limited overview of the content of the data set and the relationships of the data dimensions to one another [6][8]. The increasing importance of intelligent, machine-learning algorithms, the analysis steps of which are no longer easy to understand, also increases the need for transparency during data analysis [8]. In the case of highly complex data sets in particular the necessary transparency is often lacking, among other things, due to the lack of visualization concepts.
In his article on the visualization of complex data, Goodman identified four challenges for (explorative) data visualization [6]: Big Data, the variety and individual challenges in analysis, interface design, and the 3D selection of subsets. In the last two points, the design of information can obviously contribute to improvement. But the first two challenges offer the opportunity to find new universal design approaches.
In the case of high-dimensional data sets, though, interactive exploratory data visualization can give far more insight than an approach where data processing and statistical analysis are followed, rather than accompanied, by visualization“, Goodman, 2012 [6].
As early as the 1970s, the American statistician John Tukey defined a methodology with his work Exploratory Data Analysis (EDA) [13] to support the ability of humans to recognize patterns. It highlights the visual exploration of data sets. Modern visualization systems allow the viewer greater interactivity with the data in order to be able to assess working hypotheses quickly and fluently by viewing data from different perspectives or highlighting different parameters of a problem.
The challenge for EDA in data-driven modern astronomy, however, is to display data sets with a large number of parameters (multi-dimensional data sets). These must be clearly translated into visual parameters, such as position, shape, color, etc., so that the viewer can recognize possible patterns. Today, statistical data mining methods are used which easily capture multidimensional parameter spaces. The results of these analyzes are often incomprehensible to the astronomer and cause a certain amount of distrust/doubt. Especially due to the steadily growing amount of data, the need for new presentation concepts has been recognized by some research groups and has been investigated in various individual projects for several years [1][9][12]. However, most on these approaches still rely on 2D traditional concepts or displays (e.g., Fig. 10).
The new digital media allow us today to go one step further than to design interactive three-dimensional applications for the screen. Both augmented reality (AR) and completely virtual reality (VR) make it possible to experience the geometry of data rooms in three-dimensional spaces anew. The question is what advantage VR and AR actually have and whether this results in an improved perception of multidimensional data sets? The potential of these technologies is certainly not yet exhausted in the field of information design and might have a huge impact for scientific data exploration.
During my years in astronomy, data visualization has been an elemental part of my research. Toward the end of my PhD, I encountered a challenge quite common in modern astronomy: understanding and visualizing information of a big dataset. Since I was also studying information design, I started my exploration into data visualizations and how it could be a tool in processing multidimensional data in science or industry. In this series of articles I will describe my adventure, which eventually led to the development of the Virtual Data Cosmos.
References:
[1] Baines, D., et al., (2017): Visualization of Multi-mission Astronomical Data with ESASky, in: Publications of the Astronomical Society of the Pacific, Vol 129:028001, S. 6
[2] Cairo, A., (2013): The Functional Art: An introduction to information graphics and visualization, New Riders (Voices That Matter), 2. Auflage
[3] Costa, J. (1998): La esquemática: visualizar la información, Barcelona: Editorial Paidós, Colección Paidós Estética 26
[4] Courtois, H. et al., (2013): Cosmography of the local Universe, in: The Astronomical Journal 146.3
[5] Damasio, A., (2010): Self comes to mind: constructing the conscious brain, New York: Pantheon Books
[6] Goodman, A. A., (2012): Principles of high-dimensional data visualization in astronomy, in: Astronomische Nachrichten, Vol. 333, Issue 5–6, S. 505
[7] Gutekunst, M., (2015): Der Blinkkomparator, http://www.sternwarte-eberfing.de/Einfuehrung/Methoden/Blinkkomperator.html (read 18.01.2020)
[8] Hassan, A. & Fluke, C. J. (2011): Scientific Visualization in Astronomy: Towards the Petascale Astronomy Era , in: Publications of the Astronomical Society of Australia, Nr. 28, S. 150–170
[9] Kent, B., (2013): Visualizing Astronomical Data with Blender, in: Publications of the Astronomical Society of the Pacific, Vol. 125, S. 731–748
[10] Longo, G. / Merényi, E. / Tiňo, P., (2019),: Foreword to the Focus Issue on Machine Intelligence in Astronomy and Astrophysics, in: Publications of the Astronomical Society of the Pacific, 131:1001– 01 (6pp)
[11] Maarten, A. / Veljanoski, B. / Veljanoski, J. (2018): Veax: Big Data exploration in the era of Gaia, in: Astronomy & Astrophysics, Vol 618, S. A13
[12] Taylor, R., (2017), Visualizing Three-dimensional Volumetric Data with an Arbitrary Coordinate System, in: Publications of the Astronomical Society of the Pacific, Vol. 129:028002, S. 6
[13] Tukey J., (1977): Exploratory data analysis , in: Addison-Wesley Series in Behavioral Science: Quantitative Methods, Reading, Mass.: Addison-Wesley, Vol. 00/1977
[14] Ware, C., (2004): Information Visualization: perception for design. San Francisco: Morgan Kauffman
Image references:
Fig. 1: https://eventhorizontelescope.org/press-release-april-10-2019-astronomers-capture-first-image-black-hole (Zugriff am 18.02.2020)
Fig. 2: https://voyager.jpl.nasa.gov/golden-record/ (Zugriff am 19.01.2020)
Fig. 3: Rendgen, S. (2019)History of Information Graphics, Taschen Verlag
Fig. 4: Archiv Uwe Reichert, in: Karten, Krater und Kanäle, Sterne und Weltraum, Dezember, 2016
Fig. 5: Wikipedia, https://commons.wikimedia.org/wiki/File:Cassini_apparent.jpg (Zugriff am 20.01.2020) Original from Encyclopaedia Britannica (1st Edition, 1771; facsimile reprint 1971), Volume 1, Fig. 2 of Plate XL facing page 449
Fig. 6: Hamburger Sternwarte Digital Plate Archives
Fig. 7: ESA Sky: https://sky.esa.int (read 12.10.2019)
Fig. 8: Cosmic Flows Project, https://www.ip2i.in2p3.fr/projet/cosmicflows/ (read 14.11.2019)
Fig. 9: Millenium Simulation, MPA Garching, https://wwwmpa.mpa-garching.mpg.de/galform/virgo/millennium/ (read 05.01.2020)
Fig. 19: Taylor, R., (2017) https://www.youtube.com/watch?v=DMgXAKS1yI8&feature=youtu.be (read 22.05.2020)
