Data Visualization Research Topic

Storytelling by Numbers

Every scientific dataset – no matter how dense – has a story to tell. Data visualization is key for communicating our science to a general audience, but finding the core narrative buried in terabytes of data can be challenging. Leveraging Big Data analytical techniques to find that central story is key.

Just as there is no aspect of space-based research that falls outside the realm of extreme engineering, the idea of satellite measurements that cannot be considered “Big Data” is a rare concept. Given the cost to deploy and operate a satellite, sponsoring organizations strive to maximize the scientific value of every instrument in space. This leads to terabytes or even petabytes of data coming down from each satellite over the course of its mission.

With the immense volume of space-based observations available, finding meaningful trends or special cases is akin to finding a needle in a haystack. Fortunately, analytical methods have been developed for the era of Big Data that can bring these needles to the surface. Data mining, machine learning, and even brute force statistical analyses have led to some interesting discoveries in the lightning field. For example, the detection of bolides (meteors) in the GLM data as they burn up in the atmosphere.

However, relying on these statistical or pattern recognition techniques can be dangerous in scientific research if they become completely divorced from the underlying physics of the problem. Just as one can find correlations between completely unrelated variables (for example, the number of civil engineering doctoral degrees awarded and the per capita consumption of mozzarella cheese ), hands-off scientific analyses can draw premature or even completely incorrect conclusions.

For this reason, I take a guided statistical approach to space-based lightning research that reconciles all available data with the physics of the lightning discharge as well as signal modifications via scattering in the cloud layer. This leads to the production of holistic lightning datasets that describe the entire problem rather than a single part (i.e., optical emission from a return stroke) and simplify the identification of curious cases or larger trends.

Once these cases are identified, I visualize them by recreating to the extent possible what an observer would witness if they were there. Thus, my visualizations use ancillary data (i.e., NASA’s Blue Marble imagery), 3D scene rendering software, and geospatial Python packages to transform scientific data into imagery that approximates the real world, and then construct standard flat of 360-degree videos of the event. The latter can be used with Virtual Reality hardware to experience nature’s fury without the hazards of storm chasing.