Abstract
Climate simulations belong to the most data-intensive scientific disciplines and are-in relation to one of humankind's largest challenges, i.e., facing anthropogenic climate change-ever more important. Not only are the outputs generated by current models increasing in size, due to an increase in resolution and the use of ensembles, but the complexity is also rising as a result of maturing models that are able to better describe the intricacies of our climate system. This article focuses on developments and trends in the scientific workflow for the analysis and visualization of climate simulation data, as well as on changes in the visualization techniques and tools that are available.
Highlights
Climate research has come a long way, from models and observations that coarsely describe Earth’s atmosphere and ocean in the past, to dense satellite data and km scale global simulations that explicitly resolve clouds and precipitation today
The output generated by climate models has grown exponentially
While a few years back 80 km global simulations had been considered state of the art, producing about 2 GB of data per simulated day, nowadays we are working on 1 km global simulations requiring to store at least 8 TB of data per simulated day with the same number of time steps[1, 2]
Summary
At the same time, the analysis of small-scale processes that can be resolved demand a higher frequency output in order to be visualized properly, further increasing the amount of data by a large factor. As storage capacities are tight, one needs to carefully choose what to store and what not Another form of large data emanates from very long simulations, such as within the German Government funded project PALMOD, which simulates the climate from the last interglacial to the Anthropocene, i.e., a complete glacial cycle of about 120 k years.[3] While the spatial resolution of the model is low, data extraction and the visualization of many thousand time steps, and especially finding key events in this long dataset, is not a trivial task (see Figure 2)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have