Climate and Health | Andy Berres

Ocean Simulation Development

The Model for Prediction Across Scales (MPAS) is an effort to model oceans, land ice, sea ice, and atmosphere for use in climate research and weather studies. The components are integrated in the Energy Exascale Earth System Model (E3SM). MPAS-Ocean is the ocean component of this model, which operates at mesoscale (sub 1km), and which offers local mesh refinement using using Spherical Centriodal Voronoi Tesselations (SCVTs) (unstructured tesselations consisting of 12 pentagons and a lot of hexagons to cover a sphere – a soccer ball is the most minimal example of such a tesselation).

Illustration beginning with simulation settings, then running the simulation and its analysis members (illustrated with rendered ocean flow), simulation output, and analysis scripts (illustrated with a line chart). — Overall MPAS-Ocean workflow from configuring settings to analysis.

During my involvement with this project, the active efforts included the development of new analysis members (simulated variables), and the addition of a mask-based method to enable the output of only a region of interest of the ocean. This can save massive amounts of time (for I/O) and storage space when running the simulation for centuries.

Most of a world map, centered on the Northern Atlantic ocean. A gray polygon covers the Northern Atlantic and parts of the Southern Atlantic. — A mask for the area containing the gulf stream.

My Contributions

Implemented “analysis members” (output variables) for MPAS-Ocean (Model for Prediction Across Scales) in Fortran90. in collaboration with physics and oceanography experts.
Converted global methods to mask-based regional methods.
Validated and visualized simulation outputs using charts and maps.

Two line charts with tolerances, each showing results for 4 different models in different colors. The top chart shows global MHT which is negative in the Southern hemisphere with a larger positive in the Northern hemisphere. The bottom chart shows Atlantic MHT which is consistently positive with maxima around 20 degrees North. — Comparison of Meridional heat transport (MHT) for the globe (top) and the Atlantic region (bottom). The model output (red) is compared to two reanalysis climatologies (blue and green).

Selected Publications

Compression of In-Situ Simulation Data

Climate research monitors many different spatial and temporal scales to assess future impacts. Centuries of climate simulations are run with very high resolution (1–10 km gridcells) ocean, sea ice, and atmosphere components. The petabytes of outputs can overload storage systems and hinder visualization and analysis. One way to mitigate this impact is to produce image databases of rendered results by performing in-situ visualization while the simulation is running. ParaView’s Catalyst component ties into the simulation to intercept intermediate results and produce these image databases. The image databases are an order of magnitude smaller than the raw simulation outputs.

The goal of this project was to assess if the output can be further compressed, and study the cognitive quality of the resulting images.

The top shows the workflow from start to finish: simulation, in situ image database, encoded into video database. From there, either extraction back into images or compression into a compressed video database and extraction back nto images. These outcomes are used to produce image quality metrics and to perform a perceptual study. The bottom shows to renderings of ocean flow in different colors. — Workflow for video compression and user testing (top) and example renderings (bottom) for different regions, colormaps, and variables (temperature, kinetic energy, and salinity).

My Contributions

Developed a workflow to connect the group’s work on energy measurements, algorithms, and cognitive evaluation.
Produced image databases of climate data using in-situ visualization during large-scale MPAS-Ocean simulation runs (ParaView/Catalyst).
Encoded and compressed image databases into MPEG-4 files.
Analyzed the impact of different traversal methods, video encoding parameters, and compression on database size, image quality metrics, and access speeds.
Supported conduction of a user study to evaluate the cognitive quality of images after encoding and compression using A/B testing.

Example video database showing kinetic energy. This video database used 198 cameras ($\varphi=18$, $\vartheta=11$).

Black and white drawing of the earth. Blue arrows labeled phi run around the globe in parallel to the equator. A red arrow labeled theta curves around the outside of the globe. — Illustration of the two angles, $\varphi$ and $\vartheta$, which are used to parametrize the different views of the simulated output.

Selected Publications

[Compressing In-Situ Simulation Data](../../publications/ berres2017vidcompression)
User Testing to Determine Cognitive Impact of Compression

Energy-Water Nexus

There are different ways to use water to produce energy (e.g., hydropower and wave power). Simultaneously, some energy production uses water for cooling (e.g., nuclear). This project’s goal was to build a web tool to analyze the interdependency of energy and water, using the World Spatiotemporal Analytics and Mapping Project (WSTAMP) framework as a basis.

Unfortunately, this project ended after one year due to funding cuts in the DOE Office of Science’s Biological and Environmental Research (BER) program.

User interface with a map view at the center and a scatterplot with a wave shaped pattern at the bottom. Both use blue yellow red colormaps and show temperatures. There’s a layer list on the left and an analytics workflow on the right. The top shows a series of bars of varying fill levels. — Main view of the user interface of the Energy-Water Nexus Knowledge Discovery Framework, displaying NCEI data on the Web World Wind Globe view. The chart at the bottom shows a timeline of temperatures. The bar at the top indicates data completeness.

My Contributions

Led interface (re)design for the web tool (Docker, Angular).
Integrated WebWorldWind as base map for visualization of data.
Implemented and coordinated data fusion for several dozen data layers.
Collaborated with contractors and staff members.

I presented this work at the American Geophysical Union (AGU) Annual Meeting in 2017.

Selected Publications

Full paper about the development of the tool

Health Data

COVID-19 Tracking and Projections

[!TODO]+ My Contributions

Selected Publications

M.Sc. Thesis: 3D In-Context Visualization for Probabilistic Tractography

Tractography is a method to visualize neural tracts using diffusion Magnetic Resonance Imaging data. Traditionally (in deterministic tractography), the nerve tracts are represented as lines of nerve bundles. On the other hand, Probabilistic tractography visualizes the probability of a nerve connection from a given point in the brain to any other point. What you get is volumetric data instead of bundles of lines.

My Master’s project and thesis were kindly co-supervised by staff at the University of Leipzig, in collaboration for the Max Planck Institute for Neurosciences in Cologne. My task was to visualize probabilistic tractography data and integrate my implementation in OpenWalnut, an open source visualization tool for medical data.

Teal and green nested surfaces + brain = surfaces rendered inside brain (with half the brain translucent.) — Visualization of just probabilistic tractography, just context, and both visualizations combined.

My Contributions

Developed a “focus and context” visualization for probabilistic tractogram, consisting of 2 components:
- a nested isosurface visualization with variable transparency for the tractography data, and
- a viewpoint-aware “glass brain” rendering of the cerebral cortex to serve as context.
Implemented both modules in C++ and GLSL, and integrated into OpenWalnut.
Wrote a M.Sc. thesis and presented the results to a colloquium audience.
Presented the results as a full paper at Visual Computing for Biology and Medicine and as a paper at IEEE BioVis.

Selected Publications