Many researchers in Earth and Atmospheric Sciences use PACE to enable their work. Here, we highlight a few examples of their accomplishments:
Studying Coral Ecology with Machine Learning
Anthropogenic stressors such as fishing, coastal development and global climate warming threaten ocean biodiversity and ecosystem functioning.
In the past 30 years, surface ocean temperatures around the Coral Triangle, the most diverse and biologically complex marine ecosystem, have risen by approximately 0.1 °C per decade and are likely to climb an additional 1-4 °C by 2100. An increase of more than 2 °C will eliminate most coral-dominated reefs, with the potential of impacting the livelihoods of 120 million people. If corals disappear, marine biodiversity as a whole will reduce greatly, because corals are foundational species.
Conservation management and mitigation strategies requires information about the stressors, and most importantly about the distinct ecoregions that demark unique assemblages of species. For example, the potential of a marine protected area to retain biodiversity and restock species beyond its border, is inherently linked to some form of regional discretization and population connectivity. In most of the world Ocean, and in the Coral Triangle, defining the ecoregions is complicated by data sparseness and by the large-scale, time-dependent dispersal of ocean currents. Indeed, ocean currents and water characteristics change in space and time, and predictions of community susceptibility to these changes remain elusive.
This knowledge is especially important when biodiversity must rebuild following a devastating damage, as for example following bleaching events, because ecoregions and connectivity contribute to colonization and resilience of a given reef. In most of the world’s Oceans direct observations of species distributions are indeed sparse in space and time. Model simulations of ocean circulation and larval transport can help, but they are limited in their spatial and temporal resolution.
In Novi and Bracco (2022), the research group of Dr. Annalisa Bracco developed a framework to help filling this fundamental knowledge gap in coral ecology and introduced an unsupervised machine learning (ML) approach that exploits the dynamical relationship between sea surface temperature anomalies and sea surface height anomalies, and therefore currents, at spatiotemporal frequencies pertinent marine communities at latitudes comprised between 45°N and 45°S. The ML framework, schematically summarized in the figure, was applied to the sea surface temperature anomalies from a high resolution reanalysis, GLORYS, from 1993 to 2017, grouped according to the phase of the El Niño Southern Oscillation (ENSO). They then introduced a biodiversity score and a recovery potential metric. Taken together, they allow for identifying appropriate temporal and spatial windows where connectivity-based restoration efforts and monitoring should be prioritized. The analysis has shown also that the extraordinary biodiversity that we observe today in the Coral Triangle is maintained and enhanced by ENSO variability. ENSO has long been associated with coral reefs mortality, due to the prolonged higher-than-normal temperatures experienced in some areas during its El Niño and La Niña phases, but the researchers demonstrated that it is also highly beneficial to biodiversity. It enhances the large-scale exchange of genetic material between the Indian Ocean and the Coral Triangle during La Niña years, and between the Coral Triangle and the area to the east in neutral conditions.
Given the societal relevance of reef ecosystems, the potential of the framework presented is profound.
Thermal Inertia in Mars's North Polar Region
Understanding the geology of a planetary surface is essential for interpreting its evolution and geologic history and critical for landing site selection. One such method of orbital data analysis is the derivation of a surface's thermal inertia. Thermal inertia is the fundamental property controlling a surface's daily temperature variation. It depends on particle size, degree of cementation, rock abundance, and bedrock or ice exposure within the subsurface' top few centimeters. Thermal inertia measures the surface's ability to store heat during the day and re-radiate it at night. On Mars, geologic investigations of this nature provide insight into past climate conditions and surface processes occurring today, such as sedimentary deposition and regolith formation. However, the accurate derivation of thermal inertia requires the combination of multiple datasets with robust thermophysical models.
To tackle this, Drs. Benjamin McKeeby and Frances Rivera-Hernandez at Georgia Institute of Technology and PACE developed a novel method combining the facilities' computing resources with existing thermophysical models. This image, computed on Georgia Tech's Phoenix cluster, represents the distribution of exposed ice (red), sand (yellow), and dust (purple) found within Olympia Undae in Mars's north polar region. Olympia Undae is a vast region of gypsum and basaltic sand dunes that ring the planet's north pole. This image's high degree of heterogeneity indicates an active surface with different geologic processes between the dune ridges and troughs.
Modeling Historical Floods to Prepare for the Future
Outburst floods from naturally dammed lakes have dramatically shaped landscapes for eons. Failure of small lakes and reservoirs are causing devastating impacts on downstream communities and infrastructure today, but much bigger floods occurred thousands of years ago when widespread glaciers and the sediment they carried dammed river flow to create huge lakes. Today, Dr. Karin Lehnigk and the research group of Dr. Karl Lang use models to recreate these ancient lakes and study the floods that occurred when they could no longer hold back the water. They want to know how big these floods were, how they interacted with sediment, and how they changed topography, so that we can be better prepared for modern extreme floods and understand the full range of what the Earth is capable of. This figure shows modeled water depth in the first 2.5 hours of a flood from the ancient 80 cubic kilometer lake formed when glaciers blocked the Yarlung-Siang river in the eastern Himalaya during the last Ice Age, hovering above the former location of the dam looking downstream; one can see flat sediments along the edges of the river channel that settled out of the lake near the bottom of the image. The flood winds its way through narrow mountain passes constrained by the tectonically active landscape, eroding bedrock and moving sediment. Models are run using the 2-phase depth-averaged shallow-wave hydrodynamic model D-Claw[1,2] run on the Partnership for an Advanced Computing Environment (PACE) at Georgia Tech; imagery and topography are from Google Earth.
[1] A depth-averaged debris-flow model that includes the effects of evolving dilatancy: 2. Numerical predictions and experimental tests. D. L. George and R.M. Iverson, 2014. Proc. R. Soc. A, 470 (2170).
[2] A depth-averaged debris-flow model that includes the effects of evolving dilatancy: 1. Physical basis. R.M. Iverson and D.L. George, 2014. Proc. R. Soc. A, 470 (2170).
