Search

This article studies the optimization of parallel clustering throughput in the context of variant-based parallelism, which exploits commonalities and reuse among variant computations for multithreading scalability. This direction is motivated by challenging scientific applications where scientists have to execute multiple runs of clustering algorithms with different parameters to determine which ones best explain phenomena observed in empirical data. To make this process more efficient, we propose a novel set of optimizations to maximize the throughput of Density-Based Spatial Clustering of Applications with Noise (DBSCAN), a frequently used algorithm for scientific data mining in astronomy, geoscience, and many other fields. Our approach executes multiple algorithm variants in parallel, computes clusters concurrently, and leverages heuristics to maximize the reuse of results from completed variants. As scientific datasets continue to grow, maximizing clustering throughput with our techniques may accelerate the search and identification of natural phenomena of interest with computational support, i.e., Computer-Aided Discovery. We present evaluations on a whole spectrum of datasets, such as geoscience data on space weather phenomena, astronomical data from the Sloan Digital Sky Survey on intermediate-redshift galaxies, as well as synthetic datasets to characterize performance properties. Selected results show a 1,115 percent performance improvement due to indexing tailored for variant-based clustering, and a 2,209 percent performance improvement when applying all of our proposed optimizations.

We used gravity mapping observations from NASA’s Gravity Recovery and Interior Laboratory (GRAIL) to detect, characterize and validate the presence of large impact craters buried beneath the lunar maria. In this paper we focus on two prominent anomalies detected in the GRAIL data using the gravity gradiometry technique. Our detection strategy is applied to both free-air and Bouguer gravity field observations to identify gravitational signatures that are similar to those observed over buried craters. The presence of buried craters is further supported by individual analysis of regional free-air gravity anomalies, Bouguer gravity anomaly maps, and forward modeling. Our best candidate, for which we propose the informal name of Earhart Crater, is approximately 200 km in diameter and forms part of the northwestern rim of Lacus Somniorum, The other candidate, for which we propose the informal name of Ashoka Anomaly, is approximately 160 km in diameter and lies completely buried beneath Mare Tranquillitatis. Other large, still unrecognized, craters undoubtedly underlie other portions of the Moon’s vast mare lavas.

NASA's GRAIL mission employed twin spacecraft in polar orbits around the Moon to measure the lunar gravity field at unprecedentedly high accuracy and resolution. The low spacecraft altitude in the extended mission enables the detection of small-scale surface or subsurface features. We analyzed these data for evidence of empty lava tubes beneath the lunar maria. We developed two methods, gradiometry and cross-correlation, to isolate the target signal of long, narrow, sinuous mass deficits from a host of other features present in the GRAIL data. Here, we report the discovery of several strong candidates that are either extensions of known lunar rilles, collocated with the recently discovered “skylight” caverns, or underlying otherwise unremarkable surfaces. Owing to the spacecraft polar orbits, our techniques are most sensitive to east-west trending near-surface structures and empty lava tubes with minimum widths of several kilometers, heights of hundreds of meters, and lengths of tens of kilometers.

The process of scientific discovery is traditionally assumed to be entirely executed by humans. This article highlights how increasing data volumes and human cognitive limits are challenging this traditional assumption. Relevant examples are found in observational astronomy and geoscience, disciplines that are undergoing transformation due to growing networks of space-based and ground-based sensors. The authors outline how intelligent systems for computer-aided discovery can routinely complement and integrate human scientists in the insight generation loop in scalable ways for next-generation science. The pragmatics of model-based computer-aided discovery systems go beyond feature detection in empirical data to answer fundamental questions, such as how empirical detections fit into hypothesized models and model variants to ease the scientist's work of placing large ensembles of detections into a theoretical context. The authors demonstrate successful applications of this paradigm in several areas, including ionospheric studies, volcanics, astronomy, and planetary landing site identification for spacecraft and robotic missions.

Analysis of transient deformation events in time series data observed via networks of continuous Global Positioning System (GPS) ground stations provide insight into the magmatic and tectonic processes that drive volcanic activity. Typical analyses of spatial positions originating from each station require careful tuning of algorithmic parameters and selection of time and spatial regions of interest to observe possible transient events. This iterative, manual process is tedious when attempting to make new discoveries and does not easily scale with the number of stations. Addressing this challenge, we introduce a novel approach based on a computer-aided discovery system that facilitates the discovery of such potential transient events. The advantages of this approach are demonstrated by actual detections of transient deformation events at volcanoes selected from the Alaska Volcano Observatory database using data recorded by GPS stations from the Plate Boundary Observatory network. Our technique successfully reproduces the analysis of a transient signal detected in the first half of 2008 at Akutan volcano and is also directly applicable to 3 additional volcanoes in Alaska, with the new detection of 2 previously unnoticed inflation events: in early 2011 at Westdahl and in early 2013 at Shishaldin. This study also discusses the benefits of our computer-aided discovery approach for volcanology in general. Advantages include the rapid analysis on multi-scale resolutions of transient deformation events at a large number of sites of interest and the capability to enhance reusability and reproducibility in volcano studies., Comment: Published in the Journal of Volcanology and Geothermal Research. 9 pages, 7 figures

Multiring basins, large impact craters characterized by multiple concentric topographic rings, dominate the stratigraphy, tectonics, and crustal structure of the Moon. Using a hydrocode, we simulated the formation of the Orientale multiring basin, producing a subsurface structure consistent with high-resolution gravity data from the Gravity Recovery and Interior Laboratory (GRAIL) spacecraft. The simulated impact produced a transient crater, ~390 kilometers in diameter, that was not maintained because of subsequent gravitational collapse. Our simulations indicate that the flow of warm weak material at depth was crucial to the formation of the basin’s outer rings, which are large normal faults that formed at different times during the collapse stage. The key parameters controlling ring location and spacing are impactor diameter and lunar thermal gradients.

This paper studies a form of parallelism termed variant-based parallelism, which exploits commonalities and reuse among variant computations in order to improve multithreading scalability. The problem is motivated by space weather studies that aim to identify changes in the Earth's ionosphere caused by auroral activity, tsunamis, and earthquakes. Today it is common to execute cluster algorithm variants with different parameters in order to determine which ones best explain phenomena in empirical data. We propose a novel approach and a set of optimizations to maximize throughput in such clustering algorithms. This is achieved by executing multiple clustering algorithm variants in parallel and developing efficient approaches to concurrently cluster data and maximize the reuse of results from completed variants. We present evaluations on real-world space weather datasets with up to 5 million ionospheric total electron content data points as well as synthetic datasets with up to a million data points. Results show a 1101% performance improvement due to indexing tailored for variant-based clustering, and a 2209% performance improvement when applying all of our proposed optimizations. Our optimizations enable new approaches in computer-aided discovery and could enable the short run times required for early warning systems for natural hazards.