# TerraWulf - Projects

## Infrastructure Projects

**TerraWulf III Development:**- May 2011: Initial planning meetings
- June-November 2011: Negotiation with vendors, minor orders
- November 2011: Final approval of funding, electrical upgrade
- November 28 2011: Server order placed
- March 16 2012: Final delivery of the main computer hardware
- May 2012: Air Conditioning Upgrade
- July 2012: Staged installation of TIII compute nodes
- 25 July 2012: TIII online
- November 2012: Data server installed

**TerraWulf II Development:**- October 2006-June 2007: Study of options and negotiation with vendor
- June 28 2007: Final approval of funding. Order placed.
- August 9 2007: Delivery of the main computer hardware
- September-November 2007: Power & Air Conditioning Upgrades
- October-December 2007: Staged installation of TII compute nodes
- January-March 2008: Commissioning and testing
- March-April 2008: Storage expansion installed
- June 2008: Public launch

## Computational Research Projects

### Study of atmospheric effects in GPS analysis

*Christopher Watson (CSIS, University of Tasmania), Paul Tregoning (ANU)*

This project involves the analysis of several years of global GPS data using different analysis strategies in order to quantify the effects and errors that arise from using different models to represent atmospheric effects. The Earth's atmosphere affects the estimation of accurate GPS site positions both by retarding the signals transmitted by the satellites as they travel to Earth and by deforming the surface of the Earth as a result of atmospheric pressure loading. These two effects interfere destructively, causing errors in the estimates of GPS heights which subsequently limit the ability to analyse and interpret the geophysical signals of interest such as glacial isostatic adjustment.

**Software:** Gamit

### Seismic tomography with later arrivals

*Juerg Hauser (ANU)*

The focus of earthquake/teleseismic tomography has largely been on first arrivals and numerous ray and grid based schemes have been developed for predicting them. However, later arrivals often contribute much to the length and shape of a recorded wavetrain, particularly in regions of complex geology. These arrivals are likely to contain additional information about seismic structure, as their two point path differs from that of the first arrival. The project consists of testing and applying new schemes for the computation of later arrivals to allow harvesting of the information contained in them and has the potential to significantly improve seismic imaging.

**Software:** Development code

### Inversion of airborne electromagnetic data

*Ross Brodie (ANU/ Geoscience Australia), Murray Richardson (Geoscience
Australia), Malcolm Sambridge (ANU), Andrew Fitzpatrick (CSIRO Exploration
and Mining)*

Testing and tuning of new inversion techniques for airborne electromagnetic (AEM) data which improve on current frequency-domain AEM practice. In current practice, the processing & calibration of datasets are carried out on each frequency separately and the inversion is carried out on each sample separately, relying on subjective and non-quantitative procedures. It is only by considering all the components together in one large model system that consistency can be guaranteed. A new inversion approach called “holistic inversion” has been developed to simultaneously process, calibrate and invert all frequencies and all samples of a frequency-domain AEM survey dataset. This allows the use of a single model of the Earth, and estimation of AEM system calibration errors that are mutually consistent with the measured AEM data and independent ground truth.

**Software:** PETSc, FFTW

### Nonlinear inversion methods

*Malcolm Sambridge (ANU)*

A program of large scale nonlinear inversions of realistic geophysical datasets are being performed using leading edge algorithms from computational mathematics and statistics to examine their applicability to geoscience problems. Inverse problems arise in many areas of the geosciences from imaging the Earth’s interior to parameter fitting in geochemistry. The most difficult cases arise when the mathematical relationships between data and unknowns are nonlinear. The aim of this project is to develop new classes of approach to these problems, especially for highly nonlinear cases, or problems with many unknowns. The focus is on direct search (derivative free) methods that parallelize well on distributed computing platforms.

**Software:** Development codes, Neighbourhood algorithm, Simulated Tempering,
Markov chain Monte Carlo samplers

### Implementation of earthquake source parameter inversion

*Hrvoje Tkalcic (ANU), Myall Hingee (ANU), Andreas Fichtner (University
of Munich), Brian Kennett (ANU)*

A series of simulations investigating the relative importance of realistic, three-dimensional structural models of the Earth versus using more simplistic one-dimensional models for which computation is faster. The accurate determination of earthquake sources is of fundamental importance for a wide range of geoscientific applications including tsunami warning, seismic tomography, subduction zone dynamics, high-pressure mineral physics and geodesy. The process of seismic wave propagation from the source to the receiver can be described by Greens functions that account for the Earth’s heterogeneous structure. We will test different methods for the computation of the large set of Greens functions needed to account for the strong lateral heterogeneity of the crust and upper mantle in the Australasian region. This will require a large number of simulations with a well-tested spectral-element code in order to cover as many potential earthquake hypocentres as possible. For the calculation of a complete, long period waveforms, we use the spectral element method developed by A. Fichtner which has the important ability to accurately account for realistic continent-ocean boundaries.

**Software:** Development code

### Ocean overturning circulation role

*Andy Hogg (ANU), Graham Hughes (ANU), Ross Griffiths (ANU)*

A series of numerical experiments to assess which energy fluxes are significant in the ocean overturning circulation. The energy budget associated with the overturning circulation of the oceans provides insights into the governing dynamics. However the budget currently contains many unknowns, the resolution of which will be helpful in addressing problems such as the response of the overturning circulation to forcing changes and in highlighting processes that are important for general circulation and climate model development. We will examine the external inputs of available potential energy from surface buoyancy forcing as a function of model parameters to clarify the role of available potential energy in governing the circulation.

**Software:** MITgcm, NetCDF

### High-resolution ambient noise tomography

*Pierre Arroucau (ANU), Nick Rawlinson (ANU), Malcolm Sambridge (ANU)*

Previous theoretical and experimental studies have shown that cross-correlation of the long-term ambient noise wavefield recorded at pairs of stations provides an estimate of the impulse response of the intervening medium. This has given rise to a new class of seismic tomography which has been successfully applied at different scales around the world. The aim of this project is to exploit ambient noise data from an extensive program of temporary short period seismic array deployments occupied over the last decade to apply the new tomography to Southeast Australia.

**Software:** sac

### Earth's inner core structure & geodynamo

*Vernon Cormier (University of Connecticut), Hrvoje Tkalcic (ANU)*

Modelling the seismological effects of the crystalline fabric of the
inner core to identify lateral variations in structure associated with
crystallization of the solid inner core and fluid motion in the liquid
outer core. Focusing on the bottom of the liquid outer core, computations
will test for the existence of a high viscosity glassy region which forms
part of a new viscosity model of the liquid outer core. This has relevance
to the laminar versus turbulent nature of flow in the outer core.

**Software: **Development code

### Characterising the noise structure within geodetic GPS time series

*Christopher Watson (CSIS, University of Tasmania), Paul Tregoning (ANU)
*

Analyses of geodetic GPS data are commonly used to derive geophysical parameters such a plate tectonic movement, periodic deformation caused by seasonal mass loads (the annual water cycle for example) and secular changes such as that caused by Glacio-Isostatic Adjustment. The estimation of these parameters and their uncertainties is confounded by the coloured or frequency dependent noise structure present within the GPS time series. This research is aimed at applying maximum likelihood techniques to quantify the noise structure present within various different global GPS time series. This will provide a more rigorous determination of the uncertainties present in GPS data processed using different analytical strategies, leading to more robust assessment of these strategies.

**Software:** GAMIT, Development code

### Monte Carlo approaches to tomographic problems

*Thomas Bodin (ANU), Malcolm Sambridge (ANU)*

The main goal of the project is to develop new tomographic imaging methods in Earth sciences. A seismic wave propagating from a source to a receiver carries information about the structure within its volume of influence. Hence, after gathering the measurements of all the seismic waves that have traveled from a large number of sources to a large number of receivers it is possible to create a tomographic algorithm which will produce an image of seismic wave velocities. The construction of the algorithm depends on the sizes of the vectors representing the data set and the Earth model, on the physical assumptions that are made about the propagation of seismic energy, and, of course, on the computational power available. As more computational power becomes available, a new class of algorithms (traditionally restricted to small problems) can now be applied to seismic tomography. Terrawulf II will allow us to investigate these new possibilities for solving the tomographic problem.

**Software: **Development code

### Hydrology studies from space geodetic observations

*Paul Tregoning (ANU) , Christopher Watson (UTAS), Guillaume Ramillien
(CNRS, Toulouse), Herb McQueen (ANU), Jason Zhang (ANU)*

Hydrological processes cause variations in gravitational potential and surface deformations, both of which are detectable using space geodetic techniques. In this study we compute elastic deformation using continental water load estimates derived from the Gravity Recovery and Climate Experiment and compare it to 3D deformation estimated from GPS observations. Initial tests show significant agreement at several scales. This suggests that the techniques are more accurate than previously thought and that a large percentage of the non-linear variations seen in our GPS time series are most likely related to geophysical processes rather than analysis error. Comprehensive comparisons using Terrawulf II will allow us to test these conclusions.

**Software: **Development code

### Crustal deformation associated with the 2004 Macquarie Island earthquake

*Paul Tregoning (ANU) , Christopher Watson (UTAS), Reed Burgette (UTAS)*

The objectives of the study are to quantify from GPS observations the co- and post-seismic deformation associated with the 2004 Mw=8.4 earthquake that occurred near Macquarie Island. This involves analysing data from a global network of sites, including all available sites operating in Antarctica, Australia and New Zealand at the time. The GPS deformations will then be inverted to solve for fault rupture parameters and slip distribution on the rupture plane.

**Software: **GAMIT, Development code

### Glacial isostatic adjustment and non-stationary signals observed by GRACE

*Paul Tregoning (ANU), Guillaume Ramillien (CNRS, Toulouse), Herb McQueen
(ANU), Dan Zwartz (ANU)*

Changes in hydrologic surface loads, glacier mass balance and glacial isostatic adjustment (GIA) have been observed using data from the Gravity Recovery and Climate Experiment (GRACE) mission. However the geophysical phenomena are often not stationary in nature or are dominated by non-stationary signals. Using the variation in linear rate estimates that arise when selecting different time intervals of GRACE solutions we are deriving more accurate estimates of stationary signals such as GIA by more accurately removing model-based hydrologic effects.

**Software: **Development code

### Assessing the impact of reference frame stabilization sites on geodetic GPS time series

* Reed Burgette (UTAS), Paul Tregoning (ANU), Christopher Watson (UTAS)*

Analyses of geodetic GPS data are commonly used to derive parameters of geophysical interest such a plate tectonic movement, periodic deformation caused by seasonal mass loads (the annual water cycle for example) and secular changes such as that caused by Glacio-isostatic Adjustment. A key step in global GPS data processing is determining site positions in a consistent terrestrial reference frame through matching the positions of a stable sub-set of stabilization sites to values determined through the synthesis of multiple space geodetic techniques. However, even the most stable sites move from their nominal coordinates due to geophysical processes. This project will assess the errors associated with using different sets of stabilization sites to transform GPS data to the terrestrial reference frame and seek an optimal strategy, using both observed and simulated GPS time series.

**Software: **Development code