[CRYOLIST] *Deadline extended* Five PhD studentships in Data Science of the Natural Environment at Lancaster University - including three Cryosphere topics

Leeson, Amber a.leeson at lancaster.ac.uk
Thu Jan 24 05:42:12 PST 2019


Applications are invited for five fully-funded PhD studentships in which you will learn to develop cutting-edge data science approaches to address key environmental science challenges. The studentships are part of the large-scale £2.6M EPSRC-funded grant “Data Science for the Natural Environment (DSNE)” (http://www.lancaster.ac.uk/dsne). This is an exciting opportunity to work at the heart of a multi-disciplinary team of researchers consisting of computer scientists, statisticians, environmental scientists and stakeholder organisations, working together to deliver methodological innovation in data science to tackle grand challenges around environmental change.



If learning to develop and deploy data science techniques to solve the biggest problems faced by humanity is an appealing next step, please send a letter of application to dsne at lancaster.ac.uk<mailto:dsne at lancaster.ac.uk> by 5pm GMT on Monday 11th February. The letter should include

  *   An ordered list of which of the PhD projects you would like to be considered for, with an explanation of your reasoning
  *   An explanation of why your skill set and previous education will allow you to besuccessful in these projects (a transcript of your undergraduate / masters degree programme is likely to be helpful)

To be able to answer these questions sensibly, it is advisable to talk to the supervisors of your desired projects in advance of submitting your application.



For a full list of projects click through link, details of Cryosphere topics copied below:



https://www.lancaster.ac.uk/data-science-of-the-natural-environment/studentships/



Non-parametric Mixture Methods for Improved Satellite Altimeter Retrievals over Ice Sheets
Supervisors: Mal McMillan, Marco Battiston, Amber Leeson & Chris Nemeth



For more information contact Mal McMillan on m.mcmillan at lancaster.ac.uk<mailto:m.mcmillan at lancaster.ac.uk>



This project offers the exciting opportunity to improve estimates of ice sheet mass loss, by developing novel approaches to processing complex satellite datasets. This project is particularly well-suited to those with a numerate background in mathematics, statistics, physics, computer science, or engineering, who want to make the step towards more applied study in the fields of environmental science and geophysics.



For the past 25 years, satellite radar altimeters have provided the longest continuous record of ice sheet response to climate change, detailing the increasing rates of ice melt and discharge into the oceans. One of the principal remaining challenges associated with processing these data is to retrieve estimates in complex topography, across areas such as rugged coastal terrain and the mountainous regions of the Antarctic Peninsula. These rapidly changing areas are amongst the most important to measure; yet the conventional retracking methods used to process altimetry data often struggle to adapt to the true complexity of the detected signal.



In this project, you will work with data from the novel Sentinel-3 Delay-Doppler altimeter, to develop new altimeter waveform processing techniques, based upon the field of Bayesian non-parametric mixture methods. These mixture methods offer a new approach for retracking altimeter echoes, because no strong assumptions are required about the topography of the ice sheet surface; instead this information is inferred directly from the data. Here we will focus specifically on the class of dependent Dirichlet process mixture models because they allow for highly non-Gaussian behaviour, skewness and multimodality in the waveform shape, all of which are expected features of the data. Following method development and proof-of-concept studies, you will apply this new approach to map contemporary rates of surface elevation change across the challenging coastal regions of Greenland and Antarctica, with the ultimate aim of delivering improved estimates of the contribution of both ice sheets to global sea level rise.



Diagnosing Antarctic ice shelf risk using coupled computational modelling
Supervisors: Amber Leeson, Andrew Orr (BAS), Gordon Blair, Chris Nemeth, Ryan Hossaini



For more information contact Amber Leeson, a.leeson at lancaster.ac.uk<mailto:a.leeson at lancaster.ac.uk>



75% of the Antarctic ice sheet is ringed by floating ice shelf, which is formed as the ice sheet spreads out over the sea. Since ice shelves are low-lying with respect to grounded ice, they are particularly vulnerable to atmospherically induced melting. In fact it is thought that surface melting induced the collapse of the Larsen B ice shelf on the east side of the Antarctic Peninsula in 2002, which resulted in a two- to six-fold increase in the flow of glaciers which formerly fed the shelf (Scambos et al., 2004). This has resulted in an ice loss to the sea equating to around one-third of the net loss from the entire region since the collapse event. The evolution of ice shelves relies on complex interactions between the atmosphere and the ice surface, and so in order to assess the degree to which the other Antarctic ice shelves are at risk a coupled modelling approach at high spatial resolution is required. However models do not yet exist which both 1) simulate climate on spatial scales of < 5km and 2) incorporate atmosphere-ice interactions deep enough into the firn pack which lies atop the ice. In this project we will develop computational infrastructure to couple the 3D Met Office Unified Model (UM) with the 1D community firn model (CfM) in order to address this gap. The model will be optimized using a fully Bayesian approach (e.g. Bobak et al., 2016) to select a model configuration and estimate the model parameters. High resolution simulations over the remaining ice shelves on the Antarctic peninsula (e.g. Larsen C, which is twice the size of Wales) will be performed under a range of future climate forcings out to 2100 and uncertainty in these simulations will be explored. This project will unite the disciplines of computer science, environmental science and statistics in order to provide a robust assessment of present and future stability of these shelves and ultimately determine which are at risk of collapse under a warming climate.



Bobak, S., Swersky, K., Wang, Z., Adams, R. P. and De Freitas, N. 2016. "Taking the human out of the loop: A review of Bayesian optimization." Proceedings of the IEEE104, no. 1: 148-175

Scambos, T. A., Bohlander, J. A., Shuman, C. A. and Skvarca P. 2004, “Glacier acceleration and thinning after ice shelf collapse in the Larsen B embayment, Antarctica”, Geophys. Res. Lett., 31, L18402, doi: 10.1029/2004GL020670.





Automated quantification of Greenland ice sheet melting using spaceborne radar data and multivariate changepoint methods
Supervisors: Amber Leeson, Rebecca Killick, Anna Hogg (CPOM)



For more information contact Amber Leeson, a.leeson at lancaster.ac.uk<mailto:a.leeson at lancaster.ac.uk>



Since the 1990s, the Greenland Ice Sheet has shifted from a state of near mass balance, to one of significant mass loss and has contributed approximately 10% to the measured global sea level rise during the last twenty years (Church, 2013). In the past decade, the rate of ice loss from Greenland has increased and the ice sheet has experienced episodes of rare and extreme surface melt. The dramatic increase in spatial and temporal coverage of satellite observations of the Greenland ice sheet offers a new opportunity to greatly improve our understanding of ice sheet melting, both in terms of these extreme melt events and under climatically ‘normal’ conditions. For example, Synthetic Aperture Radar (SAR) data acquired by the Sentinel series of satellites have previous been used to map and monitor snowmelt in the European Alps (e.g. Nagler et al., 2016). Here we propose to develop an image segmentation approach, based on changepoint methods, to derive melt area and volume from Sentinel 1a SAR backscatter images of the Greenland ice sheet from 2014 up to now. In the univariate and multivariate time series context, changepoint methods are commonly used to segment ordered data (e.g. Truong et al., 2018). This project will extend these ideas to the Sentinel 1a SAR data thereby creating a new framework for segmenting images using changepoint methods in addition to improving our understanding of Greenland ice sheet melting. The project will be based in Lancaster and there will be an opportunity for the student to pay an extended visit to the external partner based at the Centre for Polar Observation and Modelling in Leeds. Through this partner, methods developed during this PhD could potentially become part of the routine processing of these satellite data and be used to monitor future changes to Greenland ice sheet melting in near-real time.



Church, J. A., et al., : Sea level change, in: Climate change 2013: the physical science basis. Contribution of working group I to the fifth assessment report of the intergovernmental panel on climate change, edited by: Stocker, T. F., et al., Cambridge United Kingdom and New York, NY, USA, Cambridge University Press, 201

Nagler, T., Rott, H., Ripper, E., Bippus, G. & Hetzenecker, M. 2016. Advancements for snowmelt monitoring by means of Sentinel-1 SAR. Remote sensing, 8, 348.

Truong, C., Oudre, L. and Vayatis, N. (2018) A review of change point detection methods. ArXiV 1801.00718

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.cryolist.org/pipermail/cryolist-cryolist.org/attachments/20190124/44f3ea7b/attachment.html>


More information about the CRYOLIST mailing list