Visit to NCAR for Statistical-based Compression of Climate Model Output

Lead Research Organisation: Newcastle University
Department Name: Mathematics and Statistics

Abstract

This travel grant aims at establishing a collaboration between the Principal Investigator, a newly appointed Lecturer in the School of Mathematics & Statistics at Newcastle University, and the National Center for Atmospheric Research (NCAR) in Boulder, USA (one of the top research centers in climate in the United States and in the world) on a project in climate data compression using statistical models. The development of compression algorithms specifically aimed at climate output is new to the scientific community, and NCAR is, at present, the only institution pursuing this innovative research area. The development of statistically-based compression algorithms, as well as methodologies to assess the quality of the compression, will allow climate scientists to generate a larger quantity of data, to improve their scientific understanding of the Earth's system and to deliver more reliable future climate projections. The PI has developed a statistical based compression algorithm for climate ensemble that has already been published in a top statistical journal, and intends to extend the methodology to multiple variables, an investigation of high interest to NCAR scientists. As part of this collaboration, a joint PhD proposal for the EPSRC Centre for Doctoral Training in Cloud Computing for Big Data at Newcastle University will be conceptualized and drafted. The visit, which will be scheduled between June and August 2015, will contribute to initiate an international collaboration with a world leading institution in climate modelling with the PI and Newcastle University that will lead to high profile publications in top Statistics and Geosciences research journals.

Planned Impact

The development of fast and reliable compression algorithms for climate model ensembles will allow climate model users to test initial scientific hypotheses with full climate model runs without downloading very large quantities of data. Further, this methodology will allow a better quantification of the uncertainty of climate model projections, which is one of the strategic goals of the United Nation's Intergovernmental Panel on Climate Change, by generating more surrogate climate model runs.
The beneficiaries are climate model users, spanning from climate scientists to economists interested in quantifying the uncertainty of climate model projections.

Publications

10 25 50
 
Description The travel grant awarded has been successfully implemented with the PI's visit to the National Center for Atmospheric Research (NCAR) in Boulder, Colorado. The most significant achievement of this short-term visit is the proposal of a joint Newcastle University-NCAR PhD proposal for the EPSRC CDT on Cloud Computing and Big Data on the topic of data compression. The project has already been chosen by a student, who is currently starting his doctoral studies on the topic. Besides the primary goal, the PI has also performed a seminar in the research center, and outlined a project on customized compression. Further, the collaborative visit has generated a manuscript led by the PI, which is currently under review.
The main objective of this Overseas Travel Grant has hence been successfully achieved, and NCAR has sponsored the student's visit over the summer 2016 as part of the joint supervision. In 2017, the same student has been awarded a fellowship at NCAR which will allow him to visit again the institution over this spring.
Exploitation Route This visit has highlighted the strong interest from NCAR to further develop this theme and to provide a new set of tools to climate model practitioners to achieve compression of climate model output at very high rates, in particular by tailoring the compression scheme to a particular scientific investigation. To this end, NCAR has made available a collection of climate model runs, we have identified a subclass of variables of interest, and outlined a class of models that will be suitable for the initial development of the compression scheme.
With the preliminary results achieved from visit, as well as with the accumulated track record of the PI on the development of the methodology, an EPSRC grant proposal is foreseen over the next year on this topic.
Sectors Environment
 
Description Since this travel grant was aimed at developing an initial exploratory collaboration with NCAR on the topic of data compression, it is expected that impact will be achieved on a long-term basis by 1) the training of a doctoral student 2) the development of a suite of compression methods in the form of software packages for widely used programming languages (R and Matlab). The PhD student who will develop this project will focus on this theme for the next two years, and will generate a long term benefit to the UK economy by bringing expertise on the theme of data compression for climate data, which is currently exclusively developed in the United States. The development of more sophisticated compression schemes and the delivery of these in the form of software packages will require an extensive methodological work which the student will only partly be able to deliver. Hence, the development of a standard EPSRC proposal emerges as a necessity to fully capitalize on this investment.
First Year Of Impact 2016
Sector Environment
Impact Types Cultural,Societal
 
Description established collaboration with the National Center for Atmospheric Reserach 
Organisation National Center for Atmospheric Research
Country United States 
Sector Public 
PI Contribution The goal of the overseas travel grant was to initiate a collaboration on the topic of data compression. The PI has provided his expertise and led an investigation on the topic while drafting the PhD proposal that was eventually chosen by a student.
Collaborator Contribution NCAR provided the PI with access to all facilities, from office to supercomputers to test some of the models as part of his research.
Impact A PhD student of the PI is currently being co-supervised with a scientist at NCAR. A manuscript led by the PI is currently under review.
Start Year 2016