IMAGe Theme Of the Year

IMAGe’s Theme Of the Year (TOY) is a series of activities that explores the opportunity to enrich both applied mathematics and the geosciences through a common scientific topic. TOY is designed to advance research and education between the mathematical and geosciences communities; it uses targeted projects for building interdisciplinary communities. The topics are selected by the IMAGe external advisory panel and coordinated by one or more visiting co-directors. The yearly TOY programs support CISL’s education imperative to integrate research and education, sparking collaborations between the mathematics community and Earth System scientists.

Effects of changes in observation methods
This figure is a slide from the talk by Lucie Vincent, (Environment Canada) given at the TOY workshop on Surface Temperature Analysis. It shows how surface temperature measurements are affected by even modest changes in the instrument location or the shelter (screen). The station is located in the southern interior of British Columbia, Canada and has been maintained for more than 100 years. Here we see clear level shifts in the annual mean temperatures based on documented events by the station observer. The challenge for data science is to infer level shifts and other biases when the information about changes at a station are incomplete or nonexistent, then attempt to adjust for biases so that regional trends can be estimated accurately. It is also a caution to the statistical community that the analysis of surface temperature measurements should be done in collaboration with climate experts who are familiar with the many kinds of biases that can enter into the observations.

The 2014 TOY was concerned with new ways of teaching data analysis and disseminating cutting-edge statistical techniques to the geoscience community. It was proposed partly as a response to the growing interest in data science as a discipline and the distinct challenges for working with "big data" problems.

Statistics is the science of interpreting data through mathematical models with an emphasis on quantifying the uncertainty in any analysis. Typically, part of a statistical analysis will also involve the use of graphics to communicate complex relationships and patterns. Traditional courses in statistics focus on developing the mathematical basis of statistical concepts, for example sampling from a population, using probability models for testing hypotheses and setting error bounds for parameter estimates. However, this view may miss the rich set of tools that can be applied to different kinds of data. Moreover, the rapid increase in computing power and the advent of the R statistical language has made statistical methods accessible to a broad scientific and engineering community. With these advances, the ability to interpret large and complex data sets has improved dramatically.

TOY’s FY2014 education activities involved teaching courses on data analysis with a focus on the kinds of methods used for weather and climate research. Besides the direct benefit of these courses for the participants, there is an added product in lecture materials, software, and data sets that will be used in future activities. The data analysis course taught by IMAGe staff during summer term A at the University of Colorado in Boulder was intended to reach both undergraduates in the mathematical and physical sciences and also graduate students in engineering and science. The course was successful in engaging students on problems that involved large data sets. For example, a subset of the NARCCAP regional climate model output was used for projects. The second event was a short course, Ecological Informatics, that focused on spatial statistical methods for ecology graduate students. This was team-taught by several outside statistics faculty members and emphasized the use of R for the analysis of large spatial data.

Developing curricula and teaching was complemented by an intensive data analysis-oriented meeting: Workshop on Surface Temperature Analysis, which was cosponsored by NCAR, the Statistical and Applied Mathematical Sciences Institute (SAMSI), and the International Surface Temperature Initiative (ISTI). The meeting brought together about 40 climate scientists and statisticians to address issues in the analysis of land surface air temperature observational records in support of the International Surface Temperature Initiative.

Participants spent the majority of time working in breakout sessions and working groups that were charged with undertaking distinct analyses of the recently released databank holdings of 32,000+ station records. This meeting was an experiment to confront climate data specialists with new statistical methods for determining station biases and inferring regional trends. It was also an opportunity for the statistical scientists to understand the unique needs of interpreting surface temperature station records. One success of these meetings was a new approach to identifying regional trends using the R package LatticeKrig, a spatial analysis method designed for large data sets. Another synergy was a statistical formulation for detecting inhomogeneities in station records by looking for change points but also adjusting for temporal correlation in the temperature series.

Other significant events hosted by IMAGe’s TOY include:

  • The Fourth International Workshop on Climate Informatics

  • Uncertainty in Climate Change Research: An Integrated Approach

  • Pattern Scaling, Climate Model Emulators, and their Application to the New Scenario Process

  • PDEs, The Workshop on Partial Differential Equations on the Sphere

Outreach activities of the Theme of the Year are supported by NSF Core funding.