CISL 2007 annual report banner

Geophysical Statistics Project

Accounting for past temperature uncertainty

This figure shows the uncertainty of decadal average (blue line with gray band) and the uncertainty of decadal maxima (purple dots and box plots), accompanied by instrumented decadal temperature (red). It is widely acknowledged that past temperature reconstructions may contain substantial uncertainty that is difficult to quantify in its full complexity, and this blurs our understanding of the temperature evolution and the significance of the recent warming. This figure results from a statistical method to reconstruct past temperatures together with their confidence ranges by keeping track of different sources of uncertainties. This method offers, for the first time, an explicit answer to questions about decadal average temperatures because we now have the ability to account for the serial correlation of uncertainty.

Analyzing model output variances

This figure shows the decomposition of the output of a particular run of a simplified regional climate model (RCM) experiment. The 30-year average summer temperature for a control run is organized into effects that can be attributed to the particular global climate model (GCM) that was used as forcing for the RCM, the particular RCM, as well as the interaction associated with the particular GCM/RCM pair. This analysis of variance (ANOVA) approach incorporates a linear combination of Gaussian processes and represents a novel statistical technique for analyzing such experiments. Further, it helps climate model researchers understand model biases and sources of variation in the model output, and it leads to improved projections of climate change that are based on multi-model ensembles.


For the past decade, the Geophysical Statistics Project (GSP) has been a leader in training and research emphasizing the synergy between the geosciences and the statistical sciences. Aside from basic methodological and theoretical statistical research, GSP has had a strong training component supporting from four to six postdoctoral visiting scientists. The postdocs are immersed in research activities that not only focus their skills as applied statisticians but also expose them to important applications in the geosciences.

In addition to these core activities, GSP also has an active visitor program providing research opportunities for visiting graduate students and junior and senior faculty members from across the nation and abroad. As with the postdoctoral training program, the goal of these programs is to foster collaboration between graduate students, postdocs, the permanent and visiting statistical staff, and the NCAR scientists. These programs, as well as the research and training aspects of GSP that emphasize the interaction between statistics and the geosciences, embody the tenets of integration, innovation, and community building within the NCAR strategic plan. Specifically, this program supports the NCAR strategic priorities of "Conducting computer science, computational science, applied mathematics, statistics, and numerical methods R&D," "Supporting and enhancing formal science education at all levels," and "Engaging a broader and more diverse community in the atmospheric and geosciences."

During FY2007, GSP researchers have been involved in numerous important projects, including:

  • Design and analysis of computer experiments, in particular focusing on regional climate models and models of the upper atmosphere and the magnetosphere
  • Developing methodology for analyzing extremes of weather and climate
  • Stochastic weather generators
  • Carbon transport
  • Modeling uncertainty in climate reconstruction
  • Impacts of climate and climate change on public health


GSP continues to develop methodology for analyzing spatial data, including nonstationary covariance models, models for spatial lattice data, multivariate spatial observations, spatial-temporal models, as well as general methodology for computational statistics and Bayesian hierarchical models.

In FY2008, the scientific focus on computer models will continue, in particular through GSP scientists being involved in such NCAR programs as the North American Regional Climate Model Assessment Program (NARCCAP) as well as in collaborations with other computer modeling groups across NCAR. Beyond computer models, GSP scientists will continue to assess the impacts of climate and climate change on public health, to develop methodology for analyzing extremes, to develop methodology for quantifying the uncertainty in climate reconstructions, and to develop statistical methodology for the analysis of complex, spatial and spatial-temporal data.

This project is made possible through NSF Core funding, as well as grants through NSF's Division of Mathematical Sciences and NSF's Collaboration in Mathematical Geosciences.