Get your brand new Wikispaces Classroom now
and do "back to school" in style.
Pages and Files
BIRS Workshop Papers
Add "All Pages"
BIRS Workshop Papers
BIRS Workshop Papers and Notes
File size for uploads is restricted to less than 10 MB.
Paul Kushner: Opening Remarks
Opening Remarks and Santer video
Ben Santer: Identifying a "discernible human influence" on global climate: A
personal perspective on the application of the Hasselmann fingerprint method
Presentation (in 3 parts because of 10 MB limit):
Opening Remarks and Santer video
David Lobell: Impacts of climate change on agriculture and opportunities for detection and attribution
This talk will try to do three things. First, I will present some examples of changes in agricultural systems and explain why attributing changes to climate is difficult (there are many contemporaneous changes, including possible adaptations), and why detection in the traditional sense is impossible (there is no clear counterfactual analogous to the “control” simulation in climate models). Second, I will give a brief summary of recent attempts to quantify impacts of climate change on agriculture, based on efforts related to the next IPCC report, with an emphasis on work at scales where DNA in climate is most plausible. Third, I will outline some options for moving forward, ranging from simple to complex. These include (i) using process understanding of crops and global food markets to identify climate variables for more traditional climate DNA to focus on; (ii) running crop models for specific crops/regions with climate model output from “control” and “ALL” simulations, to quantify impacts of emissions of various gases and aerosols; (iii) go for the full monty: running historical simulations regionally or globally with and without anthropogenic emissions and defining fingerprints to compare with observations. In my opinion, the lowest hanging fruit is probably to focus on the first, with an emphasis on large spatial scales and the chances for clusters of extremes events in major production regions.
Notes from Monday morning session (A. Braverman):
Nathan Gillett: An introduction to climate change detection and attribution.
This presentation will build on the first introductory lecture on climate change detection to consider studies which explicitly separate the responses to multiple climate forcings, for example anthropogenic forcings and natural forcings. Such studies typically relate observed changes in a variable of interest to simulated responses to a set of forcings using a multiple linear regression equation. Rather than assuming that the response to each set of forcings consists of a linear trend, such studies often use spatio-temporal patterns of response in order to better differentiate between the responses to different forcings. This talk will introduce regression-based approaches to detection and attribution, consider signal-to-noise opimisation, approaches which account for uncertainty in the simulated response patterns, multi-model approaches, the use of results to constrain projections, and other innovations in methodology. Applications to a range of variables will be discussed, including recent attribution results for surface temperature change and sea level pressure change derived using CMIP5 simulations. The limitations of such approaches will also be discussed.
Chris Paciorek (for Francesca Domenici): Human health impacts of climate change
Francis Zwiers: Progress in Detecting Anthropogenic Influence on Temperature and Precipitation Extremes
There is a well established approach to detecting and attributing the causes of observed changes in mean climatic conditions that has been applied progressively from global scales to regional scales to temperature, precipitation and other climate variables. While this research has provided a great deal of useful information about the causes of climate change observed during the past century or more, policy makers and others have also been demanding answers about whether there are attributable changes in the frequency and/or intensity of extreme weather and climate events. The statistical techniques required to respond to these questions are only now begin developed. This talk will describe several approaches that have been proposed to assess whether there is a detectable human influence in the far tails of the distribution of climate variables such as daily maximum air temperature or daily precipitation amount. We also describe initial applications of these approaches, and discuss limitations and further areas of improvement. These applications suggest that human influence on the climate system has affected the extremes of daily maximum and minimum temperatures, and extreme daily precipitation amounts, altering the waiting times for events of a fixed amplitude.
Richard Smith: DA of Extremes/Climate Events
Panel discussion: theory and methodology (Guttorp et al.):
Randall Dole: Event Attribution: Physical-Diagnostic Approach
Presentation: Part 1
2012-06-01 BIRS_Dole_attribution part1.pdf
. Part 2:
2012-06-01 BIRS_Dole_attribution part 2.pdf
Notes from breakout discussion on extremes and impacts (T. Greasby):
Peter Thorne: Making sense of the uncertainty in in-situ climate data
Presentation: Part 1:
Carl Mears: Realistic and Easy-to-Use Uncertainty Analyses for Climate Observations. A Case Study.
All too often, researchers assign a simple error bars to each datum they produce, and then conclude that the error analysis is complete. For Earth observations intended for use in climate change analysis, this simple approach falls far short of what is needed, since it ignores the strong spatial/temporal correlations that are present in (and often dominate) the uncertainty in global-scale Earth observations. There are a plethora of causes of such correlated errors -- changes in instrumentation over time, changes in instrumentation between organizations, drifts in satellite calibration or drifts in local observation time -- to name a few. In this presentation I will show how we calculated uncertainty estimates for an example of a satellite-derived climate data record, the atmospheric temperature data record from the Microwave Sounding Unit (MSU) and the Advanced Microwave Sounding Unit (AMSU), which is assembled by combining measurements from 15 different satellites. Our analysis is based on realistic set of estimated input errors. These input errors interact with the intersatellite merging procedure in a complicated way, so that their combined influence is most easily found using a Monte-Carlo approach. Even after such an analysis is complete, it remains challenging to convince researchers to use the uncertainty estimates in their own work. I will discuss several options for presenting the information in an easy-to-use way.
Martin Tingley: Arctic temperature extremes over the last 600 years
Peter Craigmile: Statistical modeling of extreme value behavior in paleoclimate proxies
Panel discussion of paleoclimate (Bala Rajaratnam et al.):
Notes from panel (Ying Sun):
Notes from Climate Impacts discussion (B. Shaby):
Ben Armstrong: Climate change, health, and detection/attribution: a broader context
William Anderegg: Detection and attribution of climate-related tree mortality rates in the western US and Canada
Amy Braverman: Likelihood-based comparison of CMIP5 decadal experiment runs and AIRS specific humidity observations
Like other scientific and engineering problems that involve physical modeling of complex systems, climate models can be evaluated and diagnosed by comparing their output to observations of the fields they attempt to simulate or predict. Remote sensing data such as those collected by NASA's AIRS instrument are particularly well-suited for this purpose because they are spatially and temporally dense enough to provide distributional information that goes beyond simple moments. Comparisons against multiple members of an ensemble of climate model runs can provide useful information about which models, and therefore which physical parameterizations and assumptions most closely match what is seen in the real world. In this talk, we propose a method for assessing how well monthly climate model output from CMIP5
decadal experiment ensembles agree with data from the AIRS monthly IPCC data set prepared for the "obs4mips" project. We estimate the likelihoods that summary statistics computed from AIRS time series arise from sampling distributions of the same statistics calculated from ensemble members' time series. The likelihood is the probability of the observed value of a selected summary statistic, given that the atmosphere behaves statistically like the climate model time series under consideration. Assuming a uniform prior distribution on the ensemble members, these empirically-derived likelihoods can be used to calculate posterior probabilities of each ensemble member given the value of the summary statistic from the AIRS time series. We report here on the results of comparisons of AIRS specific humidity to that obtained from decadal experiment runs of the CanCM4 and CNRM-CM5 models. The comparisons are performed for three vertical levels (850, 700, and 500 hPa), in a representative set of grid cells in the tropical western pacific, and for a variety of summary statistics including quartiles, extremes, and lag-autocorrelations. We find that for some statistics, certain ensemble members out-perform others convincingly.
Jenny Brynjarsdottir: Downscaling temperatures over Antarctica - Dimension reduced spatio-temporal modeling with Maximum Covariance Patterns
Dimension reduced approaches to spatio-temporal modeling are often based on modeling the spatial structure in terms of a low number of specified basis functions. The temporal evolution of the space-time process is then modeled through the amplitudes of the basis functions. A common choice of basis are data-dependent basis vectors such as Empirical Orthogonal Functions (EOFs), also known as Principal Components. I will show ways to extend these ideas to modeling of two spatio-temporal processes where the primary goal is to predict one process from the other. I incorporate these methods in a Bayesian hierarchical model and show an example of downscaling temperatures over the Antarctic.
Tim Delsole: Robust Multi-year Predictability on Continental Scales
Tim Delsole: Optimizing Detectability on Continental Scales
Alexis Hannart: Approaching D&A as an inverse problem: data Assimilation in a toy model of stratospheric cooling
D&A can be seen as an inverse problem. An interest of taking this perspective is that it gives access to many methods that are not currently used in D&A. With that said, is there anything in the inverse problem toolbox that is valuable to tackle the present challenges of D&A ? We start exploring this question by implementing a Kalman filter inversion of an idealized radiative atmospheric model. We show that such a data assimilation procedure is able to reconstruct the contributions of the different radiative forcings from climate observations. We discuss further possible extensions of this data assimilation approach of D&A ('DADA').
Jara Imbers: Sensitivity of detection and attribution to simulated internal variability
The IPCC very likely statement that anthropogenic emissions are affecting climate is based on a statistical detection and attribution methodology, heavily dependent on the characterisation of internal climate variability. In this paper, we ask how robust is the IPCC statement in the event of varying the representation of such variability? We do this conceptually with a simple energy balance model of the climate system forced with dierent natural and anthropogenic forcings, a stochastic short memory process (AR1), and a stochastic long memory process, (FD). We nd that independently of the representation the greenhouse gas signal in particular remains statistically signicant. We present this paper as additional evidence that the IPCC statement is robust but also, we claim the need to update the methodologies to assess GCMs simulations compared to the observed record.
Paul Kushner (work by Vyushin, Kushner, and Zwiers): Statistical representations of climate persistence
In this work, two parsimonious statistical representations of climate variability on interannual to multidecadal time scales are compared: the short-memory first order autoregressive representation (AR1) and the long-memory ``power-law" representation. Parameters for each statistical representation are fitted to observed surface air temperature at each spatial point. The parameter estimates from observations are found in general to be captured credibly in the Coupled Model Intercomparison Project 3 (CMIP3) simulations. The power-law representation provides an upper bound and the AR1 representation provides a lower bound on persistence as measured by the lag-one autocorrelation. Both representations fit the data equally well according to goodness-of-fit-tests. Comparing simulations with and without external radiative forcings shows that anthropogenic forcing has little effect on the measures of persistence considered (for detrended data). Given that local interannual to multidecadal climate variability appears to be more persistent than an AR1 process and less persistent than a power-law process, it is concluded that both representations are potentially useful for statistical applications. It is also concluded that current climate simulations can well represent interannual to multidecadal internal climate persistence in the absence of natural and anthropogenic radiative forcing, at least to within observational uncertainty.
Pardeep Pall: Anthropogenic greenhouse gas contribution to flood risk in England and Wales in Autumn 2000
Aurélien Ribes: D&A on global mean temperature based on CMIP5 models
We wonder to what extent the responses to greenhouse gases (GHG), anthropogenic aerosols (SULF) and natural (NAT) forcings may be distinguished based only on the global mean temperature. We apply a classical D&A analysis to the global mean temperature time-series over the 1901-2010 period. The CMIP5 archive provides the opportunity to use the outputs from 9 models into this analysis. We find that the GHG and SULF responses are often hard to distinguish because of being collinear. Then, the models simulating a less collinear response are also showing an accelerated warming over the last decade, which is maybe not realistic
Aurélien Ribes: Regularised Optimal Fingerprint for attribution
The Regularised Optimal Fingerprint had been introduced for detection in order to avoid EOFs projection before applying optimal fingerprints algorithm. It is here extend to attribution, and adapted to the OLS and TLS algorithms. Monte Carlo simulations and perfect framework analysis supports the conclusion that it allows a more objective and a little more efficient implementation of optimal fingerprinting.
Ben Shaby: Bayesian spatial extreme value analysis to assess the changing risk of widespread crop failure in Europe
There is strong evidence that extremely high temperatures are detrimental to the yield and quality of many economically and socially critical crops. Fortunately, the most deleterious conditions for agriculture occur rarely. We wish to assess the risk of the catastrophic scenario in which large areas of croplands simultaneously experience extreme heat stress. Applying a hierarchical Bayesian spatial extreme value model that allows the distribution of extreme temperatures to change in time both marginally and in spatial coherence, we examine whether the risk of widespread crop failure due to high temperatures has increased in Europe over the last century.
Ying Sun: Function Median Polish with Climate Applications
We propose functional median polish, an extension of univariate median polish, for one-way and two-way functional analysis of variance (ANOVA). The functional median polish estimates the functional grand effect and functional main factor effects based on functional medians in an additive functional ANOVA model assuming no interaction among factors. A functional rank test is used to assess whether the functional main factor effects are significant. The robustness of the functional median polish is demonstrated by comparing its performance with the traditional functional ANOVA fitted by means under different outlier models in simulation studies. The functional median polish is illustrated on various applications in climate science, including one-way and two-way ANOVA when functional data are either curves or images. Specifically, U.S. precipitation observations and outputs of global and regional climate models are considered.
Peter Thorne: The International Surface Temperature Initiative
help on how to format text
Turn off "Getting Started"