Model performance in spatiotemporal patterns of precipitation: New methods for identifying value added by a regional climate model
This paper develops two approaches to evaluate how well downscaled reginal climate models capture spatiotemporal relationships by comparing RCMs directly against observational data via the spatial variogram and the spatialtemporal correlation. These measures are applied to a high-resolution RCM and to the National Center for Environmental Prediction (NCEP)-U.S. Department of Energy Atmospheric Model Intercomparison Project II reanalysis data (NCEP-R2).
Authors: Jiali Wang, F.N.U. Swati, Michael L. Stein, and V. Rao Kotamarthi
Full text: link
Published in: Journal of Geophysical Research: Atmospheres, Vol 120, Issue 4, pg 1239-1259
Consequences of kriging and land use regression for PM2.5 predictions in epidemiologic analyses: insights into spatial variability using high-resolution satellite data
The paper describes a novel study of air pollution exposures that goes beyond simple spatial models by assuming a realistic, complex exposure surface derived from fine-scale remote-sensing satellite data.
The accuracy of epidemiological health effect estimates in linear and logistic regression using spatial air pollution predictions from kriging and land use regression models is evaluated by way of simulations. Substantial bias from often used spatial models was uncovered in the study.
Authors: S.E. Alexeeff, J. Schwartz, I. Kloog, A. Chudnovsky, P. Koutrakis, B.A. Coull
Full text: link
Published in: Journal of Exposure Science and Environmental Epidemiology 25, 138-144
Statistical Prediction of Global Sea Level from Global Temperature
This analysis of statistical climate projections of global mean sea level from global mean temperature provides a statistically defensible uncertainty analysis. It shows that methods, which in the climate science literature have been considered substantially different, have overlapping confidence sets when using this uncertainty analysis. The confidence sets also overlap the latest IPCC projections. The analysis uses a novel approach to calculating simultaneous confidence intervals for mixtures of normal distributions. The statistical model is sensitive to choices of temperature and sea level data.
Node: University of Washington
Authors: David Bolin, Peter Guttorp, Alex Januzzi, Daniel Jones, Marie Novak, Harry Podschwit, Lee Richardson, Aila Sarkka, Colin Sowder and Aaron Zimmerman
Preprint full text available: here
To appear in: Statistica Sinica
Simulation of Future Climate Under Changing Temporal Covariance Structures
A growing body of evidence indicates that anthropogenic greenhouse gases are changing Earth’s climate, and those changes may involve not only changes in climatic means but also in variability. Climate models may be informative about these future changes, but use is complicated by the fact that they do not capture variability in current climate well. Many methods have therefore been developed to combine models and data in simulations of future climate, but current methods generally account only for changes in marginal variation and do not capture projected changes in correlation (spatial, temporal, spatio-temporal). We develop here a procedure to simulate future daily mean temperature that modifies climate observations based on changes in the mean and spectral density suggested by climate model output, and illustrate our methodology with projections from the CCSM3 climate model. We are able to simulate a future climate with changing temporal covariance, while largely retaining non-Gaussian features of observations. Our results suggest that in CCSM3, at most locations and most time scales, variability in daily mean temperature decreases under anthropogenic warming. The methodology presented here applies only to fully equilibrated future climate states, but may be extended to simulating transient states as well.
Node: University of Chicago
Authors: W.B. Leeds, E.J. Moyer, M.L. Stein
Preprint full-text available: here.
Supplemental material available here.
Preprint Submitted to the Journal of the American Statistical Association
Quantifying the human and natural contributions to observed climate change
There are multiple lines of observational evidence that indicate unequivocally that the global climate has warmed over the past century. External factors that affect the climate, such as the atmospheric concentration of greenhouse gases, have also changed over this period. A key problem that is of considerable policy relevance is to quantify how much of the observed climate change can be attributed to external factors. This paper describes the dominant approach to climate change detection and attribution that climate science has used for this purpose. It shows that while detection and attribution analyses involve complex physical reasoning and a series of complex data processing decisions, relatively simple statistical models are used to make inferences about the presence and magnitude of externally forced effects on our climate.
Node: Pacific Institute for Mathematical Science
Authors: Zwiers, F.W., G.C. Hegerl, X. Zhang, Q. Wen
Full text: link
Published in: Statistics in Action: A Canadian Perspective, J. Lawless, ed. CRC-Chapman, p321-340.
Flexible regression models over river networks
A spatio-temporal model was developed to model diffuse pollutants in a river network. This work was motivated by the Water Framework directive of the EU, which requires management of water bodies at river basin level. Diffuse pollutants including nitrates are monitored routinely across a network and respond seasonally to agricultural and meteorological conditions. We developed a smooth spatio-temporal model in a regression framework using penalised splines. Several interaction terms are included capturing different temporal trends in space as well as seasonal interactions. The construction of a basis set over a network faces the difficulty of combining the basis components in a suitable manner at the confluence points. This is possible, but slightly awkward, with the usual pattern of smooth overlapping functions, so instead the network was divided into a large number of small pieces. In the river setting, this arises naturally through the identification by the Scottish Environment Protection Agency of ‘stream units’ corresponding to short water stretches which are judged to be relatively constant in terms of environmental conditions.
Node: University of Glasgow
Authors: David O’Donnell, Alastair Rushworth, Adrian W. Bowman and E. Marian Scott
Full text: link
Published in: Journal of the Royal Statistical Society: Series C (Applied Statistics) Volume 63, Issue 1, pages 47–63
Spatial extreme value analysis to project extremes of large-scale indicators for severe weather
Realistic meteorological application of the Heffernan and Tawn approach to multivariate extreme value analysis of large-scale severe storm environments. The method allows for analyzing spatial patterns of the storm environments conditioned on when extreme energy in the field exists. It also provides a method for investigating how severe weather patterns may (or may not) be changing with time, as well as a way to analyze climate model output to analyze future patterns.
Authors: Eric Gilleland, Barbara G. Brown and Caspar M. Ammann
Year: 2013Full-text: link
Published in: Environmetrics, Volume 24, Issue 6, pages 418–432
Fragmentation and thermal risks from climate change interact to affect persistence of native trout in the Colorado River basin
Impending changes in climate will interact with other stressors to threaten aquatic ecosystems and their biota. Native Colorado River cutthroat trout (CRCT; Oncorhynchus clarkii pleuriticus) are now relegated to 309 isolated high-elevation (>1700 m) headwater stream fragments in the Upper Colorado River Basin, owing to past nonnative trout invasions and habitat loss. Predicted changes in climate (i.e., temperature and precipitation) and resulting changes in stochastic physical disturbances (i.e., wildfire, debris flow, and channel drying and freezing) could further threaten the remaining CRCT populations. We developed an empirical model to predict stream temperatures at the fragment scale from downscaled climate projections along with geomorphic and landscape variables. We coupled these spatially explicit predictions of stream temperature with a Bayesian Network (BN) model that integrates stochastic risks from fragmentation to project persistence of CRCT populations across the upper Colorado River basin to 2040 and 2080. Overall, none of the populations are at risk from acute mortality resulting from high temperatures during the warmest summer period. In contrast, only 37% of populations have a ≥90% chance of persistence for 70 years (similar to the typical benchmark for conservation), primarily owing to fragmentation. Populations in short stream fragments <7 km long, and those at the lowest elevations, are at the highest risk of extirpation. Therefore, interactions of stochastic disturbances with fragmentation are projected to be greater threats than warming for CRCT populations. The reason for this paradox is that past nonnative trout invasions and habitat loss have restricted most CRCT populations to high-elevation stream fragments that are buffered from the potential consequences of warming, but at risk of extirpation from stochastic events. The greatest conservation need is for management to increase fragment lengths to forestall these risks.