University of Cape Town press release, LBNL Computing Sciences press release, NERSC science feature

Seasonal extreme daily precipitation is analyzed in the ensemble of NARCAPP regional climate models. Significant variation in these models’ abilities to reproduce observed precipitation extremes over the contiguous United States is found. Model performance metrics are introduced to characterize overall biases, seasonality, spatial extent and the shape of the precipitation distribution. Comparison of the models to gridded observations that include an elevation correction is found to be better than to gridded observations without this correction. A complicated model weighting scheme based on model performance in simulating observations is found to cause significant improvements in ensemble mean skill only if some of the models are poorly performing outliers. The effect of lateral boundary conditions are explored by comparing the integrations driven by reanalysis to those driven by global climate models. Projected mid-century future changes in seasonal precipitation means and extremes are presented and discussions of the sources of uncertainty and the mechanisms causing these changes are presented.

Extreme precipitation is generally underestimated by current climate models relative to observations of present-day rainfall distributions. Possible causes of this systematic error include the convective parameterization in these models that have been designed to reproduce measurements of climatological mean precipitation. One possible approach to improve the interaction of subgrid-scale physical processes and large-scale climate is to replace the conventional convective parameterizations with a high-resolution cloud-system resolving model. A “super-parameterized” Community Atmosphere Model (SP-CAM) utilizing this approach is used in this study to investigate the distribution of extreme precipitation in the United States. Results show that SP-CAM better simulates the distributions of both light and intense precipitation compared to the standard version of CAM based upon conventional parameterizations. The improvements are mostly seen in regions dominated by convective precipitation, suggesting that super-parameterization provides a better representation of subgrid convective processes.

The state of knowledge regarding trends and an understanding of their causes is presented for a specific subset of extreme weather and climate types. For severe convective storms (tornadoes, hail storms, and severe thunderstorms), differences in time and space of practices of collecting reports of events make the use of the reporting database to detect trends extremely difficult. Overall, changes in the frequency of environments favorable for severe thunderstorms have not been statistically significant. For extreme precipitation, there is strong evidence for a nationally-averaged upward trend in the frequency and intensity of events. The causes of the observed trends have not been determined with certainty, although there is evidence that increasing atmospheric water vapor may be one factor. For hurricanes and typhoons, robust detection of trends in Atlantic and western North Pacific tropical cyclone (TC) activity is significantly constrained by data heterogeneity and deficient quantification of internal variability. Attribution of past TC changes is further challenged by a lack of consensus on the physical linkages between climate forcing and TC activity. As a result, attribution of trends to anthropogenic forcing remains controversial. For severe snowstorms and ice storms, the number of severe regional snowstorms that occurred since 1960 was more than twice that of the preceding 60 years. There are no significant multi-decadal trends in the areal percentage of the contiguous U.S. impacted by extreme seasonal snowfall amounts since 1900. There is no distinguishable trend in the frequency of ice storms for the U.S. as a whole since 1950.

One key question regarding current climate models is whether the projection of climate extremes converges to a realistic representation as the spatial and temporal resolutions of the model are increased. Ideally the model extreme statistics should approach a fixed distribution once the resolutions are commensurate with the characteristic length and time scales of the processes governing the formation of the extreme phenomena of interest. In this study, a series of AGCM runs with idealized ‘aquaplanet-steady-state’ boundary conditions have been performed with the Community Atmosphere Model CAM3 to investigate the effect of horizontal resolution on climate extreme simulations. The use of the aquaplanet framework highlights the roles of model physics and dynamics and removes any apparent convergence in extreme statistics due to better resolution of surface boundary conditions and other external inputs. Assessed at a same large spatial scale, the results show that the horizontal resolution and time step have strong effects on the simulations of precipitation extremes. The horizontal resolution has a much stronger impact on precipitation extremes than on mean precipitation. Updrafts are strongly correlated with extreme precipitation at tropics at all the resolutions, while positive low-tropospheric temperature anomalies are associated with extreme precipitation at mid-latitudes.

The tropical Atlantic interhemispheric gradient in sea surface temperature significantly influences the rainfall climate of the tropical Atlantic sector, including droughts over West Africa and Northeast Brazil. This gradient exhibits a secular trend from the beginning of the twentieth century until the 1980s, with stronger warming in the south relative to the north. This trend behavior is on top of a multidecadal variation associated with the Atlantic multidecadal oscillation. A similar long-term forced trend is found in a multimodel ensemble of forced twentieth-century climate simulations. Through examining the distribution of the trend slopes in the multimodel twentieth-century and preindustrial models, the authors conclude that the observed trend in the gradient is unlikely to arise purely from natural variations; this study suggests that at least half the observed trend is a forced response to twentieth-century climate forcings. Further analysis using twentieth-century single-forcing runs indicates that sulfate aerosol forcing is the predominant cause of the multimodel trend. The authors conclude that anthropogenic sulfate aerosol emissions, originating predominantly from the Northern Hemisphere, may have significantly altered the tropical Atlantic rainfall climate over the twentieth century.

Using the Palmer drought severity index, the ability of 19 state-of-the-art climate models to reproduce observed statistics of drought over North America is examined. It is found that correction of substantial biases in the models’ surface air temperature and precipitation fields is necessary. However, even after a bias correction, there are significant differences in the models’ ability to reproduce observations. Using metrics based on the ability to reproduce observed temporal and spatial patterns of drought, the relationship between model performance in simulating present-day drought characteristics and their differences in projections of future drought changes is investigated. It is found that all models project increases in future drought frequency and severity. However, using the metrics presented here to increase confidence in the multimodel projection is complicated by a correlation between models’ drought metric skill and climate sensitivity. The effect of this sampling error can be removed by changing how the projection is presented, from a projection based on a specific time interval to a projection based on a specified temperature change. This modified class of projections has reduced intermodel uncertainty and could be suitable for a wide range of climate change impacts projections.

Current climate models produce quite heterogeneous projections for the responses of precipitation extremes to future climate change. To help understand the range of projections from multimodel ensembles, a series of idealized ‘aquaplanet’ Atmospheric General Circulation Model (AGCM) runs have been performed with the Community Atmosphere Model CAM3. These runs have been analysed to identify the effects of horizontal resolution on precipitation extreme projections under two simple global warming scenarios. We adopt the aquaplanet framework for our simulations to remove any sensitivity to the spatial resolution of external inputs and to focus on the roles of model physics and dynamics. Results show that a uniform increase of sea surface temperature (SST) and an increase of low-to-high latitude SST gradient both lead to increase of precipitation and precipitation extremes for most latitudes. The perturbed SSTs generally have stronger impacts on precipitation extremes than on mean precipitation. Horizontal model resolution strongly affects the global warming signals in the extreme precipitation in tropical and subtropical regions but not in high latitude regions. This study illustrates that the effects of horizontal resolution have to be taken into account to develop more robust projections of precipitation extremes.

We investigate three sources of uncertainty in the calculation of extreme
value statistics for observed and modeled climate data. Inter-model differences in
formulation, unforced internal variability and choice of statistical model all contribute
to uncertainty. Using fits to the GEV distribution to obtain 20 year return
values, we quantify these uncertainties for the annual maximum daily mean surface
air temperatures of pre-industrial control runs from 15 climate models in the CMIP3
dataset.

We investigate the ability of a global atmospheric general circulation model (AGCM) to reproduce observed 20 year return values of the annual maximum daily precipitation totals over the continental United States as a function of horizontal resolution. We find that at the high resolutions enabled by contemporary supercomputers, the AGCM can produce values of comparable magnitude to high quality observations. However, at the resolutions typical of the coupled general circulation models used in the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, the precipitation return values are severely underestimated.

High-performance computing for such things as climate modeling is not going to advance at anything like the pace it has during the last two decades unless we apply fundamentally new ideas. Here we describe one possible approach. Rather than constructing supercomputers from the kinds of microprocessors found in fast desktop computers or servers, we propose adopting designs and design principles drawn, oddly enough, from the portable-electronics marketplace.

 Numerous websites, blogs and articles in the media have claimed that the climate is no longer warming, and is now cooling. Here we show that periods of no trend or even cooling of the globally averaged surface air temperature are found in the last 34 years of the observed record, and in climate model simulations of the 20^th and 21^st century forced with increasing greenhouse gases. We show that the climate over the 21^st century can and likely will produce periods of a decade or two where the globally averaged surface air temperature shows no trend or even slight cooling in the presence of longer-term warming.

In a recent multimodel detection and attribution (D&A) study using the pooled results from 22 different climate models, the simulated “fingerprint” pattern of anthropogenically caused changes in water vapor was identifiable with high statistical confidence in satellite data. Each model received equal weight in the D&A analysis, despite large differences in the skill with which they simulate key aspects of observed climate. Here, we examine whether water vapor D&A results are sensitive to model quality. The “top 10” and “bottom 10” models are selected with three different sets of skill measures and two different ranking approaches. The entire D&A analysis is then repeated with each of these different sets of more or less skillful models. Our performance metrics include the ability to simulate the mean state, the annual cycle, and the variability associated with El Niño. We find that estimates of an anthropogenic water vapor fingerprint are insensitive to current model uncertainties, and are governed by basic physical processes that are well-represented in climate models. Because the fingerprint is both robust to current model uncertainties and dissimilar to the dominant noise patterns, our ability to identify an anthropogenic influence on observed multidecadal changes in water vapor is not affected by “screening” based on model quality.