Twentieth century north atlantic climate change. Part I: assessing determinism

Boreal winter North Atlantic climate change since 1950 is well described by a trend in the leading spatial structure of variability, known as the North Atlantic Oscillation (NAO). Through diagnoses of ensembles of atmospheric general circulation model (AGCM) experiments, we demonstrate that this climate change is a response to the temporal history of sea surface temperatures (SSTs). Specifically, 58 of 67 multi-model ensemble members (87%), forced with observed global SSTs since 1950, simulate a positive trend in a winter index of the NAO, and the spatial pattern of the multi-model ensemble mean trend agrees with that observed. An ensemble of AGCM simulations with only tropical SST forcing further suggests that variations in these SSTs are of primary importance. The probability distribution function (PDF) of 50-year NAO index trends from the forced simulations are, moreover, appreciably different from the PDF of a control simulation with no interannual SST variability, although chaotic atmospheric variations are shown to yield substantial 50-year trends. Our results thus advance the view that the observed linear trend in the winter NAO index is a combination of a strong tropically forced signal and an appreciable noise component of the same phase. The changes in tropical rainfall of greatest relevance include increased rainfall over the equatorial Indian Ocean, a change that has likely occurred in nature and is physically consistent with the observed, significant warming trend of the underlying sea surface.     

Comparison of four ensemble methods combining regional climate simulations over Asia

A number of uncertainties exist in climate simulation because the results of climate models are influenced by factors such as their dynamic framework, physical processes, initial and driving fields, and horizontal and vertical resolution. The uncertainties of the model results may be reduced, and the credibility can be improved by employing multi-model ensembles. In this paper, multi-model ensemble results using 10-year simulations of five regional climate models (RCMs) from December 1988 to November 1998 over Asia are presented and compared. The simulation results are derived from phase II of the Regional Climate Model Inter-comparison Project (RMIP) for Asia. Using the methods of the arithmetic mean, the weighted mean, multivariate linear regression, and singular value decomposition, the ensembles for temperature, precipitation, and sea level pressure are carried out. The results show that the multi-RCM ensembles outperform the single RCMs in many aspects. Among the four ensemble methods used, the multivariate linear regression, based on the minimization of the root mean square errors, significantly improved the ensemble results. With regard to the spatial distribution of the mean climate, the ensemble result for temperature was better than that for precipitation. With an increasing number of models used in the ensembles, the ensemble results were more accurate. Therefore, a multi-model ensemble is an efficient approach to improve the results of regional climate simulations.     

Effect of land model ensemble versus coupled model ensemble on the simulation of precipitation climatology and variability

Through a series of model simulations with an atmospheric general circulation model coupled to three different land surface models, this study investigates the impacts of land model ensembles and coupled model ensemble on precipitation simulation. It is found that coupling an ensemble of land models to an atmospheric model has a very minor impact on the improvement of precipitation climatology and variability, but a simple ensemble average of the precipitation from three individually coupled land-atmosphere models produces better results, especially for precipitation variability. The generally weak impact of land processes on precipitation should be the main reason that the land model ensembles do not improve precipitation simulation. However, if there are big biases in the land surface model or land surface data set, correcting them could improve the simulated climate, especially for well-constrained regional climate simulations.     

On the generation of climate model ensembles

Climate model ensembles are used to estimate uncertainty in future projections, typically by interpreting the ensemble distribution for a particular variable probabilistically. There are, however, different ways to produce climate model ensembles that yield different results, and therefore different probabilities for a future change in a variable. Perhaps equally importantly, there are different approaches to interpreting the ensemble distribution that lead to different conclusions. Here we use a reduced-resolution climate system model to compare three common ways to generate ensembles: initial conditions perturbation, physical parameter perturbation, and structural changes. Despite these three approaches conceptually representing very different categories of uncertainty within a modelling system, when comparing simulations to observations of surface air temperature they can be very difficult to separate. Using the twentieth century CMIP5 ensemble for comparison, we show that initial conditions ensembles, in theory representing internal variability, significantly underestimate observed variance. Structural ensembles, perhaps less surprisingly, exhibit over-dispersion in simulated variance. We argue that future climate model ensembles may need to include parameter or structural perturbation members in addition to perturbed initial conditions members to ensure that they sample uncertainty due to internal variability more completely. We note that where ensembles are over- or under-dispersive, such as for the CMIP5 ensemble, estimates of uncertainty need to be treated with care.     

Evaluating uncertainties in regional climate simulations over South America at the seasonal scale

This work focuses on the evaluation of different sources of uncertainty affecting regional climate simulations over South America at the seasonal scale, using the MM5 model. The simulations cover a 3-month period for the austral spring season. Several four-member ensembles were performed in order to quantify the uncertainty due to: the internal variability; the definition of the regional model domain; the choice of physical parameterizations and the selection of physical parameters within a particular cumulus scheme. The uncertainty was measured by means of the spread among individual members of each ensemble during the integration period. Results show that the internal variability, triggered by differences in the initial conditions, represents the lowest level of uncertainty for every variable analyzed. The geographic distribution of the spread among ensemble members depends on the variable: for precipitation and temperature the largest spread is found over tropical South America while for the mean sea level pressure the largest spread is located over the southeastern Atlantic Ocean, where large synoptic-scale activity occurs. Using nudging techniques to ingest the boundary conditions reduces dramatically the internal variability. The uncertainty due to the domain choice displays a similar spatial pattern compared with the internal variability, except for the mean sea level pressure field, though its magnitude is larger all over the model domain for every variable. The largest spread among ensemble members is found for the ensemble in which different combinations of physical parameterizations are selected. The perturbed physics ensemble produces a level of uncertainty slightly larger than the internal variability. This study suggests that no matter what the source of uncertainty is, the geographical distribution of the spread among members of the ensembles is invariant, particularly for precipitation and temperature.     

Interpreting observed northern hemisphere snow trends with large ensembles of climate simulations

Simulated variability and trends in Northern Hemisphere seasonal snow cover are analyzed in large ensembles of climate integrations of the National Center for Atmospheric Research’s Community Earth System Model. Two 40-member ensembles driven by historical radiative forcings are generated, one coupled to a dynamical ocean and the other driven by observed sea surface temperatures (SSTs) over the period 1981–2010. The simulations reproduce many aspects of the observed climatology and variability of snow cover extent as characterized by the NOAA snow chart climate data record. Major features of the simulated snow water equivalent (SWE) also agree with observations (GlobSnow Northern Hemisphere SWE data record), although with a lesser degree of fidelity. Ensemble spread in the climate response quantifies the impact of natural climate variability in the presence and absence of coupling to the ocean. Both coupled and uncoupled ensembles indicate an overall decrease in springtime snow cover that is consistent with observations, although springtime trends in most climate realizations are weaker than observed. In the coupled ensemble, a tendency towards excessive warming in wintertime leads to a strong wintertime snow cover loss that is not found in observations. The wintertime warming bias and snow cover reduction trends are reduced in the uncoupled ensemble with observed SSTs. Natural climate variability generates widely different regional patterns of snow trends across realizations; these patterns are related in an intuitive way to temperature, precipitation and circulation trends in individual realizations. In particular, regional snow loss over North America in individual realizations is strongly influenced by North Pacific SST trends (manifested as Pacific Decadal Oscillation variability) and by sea level pressure trends in the North Pacific/North Atlantic sectors.     

Scale-dependent regional climate predictability over North America inferred from CMIP3 and CMIP5 ensemble simulations

Through the analysis of ensembles of coupled model simulations and projections collected from CMIP3 and CMIP5, we demonstrate that a fundamental spatial scale limit might exist below which useful additional refinement of climate model predictions and projections may not be possible. That limit varies among climate variables and from region to region. We show that the uncertainty (noise) in surface temperature predictions (represented by the spread among an ensemble of global climate model simulations) generally exceeds the ensemble mean (signal) at horizontal scales below 1000 km throughout North America, implying poor predictability at those scales. More limited skill is shown for the predictability of regional precipitation. The ensemble spread in this case tends to exceed or equal the ensemble mean for scales below 2000 km. These findings highlight the challenges in predicting regionally specific future climate anomalies, especially for hydroclimatic impacts such as drought and wetness.     

A multi-physics ensemble of present-day climate regional simulations over the Iberian Peninsula

This work assesses the influence of the model physics in present-day regional climate simulations. It is based on a multi-phyiscs ensemble of 30-year long MM5 hindcasted simulations performed over a complex and climatically heterogeneous domain as the Iberian Peninsula. The ensemble consists of eight members that results from combining different parametrization schemes for modeling the Planetary Boundary Layer, the cumulus and the microphysics processes. The analysis is made at the seasonal time scale and focuses on mean values and interannual variability of temperature and precipitation. The objectives are (1) to evaluate and characterize differences among the simulations attributable to changes in the physical options of the regional model, and (2) to identify the most suitable parametrization schemes and understand the underlying mechanisms causing that some schemes perform better than others. The results confirm the paramount importance of the model physics, showing that the spread among the various simulations is of comparable magnitude to the spread obtained in similar multi-model ensembles. This suggests that most of the spread obtained in multi-model ensembles could be attributable to the different physical configurations employed in the various models. Second, we obtain that no single ensemble member outperforms the others in every situation. Nevertheless, some particular schemes display a better performance. On the one hand, the non-local MRF PBL scheme reduces the cold bias of the simulations throughout the year compared to the local Eta model. The reason is that the former simulates deeper mixing layers. On the other hand, the Grell parametrization scheme for cumulus produces smaller amount of precipitation in the summer season compared to the more complex Kain-Fritsch scheme by reducing the overestimation in the simulated frequency of the convective precipitation events. Consequently, the interannual variability of precipitation (temperature) diminishes (increases), which implies a better agreement with the observations in both cases. Although these features improve in general the accuracy of the simulations, controversial nuances are also highlighted.     

Observation-based blended projections from ensembles of regional climate models

We consider the problem of projecting future climate from ensembles of regional climate model (RCM) simulations using results from the North American Regional Climate Change Assessment Program (NARCCAP). To this end, we develop a hierarchical Bayesian space-time model that quantifies the discrepancies between different members of an ensemble of RCMs corresponding to present day conditions, and observational records. Discrepancies are then propagated into the future to obtain high resolution blended projections of 21st century climate. In addition to blended projections, the proposed method provides location-dependent comparisons between the different simulations by estimating the different modes of spatial variability, and using the climate model-specific coefficients of the spatial factors for comparisons. The approach has the flexibility to provide projections at customizable scales of potential interest to stakeholders while accounting for the uncertainties associated with projections at these scales based on a comprehensive statistical framework. We demonstrate the methodology with simulations from the Weather Research & Forecasting regional model (WRF) using three different boundary conditions. We use simulations for two time periods: current climate conditions, covering 1971 to 2000, and future climate conditions under the Special Report on Emissions Scenarios (SRES) A2 emissions scenario, covering 2041 to 2070. We investigate and project yearly mean summer and winter temperatures for a domain in the South West of the United States.     

20个CMIP5模式对中亚地区年平均气温模拟能力评估

利用CMIP5中20个模式的历史模拟结果和英国东英格利亚大学CRU观测数据,采用泰勒图、趋势分析、滑动平均和EOF等方法从气候态和气候变率两方面检验各个模式对中亚年平均气温的模拟能力。结果表明：各模式能较好地模拟1951—2005年中亚地区显著增温趋势和年平均气温的空间分布特点,尤其是高、低值中心和等值线数值分布。泰勒图分析显示,大部分模式的均方根误差在0.5左右,空间相关系数在0.85～0.90之间,标准差在0.5～1.0之间。EOF分析结果表明,模式集合平均能够较好地表现出中亚气温一致升高以及南北反位相波动这两个主要模态的时空变化特征。     

Investigation of regional climate models’ internal variability with a ten-member ensemble of 10-year simulations over a large domain

Previous investigations on regional climate models’ (RCM) internal variability (IV) were limited owing to small ensembles, short simulations and small domains. The present work extends previous studies with a ten-member ensemble of 10-year simulations performed with the Canadian Regional Climate Model over a large domain covering North America. The results show that the IV has no long-term tendency but rather fluctuates in time following the synoptic situation within the domain. The IV of mean-sea-level pressure (MSLP) and screen temperature (ST) show a small annual cycle with larger values in spring, which differs from previous studies. For precipitation (PCP), the IV shows a clear annual cycle with larger values in summer, as previously reported. The 10-year climatology of the IV for MSLP and ST shows a well-defined spatial distribution with larger values in the northeast of the domain, near the outflow boundary. A comparison of the IV of MSLP and ST in summer with the transient-eddy variance reveals that the IV is close to its maximum in a small region near the outflow boundary. Same analysis for PCP in summer shows that the IV reaches its maximum in most parts of the domain, except for a small region on the western side near the inflow boundary. Finally, a comparison of the 10-year climate of each simulation of the ensemble showed that the IV may have a significant impact on the climatology of some variables.     

Do weak AR4 models bias projections of future climate changes over Australia?

Regional climate projections using climate models commonly use an “all-model” ensemble based on data sets such as the Intergovernmental Panel on Climate Change’s (IPCC) 4th Assessment (AR4). Some regional assessments have omitted models based on specific criteria. We use a criteria based on the capacity of climate models to simulate the observed probability density function calculated using daily data, model-by-model and region-by-region for each of the AR4 models over Australia. We demonstrate that by omitting those climate models with relatively weak skill in simulating the observed probability density functions of maximum and minimum temperature and precipitation, different regional projections are obtained. Differences include: larger increases in the mean maximum and mean minimum temperatures, but smaller increases in the annual maximum and minimum temperatures. There is little impact on mean precipitation but the better models simulate a larger increase in the annual rainfall event combined with a larger decrease in the number of rain days. The weaker models bias the amount of mean warming towards lower increases, bias annual maximum temperatures to excessive warming and bias precipitation such that the amount of the annual rainfall event is under-estimated. We suggest that omitting weak models from regional scale estimates of future climate change helps clarify the nature and scale of the projected impacts of global warming.     

Probabilistic multimodel ensemble prediction of decadal variability of East Asian surface air temperature based on IPCC-AR5 near-term climate simulations

Based on near-term climate simulations for IPCC-AR5 (The Fifth Assessment Report), probabilistic multimodel ensemble prediction (PMME) of decadal variability of surface air temperature in East Asia (20°-50 °N, 100°-145°E) was conducted using the multivariate Gaussian ensemble kernel dressing (GED) methodology. The ensemble system exhibited high performance in hindcasting the decadal (1981-2010) mean and trend of temperature anomalies with respect to 1961-90, with a RPS of 0.94 and 0.88 respectively. The interpretation of PMME for future decades (2006-35) over East Asia was made on the basis of the bivariate probability density of the mean and trend. The results showed that, under the RCP4.5 (Representative Concentration Pathway 4.5 W m-2 ) scenario, the annual mean temperature increases on average by about 1.1-1.2 K and the temperature trend reaches 0.6-0.7 K (30 yr)-1 . The pattern for both quantities was found to be that the temperature increase will be less intense in the south. While the temperature increase in terms of the 30-yr mean was found to be virtually certain, the results for the 30-yr trend showed an almost 25% chance of a negative value. This indicated that, using a multimodel ensemble system, even if a longer-term warming exists for 2006-35 over East Asia, the trend for temperature may produce a negative value. Temperature was found to be more affected by seasonal variability, with the increase in temperature over East Asia more intense in autumn (mainly), faster in summer to the west of 115°E, and faster still in autumn to the east of 115°E.     

Climate model dependence and the replicate Earth paradigm

Multi-model ensembles are commonly used in climate prediction to create a set of independent estimates, and so better gauge the likelihood of particular outcomes and better quantify prediction uncertainty. Yet researchers share literature, datasets and model code—to what extent do different simulations constitute independent estimates? What is the relationship between model performance and independence? We show that error correlation provides a natural empirical basis for defining model dependence and derive a weighting strategy that accounts for dependence in experiments where the multi-model mean would otherwise be used. We introduce the “replicate Earth” ensemble interpretation framework, based on theoretically derived statistical relationships between ensembles of perfect models (replicate Earths) and observations. We transform an ensemble of (imperfect) climate projections into an ensemble whose mean and variance have the same statistical relationship to observations as an ensemble of replicate Earths. The approach can be used with multi-model ensembles that have varying numbers of simulations from different models, accounting for model dependence. We use HadCRUT3 data and the CMIP3 models to show that in out of sample tests, the transformed ensemble has an ensemble mean with significantly lower error and much flatter rank frequency histograms than the original ensemble.     

Uncertainties in regional climate change prediction: a regional analysis of ensemble simulations with the HADCM2 coupled AOGCM

We analyze ensembles (four realizations) of historical and future climate transient experiments carried out with the coupled atmosphere-ocean general circulation model (AOGCM) of the Hadley Centre for Climate Prediction and Research, version HADCM2, with four scenarios of greenhouse gas (GHG) and sulfate forcing. The analysis focuses on the regional scale, and in particular on 21 regions covering all land areas in the World (except Antarctica). We examine seasonally averaged surface air temperature and precipitation for the historical period of 1961–1990 and the future climate period of 2046–2075. Compared to previous AOGCM simulations, the HADCM2 model shows a good performance in reproducing observed regional averages of summer and winter temperature and precipitation. The model, however, does not reproduce well observed interannual variability. We find that the uncertainty in regional climate change predictions associated with the spread of different realizations in an ensemble (i.e. the uncertainty related to the internal model variability) is relatively low for all scenarios and regions. In particular, this uncertainty is lower than the uncertainty due to inter-scenario variability and (by comparison with previous regional analyses of AOGCMs) with inter-model variability. The climate biases and sensitivities found for different realizations of the same ensemble were similar to the corresponding ensemble averages and the averages associated with individual realizations of the same ensemble did not differ from each other at the 5% confidence level in the vast majority of cases. These results indicate that a relatively small number of realizations (3 or 4) is sufficient to characterize an AOGCM transient climate change prediction at the regional scale. Received: 12 January 1998 / Accepted: 7 July 1999     

The internal climate variability of HadCM3, a version of the Hadley Centre coupled model without flux adjustments

We examine the internal climate variability of a 1000?year long integration of the third version of the Hadley Centre coupled model (HadCM3). The model requires no flux adjustment, needs no spin up procedure prior to coupling and has a stable climate in the global mean. The principal aims are (1) to validate the internal climate variability against observed climate variability, (2) to examine the model for any periodic modes of variability, (3) to use the model estimate of internal climate variability to asses the probability of occurrence of observed trends in climate variables, and (4) to compare HadCM3 with the previous version of the Hadley Centre model, HadCM2. The magnitude and frequency characteristics of the variability of the global mean surface temperature of HadCM3 on annual to decadal time scales is in good agreement with the observations. Observed upward trends in temperature over the last 20?years and longer are inconsistent with the internal variability of the model. The simulated spatial pattern of surface temperature variability is qualitatively similar to that observed, although there is an overestimation of the land temperature variability and regional errors in ocean temperature variability. The model simulates an El Niño Southern Oscillation with an irregular 3–4?year cycle, and with a teleconnection pattern which is much more like the observations than was found in HadCM2. The interdecadal variability of the model ocean in the tropical Pacific, North Pacific and North Atlantic is broadly similar to that in the real world with none of the simulated patterns having any periodic behaviour. HadCM3 simulates an Arctic Oscillation/North Atlantic Oscillation (NAO) in Northern Hemisphere winter which has a spatial pattern consistent with the observations in the Atlantic region, but has too much teleconnection with the North Pacific. The recent observed upward trend in the NAO index is inconsistent with the model internal variability. The variability of the simulated zonal mean atmospheric temperature shows some marked differences to the observed zonal mean temperature variability, although the comparison is confounded by the sparse observational network and its possible contamination by a climate change signal.     

Very high-resolution climate simulation over Fiji using a global variable-resolution model

A high-resolution climate model simulation has been performed for the first time for Fiji’s climatology. The simulation involved a numerical experiment for a 10-year period (1975–1984), and was conducted at a horizontal resolution of 8 km in a stretched-grid configuration, which is currently the highest resolution at which a global climate model has been applied for regional climatological simulations. Analysis of model-generated data demonstrates a fairly good skill of the CSIRO Conformal-Cubic Atmospheric Model (C-CAM) in the simulation of the annual cycles of maximum and minimum temperatures and rainfall at selected locations in Fiji. The model has also successfully reproduced the pattern of maximum and minimum surface air temperatures between the western and central divisions of Fiji. Model simulation of spatial and temporal distribution of monthly total rainfall (10-year mean) over the main island of Viti Levu in Fiji shows that it reproduces the observed intraseasonal and interannual variability; the influence of the El Niño phenomena has also been captured well in the model-simulated rainfall.     

Climate change projections over India by a downscaling approach using PRECIS

This study presents a comprehensive assessment of the possible regional climate change over India by using Providing REgional Climates for Impacts Studies (PRECIS), a regional climate model (RCM) developed by Met Office Hadley Centre in the United Kingdom. The lateral boundary data for the simulations were taken from a sub-set of six members sampled from the Hadley Centre’s 17- member Quantified Uncertainty in Model Projections (QUMP) perturbed physics ensemble. The model was run with 25 km × 25 km resolution from the global climate model (GCM) - HadCM3Q at the emission rate of special report on emission scenarios (SRES) A1B scenarios. Based on the model performance, six member ensembles running over a period of 1970-2100 in each experiment were utilized to predict possible range of variations in the future projections for the periods 2020s (2005-2035), 2050s (2035-2065) and 2080s (2065-2095) with respect to the baseline period (1975-2005). The analyses concentrated on maximum temperature, minimum temperature and rainfall over the region. For the whole India, the projections of maximum temperature from all the six models showed an increase within the range 2.5°C to 4.4°C by end of the century with respect to the present day climate simulations. The annual rainfall projections from all the six models indicated a general increase in rainfall being within the range 15-24%. Mann-Kendall trend test was run on time series data of temperatures and rainfall for the whole India and the results from some of the ensemble members indicated significant increasing trends. Such high resolution climate change information may be useful for the researchers to study the future impacts of climate change in terms of extreme events like floods and droughts and formulate various adaptation strategies for the society to cope with future climate change.     

Developing a likely climate scenario from multiple regional climate model simulations with an optimal weighting factor

This study presents a performance-based comprehensive weighting factor that accounts for the skill of different regional climate models (RCMs), including the effect of the driving lateral boundary condition coming from either atmosphere–ocean global climate models (AOGCMs) or reanalyses. A differential evolution algorithm is employed to identify the optimal relative importance of five performance metrics, and corresponding weighting factors, that include the relative absolute mean error (RAME), annual cycle, spatial pattern, extremes and multi-decadal trend. Based on cumulative density functions built by weighting factors of various RCMs/AOGCMs ensemble simulations, current and future climate projections were then generated to identify the level of uncertainty in the climate scenarios. This study selected the areas of southern Ontario and Québec in Canada as a case study. The main conclusions are as follows: (1) Three performance metrics were found essential, having the greater relative importance: the RAME, annual variability and multi-decadal trend. (2) The choice of driving conditions from the AOGCM had impacts on the comprehensive weighting factor, particularly for the winter season. (3) Combining climate projections based on the weighting factors significantly increased the consistency and reduced the spread among models in the future climate changes. These results imply that the weighting factors play a more important role in reducing the effects of outliers on plausible future climate conditions in regions where there is a higher level of variability in RCM/AOGCM simulations. As a result of weighting, substantial increases in the projected warming were found in the southern part of the study area during summer, and the whole region during winter, compared to the simple equal weighting scheme from RCM runs. This study is an initial step toward developing a likelihood procedure for climate scenarios on a regional scale using equal or different probabilities for all models.     

Using climate impacts indicators to evaluate climate model ensembles: temperature suitability of premium winegrape cultivation in the United States

We explore the potential to improve understanding of the climate system by directly targeting climate model analyses at specific indicators of climate change impact. Using the temperature suitability of premium winegrape cultivation as a climate impacts indicator, we quantify the inter- and intra-ensemble spread in three climate model ensembles: a physically uniform multi-member ensemble consisting of the RegCM3 high-resolution climate model nested within the NCAR CCSM3 global climate model; the multi-model NARCCAP ensemble consisting of single realizations of multiple high-resolution climate models nested within multiple global climate models; and the multi-model CMIP3 ensemble consisting of realizations of multiple global climate models. We find that the temperature suitability for premium winegrape cultivation is substantially reduced throughout the high-value growing areas of California and the Columbia Valley region (eastern Oregon and Washington) in all three ensembles in response to changes in temperature projected for the mid-twenty first century period. The reductions in temperature suitability are driven primarily by projected increases in mean growing season temperature and occurrence of growing season severe hot days. The intra-ensemble spread in the simulated climate change impact is smaller in the single-model ensemble than in the multi-model ensembles, suggesting that the uncertainty arising from internal climate system variability is smaller than the uncertainty arising from climate model formulation. In addition, the intra-ensemble spread is similar in the NARCCAP nested climate model ensemble and the CMIP3 global climate model ensemble, suggesting that the uncertainty arising from the model formulation of fine-scale climate processes is not smaller than the uncertainty arising from the formulation of large-scale climate processes. Correction of climate model biases substantially reduces both the inter- and intra-ensemble spread in projected climate change impact, particularly for the multi-model ensembles, suggesting that—at least for some systems—the projected impacts of climate change could be more robust than the projected climate change. Extension of this impacts-based analysis to a larger suite of impacts indicators will deepen our understanding of future climate change uncertainty by focusing on the climate phenomena that most directly influence natural and human systems.