The Atlantic Multidecadal Oscillation in twentieth century climate simulations: uneven progress from CMIP3 to CMIP5

Decadal variability in the climate system from the Atlantic Multidecadal Oscillation (AMO) is one of the major sources of variability at this temporal scale that climate models must properly incorporate because of its climate impact. The current analysis of historical simulations of the twentieth century climate from models participating in the CMIP3 and CMIP5 projects assesses how these models portray the observed spatiotemporal features of the sea surface temperature (SST) and precipitation anomalies associated with the AMO. A short sample of the models is analyzed in detail by using all ensembles available of the models CCSM3, GFDL-CM2.1, UKMO-HadCM3, and ECHAM5/MPI-OM from the CMIP3 project, and the models CCSM4, GFDL-CM3, UKMO-HadGEM2-ES, and MPI-ESM-LR from the CMIP5 project. The structure and evolution of the SST anomalies of the AMO have not progressed consistently from the CMIP3 to the CMIP5 models. While the characteristic period of the AMO (smoothed with a binomial filter applied fifty times) is underestimated by the three of the models, the e-folding time of the autocorrelations shows that all models underestimate the 44-year value from observations by almost 50 %. Variability of the AMO in the 10–20/70–80 year ranges is overestimated/underestimated in the models and the variability in the 10–20 year range increases in three of the models from the CMIP3 to the CMIP5 versions. Spatial variability and correlation of the AMO regressed precipitation and SST anomalies in summer and fall indicate that models are not up to the task of simulating the AMO impact on the hydroclimate over the neighboring continents. This is in spite of the fact that the spatial variability and correlations in the SST anomalies improve from CMIP3 to CMIP5 versions in two of the models. However, a multi-model mean from a sample of 14 models whose first ensemble was analyzed indicated there were no improvements in the structure of the SST anomalies of the AMO or associated regional precipitation anomalies in summer and fall from CMIP3 to CMIP5 projects.     

An ENSO stability analysis. Part I: results from a hybrid coupled model

In this study, we use the Bjerknes stability (BJ) index as a tool to investigate overall El Niño-Southern Oscillation (ENSO) stability in a hybrid-coupled model (HCM) with various atmosphere and ocean background states. This HCM shows that ENSO growth rates as measured by the BJ index and linear growth rates estimated directly with a time series of the Niño 3.4 indices from the coupled model simulations exhibit similar dependence on background states, coupling strength, and thermodynamic damping intensity. That is, the BJ index and linear growth rates increase with a decrease in the intensity of the background wind, an increase in coupling strength, or a decrease in the intensity of thermodynamic damping, although the BJ index tends to overestimate the growth rate. A detailed analysis of the components of the BJ index formula suggests the importance of model climatological background states and oceanic dynamic parameters in determining ENSO stability. We conclude that the BJ index may serve as a useful tool for qualitatively evaluating the overall ENSO stability in coupled models or in observations without a detailed eigen-analysis that is difficult to perform in models more complex than relatively simple models.     

ENSO representation in climate models: from CMIP3 to CMIP5

We analyse the ability of CMIP3 and CMIP5 coupled ocean–atmosphere general circulation models (CGCMs) to simulate the tropical Pacific mean state and El Niño-Southern Oscillation (ENSO). The CMIP5 multi-model ensemble displays an encouraging 30 % reduction of the pervasive cold bias in the western Pacific, but no quantum leap in ENSO performance compared to CMIP3. CMIP3 and CMIP5 can thus be considered as one large ensemble (CMIP3 + CMIP5) for multi-model ENSO analysis. The too large diversity in CMIP3 ENSO amplitude is however reduced by a factor of two in CMIP5 and the ENSO life cycle (location of surface temperature anomalies, seasonal phase locking) is modestly improved. Other fundamental ENSO characteristics such as central Pacific precipitation anomalies however remain poorly represented. The sea surface temperature (SST)-latent heat flux feedback is slightly improved in the CMIP5 ensemble but the wind-SST feedback is still underestimated by 20–50 % and the shortwave-SST feedbacks remain underestimated by a factor of two. The improvement in ENSO amplitudes might therefore result from error compensations. The ability of CMIP models to simulate the SST-shortwave feedback, a major source of erroneous ENSO in CGCMs, is further detailed. In observations, this feedback is strongly nonlinear because the real atmosphere switches from subsident (positive feedback) to convective (negative feedback) regimes under the effect of seasonal and interannual variations. Only one-third of CMIP3 + CMIP5 models reproduce this regime shift, with the other models remaining locked in one of the two regimes. The modelled shortwave feedback nonlinearity increases with ENSO amplitude and the amplitude of this feedback in the spring strongly relates with the models ability to simulate ENSO phase locking. In a final stage, a subset of metrics is proposed in order to synthesize the ability of each CMIP3 and CMIP5 models to simulate ENSO main characteristics and key atmospheric feedbacks.     

Variability and trends of sub-continental scale surface climate in the twentieth century. Part II: AOGCM simulations

This work presents an analysis of simulated temperature and precipitation variability and trends throughout the twentieth century over 22 land regions of sub-continental scale in the HADCM3 and HADCM2 (two realizations) coupled models. Regional temperature biases in the HADCM3 and HADCM2 are mostly in the range of -5 K to +3 K for the seasonal averages and -3 K to +2 K for the annual average. Seasonal precipitation biases are mostly in the range of -50% to 75% of present day precipitation, with a tendency in both models to overpredict cold season precipitation. Except for cold season temperature in mid- and high-latitude Northern Hemisphere regions, the average climatology of the HADCM2 and HADCM3 is of comparable quality despite the lack of an ocean flux adjustment in the HADCM3. Both models show warming trends of magnitude in line with observations, although the observed inter-regional patterns of warming trend are not well reproduced. Measures of temperature and precipitation interannual to interdecadal variability in the models are in general agreement with observations except for Northern Hemisphere summer temperature variability, which is overestimated. The models somewhat underestimate the inter-decadal variations in interannual variability measures observed during the century and overestimate the range of anomalies. Both models tend to overpredict the occurrences of short persistences (1-3 years) and underpredict the occurrence and maximum length of long persistences (greater than three years), which is an indication of a deficiency in the simulation of long-lived anomaly regimes. Compared to observations, the models produce a higher magnitude of temporal anomaly correlation across regions and correlation between temperature and precipitation anomalies for a given region. This suggests that local processes that may be effective in decoupling the observed regional anomalies are not captured well. Overall, the variability measures in the HADCM2 and HADCM3 are of similar quality, indicating that the use of a flux correction in the HADCM2 does not strongly affect the regional variability characteristics of the model.     

CMIP5模式对ENSO现象的模拟能力评估

针对参与耦合模式比较计划(CMIP5)的17个海-气耦合模式对20世纪气候的模拟结果,从热带太平洋海表温度和大气海平面气压变化的综合分析角度较详细评估了模式对厄尔尼诺-南方涛动(ENSO)现象的模拟能力。结果表明,这些模式基本上能模拟出ENSO现象的一些主要特征,包括热带太平洋海温的空间分布及其时空演变特征、与海平面气压变化的关联、ENSO周期变化及锁相特征等,但不同模式的模拟结果仍然差异较大。(1)从模拟的热带太平洋年平均海温的偏差来看,多模式集合平均值与观测的均方根误差小于1.0℃,但单个模式的误差相对要大一些。误差较小的为1.2—1.3℃,多数模式在1.6℃以下,但也有个别模式的误差超过2.0℃。(2)从经验正交函数分解结果来看,热带太平洋实测月平均海表温度距平和海平面气压距平的年际尺度变化第1模态主要表现为ENSO变化特征,第2模态反映的是海温的长期变化趋势。只有少数几个CMIP5模式能够再现这种特征,多数模式所模拟的海温距平/海平面气压距平时空变化的第1、第2特征向量分布顺序与观测分析正好相反,ENSO变成了第2模态,趋势成了最主要的模态。尽管如此,所有模式都能模拟出南方涛动变化与热带太平洋海温距平时空变化的密切关联,无论是作为第1还是第2特征模态,所有模式模拟的南方涛动与热带太平洋海温距平时空变化都有密切相关。(3)谱分析结果表明,ENSO现象具有2—7年的周期,其中,4年的周期最明显。大多数模式模拟的ENSO周期在此范围内,但有些模式的主要周期偏短,为2年左右。个别模式的ENSO主要周期为11年,已超出2—7年的范围。(4)多数模式模拟的厄尔尼诺及拉尼娜的峰值出现在冬季(11—2月),与观测基本吻合。另有少数模式模拟的峰值出现在9—10月,比观测略提前。只有个别模式模拟的峰值出现在夏季,与观测相差太大。     

ENSO phase-locking to the boreal winter in CMIP3 and CMIP5 models

In this study, the El Nino-Southern Oscillation (ENSO) phase-locking to the boreal winter in CMIP3 and CMIP5 models is examined. It is found that the models that are poor at simulating the winter ENSO peak tend to simulate colder seasonal-mean sea-surface temperature (SST) during the boreal summer and associated shallower thermocline depth over the eastern Pacific. These models tend to amplify zonal advection and thermocline depth feedback during boreal summer. In addition, the colder eastern Pacific SST in the model can reduce the summertime mean local convective activity, which tends to weaken the atmospheric response to the ENSO SST forcing. It is also revealed that these models have more serious climatological biases over the tropical Pacific, implying that a realistic simulation of the climatological fields may help to simulate winter ENSO peak better. The models that are poor at simulating ENSO peak in winter also show excessive anomalous SST warming over the western Pacific during boreal winter of the El Nino events, which leads to strong local convective anomalies. This prevents the southward shift of El Nino-related westerly during boreal winter season. Therefore, equatorial westerly is prevailed over the western Pacific to further development of ENSO-related SST during boreal winter. This bias in the SST anomaly is partly due to the climatological dry biases over the central Pacific, which confines ENSO-related precipitation and westerly responses over the western Pacific.     

The role of atmosphere feedbacks during ENSO in the CMIP3 models. Part II: using AMIP runs to understand the heat flux feedback mechanisms

Several studies using ocean?Catmosphere general circulation models (GCMs) suggest that the atmospheric component plays a dominant role in the modelled El Ni?o-Southern Oscillation (ENSO). To help elucidate these findings, the two main atmosphere feedbacks relevant to ENSO, the Bjerknes positive feedback (??) and the heat flux negative feedback (??), are here analysed in nine AMIP runs of the CMIP3 multimodel dataset. We find that these models generally have improved feedbacks compared to the coupled runs which were analysed in part I of this study. The Bjerknes feedback,???, is increased in most AMIP runs compared to the coupled run counterparts, and exhibits both positive and negative biases with respect to ERA40. As in the coupled runs, the shortwave and latent heat flux feedbacks are the two dominant components of ?? in the AMIP runs. We investigate the mechanisms behind these two important feedbacks, in particular focusing on the strong 1997?C1998 El Ni?o. Biases in the shortwave flux feedback, ?? _SW, are the main source of model uncertainty in ??. Most models do not successfully represent the negative ??_SW in the East Pacific, primarily due to an overly strong low-cloud positive feedback in the far eastern Pacific. Biases in the cloud response to dynamical changes dominate the modelled ?? _SW biases, though errors in the large-scale circulation response to sea surface temperature (SST) forcing also play a role. Analysis of the cloud radiative forcing in the East Pacific reveals model biases in low cloud amount and optical thickness which may affect ?? _SW. We further show that the negative latent heat flux feedback, ?? _LH, exhibits less diversity than ?? _SW and is primarily driven by variations in the near-surface specific humidity difference. However, biases in both the near-surface wind speed and humidity response to SST forcing can explain the inter-model ??_LH differences.     

The Asian summer monsoon: an intercomparison of CMIP5 vs. CMIP3 simulations of the late 20th century

The boreal summer Asian monsoon has been evaluated in 25 Coupled Model Intercomparison Project-5 (CMIP5) and 22 CMIP3 GCM simulations of the late twentieth Century. Diagnostics and skill metrics have been calculated to assess the time-mean, climatological annual cycle, interannual variability, and intraseasonal variability. Progress has been made in modeling these aspects of the monsoon, though there is no single model that best represents all of these aspects of the monsoon. The CMIP5 multi-model mean (MMM) is more skillful than the CMIP3 MMM for all diagnostics in terms of the skill of simulating pattern correlations with respect to observations. Additionally, for rainfall/convection the MMM outperforms the individual models for the time mean, the interannual variability of the East Asian monsoon, and intraseasonal variability. The pattern correlation of the time (pentad) of monsoon peak and withdrawal is better simulated than that of monsoon onset. The onset of the monsoon over India is typically too late in the models. The extension of the monsoon over eastern China, Korea, and Japan is underestimated, while it is overestimated over the subtropical western/central Pacific Ocean. The anti-correlation between anomalies of all-India rainfall and Niño3.4 sea surface temperature is overly strong in CMIP3 and typically too weak in CMIP5. For both the ENSO-monsoon teleconnection and the East Asian zonal wind-rainfall teleconnection, the MMM interannual rainfall anomalies are weak compared to observations. Though simulation of intraseasonal variability remains problematic, several models show improved skill at representing the northward propagation of convection and the development of the tilted band of convection that extends from India to the equatorial west Pacific. The MMM also well represents the space–time evolution of intraseasonal outgoing longwave radiation anomalies. Caution is necessary when using GPCP and CMAP rainfall to validate (1) the time-mean rainfall, as there are systematic differences over ocean and land between these two data sets, and (2) the timing of monsoon withdrawal over India, where the smooth southward progression seen in India Meteorological Department data is better realized in CMAP data compared to GPCP data.     

The Indo-Australian monsoon and its relationship to ENSO and IOD in reanalysis data and the CMIP3/CMIP5 simulations

A large spread exists in both Indian and Australian average monsoon rainfall and in their interannual variations diagnosed from various observational and reanalysis products. While the multi model mean monsoon rainfall from 59 models taking part in the Coupled Model Intercomparison Project (CMIP3 and CMIP5) fall within the observational uncertainty, considerable model spread exists. Rainfall seasonality is consistent across observations and reanalyses, but most CMIP models produce either a too peaked or a too flat seasonal cycle, with CMIP5 models generally performing better than CMIP3. Considering all North-Australia rainfall, most models reproduce the observed Australian monsoon-El Niño Southern Oscillation (ENSO) teleconnection, with the strength of the relationship dependent on the strength of the simulated ENSO. However, over the Maritime Continent, the simulated monsoon-ENSO connection is generally weaker than observed, depending on the ability of each model to realistically reproduce the ENSO signature in the Warm Pool region. A large part of this bias comes from the contribution of Papua, where moisture convergence seems to be particularly affected by this SST bias. The Indian summer monsoon-ENSO relationship is affected by overly persistent ENSO events in many CMIP models. Despite significant wind anomalies in the Indian Ocean related to Indian Ocean Dipole (IOD) events, the monsoon-IOD relationship remains relatively weak both in the observations and in the CMIP models. Based on model fidelity in reproducing realistic monsoon characteristics and ENSO teleconnections, we objectively select 12 “best” models to analyze projections in the rcp8.5 scenario. Eleven of these models are from the CMIP5 ensemble. In India and Australia, most of these models produce 5–20 % more monsoon rainfall over the second half of the twentieth century than during the late nineteenth century. By contrast, there is no clear model consensus over the Maritime Continent.     

European climate in the late twenty-first century: regional simulations with two driving global models and two forcing scenarios

A basic analysis is presented for a series of regional climate change simulations that were conducted by the Swedish Rossby Centre and contribute to the PRUDENCE (Prediction of Regional scenarios and Uncertainties for Defining EuropeaN Climate change risks and Effects) project. For each of the two driving global models HadAM3H and ECHAM4/OPYC3, a 30-year control run and two 30-year scenario runs (based on the SRES A2 and B2 emission scenarios) were made with the regional model. In this way, four realizations of climate change from 1961–1990 to 2071–2100 were obtained. The simulated changes are larger for the A2 than the B2 scenario (although with few qualitative differences) and in most cases in the ECHAM4/OPYC3-driven (RE) than in the HadAM3H-driven (RH) regional simulations. In all the scenario runs, the warming in northern Europe is largest in winter or late autumn. In central and southern Europe, the warming peaks in summer when it locally reaches 10 °C in the RE-A2 simulation and 6–7 °C in the RH-A2 and RE-B2 simulations. The four simulations agree on a general increase in precipitation in northern Europe especially in winter and on a general decrease in precipitation in southern and central Europe in summer, but the magnitude and the geographical patterns of the change differ markedly between RH and RE. This reflects very different changes in the atmospheric circulation during the winter half-year, which also lead to quite different simulated changes in windiness. All four simulations show a large increase in the lowest minimum temperatures in northern, central and eastern Europe, most likely due to reduced snow cover. Extreme daily precipitation increases even in most of those areas where the mean annual precipitation decreases.     

Future projections of synoptic weather types over the Arabian Peninsula during the twenty-first century using an ensemble of CMIP5 models

An assessment of future change in synoptic conditions over the Arabian Peninsula throughout the twenty-first century was performed using 20 climate models from the Coupled Model Intercomparison Project Phase 5 (CMIP5) database. We employed the mean sea level pressure (SLP) data from model output together with NCEP/NCAR reanalysis data and compared the relevant circulation types produced by the Lamb classification scheme for the base period 1975–2000. Overall, model results illustrated good agreement with the reanalysis, albeit with a tendency to underestimate cyclonic (C) and southeasterly (SE) patterns and to overestimate anticyclones and directional flows. We also investigated future projections for each circulation-type during the rainy season (December–May) using three Representative Concentration Pathways (RCPs), comprising RCP2.6, RCP4.5, and RCP8.5. Overall, two scenarios (RCP4.5 and RCP 8.5) revealed a statistically significant increase in weather types favoring above normal rainfall in the region (e.g., C and E-types). In contrast, weather types associated with lower amounts of rainfall (e.g., anticyclones) are projected to decrease in winter but increase in spring. For all scenarios, there was consistent agreement on the sign of change (i.e., positive/negative) for the most frequent patterns (e.g., C, SE, E and A-types), whereas the sign was uncertain for less recurrent types (e.g., N, NW, SE, and W). The projected changes in weather type frequencies in the region can be viewed not only as indicators of change in rainfall response but may also be used to inform impact studies pertinent to water resource planning and management, extreme weather analysis, and agricultural production.     

Climate of Russia in the 21st century. Part 3. Future climate changes calculated with an ensemble of coupled atmosphere-ocean general circulation CMIP3 models

Probable climate changes in Russia in the 21st century are considered based on the results of global climate simulations with an ensemble of coupled atmosphere-ocean CMIP3 models. The future changes in the surface air temperature, atmospheric pressure, cloud amount, atmospheric precipitation, snow cover, soil water content, and annual runoff in Russia and some of its regions in the early, middle, and late 21st century are analyzed using the A2 scenario of the greenhouse gas and aerosol emission. Future changes in the yearly highest and lowest surface air temperatures and in summer precipitation of high intensity are estimated for Russia. Possible oscillations of the Caspian Sea level associated with the expected global climate warming are estimated. In addition to the estimates of the ensemble mean changes in climatic characteristics, the information about standard deviations and statistical significance of the corresponding climate changes is given.     

ENSO modes of the equatorial Pacific Ocean in observations and CMIP5 models

El Niño/Southern Oscillation (ENSO) is the predominant interannual variability of the global climate system. How might ENSO change in a warmer world? The dominant two Combined Empirical Orthogonal Functions (CEOF) of the equatorial ocean temperature and zonal and vertical motion identify two modes that shown a transition in the eastern Pacific from a warming eastward/downward motion to a cooling westward/upward flow. These results also suggest consistent changes to the west and at depths down to 300 m. These dominate CEOFs provide a compact tool for assessing Coupled Model Intercomparison Project Phase 5 ocean model output for both the recent historical period and for the latter part of the twenty first century. Most of the analyzed models replicate well the spatial patterns of the dominant observational CEOF modes, but nearly always underestimate the magnitudes. Comparing model output for the twentieth and twenty first centuries there is very little change between the spatial patterns of the ENSO modes of the two periods. This lack of response to climate change is shown to be partly related to competing influences of climatic changes in the mean ocean circulation.     

Simple indices of global climate variability and change Part II: attribution of climate change during the twentieth century

Five simple indices of surface temperature are used to investigate the influence of anthropogenic and natural (solar irradiance and volcanic aerosol) forcing on observed climate change during the twentieth century. These indices are based on spatial fingerprints of climate change and include the global-mean surface temperature, the land-ocean temperature contrast, the magnitude of the annual cycle in surface temperature over land, the Northern Hemisphere meridional temperature gradient and the hemispheric temperature contrast. The indices contain information independent of variations in global-mean temperature for unforced climate variations and hence, considered collectively, they are more useful in an attribution study than global mean surface temperature alone. Observed linear trends over 1950–1999 in all the indices except the hemispheric temperature contrast are significantly larger than simulated changes due to internal variability or natural (solar and volcanic aerosol) forcings and are consistent with simulated changes due to anthropogenic (greenhouse gas and sulfate aerosol) forcing. The combined, relative influence of these different forcings on observed trends during the twentieth century is investigated using linear regression of the observed and simulated responses of the indices. It is found that anthropogenic forcing accounts for almost all of the observed changes in surface temperature during 1946–1995. We found that early twentieth century changes (1896–1945) in global mean temperature can be explained by a combination of anthropogenic and natural forcing, as well as internal climate variability. Estimates of scaling factors that weight the amplitude of model simulated signals to corresponding observed changes using a combined normalized index are similar to those calculated using more complex, optimal fingerprint techniques.