Search

Large-scale and persistent heat waves affecting central-eastern China are investigated in 40 different simulations of sea surface temperature driven global atmospheric models. The different models are compared with results from reanalysis and ground station datasets. It is found that the dynamics of heat-wave events is well reproduced by the models. However, they tend to produce too-persistent heat-wave events (lasting more than 20 days), and several hypotheses were tested to explain this bias. The daily variability of the temperatures or the seasonal signal did not explain the persistence. However, interannual variability of the temperatures in the models, and especially the sharp transition in the mid-1990s, has a large impact on the duration of heat waves. A filtering method was applied to select the models closest to the observations in terms of events persistence. The selected models do not show a significant difference from the other models for the long-term trends. Thus, the bias on the duration of the events does not impact the reliability of the model positive trends, which is mainly controlled by the changes in mean temperatures.

Degree-days are a temperature index used for understanding the impact of climate change. Different methods to deal with climate model biases, termed bias correction or more generally calibration, yield different projections of such indices, something not widely understood for temperature indices in many impact sectors. An analytical expression is derived for the expected value of degree-days given parameters of the underlying statistical distribution (assumed to be Gaussian). It is demonstrated that the uncertainty introduced by calibration methodology is driven by the magnitude of the nonlinearity in this expression. In a climate where mean temperature is, and remains, far from (approximately three standard deviations) the threshold used in defining the index, the equation is approximately linear, and methodological choice makes little difference relative to the absolute number of degree-days. However, case studies for U.K. cities London and Glasgow for heating and cooling degree-days (HDD and CDD; these are degree-day indices used in the estimation of energy use for heating and cooling buildings) demonstrate that, when temperatures are close to the threshold, unrealistic results may arise if appropriate calibration is not performed. Seasonally varying temperature biases in the 11-member perturbed parameter ensemble HadRM3 are discussed, and different calibration strategies are applied to this ensemble. For projections of U.K. HDD, the difference between results from simple and advanced methodologies is relatively small, as the expression for HDD is approximately linear in many months and locations. For U.K. CDD, an inappropriate method has a large relative impact on projections because of the proximity to the threshold. In both cases, the uncertainty caused by methodology is comparable to that caused by ensemble spread.

Reconstructions of past climate show notable temperature variability over the past millennium, with relatively warm conditions during the Medieval Climate Anomaly (MCA) and a relatively cold Little Ice Age (LIA). Multimodel simulations of the past millennium are used together with a wide range of reconstructions of Northern Hemispheric mean annual temperature to separate climate variability from 850 to 1950 CE into components attributable to external forcing and internal climate variability. External forcing is found to contribute significantly to long-term temperature variations irrespective of the proxy reconstruction, particularly from 1400 onward. Over the MCA alone, however, the effect of forcing is only detectable in about half of the reconstructions considered, and the response to forcing in the models cannot explain the warm conditions around 1000 CE seen in some reconstructions. The residual from the detection analysis is used to estimate internal variability independent from climate modeling, and it is found that the recent observed 50- and 100-yr hemispheric temperature trends are substantially larger than any of the internally generated trends even using the large residuals over the MCA. Variations in solar output and explosive volcanism are found to be the main drivers of climate change from 1400 to 1900, but for the first time a significant contribution from greenhouse gas variations to the cold conditions during 1600–1800 is also detected. The proxy reconstructions tend to show a smaller forced response than is simulated by the models. This discrepancy is shown, at least partly, to be likely associated with the difference in the response to large volcanic eruptions between reconstructions and model simulations.

Clear-sky brightness temperature measurements from the High-Resolution Infrared Radiation Sounder (HIRS) are simulated with two climate models via a radiative transfer code. The models are sampled along the HIRS orbit paths to derive diurnal climatologies of simulated brightness temperature analogous to an existing climatology based on HIRS observations. Simulated and observed climatologies are compared to assess model performance and the robustness of the observed climatology. Over land, there is good agreement between simulations and observations, with particularly high consistency for the tropospheric temperature channels. Diurnal cycles in the middle- and upper-tropospheric water vapor channels are weak in both simulations and observations, but the simulated diurnal brightness temperature ranges are smaller than are observed with different phase and there are also intermodel differences. Over sea, the absence of diurnal variability in the models’ sea surface temperatures causes an underestimate of the small diurnal cycles measured in the troposphere. The simulated and observed climatologies imply similar diurnal sampling biases in the HIRS record for the tropospheric temperature channels, but for the upper-tropospheric water vapor channel, differences in the contributions of the 24- and 12-hourly diurnal harmonics lead to differences in the implied bias. Comparison of diurnal cycles derived from HIRS-like and full model sampling suggests that the HIRS measurements are sufficient to fully constrain the diurnal behavior. Overall, the results suggest that recent climate models well represent the major processes driving the diurnal behavior of clear-sky brightness temperature in the HIRS channels. This encourages further studies of observed and simulated climate trends over the HIRS era.

Uncertainties in observed records of atmospheric temperature aloft remain poorly quantified. This has resulted in considerable controversy regarding signals of climate change over recent decades from temperature records of radiosondes and satellites. This work revisits the problems associated with the removal of inhomogeneities from the historical radiosonde temperature records, and provides a method for quantifying uncertainty in an adjusted radiosonde climate record due to the subjective choices made during the data homogenization. This paper presents an automated homogenization method designed to replicate the decisions made by manual judgment in the generation of an earlier radiosonde dataset [i.e., the Hadley Centre radiosonde temperature dataset (HadAT)]. A number of validation experiments have been conducted to test the system performance and impact on linear trends. Using climate model data to simulate biased radiosonde data, the authors show that limitations in the homogenization method are sufficiently large to explain much of the tropical trend discrepancy between HadAT and estimates from satellite platforms and climate models. This situation arises from the combination of systematic (unknown magnitude) and random uncertainties (of order 0.05 K decade−1) in the radiosonde data. Previous assessment of trends and uncertainty in HadAT is likely to have underestimated the systematic bias in tropical mean temperature trends. This objective assessment of radiosonde homogenization supports the conclusions of the synthesis report of the U.S. Climate Change Science Program (CCSP), and associated research, regarding potential bias in tropospheric temperature records from radiosondes.

Simulations of the last 500 yr carried out using the Third Hadley Centre Coupled Ocean–Atmosphere GCM (HadCM3) with anthropogenic and natural (solar and volcanic) forcings have been analyzed. Global-mean surface temperature change during the twentieth century is well reproduced. Simulated contributions to global-mean sea level rise during recent decades due to thermal expansion (the largest term) and to mass loss from glaciers and ice caps agree within uncertainties with observational estimates of these terms, but their sum falls short of the observed rate of sea level rise. This discrepancy has been discussed by previous authors; a completely satisfactory explanation of twentieth-century sea level rise is lacking. The model suggests that the apparent onset of sea level rise and glacier retreat during the first part of the nineteenth century was due to natural forcing. The rate of sea level rise was larger during the twentieth century than during the previous centuries because of anthropogenic forcing, but decreasing natural forcing during the second half of the twentieth century tended to offset the anthropogenic acceleration in the rate. Volcanic eruptions cause rapid falls in sea level, followed by recovery over several decades. The model shows substantially less decadal variability in sea level and its thermal expansion component than twentieth-century observations indicate, either because it does not generate sufficient ocean internal variability, or because the observational analyses overestimate the variability.

A new flexible gridded dataset of sea surface temperature (SST) since 1850 is presented and its uncertainties are quantified. This analysis [the Second Hadley Centre Sea Surface Temperature dataset (HadSST2)] is based on data contained within the recently created International Comprehensive Ocean–Atmosphere Data Set (ICOADS) database and so is superior in geographical coverage to previous datasets and has smaller uncertainties. Issues arising when analyzing a database of observations measured from very different platforms and drawn from many different countries with different measurement practices are introduced. Improved bias corrections are applied to the data to account for changes in measurement conditions through time. A detailed analysis of uncertainties in these corrections is included by exploring assumptions made in their construction and producing multiple versions using a Monte Carlo method. An assessment of total uncertainty in each gridded average is obtained by combining these bias-correction-related uncertainties with those arising from measurement errors and undersampling of intragrid box variability. These are calculated by partitioning the variance in grid box averages between real and spurious variability. From month to month in individual grid boxes, sampling uncertainties tend to be most important (except in certain regions), but on large-scale averages bias-correction uncertainties are more dominant owing to their correlation between grid boxes. Changes in large-scale SST through time are assessed by two methods. The linear warming between 1850 and 2004 was 0.52° ± 0.19°C (95% confidence interval) for the globe, 0.59° ± 0.20°C for the Northern Hemisphere, and 0.46° ± 0.29°C for the Southern Hemisphere. Decadally filtered differences for these regions over this period were 0.67° ± 0.04°C, 0.71° ± 0.06°C, and 0.64° ± 0.07°C.

Validation of the decadal to centennial timescale variability of coupled climate models is limited by the scarcity of long observational records. Proxy indicators of climate, such as tree rings, ice cores, etc., can be utilized for this purpose. This study presents a quantitative comparison of the variability of the third version of the Hadley Centre ocean‐atmosphere coupled model with a network of temperature-sensitive tree-ring densities covering the northern high latitudes. The tree-ring density records are up to 600 years long, and temperature reconstructions based on two different methods of removing the bias due to changing tree age are used. The first is a standard method that may remove low-frequency variability on timescales of the order of the tree life span (i.e., multidecadal to century timescales). The second (age-band decomposition) maintains low-frequency variability by only comparing similar age tree rings at each site, thus avoiding the need to remove the age effect (but at the cost of greater uncertainty in the earlier years when fewer tree cores are available). The variability of the model control simulation, which represents only the internal variability of the climate system, agrees reasonably well with the tree-ring reconstructions using the standard method at the regional level, although the model may underestimate the variance of mean Northern Hemisphere land temperature by as much as a factor of 1.8 on all timescales if one takes account of the uncertainty in the tree-ring reconstructions. Agreement with the ageband decomposition tree-ring reconstructions is less good with the model underestimating the hemispheric variance by as much as a factor of 2.1 on all timescales and by as much as a factor of 3.0 on decadal to centennial timescales. Underestimation of the natural variability of climate by the model would be serious as it may lead to false detections of climate change or erroneously low uncertainty estimates in future climate predictions. However, it is shown that some of this underestimation may be due to the lack of natural climate forcing in the model control simulation due, for example, to solar variability and volcanic eruptions. The study suggests that further quantification of the uncertainties in the proxy data, and inclusion of natural climate forcings in the model simulations, are important steps in making comparisons of climate models with the proxy record over the last 1000 years.

This paper studies the variability of deep-ocean mass transport using four 1000-yr integrations performed with coupled general circulation models. Statistics describing the spectral and spatial features are considered. It is shown that these features depend crucially on the time-mean state. For the transport of tropical and subtropical water masses in three of the integrations, the spectral levels continually increase with decreasing frequency and do not show isolated peaks at low frequencies. The slope of the low-frequency spectrum (in a log–log plot) changes with increasing depth. It has values of about 0 near the surface, about −1 at intermediate depth, and about −2 at or near the bottom. The result indicates that the maximal memory timescale for deep-ocean mass transport is longer than a few centuries. The situation is different in the fourth integration, which has a different mean circulation pattern. In this case, the low-frequency spectrum is more or less flat in the tropical and subtropic...

This study compares the variability of surface air temperature in three long coupled ocean‐atmosphere general circulation model integrations. It is shown that the annual mean climatology of the surface air temperatures (SAT) in all three models is realistic and the linear trends over the 1000-yr integrations are small over most areas of the globe. Second, although there are notable differences among the models, the models’ SAT variability is fairly realistic on annual to decadal timescales, both in terms of the geographical distribution and of the global mean values. A notable exception is the poor simulation of observed tropical Pacific variability. In the HadCM2 model, the tropical variability is overestimated, while in the GFDL and HAM3L models, it is underestimated. Also, the ENSO-related spectral peak in the globally averaged observed SAT differs from that in any of the models. The relatively low resolution required to integrate models for long time periods inhibits the successful simulation of the variability in this region. On timescales longer than a few decades, the largest variance in the models is generally located near sea ice margins in high latitudes, which are also regions of deep oceanic convection and variability related to variations in the thermohaline circulation. However, the exact geographical location of these maxima varies from model to model. The preferred patterns of interdecadal variability that are common to all three coupled models can be isolated by computing empirical orthogonal functions (EOFs) of all model data simultaneously using the common EOF technique. A comparison of the variance each model associated with these common EOF patterns shows that the models generally agree on the most prominent patterns of variability. However, the amplitudes of the dominant modes of variability differ to some extent between the models and between the models and observations. For example, two of the models have a mode with relatively large values of the same sign over most of the Northern Hemisphere midlatitudes. This mode has been shown to be relevant for the separation of the temperature response pattern due to sulfate aerosol forcing from the response to greenhouse gas forcing. This indicates that the results of the detection of climate change and its attribution to different external forcings may differ when unperturbed climate variability in surface air temperature is estimated using different coupled models. Assuming that the simulation of variability of the global mean SAT is as realistic on longer timescales as it is for the shorter timescales, then the observed warming of more than 0.5 K of the SAT in the last 110 yr is not likely to be due to internally generated variability of the coupled atmosphere‐ocean‐sea ice system. Instead, the warming is likely to be due to changes in the radiative forcing of the climate system, such as the forcing associated with increases in greenhouse gases.

Spatially and temporally dependent fingerprint patterns of near-surface temperature change are derived from transient climate simulations of the second Hadley Centre coupled ocean–atmosphere GCM (HADCM2). Trends in near-surface temperature are calculated from simulations in which HADCM2 is forced with historical increases in greenhouse gases only and with both greenhouse gases and anthropogenic sulfur emissions. For each response an ensemble of four simulations is carried out. An estimate of the natural internal variability of the ocean–atmosphere system is taken from a long multicentury control run of HADCM2. The aim of the study is to investigate the spatial and temporal scales on which it is possible to detect a significant change in climate. Temporal scales are determined by taking temperature trends over 10, 30, and 50 yr using annual mean data, and spatial scales are defined by projecting these trends onto spherical harmonics. Each fingerprint pattern is projected onto the recent observed p...

A large number of perturbed-physics simulations of version 3 of the Hadley Centre Atmosphere Model (HadAM3) were compared with the Clouds and the Earth's Radiant Energy System (CERES) estimates of outgoing longwave radiation (OLR) and reflected shortwave radiation (RSR) as well as OLR and RSR from the earlier Earth Radiation Budget Experiment (ERBE) estimates. The model configurations were produced from several independent optimization experiments in which four parameters were adjusted. Model–observation uncertainty was estimated by combining uncertainty arising from satellite measurements, observational radiation imbalance, total solar irradiance, radiative forcing, natural aerosol, internal climate variability, and sea surface temperature and that arising from parameters that were not varied. Using an emulator built from 14 001 “slab” model evaluations carried out using the climateprediction.net ensemble, the climate sensitivity for each configuration was estimated. Combining different prior probabilities for model configurations with the likelihood for each configuration and taking account of uncertainty in the emulated climate sensitivity gives, for the HadAM3 model, a 2.5%–97.5% range for climate sensitivity of 2.7–4.2 K if the CERES observations are correct. If the ERBE observations are correct, then they suggest a larger range, for HadAM3, of 2.8–5.6 K. Amplifying the CERES observational covariance estimate by a factor of 20 brings CERES and ERBE estimates into agreement. In this case the climate sensitivity range is 2.7–5.4 K. The results rule out, at the 2.5% level for HadAM3 and several different prior assumptions, climate sensitivities greater than 5.6 K.

A 75-year integration of a coupled atmosphere–ocean model is examined for tropical interannual variability. The atmospheric model has interactive cloud and a seasonal cycle. The fluxes of heat and salinity into the ocean component of the model are flux corrected. The model has tropical variability that is qualitatively similar to that of the observed El Nino/Southern Oscillation (ENSO). The maximum amplitude of the model Nino3 signal is approximately half that observed and the modeled ENSO timescale is greater than that observed. In the first 50 years of the integration the model has eight warm events. Each event is one of two types: one characterized by a standing SST anomaly in the central and eastern Pacific and the other by a westward propagating sea surface temperature anomaly. The majority of the model warm events are of the first type. The first type of event is triggered by the eastward propagation of Kelvin waves across the Pacific, and the second by westward propagation of warm temperat...