Large Model Ensembles Research Articles

Current and future wind energy resources in the North Sea according to CMIP6

Abstract. We explore the changes in wind energy resources in northern Europe using output from historical to mid-21st century CMIP6 simulations and the high-emission SSP5-8.5 scenario. This study improves upon many assumptions made in the past. First, we interpolate the winds to hub height using model-level raw data; second, we use a large ensemble of CMIP6 models; third, we consider the possible wake effects on the annual energy production of a large wind farm cluster proposed for the North Sea. The common practice of extrapolating 10 m wind speeds to turbine height using the power law with a constant shear exponent is often a poor approximation of the actual turbine-height wind speed. This approximation can exaggerate the future changes in wind resources and ignore possible surface roughness and atmospheric stability changes. The evaluation of the wind climatologies in the CMIP6 models over the North Sea for the historical period shows good correspondence with measurements from tall masts and three reanalysis data points for 16 of the 18 models. Some of the models run at relatively high spatial resolution are as good as the reanalyses at representing the wind climate in this region. Our results show that annual mean wind speed and wind resources in northern Europe are not particularly affected by climate change in 2031–2050 relative to 1995–2014, according to a subset of 16 models in the CMIP6 collection. However, the seasonal distribution of these resources is significantly altered. Most models agree on reductions in the future wind in summer in a band that extends from the British Isles to the Baltic Sea and on increases in winter in the southern Baltic Sea. The energy production calculations show that summer energy production in a planned large wind farm cluster in the North Sea could be reduced by a median of 6.9 % during 2031–2050 when taking into account the wind farm wakes (that accounts for −0.7 %) and the changes in air density (that account for −0.9 %).

What Controls the Interannual Variability of the Boreal Winter Atmospheric River Activities over the Northern Hemisphere?

Abstract Interannual variability of the winter atmospheric river (AR) activities over the Northern Hemisphere is investigated. The leading modes of AR variability over the North Pacific and North Atlantic are first identified and characterized. Over the Pacific, the first mode is characterized by a dipole structure with enhanced AR frequency along the AR peak region at about 30°N and reduced AR frequency farther north. The second mode exhibits a tripole structure with a narrow band of positive AR anomalies at about 30°N and sandwiched by negative anomalies. Over the Atlantic, the first mode exhibits an equatorward shift of the ARs with positive anomalies and negative anomalies located on the equatorward and poleward side of the AR peak region at about 40°N, respectively. The second mode is associated with the strengthening and eastward extension of the AR peak region, which is sandwiched by negative anomalies. A large ensemble of atmospheric global climate models from phase 6 of the Coupled Model Intercomparison Project (CMIP6), which shows high skill in simulating these modes, is then used to quantify the roles of sea surface temperature (SST) forcing versus internal atmospheric variability in driving the formation of these modes. Results show that SST forcing explains about half of the variance for the Pacific leading modes, while that number drops to about a quarter for the Atlantic leading modes, suggesting higher predictability for the Pacific AR variability. An additional ensemble driven only by observed tropical SST is further utilized to demonstrate the more important role that tropical SST plays in controlling the Pacific AR variability while both tropical and extratropical SST exert comparable influences on the Atlantic AR variability.

Assessing Internal Variability of Global Mean Surface Temperature From Observational Data and Implications for Reaching Key Thresholds

AbstractObserved global mean surface temperature (GMST) combines a forced response with internal variability, and quantifying internal variability is important in assessing the reaching of key thresholds, such as the 1.5°C warming threshold in the Paris Agreement. This study uses observational data to estimate internal variability. Since the current period of warming began in the 1970s, the 10‐year mean of GMST has been very close to the 30‐year mean for the period it is centred in and can therefore be considered as a robust indicator of the recent state of the climate. The range between the 5th and 95th percentile of annual residuals of observed GMST is 0.319°C, substantially less than the corresponding range of 0.4°C–0.6°C in large model ensembles, implying that the first individual year above 1.5°C may occur later than indicated by climate models. The largest annual residuals are mostly associated with large‐amplitude El Niño‐Southern Oscillation (ENSO) events or major volcanic eruptions, with the relationship between cool years and La Niña more consistent than that between warm years and El Niño. The relationship between multi‐year GMST means for differing periods indicates that the probability that the 1.5°C threshold has been crossed (using the IPCC definition of the midpoint of the first 20‐year period above the threshold) exceeds 50% once the most recent observed 11‐year mean reaches 1.43°C.

Increasing precipitation whiplash in climate change hotspots

Throughout the world, the hydrologic cycle is projected to become more variable due to climate change, posing challenges in semi-arid regions with high water resource vulnerability. Precipitation whiplash results from hydrologic variability, and refers to interannual shifts between wet (⩾80th historical percentile) and dry (⩽20th historical percentile) years. Using five model large ensembles, we show that whiplash is projected to increase in frequency (25%–60%) and intensity (30%–100%) by 2100 across several semi-arid regions of the globe, including Western North America and the Mediterranean. These changes can be driven by increases in the frequency of wet years or dry years, or both, depending on the region. Moisture budget calculations in these regions illuminate the physical mechanisms behind increased whiplash. Thermodynamic changes generally dominate, with modulations by dynamics, evaporation, and eddies on regional or global scales. These findings highlight increasingly volatile hydrology in semi-arid regions as the 21st Century progresses.

Machine-learning-based downscaling of modelled climate change impacts on groundwater table depth

Abstract. There is an urgent demand for assessments of climate change impacts on the hydrological cycle at high spatial resolutions. In particular, the impacts on shallow groundwater levels, which can lead to both flooding and drought, have major implications for agriculture, adaptation, and urban planning. Predicting such hydrological impacts is typically performed using physically based hydrological models (HMs). However, such models are computationally expensive, especially at high spatial resolutions. This study is based on the Danish national groundwater model, set up as a distributed, integrated surface–subsurface model at a 500 m horizontal resolution. Recently, a version at a higher resolution of 100 m was created, amongst others, to better represent the uppermost groundwater table and to meet end-user demands for water management and climate adaptation. However, the increase in resolution of the hydrological model also increases computational bottleneck. To evaluate climate change impacts, a large ensemble of climate models was run with the 500 m hydrological model, while performing the same ensemble run with the 100 m resolution nationwide model was deemed infeasible. The desired outputs at the 100 m resolution were produced by developing a novel, hybrid downscaling method based on machine learning (ML). Hydrological models for five subcatchments, covering around 9 % of Denmark and selected to represent a range of hydrogeological settings, were run at 100 m resolutions with forcings from a reduced ensemble of climate models. Random forest (RF) algorithms were established using the simulated climate change impacts (future – present) on water table depth at 100 m resolution from those submodels as training data. The trained downscaling algorithms were then applied to create nationwide maps of climate-change-induced impacts on the shallow groundwater table at 100 m resolutions. These downscaled maps were successfully validated against results from a validation submodel at a 100 m resolution excluded from training the algorithms, and compared to the impact signals from the 500 m HM across Denmark. The suggested downscaling algorithm also opens for the spatial downscaling of other model outputs. It has the potential for further applications where, for example, computational limitations inhibit running distributed HMs at fine resolutions.

Constraining extreme precipitation projections using past precipitation variability

Projected changes of future precipitation extremes exhibit substantial uncertainties among climate models, posing grand challenges to climate actions and adaptation planning. Practical methods for narrowing the projection uncertainty remain elusive. Here, using large model ensembles, we show that the uncertainty in projections of future extratropical extreme precipitation is significantly correlated with the model representations of present-day precipitation variability. Models with weaker present-day precipitation variability tend to project larger increases in extreme precipitation occurrences under a given global warming increment. This relationship can be explained statistically using idealized distributions for precipitation. This emergent relationship provides a powerful constraint on future projections of extreme precipitation from observed present-day precipitation variability, which reduces projection uncertainty by 20–40% over extratropical regions. Because of the widespread impacts of extreme precipitation, this has not only provided useful insights into understanding uncertainties in current model projections, but is also expected to bring potential socio-economic benefits in climate change adaptation planning.

The Forced Response and Decadal Predictability of the North Atlantic Oscillation: Nonstationary and Fragile Skills

Abstract We investigate the forced response of the North Atlantic Oscillation (NAO)—calculated as the ensemble mean—in different large ensembles of climate models including simulations with historical forcings and initialized decadal hindcasts. The forced NAO in the CMIP6 historical ensemble correlates significantly with observations after 1970. However, the forced NAO shows an apparent nonstationarity with significant correlations to observations only in the period after 1970 and in the period before 1890. We demonstrate that such apparent nonstationarity can be due to chance even when models and observations are independent. For the period after 1970 the correlation to the observed NAO continues to increase while the amplitude of the forced signal continues to decrease—although both with some signs of saturation—when the ensemble size grows. This behavior can be explained by a simple statistical model assuming a very small signal-to-noise ratio in the models. We find only rather weak evidence that initialization improves the skill of the NAO on decadal time scales. The NAO in the historical ensembles including only natural forcings, well-mixed greenhouse gases, or anthropogenic aerosols show skill that is not significantly different from zero. The same holds for a large single-model ensemble. The skills of these ensembles, except for the well-mixed greenhouse gas ensemble, are also significantly different from the skill of the larger full historical ensemble even though their ensemble sizes are smaller. Taken together, our results challenge the possibility of useful NAO predictions on decadal time scales.

Systematic Climate Model Biases in the Large‐Scale Patterns of Recent Sea‐Surface Temperature and Sea‐Level Pressure Change

AbstractObserved surface temperature trends over recent decades are characterized by (a) intensified warming in the Indo‐Pacific Warm Pool and slight cooling in the eastern equatorial Pacific, consistent with Walker circulation strengthening, and (b) Southern Ocean cooling. In contrast, state‐of‐the‐art coupled climate models generally project enhanced warming in the eastern equatorial Pacific, Walker circulation weakening, and Southern Ocean warming. Here we investigate the ability of 16 climate model large ensembles to reproduce observed sea‐surface temperature and sea‐level pressure trends over 1979–2020 through a combination of externally forced climate change and internal variability. We find large‐scale differences between observed and modeled trends that are very unlikely (<5% probability) to occur due to internal variability as represented in models. Disparate trends in the ratio of Indo‐Pacific Warm Pool to tropical‐mean warming, which shows little multi‐decadal variability in models, hint that model biases in the response to historical forcing constitute part of the discrepancy.

Mid-Pliocene El Niño/Southern Oscillation suppressed by Pacific intertropical convergence zone shift

The El Niño/Southern Oscillation (ENSO), the dominant driver of year-to-year climate variability in the equatorial Pacific Ocean, impacts climate pattern across the globe. However, the response of the ENSO system to past and potential future temperature increases is not fully understood. Here we investigate ENSO variability in the warmer climate of the mid-Pliocene (~3.0–3.3 Ma), when surface temperatures were ~2–3 °C above modern values, in a large ensemble of climate models—the Pliocene Model Intercomparison Project. We show that the ensemble consistently suggests a weakening of ENSO variability, with a mean reduction of 25% (±16%). We further show that shifts in the equatorial Pacific mean state cannot fully explain these changes. Instead, ENSO was suppressed by a series of off-equatorial processes triggered by a northward displacement of the Pacific intertropical convergence zone: weakened convective feedback and intensified Southern Hemisphere circulation, which inhibit various processes that initiate ENSO. The connection between the climatological intertropical convergence zone position and ENSO we find in the past is expected to operate in our warming world with important ramifications for ENSO variability. Suppressed El Niño/Southern Oscillation variability during the mid-Pliocene Warm Period was caused mainly by a northward displacement of the intertropical convergence zone, according to an analysis of a large ensemble of climate model simulations.

Comparison of Climate Model Large Ensembles With Observations in the Arctic Using Simple Neural Networks

AbstractEvaluating historical simulations from global climate models (GCMs) remains an important exercise for better understanding future projections of climate change and variability in rapidly warming regions, such as the Arctic. As an alternative approach for comparing climate models and observations, we set up a machine learning classification task using a shallow artificial neural network (ANN). Specifically, we train an ANN on maps of annual mean near‐surface temperature in the Arctic from a multi‐model large ensemble archive in order to classify which GCM produced each temperature map. After training our ANN on data from the large ensembles, we input annual mean maps of Arctic temperature from observational reanalysis and sort the prediction output according to increasing values of the ANN's confidence for each GCM class. To attempt to understand how the ANN is classifying each temperature map with a GCM, we leverage a feature attribution method from explainable artificial intelligence. By comparing composites from the attribution method for every GCM classification, we find that the ANN is learning regional temperature patterns in the Arctic that are unique to each GCM relative to the multi‐model mean ensemble. In agreement with recent studies, we show that ANNs can be useful tools for extracting regional climate signals in GCMs and observations.

The Evolving Role of External Forcing in North Atlantic SST Variability over the Last Millennium

AbstractAtlantic multidecadal variability (AMV) impacts temperature, precipitation, and extreme events on both sides of the Atlantic Ocean basin. Previous studies with climate models have suggested that when external radiative forcing is held constant, the large-scale ocean and atmosphere circulation are associated with sea surface temperature (SST) anomalies that have similar characteristics to the observed AMV. However, there is an active debate as to whether these internal fluctuations driven by coupled atmosphere–ocean variability remain influential to the AMV on multidecadal time scales in our modern, anthropogenically forced climate. Here we provide evidence from multiple large ensembles of climate models, paleoreconstructions, and instrumental observations of a growing role for external forcing in the AMV. Prior to 1850, external forcing, primarily from volcanoes, explains about one-third of AMV variance. Between 1850 and 1950, there is a transitional period, where external forcing explains one-half of AMV variance, but volcanic forcing only accounts for about 10% of that. After 1950, external forcing explains three-quarters of AMV variance. That is, the role for external forcing in the AMV grows as the variations in external forcing grow, even if the forcing is from different sources. When forcing is relatively stable, as in earlier modeling studies, a higher percentage of AMV variations are internally generated.

Influences of atmospheric blocking on North American summer heatwaves in a changing climate: a comparison of two Canadian Earth system model large ensembles

As summer heatwaves have severe adverse impacts on human society and ecosystems, there is need to better understand their meteorological drivers and future projections under climate change. This study investigates the linkage between atmospheric blocking and summer (June–August) heatwaves over North America using two reanalysis datasets (ERA-Interim and NCEP-DOE-R2) and two large-ensembles of Canadian Earth System Models (CanESM2 and CanESM5) for the 1981–2010 baseline period as well as projected changes under high-emission scenarios out to 2071–2100. Compared to NCEP-DOE-R2, both ensembles underestimate summer blocking frequency in the north Pacific, Alaska, and western Canada (by − 37%), while CanESM2 ensemble also underestimates blocking frequency in central and eastern Canada (by − 36%). CanESM5 generally shows better performance than CanESM2 in its reproduction of blocking frequency over central and eastern Canada, which is consistent with its overall improvements in simulating large-scale climate patterns. The two ensembles, however, agree with the reanalyses in their blocking-heatwave linkages. Above-normal heatwave frequency occurs in the blocking core and its surroundings due to positive heat flux anomalies, while below-normal frequency occurs at remote locations on the eastern and/or southern flanks of the blocking core due to cold air temperature advection anomalies. Future projections in central Canada differ between the models, largely due to the significant under-representation of blocking frequency by CanESM2. However, the two ensembles generally project similar behavior between the baseline and future period for spatial distributions of blocking-heatwave linkages, indicating blocking will continue to play an important role in the development of summer heatwaves in the future.

Climate change increases risk of extreme rainfall following wildfire in the western United States.

Post-wildfire extreme rainfall events can have destructive impacts in the western United States. Using two climate model large ensembles, we assess the future risk of extreme fire weather events being followed by extreme rainfall in this region. By mid-21st century, in a high warming scenario (RCP8.5), we report large increases in the number of extreme fire weather events followed within 1 year by at least one extreme rainfall event. By 2100, the frequency of these compound events increases by 100% in California and 700% in the Pacific Northwest in the Community Earth System Model v1 Large Ensemble. We further project that more than 90% of extreme fire weather events in California, Colorado, and the Pacific Northwest will be followed by at least three spatially colocated extreme rainfall events within five years. Our results point to a future with substantially increased post-fire hydrologic risks across much of the western United States.

Precipitation trends determine future occurrences of compound hot\u2013dry events

Compound hot–dry events—co-occurring hot and dry extremes—frequently cause damages to human and natural systems, often exceeding separate impacts from heatwaves and droughts. Strong increases in the occurrence of these events are projected with warming, but associated uncertainties remain large and poorly understood. Here, using climate model large ensembles, we show that mean precipitation trends exclusively modulate the future occurrence of compound hot–dry events over land. This occurs because local warming will be large enough that future droughts will always coincide with at least moderately hot extremes, even in a 2 °C warmer world. By contrast, precipitation trends are often weak and equivocal in sign, depending on the model, region and internal climate variability. Therefore, constraining regional precipitation trends will also constrain future compound hot–dry events. These results help to assess future frequencies of other compound extremes characterized by strongly different trends in the drivers.

Upscaling and downscaling Monte Carlo ensembles with generative models

SUMMARYMonte Carlo methods are widespread in geophysics and have proved to be powerful in non-linear inverse problems. However, they are associated with significant practical challenges, including long calculation times, large output ensembles of Earth models, and difficulties in the appraisal of the results. This paper addresses some of these challenges using generative models, a family of tools that have recently attracted much attention in the machine learning literature. Generative models can, in principle, learn a probability distribution from a set of given samples and also provide a means for rapid generation of new samples which follow that approximated distribution. These two features make them well suited for application to the outputs of Monte Carlo algorithms. In particular, training a generative model on the posterior distribution of a Bayesian inference problem provides two main possibilities. First, the number of parameters in the generative model is much smaller than the number of values stored in the ensemble, leading to large compression rates. Secondly, once trained, the generative model can be used to draw any number of samples, thereby eliminating the dependence on an often large and unwieldy ensemble. These advantages pave new pathways for the use of Monte Carlo ensembles, including improved storage and communication of the results, enhanced calculation of numerical integrals, and the potential for convergence assessment of the Monte Carlo procedure. Here, these concepts are initially demonstrated using a simple synthetic example that scales into higher dimensions. They are then applied to a large ensemble of shear wave velocity models of the core–mantle boundary, recently produced in a Monte Carlo study. These examples demonstrate the effectiveness of using generative models to approximate posterior ensembles, and indicate directions to address various challenges in Monte Carlo inversion.

Increasing spatiotemporal proximity of heat and precipitation extremes in a warming world quantified by a large model ensemble

Increases in climate hazards and their impacts mark one of the major challenges of climate change. Situations in which hazards occur close enough to one another to result in amplified impacts, because systems are insufficiently resilient or because hazards themselves are made more severe, are of special concern. We consider projected changes in such compounding hazards using the Max Planck Institute Grand Ensemble under a moderate (RCP4.5) emissions scenario, which produces warming of about 2.25 °C between pre-industrial (1851–1880) and 2100. We find that extreme heat events occurring on three or more consecutive days increase in frequency by 100%–300%, and consecutive extreme precipitation events increase in most regions, nearly doubling for some. The chance of concurrent heat and drought leading to simultaneous maize failures in three or more breadbasket regions approximately doubles, while interannual wet-dry oscillations become at least 20% more likely across much of the subtropics. Our results highlight the importance of taking compounding climate extremes into account when looking at possible tipping points of socio-environmental systems.

Attributing Compound Events to Anthropogenic Climate Change

Abstract Extreme event attribution answers the question of whether and by how much anthropogenic climate change has contributed to the occurrence or magnitude of an extreme weather event. It is also used to link extreme event impacts to climate change. Impacts, however, are often related to multiple compounding climate drivers. Because extreme event attribution typically focuses on univariate assessments, these assessments might only provide a partial answer to the question of anthropogenic influence to a high-impact event. We present a theoretical extension to classical extreme event attribution for certain types of compound events. Based on synthetic data, we illustrate how the bivariate fraction of attributable risk (FAR) differs from the univariate FAR depending on the extremeness of the event as well as the trends in and dependence between the contributing variables. Overall, the bivariate FAR is similar in magnitude or smaller than the univariate FAR if the trend in the second variable is comparably weak and the dependence between both variables is moderate or high, a typical situation for temporally co-occurring heat waves and droughts. If both variables have similarly large trends or the dependence between both variables is weak, bivariate FARs are larger and are likely to provide a more adequate quantification of the anthropogenic influence. Using multiple climate model large ensembles, we apply the framework to two case studies, a recent sequence of hot and dry years in the Western Cape region of South Africa and two spatially co-occurring droughts in crop-producing regions in South Africa and Lesotho.

Fast Screening Tool for Assessing the Impact of Poro-Mechanics on Fractured Reservoirs Using Dual-Porosity Flow Diagnostics

Summary Accounting for poro-mechanical effects in full-field reservoir simulation studies and uncertainty quantification workflows is still limited, mainly because of their high computational cost. We introduce a new approach that couples hydrodynamics and poro-mechanics with dual-porosity flow diagnostics to analyze how poro-mechanics could affect reservoir dynamics in naturally fractured reservoirs without significantly increasing computational overhead. Our new poro-mechanically informed dual-porosity flow diagnostics account for steady-state and single-phase flow conditions in the fractured medium while the fracture-matrix fluid exchange is approximated using a physics-based transfer rate coefficient, which models two-phase flow using an analytical solution for spontaneous imbibition or gravity drainage. The deformation of the system is described by the dual-porosity poro-elastic theory, which is based on mixture theory and micromechanics to compute the effective stresses and strains of the rock matrix and fractures. The solutions to the fluid flow and rock deformation equations are coupled sequentially. The governing equations for fluid flow are discretized using a finite-volume method with two-point flux-approximation while the governing equations for poro-mechanics are discretized using the virtual element method. The solution of the coupled system considers stress-dependent permeabilities for fractures and matrix. Our framework is implemented in the open-source MATLAB Reservoir Simulation Toolbox (MRST). We present a case study using a fractured carbonate reservoir analog to illustrate the integration of poro-mechanics within the dual-porosity flow diagnostics framework. The extended flow diagnostics calculations enable us to quickly screen how the dynamics in fractured reservoirs (e.g., reservoir connectivity, sweep efficiency, and fracture-matrix transfer rates) are affected by the complex interactions between poro-mechanics and fluid flow where changes in pore pressure and effective stress modify petrophysical properties and hence affect reservoir dynamics. Because of the steady-state nature of the calculations and the effective coupling strategy, these calculations do not incur significant computational overheads. They provide an efficient complement to traditional reservoir simulation and uncertainty quantification workflows because they enable us to assess a broader range of reservoir uncertainties (e.g., geological, petrophysical, and hydromechanical uncertainties). The capability of studying a much broader range of uncertainties allows the comparison and ranking from a large ensemble of reservoir models and select individual candidates for more detailed full-physics reservoir simulation studies without compromising on assessing the range of uncertainties inherent to fractured reservoirs.

Projections of indices of daily temperature and precipitation based on bias-adjusted CORDEX-Africa regional climate model simulations

We present a dataset of daily, bias-adjusted temperature and precipitation projections for continental Africa based on a large ensemble of regional climate model simulations, which can be useful for climate change impact studies in several sectors. We provide guidance on the benefits and caveats of using the dataset by investigating the effect of bias-adjustment on impact-relevant indices (both their future absolute value and change). Extreme threshold-based temperature indices show large differences between original and bias-adjusted values at the end of the century due to the general underestimation of temperature in the present climate. These results indicate that when biases are accounted for, projected risks of extreme temperature-related hazards are higher than previously found, with possible consequences for the planning of adaptation measures. Bias-adjusted results for precipitation indices are usually consistent with the original results, with the median change preserved for most regions and indices. The interquartile and full range of the original model ensemble is usually well preserved by bias-adjustment, with the exception of maximum daily precipitation, whose range is usually greatly reduced by the bias-adjustment. This is due to the poor simulation and extremely large model range for this index over the reference period; when the bias is reduced, most models converge in projecting a similar change. Finally, we provide a methodology to select a small subset of simulations that preserves the overall uncertainty in the future projections of the large model ensemble. This result can be useful in practical applications when process-based impact models are too expensive to be run with the full ensemble of model simulations.

Canadian Large Ensembles Adjusted Dataset version 1 (CanLEADv1): Multivariate bias‐corrected climate model outputs for terrestrial modelling and attribution studies in North America

AbstractThe Canadian Large Ensembles Adjusted Dataset version 1 (CanLEADv1) contains 50‐member ensembles of bias‐adjusted near‐surface global and regional climate model variables on a 0.5° grid over North America for historical and future scenarios (1950–2100). Canadian Earth System Model Large Ensembles (CanESM2 LE) and Canadian Regional Climate Model Large Ensemble (CanRCM4 LE) datasets are bias‐corrected using a multivariate quantile‐mapping algorithm for statistical consistency – in terms of marginal distributions and multivariate dependence structure – with two observationally constrained historical meteorological forcing datasets. For each observational dataset, bias‐adjusted variables are provided for two sets of 50‐member initial‐condition CanESM2 ensembles (historical plus RCP8.5 scenarios, 1950–2005 and 2006–2100, respectively; and historicalNAT scenario, 1950–2020, which excludes anthropogenic forcings), and one 50‐member CanRCM4 ensemble (historical plus RCP8.5). The archive includes daily minimum temperature, maximum temperature, precipitation, relative humidity, surface pressure, wind speed, incoming shortwave radiation and incoming longwave radiation. Intended uses include hydrological and land surface impact modelling, as well as related event attribution studies.