Search

We present an ensemble of last glacial inception (LGI) simulations for the Northern Hemisphere that captures a significant fraction of inferred ice volume changes within proxy uncertainties. This ensemble was performed with LCice 1.0, a coupled ice sheet and climate model, varying parameters of both climate and ice sheet components, as well as the coupling between them. Certain characteristics of the spatiotemporal pattern of ice growth and subsequent retreat in both North America (NA) and Eurasia (EA) are sensitive to parameter changes while others are not. We find that the initial inception of ice over NA and EA is best characterized by the nucleation of ice at high-latitude and high-elevation sites. Subsequent spreading and merger along with large-scale conversion of snowfields dominate in different sectors. The latter plays an important role in the merging of eastern and western ice regions in NA. The inception peak ice volume in the ensemble occurs approximately at 111 ka and therefore lags the summer 60∘ N insolation minimum by more than 3 kyr. Ice volumes consistently peak earlier over EA than NA. The inception peak in North America is characterized by a merged Laurentide and Cordilleran ice sheet, with the Davis Strait covered in ice in ∼80 % of simulations. Ice also bridges Greenland and Iceland in all runs by 114 ka and therefore blocks the Denmark Strait. This latter feature would thereby divert the East Greenland Current and Denmark Strait overflow with a potentially significant impact on ocean circulation. The Eurasian ice sheet at its inception peak varies across ensemble runs between a continuous ice sheet and multiple smaller ice caps. In both continents, the colder high latitudes (i.e. Ellesmere and Svalbard) tend to grow ice through the entire simulation (to 102 ka), while lower latitudes lose ice after ∼110 ka. We find temperature decreases over the initial phases of the inception lead to the expansion of NA ice sheet area and that subsequent precipitation increases contribute to its thickening. EA ice sheet area also expands with decreasing temperatures, but sea ice limits any increases in precipitation, leading to an earlier retreat away from the EA maximum ice sheet volume. We also examine the extent to which the capture of both LGI ice growth and retreat constrains the coupled ice–climate model sensitivity to changing atmospheric pCO2. The 55-member sub-ensemble that meets our criteria for “acceptable” ice growth and retreat has an equilibrium climate sensitivity lower bound that is 0.3 ∘C higher than that of the full ensemble. This suggests some potential value of fully coupled ice–climate modelling of the last glacial inception to constrain future climate change.

Projections of anthropogenic climate change suggest possible surface temperature increases similar to those during past major shifts of the mean climate like the Last Deglaciation. Such shifts do not only affect the mean but rather the full probability distributions of climatic variables such as temperature and precipitation. Changes to their distributions can thus be expected for the future as well and need to be constrained. To this end, we examine transient simulations of the Last Deglaciation from a hierarchy of climate models, ranging from an energy balance model to state-of-the-art Earth System Models. Besides the mean, we use the higher moments of variability – variance, skewness, and kurtosis – to characterize changes of the distribution. The analysis covers annual to millennial timescales and examines how patterns vary with timescale and region in response to warming. Furthermore, we evaluate how the changes of the distributions affect the occurrence of extremes. To analyze the influence of forcings on the distributions, we compare the patterns of the fully-forced simulations to those in sensitivity experiments that isolate the effects of individual forcings. In particular, the effect of volcanism is examined across the hierarchy, as well as changes in ice cover, freshwater input, CO2, and orbit. While large-scale global patterns can be found, there are significant regional differences as well as differences between simulations, relating for example to differences in the modelling of ice cover changes and freshwater input. Finally, we investigate whether climate model projections show the same trends with respect to the change in moments as those found in the deglacial simulations and thus whether the patterns found might hold for future climate.

An increasing number of climate model simulations is becoming available for the transition from the Last Glacial Maximum to the Holocene. Assessing the simulations’ reliability requires benchmarking against environmental proxy records. To date, no established method exists to compare these two data sources in space and time over a period with changing background conditions. Here, we develop a new algorithm to rank simulations according to their deviation from reconstructed magnitudes and temporal patterns of orbital- as well as millennial-scale temperature variations. The use of proxy forward modeling avoids the need to reconstruct gridded or regional mean temperatures from sparse and uncertain proxy data. First, we test the reliability and robustness of our algorithm in idealized experiments with prescribed deglacial temperature histories. We quantify the influence of limited temporal resolution, chronological uncertainties, and non-climatic processes by constructing noisy pseudo-proxies. While model-data comparison results become less reliable with increasing uncertainties, we find that the algorithm discriminates well between simulations under realistic non-climatic noise levels. To obtain reliable and robust rankings, we advise spatial averaging of the results for individual proxy records. Second, we demonstrate our method by quantifying the deviations between an ensemble of transient deglacial simulations and a global compilation of sea surface temperature reconstructions. The ranking of the simulations differs substantially between the considered regions and timescales. We attribute this diversity in the rankings to more regionally confined temperature variations in reconstructions than in simulations, which could be the result of uncertainties in boundary conditions, shortcomings in models, or regionally varying characteristics of reconstructions such as recording seasons and depths. Future work towards disentangling these potential reasons can leverage the flexible design of our algorithm and its demonstrated ability to identify varying levels of model-data agreement.

Abrupt climate shifts of large amplitudes were common features of the Earth's climate as it transitioned into and out of the last full glacial state approximately 20 000 years ago, but their causes are not yet established. Midlatitude atmospheric dynamics may have played an important role in these climate variations through their effects on heat and precipitation distributions, sea ice extent, and wind-driven ocean circulation patterns. This study characterizes deglacial winter wind changes over the North Atlantic (NAtl) in a suite of transient deglacial simulations using the PlaSim Earth system model (run at T42 resolution) and the TraCE-21ka (T31) simulation. Though driven with yearly updates in surface elevation, we detect multiple instances of NAtl jet transitions in the PlaSim simulations that occur within 10 simulation years and a sensitivity of the jet to background climate conditions. Thus, we suggest that changes to the NAtl jet may play an important role in abrupt glacial climate changes. We identify two types of simulated wind changes over the last deglaciation. Firstly, the latitude of the NAtl eddy-driven jet shifts northward over the deglaciation in a sequence of distinct steps. Secondly, the variability in the NAtl jet gradually shifts from a Last Glacial Maximum (LGM) state with a strongly preferred jet latitude and a restricted latitudinal range to one with no single preferred latitude and a range that is at least 11∘ broader. These changes can significantly affect ocean circulation. Changes to the position of the NAtl jet alter the location of the wind forcing driving oceanic surface gyres and the limits of sea ice extent, whereas a shift to a more variable jet reduces the effectiveness of the wind forcing at driving surface ocean transports. The processes controlling these two types of changes differ on the upstream and downstream ends of the NAtl eddy-driven jet. On the upstream side over eastern North America, the elevated ice sheet margin acts as a barrier to the winds in both the PlaSim simulations and the TraCE-21ka experiment. This constrains both the position and the latitudinal variability in the jet at LGM, so the jet shifts in sync with ice sheet margin changes. In contrast, the downstream side over the eastern NAtl is more sensitive to the thermal state of the background climate. Our results suggest that the presence of an elevated ice sheet margin in the south-eastern sector of the North American ice complex strongly constrains the deglacial position of the jet over eastern North America and the western North Atlantic as well as its variability.

Freshwater, in the form of glacial runoff, is hypothesized to play a critical role in centennial to millennial scale climate variability, eg. the Younger Dryas and Dansgaard Oeschger events. Freshwater injection, or hosing, model experiments demonstrate that freshwater has the capability to generate abrupt climate transitions. However, in an attempt to mitigate the inability of most models to resolve the smaller-scale features relevant to freshwater transport (such as boundary currents and mesoscale eddies), these hosing experiments commonly apply the entirety of the freshwater directly to the regions of deepwater formation (DWF). Our results indicate that this can inflate the freshwater signal in those regions by as much as four times. We propose a novel method of freshwater injection for such low-resolution models that spatially distributes the freshwater in accord with the results of eddy-permitting modelling. Furthermore, this “freshwater fingerprint” method not only impacts the timing of simulated climate transitions but also can allow us to evaluate how much we are overestimating the effects of freshwater when injected directly into sites of DWF. The freshwater fingerprints we develop are based on a suite of freshwater injection experiments performed using an eddy permitting Younger Dryas configuration of the MITGCM. Freshwater injection locations include the Mackenzie River, Gulf of St. Lawrence, Gulf of Mexico and a location off the coast of Norway, with flux amounts bounded by glacial reconstructions. These simulations indicate that freshwater from the Mackenzie River and Fennoscandia have the largest impact on salinity in most of the conventional sites of DWF (GIN and Labrador Seas, and in these simulations, predominantly the northern North Atlantic due to extensive sea ice), while freshwater from the Gulf of St. Lawrence is effective at freshening only the northern North Atlantic. The Gulf of Mexico has little impact on any DWF region we consider, mostly because the lower but continual flux in our simulations does not allow freshwater to penetrate northward past the Gulf Stream. The dilution of the freshwater signal as it is transported from the site of injection to the DWF zones leads to a reduction in the effective freshwater forcing, making hosing directly over DWF zones even with realistic freshwater amounts unrealistic. Thus, we construct freshwater fingerprints from these simulations by extracting the freshwater anomaly spatial pattern averaged over the last 5 simulation years, vertically integrating the field and normalizing it.The freshwater fingerprint is then implemented in the COSMOS Earth Systems Model, which is run at resolutions typical for paleoclimate simulations (non-eddy permitting). Initial results show that freshwater from the Mackenzie River using our fingerprint method leads to a more gradual cooling than if the meltwater is released directly over the hosing region (50-70N). We conclude that hosing over DWF zones, even with realistic freshwater amounts, produces an unrealistically large climate response. Additional results for the remaining injection locations and with the fingerprint implemented in a simpler climate model will be presented.

Much about the response of temperature variability to a change in the climate's mean state, as the one projected for the current century, remains uncertain. These uncertainties include spatiotemporal patterns, the magnitude, and, in some cases, even the sign. For the last Deglaciation, - the last change in global mean temperature of a similar degree to that expected in projections - variability analyses of climate model simulations and temperature proxies produce conflicting results. Here, we build a hierarchy of transient simulations covering the period since the Last Glacial Maximum about 26k years ago. We include a range of climate models, from conceptual to complex Earth System Models. The simulations cover a variety of temporal and spatial resolutions, parameterizations, and modeled processes. For annual to multi-millennial temporal as well as regional to global spatial scales, we compare variability patterns and power spectra and analyze how they relate to model properties and the background state of Earth's climate. This allows for the examination of regional temperature differences between low, middle, and high latitudes and at locations of available paleoclimate proxy records. For sets of sensitivity experiments, we investigate effects of changes to ice sheets, sea ice, and in volcanic, solar, greenhouse, and orbital forcing on modeled climate variability. Thus, our analysis provides insights into when and how models disagree with each other and with proxies, and what differences arise due to specific models, simulation setups, and boundary conditions. Based on these results, we can then gauge the degree of complexity which is required to reproduce past temperature variability and predict its changes in the future.

One of the main contributors to palaeoclimate variability on millennial timescales is understood to be Dansgaard-Oeschger (D-O) cycles. Our awareness of these phenomena arises primarily from quasi-periodic, abrupt transitions of large magnitude detected in δ18O records from Greenland ice cores (e.g. Dansgaard et al, 1982; Johnsen et al, 1992), although there is evidence of similar variability in other archives and regions. D-O cycles have plenty to capture the imagination: the strength and rapidity of climate changes over Greenland, their regularity throughout MIS3 (~60 to 30 thousand years before present) and occurrence during the last deglaciation contrasting with their relative absence during the Last Glacial Maximum and Holocene, their opposed characteristics in Greenland and Antarctica, and that different models require different boundary conditions to reproduce this phenomena, if they can reproduce it at all. This talk characterises D-Olike cycles in two different models: Planet Simulator (PlaSim, an Earth System Model with simplified atmospheric physics, thermodynamic sea ice, and simplified ocean dynamics), and COSMOS (a CMIP3-era ESM). We identify four phases to D-O cycles and commonalities and differences in their representations in these models. Finally, we examine which phases of this type of variability continue to contribute to climate variability today and what that looks like.

The global hydrological cycle is of crucial importance for life on Earth. Hence, it is a focus of both future climate projections and paleoclimate modeling. The latter typically requires long integrations or large ensembles of simulations, and therefore models of reduced complexity are needed to reduce the computational cost. Here, we study the hydrological cycle of the the Planet Simulator (PlaSim) [1], a general circulation model (GCM) of intermediate complexity, which includes evaporation, precipitation, soil hydrology, and river advection.Using published parameter configurations for T21 resolution [2, 3], PlaSim strongly underestimates precipitation in the mid-latitudes as well as global atmospheric water compared to ERA5 reanalysis data [4]. However, the tuning of PlaSim has been limited to optimizing atmospheric temperatures and net radiative fluxes so far [3].Here, we present a different approach by tuning the model’s atmospheric energy balance and water budget simultaneously. We argue for the use of the globally averaged mean absolute error (MAE) for 2 m temperature, net radiation, and evaporation in the objective function. To select relevant model parameters, especially with respect to radiation and the hydrological cycle, we perform a sensitivity analysis and evaluate the feature importance using a Random Forest regressor. An optimal set of parameters is obtained via Bayesian optimization.Using the optimized set of parameters, the mean absolute error of temperature and cloud cover is reduced on most model levels, and mid-latitude precipitation patterns are improved. In addition to annual zonal-mean patterns, we examine the agreement with the seasonal cycle and discuss regions in which the bias remains considerable, such as the monsoon region over the Pacific.We discuss the robustness of this tuning with regards to resolution (T21, T31, and T42), and compare the atmosphere-only results to simulations with a mixed-layer ocean. Finally, we provide an outlook on the applicability of our parametrization to climate states other than present-day conditions.[1] K. Fraedrich et al., Meteorol. Z. 14, 299–304 (2005)[2] F. Lunkeit et al., Planet Simulator User’s Guide Version 16.0 (University of Hamburg, 2016)[3] G. Lyu et al., J. Adv. Model. Earth Syst. 10, 207–222 (2018)[4] H. Hersbach et al., Q. J. R. Meteorol. Soc. 146, 1999–2049 (2020)

Freshwater, in the form of glacial runoff, is hypothesized to play a critical role in centennial to millennial scale climate variability such as the Younger Dryas and Dansgaard-Oeschger Events. Indeed, freshwater injection/hosing experiments with climate models have long shown that freshwater has the capability of generating such abrupt climate transitions. However, the relationship between freshwater and abrupt climate transitions is not straightforward. Large-scale glacial runoff events, such as Meltwater Pulse 1A, are not always temporally proximal to subsequent large-scale cooling. As well, the typical design of hosing experiments tends to artificially amplify the climate response. This study explores the impact that limitations in the representation of runoff in conventional hosing simulations has on our understanding of this relationship and addresses the more fundamental question of where coastally released freshwater is transported when it reaches the ocean. We focus particularly on the prior use of excessive freshwater volumes (often by a factor of 5) and present-day (rather than paleo) ocean gateways, as well as the injection of freshwater directly over sites of deep-water formation (DWF) rather than at runoff locations. We track the routing of glaciologically-constrained freshwater volumes from four different plausible injection locations in a suite of eddy-permitting glacial ocean simulations using MITGCM under both open and closed Bering Strait conditions. Restricting freshwater forcing values to realistic ranges results in less spreading of freshwater across the North Atlantic and indicates that the response of DWF depends strongly on the geographical location of meltwater input. In particular, freshwater released into the Gulf of Mexico has little impact on DWF regions as a result of turbulent mixing by the Gulf Stream. In contrast, freshwater released from the Eurasian Ice sheet or initially into the Arctic is found to have the largest impact on DWF in the North Atlantic and GIN seas. Additional experiments show that when the Bering Strait is open, much like present-day, the Mackenzie River source exhibits twice as much freshening of the Labrador sea as a closed Bering Strait. Finally, our results illustrate that applying a freshwater hosing directly into the North Atlantic with even realistic freshwater amounts still over-estimates the effect of terrestrial runoff on ocean circulation.

The increasing number of Earth system model simulations that try to simulate the climate during the last deglaciation (ca 20 to 10 thousand years ago) creates a demand for benchmarking against environmental proxy records synthesized for the same time period. Comparing these two data sources over a period with changing background conditions requires new methods for model-data comparison that incorporate multiple types and sources of uncertainty. Natural archives of past reality are distributed sparsely and non-uniformly in space and time. Signals that can be obtained are in addition perturbed by uncertainties related to dating, the relationship between the proxy sensor and environmental fields, the archive build-up, and measurement. On the other hand, paleoclimate simulations are four-dimensional, complete, and physically consistent representations of the climate. However, they are subject to errors due to model inadequacies and sensitivity to the forcing protocol, and will not reproduce any particular history of unforced variability. We present a method for probabilistic, multivariate quantification of the deviation between paleo-data and paleoclimate simulations that draws on the strengths of both sources of information and accounts for the aforementioned uncertainties. We compare the shape and magnitude of orbital- and millennial-scale temperature fluctuations during the last deglaciation and compute metrics of regional and global model-data mismatches. We test our algorithm with an ensemble of published simulations of the deglaciation and simulations from the ongoing PalMod project, which aims at the simulation of the last glacial cycle with comprehensive Earth system models. These are evaluated against a compilation of temperature reconstructions from multiple archives. Our work aims for a standardized model-data comparison workflow that will be used in PalMod. This workflow can be extended subsequently with additional proxy data, new simulations, and improved representations of proxy uncertainties.

This archive contains a high variance subset of last glacial inception simulations from LCice (coupled ice and climate model). Details in contained tableOfRuns.org and associated journal submission under review: Bahadory, T., Tarasov, L., and Andres, H.: The phase space of last glacial inception for the Northern Hemisphere from coupled ice and climate modelling, Clim. Past, https://doi.org/10.5194/cp-2020-1, in review, 2020. Contact: lev@mun.ca, {"references":["Bahadory, T., Tarasov, L., and Andres, H.: The phase space of last glacial inception for the Northern Hemisphere from coupled ice and climate modelling, Clim. Past, https://doi.org/10.5194/cp-2020-1, in review, 2020"]}

We present an ensemble of Last Glacial Inception (LGI) simulations for the Northern Hemisphere that largely captures inferred ice volume changes within proxy uncertainties. This ensemble was performed with LCice 1.0, a coupled ice sheet and climate model, varying parameters of both climate and ice sheet components, as well as the coupling between them. Certain characteristics of the spatio-temporal pattern of ice growth and subsequent retreat in both North America (NA) and Eurasia (EA) are sensitive to parameter changes, especially with respect to regional rates of ice growth and retreat. We find that the initial inception of ice over NA and EA is best characterized by the nucleation of ice at high latitude and high elevation sites. Subsequent spreading and merger along with large-scale conversion of snow fields dominate in different sectors. The latter plays an important role in the merging of eastern and western ice regions in NA. The inception peak ice volume in the ensemble occurs approximately at 111 ka and therefore lags the summer 60° N insolation minimum by more than 3 kyr. Ice volumes consistently peak earlier over EA than NA. The inception peak in North America is characterized by a merged Laurentide and Cordilleran ice sheet, with Davis Strait covered in ice in 80 % of simulations. Ice also bridges Greenland and Iceland in all runs by 114 ka and therefore blocks Denmark Strait. This latter feature would thereby divert the East Greenland Current and Denmark Strait overflow and thereby potentially have a significant impact on ocean circulation. The Eurasian ice sheet at its inception peak varies across ensemble runs between a continuous ice sheet to multiple smaller ice caps. In both continents, the colder high latitudes (Ellsmere and Svalbard) tend to grow ice through the entire simulation (to 102 ka), while lower latitudes lose ice after 110 ka. We find temperature decreases over the initial phases of the inception lead to the expansion of NA ice sheet area, and that subsequent precipitation increases contribute to its thickening. EA ice sheet area also expands with decreasing temperatures, but sea ice limits any increases in precipitation, leading to an earlier retreat away from the EA maximum ice sheet volume. We also examine the extent to which the capture of both LGI ice growth and retreat constrains the coupled ice/climate model sensitivity to changing atmospheric pCO2. For a standard transient climate response experiment (1 % increase in pCO2 until doubled), warming ranges between 0.6–2.0 °C for our initial set of 500 simulations without LGI constraint. The warming is reduced to 0.7–1.4 °C for the 55 member ensemble that captures both LGI ice growth and retreat. This therefore underlines the potential value of fully coupled ice/climate modelling of last glacial inception to constrain future climate change.

Understanding changes in the climate of the late Pleistocene and the Holocene has long been a research topic. Studies rely on different sources of information, ranging from terrestrial and marine archives to a hierarchy of climate modeling activities. In contrast to the climate of the last millennium, novel approaches are necessary to bridge the different temporal and spatial representations of the various archives and the climate models, and to achieve a robust understanding of climate variability and climate processes on centennial-to-millennial time scales.

Comparing climate proxy and simulation data is fraught with challenges: age and calibration uncertainties in climate proxies, missing or incomplete processes and uncertain boundary conditions for climate models, and differences between gridded and site data are just a few examples. For the climate of the Common Era, multiple initiatives have already addressed these issues (e.g. the PAGES 2k Network regional working groups). On transient timescales beyond the late Holocene, there have been only a few integrated activities. Comparisons on these longer time scales involve large-scale changes in climate states without an equivalent during the Holocene. As such, they require methods that address both the amplitude and timing of background climate changes and account for additional processes. For example, comprehensive Earth System Models need to include changes in ice sheets and related ocean circulation changes during deglaciation. Likewise, proxy data for this period, such as lake or marine sediments, are generally less well replicated than their late Holocene counterparts (e.g. tree rings and historical documents), resulting in more uncertain climate signals (Laepple et al. 2017).