The PMIP4 contribution to CMIP6 - Part Two interglacials, scientific objective and experimental design for Holocene and Last Interglacial simulations. Geoscientific Model

. Two interglacial epochs are included in the suite of Paleoclimate Modeling Intercomparison Project (PMIP4) simulations in the Coupled Model Intercomparison Project (CMIP6). The experimental protocols for simulations of the mid-Holocene ( midHolocene , 6000 years before present) and the Last Interglacial ( lig127k , 127 000 years before present) are described here. These equilibrium simulations are designed to examine the impact of changes in orbital forcing at times when atmospheric greenhouse gas levels were similar to those of the preindustrial period and the continental conﬁgurations were almost identical to modern ones. These simulations test our understanding of the interplay between radiative forcing and atmospheric circulation, and the connections among large-scale and regional climate changes giving rise to phenomena such as land–sea contrast and high-latitude ampliﬁcation in temperature changes, and responses of the monsoons, as compared to today. They also provide an opportunity, through carefully designed additional sensitivity experiments, to quantify the strength of atmosphere, ocean, cryosphere, and land-surface feedbacks. Sensitivity experiments are proposed to investigate the role of freshwater forcing in triggering abrupt climate changes within interglacial epochs. These feedback experiments naturally lead to a focus on climate evolution during interglacial periods, which will be examined through transient experiments. Analyses of the sensitivity simulations will also focus on interactions between extratropical and tropical circulation, and the relationship between changes in mean climate state and climate variability on annual to multi-decadal timescales. The com-parative abundance of paleoenvironmental data and of quantitative climate reconstructions for the Holocene


Introduction
The modeling of paleoclimate, using physically based tools, has long been used to understand and explain past environmental and climate changes (Kutzbach and Street-Perrott, 1985), and is increasingly seen as a strong out-of-sample test of the models that are used for the projection of future climate changes (Braconnot et al., 2012;Harrison et al., 2014Harrison et al., , 2015Schmidt et al., 2014). The Paleoclimate Modelling Intercomparison Project (PMIP) has served to coordinate paleoclimate experiments and data-model comparisons for several decades (Braconnot et al., 2007a(Braconnot et al., , b, 2012Joussaume and Taylor, 1995;Joussaume et al., 1999), and now spearheads the paleoclimate contribution to the current phase of the Coupled Model Intercomparison Project (CMIP6, Eyring et al., 2016). This paper is part of a suite of five documenting the PMIP4 contributions to CMIP6.  provide an overview of the five selected time periods and the experiments. More specific information is given in the contributions for the last millennium (past1000) by Jungclaus et al. (2017), for the last glacial maximum (lgm) by Kageyama et al. (2017), for the mid-Pliocene warm period (midPliocene-eoi400) by Haywood et al. (2016), and the present paper the mid-Holocene (midHolocene) and the previous interglacial (lig127k). PMIP4 has adopted the CMIP6 categorization where the highest-priority experiments are classified as Tier 1, whereas additional sensitivity experiments or dedicated studies are Tier 2 or Tier 3. The standard experiments for the five periods are all ranked Tier 1. Tier 2 and 3 experiments absolutely require the corresponding Tier 1 experiment for their analysis, so the groups must perform the Tier 1 experiment first. Modeling groups are not obliged to run all PMIP4-CMIP6 experiments. It is mandatory, however, for all participating groups to run at least one of the experiments that were run in previous phases of PMIP (i.e., midHolocene or lgm).
The two experiments described here focus on comparing the most recent interglacial epochs and specifically the current interglacial (the Holocene) and the previous interglacial (the Last Interglacial, LIG) periods (Fig. 1). These two experiments are of interest because they examine the response of the climate system to relatively simple changes in forcing compared to present. The main difference in forcing from present was in the latitudinal and seasonal distribution of incoming solar radiation (insolation) caused by known changes in the Earth's orbit; greenhouse gas (GHG) concentrations were similar to those of the preindustrial period and the continental configurations were also very similar to modern ones. Differences in orbital configuration between the two interglacial periods (Berger, 1978) mean that the insolation changes are stronger in the LIG than in the Holocene, but the observational basis for evaluating model simulations is more extensive in the Holocene than the LIG because of preserva-  (Siegenthaler et al., 2005;Schneider et al., 2013;Monnin et al., 2001Monnin et al., , 2004; (e) and (f): atmospheric CH 4 concentration (Loulergue et al., 2008); (g) and (h): Antarctic surface air temperature reconstruction (Jouzel et al., 2007); (i) and (j): Greenland ice δ 18 O: from the NEEM ice core (NEEM Community Members, 2013) in dark grey and from the NGRIP ice core (NorthGRIP Community Members, 2004) in black. Note that NEEM ice δ 18 O is shifted by +2 ‰. (k) LIG maximum global mean sea level (GMSL) relative to the present day; uncertainties in the amplitude are indicated by the shading (see Dutton et al., 2015a, for a review). Time of maximum varies between reconstructions. No significant sea-level variations are reported throughout the Holocene compared to the present day. NGRIP ice δ 18 O is displayed on the GICC05 annual layer-counted timescale (Svensson et al., 2008) over the last 20 ka and on the AICC2012 chronology Veres et al., 2013) across the 119-110 ka time interval. All other ice core records are displayed on the AICC2012 chronology, which is coherent, by construction with the GICC05 timescale over the last 60 ka Veres et al., 2013). Vertical yellow lines indicate 127 and 6 ka, the time intervals chosen to run the coordinated PMIP4-CMIP6 lig127k and midHolocene simulations. tion issues. Taken together, these two interglacial periods are good test cases of our mechanistic understanding of the interplay between radiative forcing and atmospheric circulation, and opportunities to examine connections among large-scale and regional climate changes which give rise to phenomena such as land-sea contrast and high-latitude amplification of temperature changes, the regulation of atmospheric CO 2 and biogeochemical cycles, and the waxing and waning of the monsoons.
The Tier 1 interglacial experiments for CMIP6 are timeslice (or equilibrium) experiments at 6000 and 127 000 years before present (where present is defined as 1950), hereafter referred to as 6 ka (midHolocene) and 127 ka (lig127k). The mid-Holocene interval has been the focus for model simulations, model-model comparisons, paleodata synthesis, and model-data comparison since the beginning of PMIP, and this work has contributed to model evaluation and understanding of climate change in the last three major assessments of the Intergovernmental Panel on Climate Change (Flato et al., 2013;Folland et al., 2001;Hegerl et al., 2007;Jansen et al., 2007;Masson-Delmotte et al., 2013). The changes in insolation are characterized by enhanced seasonal contrast in the Northern Hemisphere (NH) (and reduced seasonal contrast in the Southern Hemisphere, SH), giving rise to warmer NH summers and a significant enhancement of the NH monsoons (COHMAP Members, 1988;Hély et al., 2014;Lezine et al., 2011;Saraswat et al., 2013;Tierney et al., 2017). Systematic benchmarking against pollen-based reconstructions of climate variables and lake-level-based waterbalance reconstructions (Braconnot et al., 2007b(Braconnot et al., , 2012Coe and Harrison, 2002;Harrison et al., 1998Harrison et al., , 2014Harrison et al., , 2015 has highlighted the fact that climate models persistently underestimate changes in the monsoon precipitation and produce too much continental drying (Harrison et al., 2015).
Given the long history of coordinated model experiments for 6 ka, this period allows us to assess whether there is an improvement in the ability of models to reproduce a climate state different from the modern one. For this reason, the Tier 1 midHolocene experiment is one of two possible "entry cards" for PMIP simulations in CMIP6 (Table 1): all modeling groups contributing to PMIP4-CMIP6 must perform either the midHolocene experiment or a simulation of the Last Glacial Maximum .
Although the LIG (129 to 116 ka) was discussed in the First Assessment Report of the IPCC (Folland et al., 1990), it gained more prominence in the IPCC Fourth and Fifth Assessment (AR4 and AR5) because of reconstructions highlighting that global mean sea level was at least 5 m higher (but probably no more than 10 m higher) than present for several thousand years (Dutton et al., 2015a;Jansen et al., 2007;Masson-Delmotte et al., 2013). Thus the LIG is recognized as an important period for testing our knowledge of climate-ice-sheet interactions in warm climate states. However, the ensemble of LIG simulations examined in the AR5  was not wholly consistent: the orbital forcing and GHG concentrations varied between the simulations. While it has been suggested that differences in regional temperatures between models might reflect differences in cryosphere feedback strength (Yin and Berger, 2012;Otto-Bliesner et al., 2013) or differences in the simulation of the Atlantic Meridional Overturning Circulation (AMOC) Masson-Delmotte et al., 2013), differences between models could also have arisen because of differences in the experimental protocols. Furthermore, the LIG simulations were mostly made with older and/or lowerresolution versions of the models than were used for future projections, making it more difficult to use the results to assess model reliability . The Tier 1 lig127k experiment (Table 1) is designed to address the climate responses to stronger orbital forcing than the midHolocene experiment using the same state-of-the-art models and following a common experimental protocol. It will provide a basis to address the linkages between ice sheets and climate change in collaboration with the Ice Sheet Model Intercomparison Project for CMIP6 (ISMIP6) (Nowicki et al., 2016).
The midHolocene and lig127k experiments are starting points for examining interglacial climates. A number of other experiments are proposed in the current phase of PMIP (PMIP4) to facilitate diagnosis of these Tier 1 experiments. The Tier 2 simulations will include sensitivity experiments to examine the impact of uncertainties in boundary conditions and the role of feedbacks in modulating the response to orbital forcing. Ocean, vegetation, and dust feedbacks, and the synergies between them, have been a focus in previous phases of PMIP Dallmeyer et al., 2010;Otto et al., 2009;Wohlfahrt et al., 2004), and this allows us to design simple experimental protocols to compare the strength of these feedbacks in different climate models. Simulations with prescribed but adjusted vegetation cover will be a major focus for both the Holocene and LIG in PMIP4, and comparison of these simulations with ESM simulations that include dynamic vegetation will allow exploration of the magnitude of land-surface biases in these latter models. Changes in vegetation and land-surface hydrology are an important control on dust emissions (Tegen et al., 2002;Engelstädter et al., 2003), which can affect the strength of the West African Monsoon (Konare et al., 2008;Pausata et al., 2016). The examination of the dust feedback will be a new focus in PMIP4. In addition, the LIG provides an ideal opportunity to examine the role of cryosphere feedbacks through sensitivity experiments, which will be a focus of additional experiments associated with both the Holocene and the LIG. One such feedback is the release of freshwater into the ocean and the role of such freshwater forcing in generating more abrupt climate changes than would be expected for the smoothly varying changes in insolation forcing during an interglacial (Goelzer et al., 2016a;Luan et al., 2015;Stone et al., 2016). Understanding the role of feedbacks in general in the generation of abrupt climate changes, and the need to understand the relationship between mean climate changes and short-term (annual to multi-decadal) climate variability, leads naturally to a desire to simulate the transient behavior of the climate system -and such transient experiments are proposed for both the Holocene and LIG time periods. New results have highlighted the possibility of using reconstruction of past interannual variability from corals and mollusc shells to assess the Holocene-simulated changes in variability at the scale of the tropical Pacific Ocean . Groups are also encouraged to run their models with an active land and ocean carbon cycle to assess terrestrial and ocean carbon storage and differences between the two interglacial periods. The aim of this paper is to present and explain the experimental designs for both the PMIP4-CMIP6 Tier 1 interglacial experiments, and for associated Tier 2 and Tier 3 sensitivity and transient experiments. Section 2 describes and discusses the PMIP4-CMIP6 midHolocene Tier 1/entry card and lig127k Tier 1 simulations. Section 3 describes Tier 2 and Tier 3 PMIP4 sensitivity studies that can be carried out to diagnose these Tier 1 simulations. Section 4 briefly describes the paleodata resources, which can be used to evaluate the simulations.
2 Experimental design for the Tier 1 PMIP4-CMIP6 midHolocene and lig127k simulations The core or Tier 1 experiments for the Holocene and the LIG are the midHolocene and lig127k simulations. The CMIP DECK (Diagnostic, Evaluation and Characterization of Klima) piControl for 1850 CE and the CMIP6 historical experiment (see Eyring et al., 2016, for a description of these experiments) are the reference simulations to which the paleo-experiments will be compared. Thus, the paleoexperiments must use the same model components and follow the same protocols for implementing external forcings as are used in the piControl and historical simulations. The midHolocene simulation is one of the PMIP entry cards in the PMIP4-CMIP6 experiments, which means that groups who run the lig127k simulation must also run either the mid-Holocene or the lgm (Last Glacial Maximum) experiment . The boundary conditions for the midHolocene, lig127k, and piControl experiments are given in Table 1, and more detailed information is given below.

Orbital configuration, solar constant, and insolation anomalies
Earth's orbital parameters (eccentricity, longitude of perihelion, and obliquity) should be prescribed following Berger and Loutre (1991). These parameters affect the seasonal and latitudinal distribution and magnitude of solar energy re-ceived at the top of the atmosphere and, in the case of obliquity, the annual mean insolation at any given latitude (Berger and Loutre, 1991). The DECK piControl simulations are to use the orbital parameters appropriate for 1850 CE (Table 1) (Eyring et al., 2016), when perihelion occurs close to the boreal winter solstice. The exact date slightly varies depending on the internal model calendar and the number of days used to define a year. Because of this and the fact that the length of the seasons varies as a function of precession and eccentricity (Joussaume and Braconnot, 1997), the vernal equinox must be set to noon on 21 March in all the simulations (piControl, midHolocene, and lig127k). The orbit at 6 ka was characterized by an eccentricity of 0.018682, similar to 1850 CE (Table 1). Obliquity was larger (24.105 • ) and perihelion at 6 ka occurred near the boreal autumn equinox. The orbit at 127 ka was characterized by larger eccentricity than at 1850 CE, with perihelion occurring close to the boreal summer solstice (Fig. 2). The tilt of the Earth's axis was maximal at 131 ka and remained higher than in 1850 CE through 125 ka; obliquity at 127 ka was 24.04 • ( Table 1). The different orbital configurations for the midHolocene and lig127k result in different seasonal and latitudinal distribution of top-of-atmosphere insolation compared to the DECK piControl (Fig. 3). Both time periods show large positive insolation anomalies during boreal summer. July-August anomalies between 40 and 50 • N reach about 55-60 W m −2 at 127 ka and 25 W m −2 at 6 ka. The higher obliquity at 127 and 6 ka contributes to a small but positive annual insolation anomaly compared to the preindustrial at high latitudes in both hemispheres and a slight insolation reduction in the tropics in the annual mean. The global difference in insolation forcing between the interglacial experiments and the preindustrial is negligible. The solar constant prescribed for the midHolocene and lig127k simulations is the same as in the DECK piControl simulation, which is fixed at the mean value for the first two solar cycles of the historical simulation (i.e., 1850-1871) (Eyring et al., 2016). This value (1360.7 W m −2 ) is lower than the value for the solar constant used by some models in PMIP3 (1365 W m −2 ) and this leads to a global reduction of incoming solar radiation compared to the PMIP3 experiments (Fig. 4). The slight differences in orbital parameters between the 1850 CE reference periods to be used for PMIP4-CMIP6 and the 1950 CE reference used for PMIP3 leads to seasonal differences in forcing with a slight decrease in boreal spring and increase in boreal autumn. The combination of the two factors leads to an overall reduction: the largest reduction occurs in boreal spring and is about 1.6 W m −2 between 10 • S and 40 • N.

Greenhouse gases
Ice-core records from Antarctica and Greenland provide measurements of the well-mixed GHGs: CO 2 , CH 4 , and N 2 O (Fig. 1). These measurements are given as mole fractions in dry air and are noted as parts per million (ppm) or parts per Orbital configurations for the piControl, midHolocene, and lig127k experiments. Note that the aspect ratio between the two axes of the ellipse has been magnified to better highlight the differences between the periods. However, the change in ratio between the different periods is proportional to the real values. In these graphs VE stands for vernal equinox, SS for summer solstice, AE for autumnal equinox, and WS for winter solstice. The numbers along the ellipse are the number of days between solstices and equinoxes. billion (ppb), respectively. For simplicity, we use the term "concentration" for these GHG levels. By 6 and 127 ka, the concentrations of atmospheric CO 2 and CH 4 had increased from their respective levels during the previous glacial periods, the Last Glacial Maximum and the penultimate glaciation, to values comparable to preindustrial levels.
midHolocene. In PMIP4-CMIP6, we use a revised version of an earlier trace gas reconstruction (Joos and Spahni, 2008). The CO 2 concentration for the mid-Holocene is derived from ice-core measurements from Dome C ( Monnin et al., 2001Monnin et al., , 2004) and dated using the AICC2012 age scale . A smoothing spline (Bruno and Joos, 1997;Enting, 1987) with a nominal cut-off period of 3000 years was used to produce a continuous CO 2 record. This yields a CO 2 concentration of 264.4 ppm at   6 ka. Methane has been measured in ice from Antarctic ice cores EPICA Dome C (Flückiger et al., 2002), EPICA Dronning Maud Land (EPICA Community Members, 2006) and Talos Dome (Buiron et al., 2011). For Greenland, methane data are from GRIP (Blunier et al., 1995;Chappellaz et al., 1997;Spahni et al., 2003), GISP2 (Brook, 2009), and GISP2D (Mitchell et al., 2013). Both are splined with a nominal cut-off period of 200 years. This results in a concentration of 574 ppb for the Antarctic ice cores, representative of high-latitude Southern Hemisphere air, and of 620 ppb for the Greenland ice cores, representative of the high-latitude Northern Hemisphere air, and an estimated global mean value of 597 ppb. The N 2 O data around 6 ka are from a compilation of published data from EPICA Dome C (Flückiger et al., 2002;Spahni et al., 2005) and new, unpublished data measured at the University of Bern using ice from Greenland (NGRIP) and Talos Dome (TALDICE). The data are splined with a nominal cut-off period of 700 years and the resulting N 2 O concentration at 6 ka is 262 ppb.
The realistic GHG concentrations used for the mid-Holocene PMIP4-CMIP6 experiment are different from those used in the PMIP3 experiments (Braconnot et al., 2012). The PMIP3 experiments were designed simply to examine the effects of changes in orbital forcing, and the CO 2 concentrations were therefore kept the same as the value typically used in preindustrial experiments (280 ppm), although other GHGs were prescribed from ice-core measurements. The use of realistic values for all the GHGs in the PMIP4-CMIP6 midHolocene experiment may improve comparisons with paleoclimate reconstructions and will ensure that the midHolocene experiment is consistent with planned transient Holocene simulations (see Sect. 3). However, the reduction in CO 2 concentration from 280 to 264.4 ppm will reduce GHG forcing by about 0.3 W m −2 (Myhre et al., 1998), which translates to a difference in global mean surface air temperature of −0.24 • C when applying an equilibrium climate sensitivity of 3 • C for a nominal doubling of CO 2 . Simulations with the IPSL model (Dufresne and co-authors, 2013) show that this change in the experimental protocol between PMIP3 and PMIP4-CMIP6 yields a global mean cooling of 0.24 ± 0.04 • C, as expected, with regional differences of up to 0.5 • C in parts of Eurasia and in South Africa (Fig. 5). Although these differences are small overall, they will need to be accounted for in comparisons between the PMIP4-CMIP6 midHolocene simulations and previous generations of PMIP 6 ka simulations.
lig127k. The LIG GHG concentrations are available solely from Antarctic ice cores. CO 2 concentrations can only be derived from Antarctic ice, because of potential in situ CO 2 production in the Greenland ice sheet (Tschumi and Stauffer, 2000). We also do not have any reliable CH 4 and N 2 O concentrations from Greenland in the LIG due to melt layers in the ice, as Greenland temperatures were significantly warmer at that time compared to the mean of the past millennium ( Fig. 1) (NEEM Community Members 2013). For the lig127k simulation (Table 1), we adopt mean values for 127.5-126.5 ka on the AICC2012 age scale  from EPICA Dome C (Bereiter et al., 2015;Schneider et al., 2013) for CO 2 , from EPICA Dome C and EPICA Dronning Maud Land (Loulergue et al., 2008;Schilt et al., 2010a) for CH 4 , as well as from EPICA Dome C and Talos Dome (Schilt et al., 2010a, b) for N 2 O. The atmospheric CO 2 and N 2 O concentrations of 275 ppm and 255 ppb, respectively, can be regarded as globally representative, while the mean ice core CH 4 concentration (662 ppb) is representative of high-latitude Southern Hemisphere air. A global mean atmospheric CH 4 concentration of 685 ppb is adopted for 127 ka, thereby assuming the same difference (23 ppb) be-tween the global mean atmospheric CH 4 and Antarctic ice core CH 4 values as for the mid-Holocene.

Paleogeography and ice sheets
midHolocene. Several lines of evidence indicate that the ice sheets had their modern characteristics by the mid-Holocene, except in a few places such as the Baffin Islands (Carlson et al., 2008b;Clark et al., 2000). While the presence of a relic of the Laurentide ice sheet may be the origin of modeldata mismatches in the climate of eastern North America (Wohlfahrt et al., 2004), the effect is local and small. Cosmogenic surface exposure ages and threshold lake records (Carlson et al., 2014;Larsen et al., 2015;Sinclair et al., 2016) also suggest that by 6 ka, the Greenland ice sheet was similar in extent to present. The ice-sheet distribution and elevations, land-sea mask, continental topography and oceanic bathymetry should all be prescribed as the same as in piControl in the midHolocene simulation (Table 1).
lig127k. Evidence for the evolution of the ice sheets during the LIG comes mainly from proximal marine records (Carlson and . The deposition of a detrital carbonate layer in the Labrador Sea, dated to around 128 ka based on geomagnetic secular variation , suggests that ice had retreated from Hudson Bay and is taken to indicate the final demise of the Laurentide ice sheet (Carlson, 2008;Nicholl et al., 2012). The disappearance of the Eurasian ice sheet is more difficult to constrain because either the proximal marine records lack benthic δ 18 O data or the benthic δ 18 O data show trends that are different from those of open ocean records during the LIG (Bauch, 2013). The cessation of deposition of ice-rafted debris (IRD) from the Eurasian ice sheet has been dated to between 128 and 126 ka using δ 18 O (Risebrobakken et al., 2006). However, sea-level data (Dutton et al., 2015b) suggest that this ice sheet disappeared earlier and was likely gone by ∼ 127 ka. Proximal marine records of the Greenland ice sheet document a gradual retreat during the LIG, with minimum extent around 120 ka (Carlson et al., 2008a;Colville et al., 2011;Stoner et al., 1995;Winsor et al., 2012). However, Greenland-sourced IRD reached a minimum similar to the Holocene before ∼ 127 ka (Colville et al., 2011;Winsor et al., 2012).
The extent of the Antarctic ice sheet is not directly constrained by data proximal to the ice sheet at 127 ka. Given higher-than-present sea levels, the gradual retreat of the Greenland ice sheet, and the lack of other NH ice sheets, it seems likely that the Antarctic ice sheet was smaller than present by ∼ 127 ka (Colville et al., 2011;Dutton et al., 2015a, b;Mercer, 1978). The existence of ∼ 250 ka Mt. Erebus ash in the ANDRILL site in the Ross Sea could indicate a smaller-than-present West Antarctic ice sheet (WAIS) some time after ∼ 250 ka (McKay et al., 2012). The ice-core record from Mount Moulton, West Antarctica, could be consistent with deglaciation of much of West Antarctica during the LIG, and likely at 130-126 ka (Steig et al., 2015). Standalone ice-sheet model simulations forced by ocean warming suggest the West Antarctic ice sheet to be a major contributor to LIG global mean sea-level rise, with contributions also coming from the marine-based portions of the East Antarctic ice sheet (DeConto and Pollard, 2016). Contributions are 6.0-7.5 m of the equivalent sea-level rise, which would explain the global mean sea level being at least +6 m by ∼ 127 ka (Dutton et al., 2015b). However, because of the difficulty in implementing ice-to-ocean changes for the WAIS and the uncertainties associated with dating the changes in the other ice sheets, the paleogeography of the lig127k simulation will be prescribed the same as in the DECK piControl simulation (Table 1). In view of the greater uncertainty associated with the prescription of ice sheets in the lig127k experiment, this aspect of the boundary conditions will be a major focus of sensitivity experiments (see Sect. 3).

Vegetation
There is abundant evidence for changes in vegetation distribution during the mid-Holocene and the LIG (Goni et al., 2005;Harrison and Bartlein, 2012;Hély et al., 2014;LIGA Members, 1991;Prentice et al., 2000). However, there is insufficient data coverage for many regions to be able to produce reliable global vegetation maps. Furthermore, given the very different levels of complexity in the treatment of vegetation properties, phenology and dynamics in the current generation of climate models, paleo-observations do not provide sufficient information to constrain their behavior in a comparable way. The treatment of natural vegetation in the midHolocene and lig127k simulations should therefore be the same as in the DECK piControl simulation. That is, depending on what is done in the DECK piControl simulation, vegetation should either be prescribed to be the same as in that simulation, or prescribed but with interactive phenology, or predicted dynamically (Table 1). Uncertainties related to the treatment of vegetation in the different simulations will be analyzed through sensitivity experiments (see Sect. 3).

Aerosols: tropospheric dust and stratospheric volcanic
Natural aerosols show large variations on glacial-interglacial timescales, with low aerosol loadings during interglacials compared to glacials, and during the peak of the interglacials compared to the present day (Albani et al., 2015;deMenocal et al., 2000;Kohfeld and Harrison, 2000;McGee et al., 2013). Atmospheric dust affects radiative forcing at a regional scale and can therefore affect precipitation and surface hydrology (Miller et al., 2004;Yoshioka et al., 2007), including the monsoons (Konare et al., 2008;Pausata et al., 2016;Vinoj et al., 2014), as well as moderate snow albedo feedbacks when sufficient dust is deposited . While model simulations that are observationally constrained by a global compilation of dust records suggest that the global dust budget was dominated by NH dynamics during the midHolocene as it is today, the total loading as well as regional patterns of dust loading were different (Albani et al., 2015). This motivates the inclusion of changes in dust loading in the midHolocene and lig127k simulations (Table 1, Fig. 6).
As in the case of vegetation, the implementation of changes in atmospheric aerosol in the midHolocene and lig127k simulations should follow the treatment used for the DECK piControl and historical simulations. Models with an interactive representation of dust should prescribe changes in soil erodibility or dust emissions to account for the changes in dust sources during the interglacials. Although the maps provided by PMIP for this purpose are for mid-Holocene conditions and from the only model simulation available (Albani et al., 2015), they should be used for both the midHolocene and lig127k simulations. For each model configuration, if atmospheric dust loading is prescribed in the DECK piControl and historical simulations, the midHolocene and lig127k simulations should use the three-dimensional monthly climatology of atmospheric dust mass concentrations or aerosol optical depths available from the same data-constrained simulations as the soil erodibility maps. Also available are data sets of the dust shortwave and longwave direct radiative forcing. If atmospheric dust loading is not represented in the DECK piControl and historical simulations, it should not be included in the midHolocene and lig127k simulations. The impact of dust on the radiation balance is sensitive to the optical properties prescribed (Perlwitz et al., 2001); it is uncertain how optical properties might change during interglacials (Potenza et al., 2016;Royer et al., 1983). Uncertainties in the protocol and in the interplay between dust and vegetation will be a focus of the analyses.
There is no observationally constrained estimate of the volcanic stratospheric aerosol for either the mid-Holocene or the LIG. The background volcanic stratospheric aerosol used in the CMIP6 DECK piControl should be used for the mid-Holocene and lig127k simulations. Other aerosols included in the DECK piControl should similarly be included in the midHolocene and lig127k simulations.

Setup and documentation of simulations
To provide initial conditions for the simulations, it is recommended that a spin-up simulation is performed departing from the CMIP6 piControl experiment. The length of this spin-up simulation will be model-and resource-dependent. However, it should be long enough to minimize at least surface climate trends.
The modeling groups are responsible for a comprehensive documentation of the model system and the experiments. Documentation should be provided via the ESDOC website and tools provided by CMIP6 (http://es-doc.org/) to facilitate communication with other CMIP6 projects. A PMIP4 special issue in GMD and Climate of the Past has been opened where the groups are encouraged to publish these documentations.
The documentation should include the following.
-The model version and specifications, like interactive vegetation or interactive aerosol modules.
-A link to the DECK experiments performed with this model version.
-Specification of the forcing data sets used and their implementation in the model. The provision of figures and tables giving monthly latitude insolation anomalies and daily incoming solar radiation at the top of the atmosphere (TOA) for 1 year should be provided because this allows the implementation of the most critical forcing to be checked.
-Information about the initial conditions and spin-up technique used.
-We request information on drift in key variables for a few hundred years at the end of the spin-up and the beginning of the actual experiment. These variables are globally and annually averaged SSTs; deep ocean temperatures (global and annual average over depths below 2500 m); deep ocean salinity (global and annual average over depths below 2500 m); top of atmosphere energy budget (global and annual average); surface energy budget (global and annual average); northern sea ice (annual average over the Northern Hemisphere); southern sea ice (annual average over the Southern Hemisphere); northern surface air temperature (annual average over the Northern Hemisphere); southern surface air temperature (annual average over the Southern Hemisphere); -Atlantic Meridional Overturning Circulation (maximum overturning stream function in the North Atlantic basin between 0 and 80 • N below 500 m depth); and carbon budget (if relevant).

PMIP4-CMIP6 Tier 2 and Tier simulations
The selection of only two intervals, midHolocene and lig127k, for PMIP4-CMIP6 interglacial experiments is designed to maximize both the multi-model ensemble size and opportunities for model evaluation, since both periods have been the focus for data synthesis. However, this means that the experiments do not sample the diversity in the transient forcings and responses during the LIG and the Holocene. Although transient simulations for these two periods are included in the suite of PMIP4 simulations (see Sect. 3.5), there is utility in examining other interglacial climates using equilibrium experiments parallel to the midHolocene and lig127k simulations, particularly in order to provide additional samples of the response of the system to insolation forcing. Additional Tier 2 experiments -the end of the LIG (116 ka) and the early Holocene (9.5 ka) (see Sect. 3.1) -are proposed to address this. Uncertainties in the boundary and initial conditions for the mid-Holocene and LIG mean that the PMIP4-CMIP6 mid-Holocene and lig127k simulations may not capture important feedbacks accurately. The major sources of uncertainty in the boundary conditions are the prescription of modern vegetation cover by some models, and the prescription of modern ice sheets in the lig127k simulation. Both sources of uncertainty can be addressed through Tier 2 sensitivity experiments (see Sect. 3.2, 3.3). The equilibrium experiments also do not address climate changes forced by the nonlinear behavior of ice-sheet-ocean coupling, or the possibility that such feedbacks could give rise to abrupt changes in climate superimposed on the more slowly varying insolation forcing during the Holocene and the LIG. This will be addressed through Tier 2 idealized simulations of specific freshwaterforcing events, specifically the Heinrich 11 event at the beginning of the LIG and the 8.2 ka event during the Holocene (see Sect. 3.4). Understanding the interplay among different components of the Earth system in determining the long-term evolution of LIG and Holocene climate is the major goal of the proposed Tier 3 transient experiments (Sect. 3.5) to be carried out during PMIP4.
Further information and access to data sets are available on the PMIP4 website and will be updated during the course of the project (https://pmip4.lsce.ipsl.fr/doku.php/exp_design: index).

Equilibrium response to alternative states of orbital forcing
hol9.5k. The maximum expression of Holocene orbitally induced differences in TOA insolation forcing from present occurred during the early part of the Holocene, but the climate at this time was still affected by the presence of a relic of the Laurentide ice sheet (Carlson et al., 2008b). As a result, summer temperatures at mid to high latitudes were cooler than during the mid-Holocene (Carlson et al., 2008b;Renssen et al., 2009Renssen et al., , 2012. The presence of the ice sheet delayed the response to insolation forcing in monsoon regions (Lezine et al., 2011;Marzin et al., 2013). It has also been suggested that the remnant ice sheet may have counteracted the reduction of ENSO variability in response to orbital forcing in the early Holocene (Carre et al., 2014;Luan et al., 2015). Protocols for early Holocene experiments were developed in previous phases of PMIP (PMIP2, PMIP3), and provide the basis for the proposed PMIP4 simulation for 9.5 ka. Since the phase of precession at 9.5 ka is similar to that of 127 ka, this experiment provides a basis for examination of the similarities in seasonal changes between the two interglacials (Braconnot et al., 2008). Following the experimental protocol for the mid-Holocene simulation, orbital parameters should be changed following Berger and Loutre (1991). The extent and topography of the ice sheet should be prescribed using either ICE-6G or GLAC-1D, as proposed by the PMIP deglaciation working group . GHG concentrations can also be prescribed from the last deglaciation experiment (Table 2). lig116k. Continental ice-sheet growth and associated sealevel lowering started at ∼ 116 ka, marking the end of the LIG (Stirling et al., 1998). Simulations with climate models that include feedbacks among the atmosphere, ocean, land, and sea ice are able to simulate sufficient cooling to initiate ice-sheet growth when forced with the 116 ka orbital conditions, reducing NH summer insolation (Herrington and Poulsen, 2012;Jochum et al., 2012). However, the result is sensitive to the atmospheric CO 2 concentration used. To test the ability of the CMIP6 and PMIP4 models to simulate glacial inception, we propose a sensitivity experiment using orbital parameters for 116 ka (lig116k). The CO 2 concentration should be prescribed as 273 ppm (Bereiter et al., 2015;Schneider et al., 2013). All other forcings and boundary conditions will remain the same as the lig127k simulation (Table 2).

Sensitivity to prescribed vegetation
Except in the case of models with dynamic vegetation, the midHolocene and lig127k simulations will be run with prescribed preindustrial vegetation cover because of the lack of a comprehensive and reliable global data set of vegetation for the two periods. However, pollen and macro-fossil evidence shows that boreal forest extended farther north than today in the mid-Holocene Prentice et al., 2000;Binney et al., 2017) and, except in Alaska and central Canada, extended to the Arctic coast during the LIG LIGA, 1991;Lozhkin and Anderson, 1995). Pollen and other biogeographical/geomorphological and paleohydrological evidence also indicates northward extension of vegetation into modern-day desert areas, particularly in northern Africa, both in the mid-Holocene (Drake et al., 2011;Hély et al., 2014;Larrasoana et al., 2013;Lezine et al., 2011;Prentice et al., 2000;Tierney et al., 2017) and during the maximum phase of the LIG (Castaneda et al., 2009;Hooghiemstra et al., 1992). Given the impact of increased woody cover on albedo and evapotranspiration, these vegetation changes should have profound impacts on the surface energy and water budgets and may help to explain mismatches between simulated and reconstructed high-latitude (Muschitiello et al., 2015) and monsoon climates Claussen and Gayler, 1997;Pausata et al., 2016) in both time periods.
We propose sensitivity experiments for the midHolocene and lig127k to explore the feedbacks between vegetation and climate. Vegetation cover at the NH high latitudes should be changed from tundra to boreal forest and the Sahara replaced by evergreen shrub to 25 • N and savanna/steppe poleward of 25 • N. Ideally, these regional changes should be made separately in order to diagnose the interaction between highlatitude and low-latitude climates, and a third experiment could be made implementing both changes. In each experiment, all other boundary conditions should be implemented as in the baseline midHolocene and lig127k simulations (Table 2). Sensitivity experiments will also be required to characterize the uncertainties related to the prescription of dust fields in the mid-Holocene and LIG simulations, but it is difficult to anticipate the form of such experiments until the Tier 1 experiments are diagnosed. A first step could be to investigate the vegetation feedback on emission in simulations with interactive dust exploiting the vegetation sensitivity analyses.

Sensitivity to prescribed ice sheets
The midHolocene and lig127k simulations will be run with prescribed modern ice sheets and paleogeography. However, it is highly likely that the Antarctic ice sheet was smaller than present by ∼ 127 ka, most probably because of the dis-appearance of the WAIS, and that the Greenland ice sheet was reduced in extent compared to present. Given that only about 3-4 m sea-level rise is covered by contributions from ocean thermal expansion (McKay et al., 2011), land-based glaciers (Marzeion et al., 2012), and melting of the Green-land ice sheet (NEEM Community Members, 2013;Masson-Delmotte et al., 2013), the remaining sea-level rise is most likely to be linked to a mass loss from the Antarctic ice sheet. We propose a sensitivity experiment (lig127k-ais) to test the impact of a smaller-than-present Antarctic ice sheet, using a reduced ice-sheet configuration obtained from offline simulations with their own models or the model results such as those from DeConto and Pollard (2016) or Sutter et al. (2016). These authors used a dynamic ice-sheet model, forced with climate model output and calibrated to reproduce LIG sea-level estimates, to simulate the Antarctic ice sheet at 128 ka. All other boundary conditions should be implemented as in the baseline lig127k simulation. An additional sensitivity simulation (lig127k-gris) to complement the lig127k simulations is proposed in which the Greenland ice sheet is configured to its minimum LIG extent is also of interest, using configurations obtained from offline simulations, for example from ISMIP6.

Freshwater forcing
Sensitivity to the H11 meltwater event during the early LIG. Heinrich layers in the North Atlantic, containing high concentrations of IRD, record multiple examples of prolonged iceberg discharge during the past 500 ka (Hemming, 2004;Marino et al., 2015;McManus et al., 1999). Heinrich event 11 (H11) is a well-documented example that occurred from ∼ 135 to 128 ka (Marino et al., 2015). The associated freshwater flux has been estimated as peaking at ∼ 0.3 Sv at ∼ 132.5 ka and tapering off thereafter (Marino et al., 2015), and is broadly consistent with an estimate of 0.19 Sv at 130 ka based on coral records (Carlson, 2008). There is also evidence of a rapid sea-level rise associated with this meltwater pulse, estimated at ∼ 70 m or 28 ± 8 m ka −1 (∼ 0.32 ± 0.09 Sv) during the deglacial transition (Grant et al., 2012). Model simulations have shown that the freshwater forcing of H11, including its cessation, may be important for explaining the evolution of climate through the early part of the LIG (Goelzer et al., 2016a, b;Holden et al., 2010;Loutre et al., 2014;Stone et al., 2016). We propose a sensitivity experiment (lig127k-H11) to examine the impact of the H11 event. The insolation anomalies at 130 ka are similar to those at 127 ka. Therefore the experiment can be made by adding a persistent flux of 0.2 Sv to the North Atlantic between 50 and 70 • N for 1000 years, with all other boundary conditions implemented as in the baseline lig127k simulation (Table 2). Sensitivity to the 8.2 ka freshwater event during the early Holocene. While the climate impact of the 8.2 ka event is well documented, the magnitude of the freshwater forcing and its duration are less well constrained. There are generally thought to be two components to the freshwater forcing in the early Holocene, a background flux from the Laurentide ice sheet (Hillaire-Marcel et al., 2007;Licciardi et al., 1999) and catastrophic flux from the drainage of Lake Agassiz (Barber et al., 1999;Clarke et al., 2004;Teller et al., 2002). Lake Agassiz appears to have drained in several flood events of relatively short duration, but with an estimated total discharge into the Labrador Sea of ca. 151 400 km 3 (Andrews et al., 1995Clarke et al., 2004Clarke et al., , 2009Ellison et al., 2006;Hillaire-Marcel et al., 2007;Kerwin, 1996;Lajeunesse and St-Onge, 2008;Lewis et al., 2012;Roy et al., 2011). The background flux is smaller (ca. 0.13 Sv) but persistent for several hundred years (Carlson et al., 2008b(Carlson et al., , 2009Clarke et al., 2009;Hillaire-Marcel et al., 2007). The proposed sensitivity experiment (hol8.2k) can use the orbital, ice-sheet, and GHG boundary conditions of the 9.5 ka experiment. The "Lake + Ice_100 years" scenario of Wagner et al. (2013) is more consistent with ice dynamics and the data of Carlson et al. (2009) than the shorter 1-year flood scenarios  and should be adopted for this sensitivity experiment. That is, modeling groups should impose a single input of 2.5 Sv for 1 year followed by a background freshwater flux of 0.13 Sv for 99 years (Table 2). This freshwater flux is added to the Labrador Sea, but modeling groups can choose whether to add it uniformly over the whole of the Labrador Sea or only over part of the area.
These simulations should be run a minimum of 100 years after the end of the pulse and if possible even longer, preferably until the Atlantic Meridional Overturning Circulation (AMOC) has recovered to its initial state.

Transient Holocene and LIG simulations
Transient simulations provide an opportunity to examine the time-dependent evolution of climate in response to forcings and feedbacks. Transient simulations of the last deglaciation are a major focus in PMIP4 . These simulations will be run for the period 21 to 9 ka with time-varying orbital forcing, greenhouse gases, ice sheets and other geographical changes. The later part of this experiment is obviously of interest for comparison with the early Holocene experiments. However, we are also proposing transient simulations focusing on the Holocene and the LIG.
Using the PMIP-CMIP6 midHolocene simulation as a starting point, we propose a transient simulation of the last 6000 years (past6k). In this simulation, both orbital parameters and GHGs will be changed following Berger and Loutre (1991) and ice-core measurements (as described in Sect. 2.2). Changes in paleotopography over the past 6 ka are small and, for simplicity and consistency with the midHolocene simulation, we propose using modern values throughout. Vegetation and aerosols will also be fixed at preindustrial values, except for groups running fully dynamic vegetation and/or aerosol models where the initial state of these components will be derived from their mid-Holocene simulation. Alternatively, some groups may start the Holocene transient simulation from the end of the last deglaciation experiment at 9 ka, incorporating changes in the evolution of ice sheets and paleotopography consistent with that experiment. A proposed LIG transient simulation (lig127to121k) will be run from 127 to 121 ka, using appropriate changes in orbital forcing and trace gases but with all other boundary conditions specified as in the lig127k simulation. These simulations as well as simulations planned by some modeling groups with climate-ice-sheet models will be important as input for addressing the role of coupling between climate and the ice sheets.

Paleoenvironmental data and climate reconstructions for comparison to model simulations
The ability to evaluate the realism of the Tier 1 PMIP4-CMIP6 simulations and the various sensitivity experiments is central to PMIP. Some paleoenvironmental observations can be used for direct comparison with model outputs, including, e.g., simulated water balance against lake-level reconstructions (e.g., Coe and Harrison, 2002) or simulated vegetation patterns against pollen-based vegetation reconstructions (e.g., Perez Sanz et al., 2014). Such qualitative comparisons are often adequate to evaluate simulations when, as is the case with regional climate changes in the mid-Holocene and LIG, the changes are large and regionally coherent . There are also quantitative reconstructions of climate variables from a wide variety of archives.
There are uncertainties associated with such reconstructions, both statistical and resulting from an incomplete understanding of the climate controls on specific types of records, and these uncertainties need to be taken into account in comparisons with simulations. However, an increasing number of process-based models can be used to translate climate model outputs into explicit simulations of specific paleo-records (Emile-Geay and Tingley, 2016; Li et al., 2014;Thompson et al., 2011), allowing uncertainties in process understanding to be made explicit. Drawing on ongoing work for the LGM and the use of ocean biochemistry, tracer and isotopic modeling, efforts will be made to isolate key features of the ocean reconstructions that should be reproduced by climate models.
The major analytical focus for the Holocene experiments is on systematic benchmarking (Harrison et al., 2015) of the midHolocene simulation, analysis of feedbacks, and elucidation of the relationship between mean climate state and interannual to centennial variability. Analysis of the mid-Holocene simulation and associated sensitivity experiments benefits from the fact that there has been a major focus on data synthesis for this time period (Bartlein et al., 2011;Bigelow and al., 2003;Daniau et al., 2012;Hessler et al., 2014;Kohfeld and Harrison, 2000;Leduc et al., 2010;Marchant et al., 2009;Marlon et al., 2013;Pickett et al., 2004;Prentice et al., 2000). Thus the number of records and spatial coverage of quantitative reconstructions are relatively extensive (Bartlein et al., 2011;Hessler et al., 2014). There are gaps in coverage from continental regions, particularly in the SH, but this situation is likely to improve in the near future (Flantua et al., 2015;Herbert and Harrison, 2016). Knowledge of ocean conditions during the mid-Holocene is poor and likely to remain so, in part because of incomplete understanding of the causes of differences between sea-surface temperature reconstructions based on different biological groups and in part because the signal-to-noise ratio in the reconstructions is small due to other methodological uncertainties Jonkers and Kucera, 2015;Rosell-Mele and Prahl, 2013).
There are several sources of information about short-term climate variability during the Holocene, including tree-ring records, speleothems, corals and molluscs. However, there are major gaps in data coverage from the tropical oceans that challenge our understanding of ENSO variability; the distribution of speleothem records is limited to karst areas; and few tropical trees show clear-cut seasonality in growth. More comprehensive syntheses of these data are needed, and there are major challenges in combining the different data sources to yield large-scale reconstructions of climate variability. It will also be necessary to develop appropriate methods to use these data for comparison with simulations, focusing on temporal statistics and teleconnection patterns Emile-Geay and Tingley, 2016).
There are many individual records documenting the evolution of climate through the Holocene, including quantitative climate reconstructions (Wanner et al., 2008). Synthetic products have either focused on reconstructions of global temperature changes Marcott et al., 2013;Shakun et al., 2012), or are available as geographically explicit data sets only for a limited number of climate variables in a few regions such as North America or Europe (Davis et al., 2003;Gajewski, 2015;Mauri et al., 2014;Viau and Gajewski, 2009;Viau et al., 2006). The only exception to this is the Global Lake Status Data Base (Kohfeld and Harrison, 2000), which provides qualitative estimates of the change in lake water balance through time globally. The reliability of global temperature estimates depends on the representativeness of the data included; this point has been made abundantly clear from comparisons of records of hemispheric temperature changes during the last millennium (Fernández-Donado et al., 2013;Moberg, 2013). Currently available reconstructions of global temperature changes during the Holocene are heavily biased towards marine records, making it imperative that the reliability of these records is assessed using continental reconstructions (Davis et al., 2015;Liu et al., 2014). The lack of geographically explicit reconstructions for tropical regions and the SH would limit analysis of the Holocene transient simulations, but efforts are underway to improve this situation.
The LIG is the most suitable of the pre-Holocene interglacial periods as a focus in PMIP4-CMIP6 because of the relative wealth of data compared to earlier interglacial periods. However, there is an order of magnitude less information than for the Holocene, and there are larger uncertainties in dating of specific events. This means that the LIG data-model comparisons will focus on large-scale features, such as the strength of the high-latitude amplification of warming and the role of snow and sea-ice feedbacks in this warming. There will also be a major focus on the tropical water cycle. These analyses will exploit available data sets for the LIG which mostly document surface sea and air temperatures across the globe (Anderson et al., 2006;Brewer et al., 2008;Capron et al., 2014;Hoffman et al., 2017;McKay et al., 2011;Turney and Jones, 2010), although recent efforts also synthesize reconstructions of sea-ice changes (Esper and Gersonde, 2014;Sime et al., 2013), of the deep ocean circulation (Oliver et al., 2010), and to a lesser extent the tropical hydrological cycle . In addition, several existing maps are reporting vegetation changes at the NH high latitudes (Bennike et al., 2001) and changes in lake area in the Sahara (Petit-Maire, 1999).
There are also syntheses of quantitative climate reconstructions for the LIG (Turney and Jones, 2010;McKay et al., 2011;Capron et al., 2014;Hoffman et al., 2017), which have been used for model evaluation (e.g., Lunt et al., 2013;Otto-Bliesner et al., 2013). A critical evaluation of these LIG data syntheses is available in Capron et al. (2017), and we summarize here key aspects of the comparison. The major limitation in using the data syntheses by Turney and Jones (2010) and McKay et al. (2011) for analysis of the lig127k simulations and associated sensitivity experiments is that they are compilations of information about the maximum warmth during the LIG. Given that warming was not synchronous globally (Bauch and Erlenkeuser, 2008;Cortese et al., 2007;NEEM Community Members, 2013;Govin et al., 2012;Masson-Delmotte et al., 2010;Mor et al., 2012;Winsor et al., 2012), these syntheses do not represent a specific time slice. A more recent compilation by Capron et al. (2014) has used harmonized chronologies for ice and marine records to produce records of the change in high-latitude temperature compared to present for four 2000-year long time slabs, and this approach has been expanded to include the fifth time slab (128-126 ka) for comparison with the lig127k simulation (Capron et al., 2017). Following a similar strategy, Hoffman et al. (2017) propose the first global marine compilation with harmonized chronologies for the LIG, with time slabs available at 129 and 125 ka but not 127 ka (note that a high-latitude subset of the Hoffman et al. (2017) compilation at 127 ka is available in Capron et al., 2017). However, even though these compilations are based on harmonized chronologies, dating uncertainties during the LIG can still be several thousand years depending on the type of archive and the dating methods (Govin et al., 2015). Furthermore, the different response scales of different components of the climate system mean that records from the 128-126 ka time slab may still bear the imprint of the previous deglaciation ( Fig. 1) (Capron et al., 2017). In any case, and as with the early Holocene experiments, the lig127k simulation will not solely reflect the insolation forcing. It is therefore recommended that data-model comparisons focus on using the temporal evolution of climate, as captured in the available Capron et al. (2014) and the Hoffman et al. (2017) time series, to assess the plausibility of the lig127k simulation.
The public-access reconstruction data sets currently available for the mid-Holocene and LIG serve different functions and address different aspects of the climate system. Modeling groups running mid-Holocene and LIG simulations, or sensitivity experiments, are encouraged to use multiple reconstruction data sets for a full diagnosis of the simulations. Many of these data sets provide measures of the uncertainty of the reconstructions and data-model comparisons should be designed to take these uncertainties into account.

Conclusions
The PMIP4-CMIP6 midHolocene and lig127k simulations provide an opportunity to examine the impact of two different changes in radiative forcing on climate at times when other forcings were relatively similar to present. Together with planned sensitivity experiments, this focus on the two interglacials will allow us to explore the role of feedbacks in the climate system and to quantify their contribution to largescale phenomena relevant to future projections such as landsea contrast and high-latitude amplification of temperature changes. They will also allow us to address the implications of changes in forcing and feedbacks for the tropical circulation and monsoons -again an issue that is relevant to interpreting future projections. Given that both periods have been foci for model-model and data-model comparisons during previous phases of PMIP, a major focus during CMIP6 will be on evaluating the realism of the midHolocene and lig127k simulations using a wide range of paleoenvironmental data and paleoclimate reconstructions. This evaluation will be a direct out-of-sample test of the reliability of state-of-the-art models to simulate climate changes, and particularly the climate warming.
Neither one of these interglacial simulations is a perfect analog for the future, and each interglacial has distinct differences in forcings and in the initial state of the climate system. In a sense this is advantageous because it allows us to investigate the response of the system under different conditions. Sensitivity studies allow us to assess which results may be directly transferrable to future climate projections. In the case of the midHolocene simulations, we have the advantageous ability to assess Earth's response to elevated boreal summer insolation alone, with which we can compare model results against a plethora of observations. Estimated boreal summer warming for this period is roughly equivalent to the summer warming simulated for the mid to late 21st century. In the case of the lig127k simulations, we have another advantageous end member where boreal summer insolation was even greater than the middle Holocene and that forced eventually higher-than-present sea levels. Higher temperatures in the polar regions, particularly during the summer months, directly influence sea ice and the ice sheets. The data evidence provides a means of evaluating whether we are capturing this sensitivity correctly in models being used for projections of future climate change. Consequently, this provides a potential imperfect analog for the end of this century and beyond. For example, Blaschek et al. (2015) found that the influence of freshwater forcing due to Greenland ice-sheet melting is the same, regardless of the background climate. In other cases, the response may be more strongly dependent upon the initial state, such as the response of polar amplification in Greenland, which was found to be sensitive to the prescribed ice-sheet elevation (Masson-Delmotte et al., 2006). PMIP4 will collaborate with other CMIP6 projects (Kagayema et al., 2016). The output from the lig127k simulation, for example, will be used to force standalone ice-sheet experiments (ism-lig127k-std) in ISMIP6. This will complement the suite of standalone ISMIP6 ice-sheet experiments (Nowicki et al., 2016; http://www.climate-cryosphere.org/ activities/targeted/ismip6) for the recent past and future and will add to increase our understanding of the ice-sheet sensitivity to climate changes. The PMIP4-CMIP6 midHolocene and lig127k simulations, and associated sensitivity experiments, are also relevant to analyses of sea-ice feedbacks to climate in SIMIP (Notz et al., 2016) and to assessments of the role of dust forcing by AerChemMIP (Collins et al., 2017). Beyond CMIP6, the planned PMIP4-CMIP6 interglacial simulations are relevant to the Grand Challenges set by the World Climate Research Programme (WCRP). Both the midHolocene and the lig127k simulations are relevant to the Grand Challenge "Clouds, Circulation and Climate Sensitivity", which has a major focus on the controls on the monsoon circulation. Also, the lig127k simulation is particularly relevant to the Grand Challenge "Melting Ice and Global Consequences", which addresses the stability of the ice sheets. Those simulations carried out with biogeochemical cycles enabled are relevant to the Grand Challenge "Carbon Feedbacks in the Climate System".
Data availability. The forcing and boundary condition data sets described in this paper are available in the PMIP4 repository (2017) https://pmip4.lsce.ipsl.fr/doku.php/exp_design:index. After final acceptance of this paper, they will be made available also through Input4MIPs (https://esgf-node.llnl.gov/projects/ input4mips/; see the "Input4MIPs summary" living document there on the progress of this process).
The Tier 1 midHolocene and lig127k simulations are part of the CMIP6 experiment family and data will be distributed through the official CMIP6 channels via the Earth System Grid Federation (ESGF, https://earthsystemcog.org/projects/wip/ CMIP6DataRequest). A minimum of 100 years of output is required to be uploaded for each simulation (usually the final 100 years of the simulation). However, given the increasing interest in analyzing multi-decadal variability (e.g., Wittenberg, 2009) and the availability of reconstructions of ENSO (El Niño-Southern Oscillation) and other modes of variability (see Sect. 4), modeling groups are encouraged to provide outputs for at least 500 years if possible.
In addition, the difference in orbital configuration between 127 ka, 6 ka and preindustrial means that there are differences in month and season length that should be accounted for in calculating seasonal changes (Kutzbach and Gallimore, 1988). To be able to account for this effect when comparing the simulations to the paleoclimate reconstructions, daily outputs of at least surface 2 m temperature (tas), precipitation (pr) and 10 m winds (uas) should be archived. If not possible, a less accurate but probably adequate approach would be to use a program that provides an approximate estimate of monthly means on the fixed-angular celestial calendar from the fixed-day calendar.
Data from PMIP4-only Tier 2 and 3 simulations must be processed following the same standards as Tier 1 for data processing (e.g., CMOR standards) and should be distributed via the PMIP4 ESGF or the CMIP6 ESGF Tier 2 and Tier 3 databases. Modeling groups producing these simulations are responsible for securing suitable space on ESGF nodes. These experiments will follow the same naming, variable convention and format, and documentation requests as the Tier 1 PMIP4-CMIP6 experiment so as to be compliant with ESGF database requirements.
The list of variables requested for the PMIP4-CMIP6 paleoclimate experiments can be found here: http://clipc-services. ceda.ac.uk/dreq/u/PMIP.html. This request is presently processed by the CMIP6 Working Group for Coupled Modeling Infrastructure Panel (WIP) into tables, which define the variables included in the data request to the modeling groups for data to be contributed to the archive. The most up-to-date list including all variables requested for CMIP6 can be found at the WIP site: http://proj.badc.rl.ac.uk/svn/exarch/CMIP6dreq/tags/ latest/dreqPy/docs/CMIP6_MIP_tables.xlsx.
The last two columns in each row list MIPs associated with each variable. The first column in this pair lists the MIPs, which request the variable in one or more experiments. The second column lists the MIPs proposing experiments in which this variable is requested.
As the Supplement to this paper we provide version 1.00.05 (April 2017). We note, however, that this document is still in development and that inconsistencies may still exist.
The only variables defined specifically in PMIP are those describing oxygen isotopes for model systems that calculate these data interactively .
Acknowledgements. Bette L. Otto-Bliesner acknowledges the funding by the U.S. National Science Foundation of the National Center for Atmospheric Research. Pascale Braconnot and Sandy P. Harrison acknowledge funding from JPI-Belmont project "Paleao-Constraints on Monsoon Evolution and Dynamics (PACMEDY)" through their national funding agencies. Sandy P. Harrison