The Last Interglacial (LIG; 130–115 ka) is an important test bed for climate science as an instance of significantly warmer than preindustrial global temperatures. However, LIG climate patterns remain poorly resolved, especially for winter, affected by a suite of strong feedbacks such as changes in sea-ice cover in the high latitudes. We present a synthesis of winter temperature and precipitation proxy data from the Atlantic seaboard of Europe, spanning from southern Iberia to the Arctic. Our data reveal distinct, opposite latitudinal climate trends, including warming winters seen in the European Arctic while cooling and drying occurred in southwest Europe over the LIG. Climate model simulations for 130 and 120 ka suggest these contrasting climate patterns were affected by a shift toward an atmospheric circulation regime with an enhanced meridional pressure gradient and strengthened midlatitude westerlies, leading to a strong reduction in precipitation across southern Europe.The Last Interglacial (LIG; ca. 130–115 ka; roughly coincident with marine isotope stage 5e) is an important case study for the response of the climate system to changes in climate forcing, with negligible human impact and only slight changes in total greenhouse gas forcing, but some of the largest anomalies in orbitally driven insolation forcing during the past million years (Berger and Loutre, 1991). This combination of factors led to the last known instance of warmer than preindustrial global temperatures in geological history, with average global annual sea-surface temperature estimated at +0.5 ± 0.3 °C above preindustrial (1870–1889 CE) levels (Hoffman et al., 2017) and summer surface warming reconstructed at up to 4–5 °C over Arctic land areas (CAPE–Last Interglacial Project Members, 2006), with the global surface temperature anomaly estimated at approximately +0.8 °C (Fischer et al., 2018).While the LIG is thus an important target for testing the sensitivity and robustness of climate model simulations under a range of forcing regimes, the evolution of LIG winter climate has been especially difficult to unravel due to the model dependence on sea-ice feedbacks, which strongly affect high-latitude winter temperatures (Bakker et al., 2013). In Europe, ensembles of transient climate model simulations show poor agreement in winter temperature trends, and often opposite trends compared to available proxy data, while for summer, individual models and proxy data show a consistent early LIG temperature maximum in mid- and high latitudes (Brewer et al., 2008; Bakker et al., 2013; Salonen et al., 2018). The modeling challenges are compounded by the paucity of proxy data from the high northern latitudes due to erosion of LIG deposits during the last glaciation (Helmens, 2014).To elucidate European winter climate evolution during the LIG, we present a data–modeling synthesis with high-resolution climate proxy archives spanning along the Atlantic seaboard from southern Iberia to the Arctic. Our data reveal strongly contrasting trends in the northern and southern European winter temperature and precipitation, consistent with a shift toward an atmospheric circulation regime with strengthened midlatitude westerlies seen in early and late LIG climate model experiments.We reconstructed southwest European winter temperature and precipitation using pollen data from three marine cores drilled from R/V Marion Dufresne during the IMAGES cruises of 1995, 1999, and 2004 (Fig. 1; for details about the marine cores, see Sánchez Goñi et al. [2018]). Cores MD99–2331 from the Galician margin (42.15°N, 9.68°W) and MD04–2845 from the Bay of Biscay (45.35°N, 5.22°W) captured pollen rain from the temperate forest of the adjacent landmass, with winter temperature as a primary ecological control. We inferred winter temperature qualitatively as the sum of temperate tree pollen and quantitatively as pollen-based mean January temperature (TJan) reconstructions. Core MD95–2042 was drilled farther south at the southwest Iberian margin (37.80°N, 10.17°W), and in the Mediterranean climate zone, with pollen data primarily influenced by winter precipitation. To represent winter precipitation variations, we used the pollen sums of Mediterranean forest taxa and semiarid taxa. Several isotopic stratigraphic events were identified and dated by Shackleton et al. (2003) in core MD95–2042 and used to develop a linear-interpolation age model. The chronologies of cores MD99–2331 and MD04–2845 were based on correlating forest increases to equivalent pollen stratigraphic features identified and dated in core MD95–2042 (Sánchez Goñi et al., 2005, 2012).For northern Europe, we used the Sokli sequence from northern Finland (67.80°N, 29.30°E), with pollen-based reconstructions of both January and July (TJul) mean temperatures (from Salonen et al., 2018) and supporting aquatic proxy data for winter conditions (Plikk et al., 2016, 2019; Kylander et al., 2018). In the Sokli age model, the interglacial onset and end identified in the pollen data are aligned with the start and end of interglacial conditions in northern and central European U/Th-dated speleothems (Salonen et al., 2018).We reconstructed TJan from cores MD99–2331 and MD04–2845 using the same method as that used earlier for Sokli (Salonen et al., 2018). This method uses an ensemble of six pollen–climate calibration models, but with the calibration data set here expanded southward to include the Mediterranean climate zone and a total of 1295 surface pollen samples (see the Supplemental Material1). Each reconstruction was summarized as the median of the six models, and with a LOESS (locally estimated scatterplot smoothing) smoother (span = 0.1) fitted to the median. The performance of the calibration models was estimated by 10-fold cross-validations, suggesting coefficients of determination (R2) ranging from 0.72 to 0.91 and reconstruction errors (root-mean-square error of prediction) from 2.08 °C to 3.67 °C (for further details, see Figs. S1 and S2 and Table S1 in the Supplemental Material).We discuss here results from two global climate model simulations for 130 ka and 120 ka, performed by Otto-Bliesner et al. (2013) using the fully coupled Community Climate System model, version 3 (CCSM3, https://www.cesm.ucar.edu/models/ccsm3.0/), with dynamic components for the atmosphere–land surface–ocean–sea ice system (Collins et al., 2006). The main difference between these two simulations concerns changes in orbital parameters, leading to a marked seasonal difference in the incoming solar insolation at the top of the atmosphere at 50°N decreasing by 20 W/m2 from 130 to 120 ka in July, but increasing by 10 W/m2 in January (Berger, 1978). In addition, the atmospheric levels of greenhouse gases CO2 and CH4 declined, as CO2 decreased from 300 ppmv (parts per million volume) at 130 ka to 272 ppmv at 120 ka, and CH4 decreased from 720 ppbv at 130 ka to 570 ppbv at 120 ka. All other boundary conditions (including land-ocean distribution, vegetation, ice sheets, solar constant) were kept fixed at modern values. The applied model version has a T85 spectral resolution in the atmosphere, corresponding to ∼1.4° latitude-longitude grid spacing. The ocean model has a horizontal resolution of ∼1° latitude-longitude.In northern Europe, the TJan reconstruction from Sokli shows a strong warming trend spanning the LIG (Fig. 2B). In the underlying pollen data, distinct plant indicators of oceanic climate like Corylus (hazel), Quercus (oak), and Osmunda (royal fern) (Birks and Paus, 1991; Salonen et al., 2012; Seppä et al., 2015) endure in the late LIG during the summer cooling trend (Fig. 2A). The mild late-LIG winters are supported by diatom data from the same sequence, indicating a reduced seasonality (Plikk et al., 2016). These multiproxy data from Sokli are further corroborated by the Mertuanoja sequence from western Finland, which, while of low resolution, shows a distinct late-LIG peak in Osmunda (Eriksson et al., 1999). Reliable precipitation or water balance proxy indicators are rare in northern Europe. At Sokli, chironomid, plant macrofossil, diatom, and geochemical data indicate increased chemical weathering and input of dissolved organic carbon during the late LIG, which may reflect increased precipitation, but which may also be related to an expansion of Picea (spruce) in the catchment (Plikk et al., 2016, 2019; Kylander et al., 2018). By contrast, in southern Europe, a winter cooling trend is indicated by the decreasing temperate forest pollen percentages and the falling reconstructed TJan in cores MD04–2845 (Figs. 2D and 2E) and MD99–2331 (Figs. 2F and 2G), while a winter drying trend is suggested by the decrease in Mediterranean forest pollen (Fig. 2H) and the increase in semiarid pollen (Fig. 2I) in core MD95–2042. In the intervening central European sector, the proxy data are more ambiguous, with the synthesis of Zagwijn (1996), which incorporated pollen and plant macrofossil data from 31 core sequences from 45°N to 56°N and used the reconstructed paleo-isotherms to extract a paleotemperature curve for one site (Amsterdam), showing a mid-LIG maximum for TJan (Fig. 2C). However, other studies from individual sites at this latitude band have indicated an early-LIG TJan maximum, followed by either moderate (Kühl and Litt, 2003) or strong (Field et al., 1994; Brewer et al., 2008) cooling.Our data thus show strong and opposite trends in northern European (warming) and southern European (cooling and drying) winter climate over the LIG. Importantly, the reconstructed early-LIG (ca. 130–125 ka) timing of the southern European TJan maximum contradicts the modeling of Bakker et al. (2013), which suggested a TJan maximum at 120–119 ka in northern Spain and western France, but with large uncertainties (standard deviation ∼4 k.y.). No timing was established by Bakker et al. (2013) for our northern European data region, as most of the ensemble models were unable to find a TJan peak outside the errors of the LIG mean.The latitudinal pattern of temperature and precipitation trends seen in our data is broadly consistent with the spatial signature of a shift in North Atlantic Oscillation (NAO)–type circulation, from a more negative (NAO–) state in the early LIG to a more positive (NAO+) state in the late LIG (Wanner et al., 2001). The simulated sea-level pressure (SLP) patterns for 130 ka (Fig. 3A) and 120 ka (Fig. 3B) show a more extensive Icelandic low, and also a more expressed Azores high at 120 ka, while the south-to-north SLP gradient becomes steeper at 120 ka (Fig. 3C). The NAO index values, calculated based on the difference of normalized SLP between Lisbon (Portugal) and Stykkisholmur/Reykjavik, show a shift from slightly negative (−0.245) at 130 ka to moderately positive (+0.829) at 120 ka. Zonal winds become stronger at 120 ka compared to 130 ka over most of northwest Europe (Fig. 3D), consistent with a steeper SLP gradient at 120 ka. For winter precipitation, the modeled trends (Fig. 3E) share some large-scale features of a NAO+ signature (Wanner et al., 2001; Gouveia et al., 2008), with a tendency toward increased precipitation over northern Europe, shifting to a drying trend at approximately 50°N and with reduced precipitation also modeled around southern Greenland. Our precipitation proxy data site, MD95–2042, is located within a broad latitudinal band marked by a drying trend, and here the modeled trend matches the proxy data (Figs. 2H and 2I). However, the LIG TJan patterns (Fig. 3F) deviate from a classical, mean NAO+ signature, which is characterized by a positive winter temperature anomaly over most of Europe north of 40°N (Wanner et al., 2001; Gouveia et al., 2008). Instead, a negative TJan trend is modeled over northwest Europe, similar to an east-west tilted NAO+ configuration, but lacking the cooling belt along the Mediterranean (Rousi et al., 2020). In consequence, both our northern (Sokli) and southern European (cores MD99–2331 and MD04–2845) temperature proxy sites are located in regions where the sign of the modeled trend changes. This is also true for the central European zone, which has a high density of LIG sites (Zagwijn, 1996) showing opposite temperature trends modeled across a southwest-northeast–trending line cutting across France and Germany. While some of the differences in published winter temperature reconstructions have been attributed to different reconstruction algorithms (Kühl and Litt, 2003), the discrepancies may thus also be underlain by real shifts in the temperature trend within central and western Europe.The change to a regime with an enhanced meridional pressure gradient and strengthened midlatitude westerlies, as suggested by our data and modeling, was likely forced by the decrease in summer insolation (Fig. 2L), which resulted in expansion of the ice sheets, a more extensive sea-ice cover and snowpack, and a shift from taiga to tundra vegetation commencing in climate modeling at ca. 122 ka (Crucifix and Loutre, 2002) and supported by European proxy data sets showing a major concurrent southward displacement in the deciduous tree line (Sánchez Goñi et al., 2005). The insolation forcing and albedo feedbacks significantly amplified the north-south thermal gradient over ca. 122–120 ka (Sánchez Goñi et al., 2005), leading to a more vigorous atmospheric circulation, implying stronger winds after the LIG temperature maximum at ca. 130–125 ka (Rind, 2000). Our data are broadly consistent with the modeled 122 ka timing of the mid-LIG transition (Crucifix and Loutre, 2002), with the best breakpoints (Gurarie, 2014) in the interglacial sections of all seven winter proxy series timed at 122.2 ± 1.6 ka (Figs. 2B, 2D–2I).The strong winter warming trend seen in the northern European proxy data could have been influenced by the Atlantic Meridional Overturning Circulation (AMOC), which, based on transient modeling experiments (Bakker et al., 2013) and proxy data, continued as vigorous or even strengthened (Sánchez Goñi et al., 2012; Galaasen et al., 2014) through the late LIG (Fig. 2J), and proximally by the warm late-LIG Nordic seas (Fig. 2K) (Zhuravleva et al., 2017). However, the reverse trends of strong winter cooling and drying in southwest Europe run against these forcing mechanisms, as well as the rising winter insolation in the midlatitudes (Fig. 2M), and so they would need an additional explanatory factor. Also, the modeled changes in TJan are within ±0.5°C (Fig. 3F) compared to the reconstructed TJan decrease of up to 10°C over the LIG (Figs. 2D and 2F). The muted modeled temperature changes could be the result of missing feedbacks due to the fixed vegetation and the fixed ice sheets in Greenland and Antarctica. Moreover, the preindustrial control run of the CCSM3 model shows a cold bias in the North Atlantic region (Lunt et al., 2013), possibly leading to a damped cooling response due to restricted capacity for warming at 130 ka. By comparison, the modeling and data agree on the southwest European precipitation trend, with a strong reduction in modeled January precipitation (Fig. 3E) likely sufficient to explain the response of the southern Iberian vegetation (Figs. 2H and 2I). In modern southwest Iberia, vegetation dynamics are directly controlled by the variation in the regional influence of the westerlies, with a more northward trajectory of the westerlies seen under modern NAO+ states inducing reduced vegetation activity (Gouveia et al., 2008), a response consistent with the reduction of the southwest Iberian Mediterranean forest in the late LIG.Salonen acknowledges funding from the Academy of Finland (project 310649). We thank Karin Helmens (Swedish Museum of Natural History) and three anonymous reviewers for comments, Anastasia Zhuravleva (GEOMAR Helmholtz Centre for Ocean Research Kiel, Germany) for help in acquiring data, and Bette Otto-Bliesner (U.S. National Center for Atmospheric Research, Colorado, USA) for making CCSM3 model output available for further analysis.

中文翻译：

---

最后一次间冰期期间北欧和南欧冬季气候趋势的对比

最后一次间冰期 (LIG; 130–115 ka) 是气候科学的重要试验台，是全球气温明显高于工业化前的一个例子。然而，LIG 气候模式仍然得不到很好的解决，尤其是在冬季，受到一系列强烈反馈的影响，例如高纬度地区海冰覆盖的变化。我们展示了欧洲大西洋沿岸冬季温度和降水代理数据的综合，从伊比利亚南部到北极。我们的数据揭示了截然不同的、相反的纬度气候趋势，包括欧洲北极地区的冬季变暖，而欧洲西南部的 LIG 地区则出现冷却和干燥。130 和 120 ka 的气候模型模拟表明，这些对比鲜明的气候模式受到大气环流模式转变的影响，该模式具有增强的经向压力梯度和增强的中纬度西风，导致整个欧洲南部的降水量大幅减少。 最后一次间冰期 (LIG) ；约 130–115 ka；大致与海洋同位素阶段 5e 一致）是气候系统对气候强迫变化的响应的一个重要案例研究，人类影响可以忽略不计，温室气体强迫总量变化很小，但有些过去一百万年中轨道驱动的日照强迫的最大异常（Berger 和 Loutre，1991）。这些因素的结合导致了地质历史上最后一次已知的比工业化前全球气温更高的例子，估计全球年平均海面温度比工业化前（公元 1870-1889 年）水平高 +0.5 ± 0.3 °C（Hoffman 等人，2017 年），北极陆地地区夏季地表升温高达 4-5 °C (CAPE–Last Interglacial Project Members, 2006)，全球地表温度异常估计约为 +0.8 °C (Fischer et al., 2018)。因此，LIG 是测试气候模型敏感性和稳健性的重要目标通过一系列强迫机制下的模拟，LIG 冬季气候的演变特别难以解开，因为模型依赖于海冰反馈，这强烈影响高纬度冬季温度（Bakker 等，2013）。在欧洲，瞬态气候模型模拟的集合显示冬季温度趋势不一致，与可用的代理数据相比，趋势往往相反，而对于夏季，个别模型和代理数据显示中高纬度地区的早期 LIG 温度最大值一致（Brewer 等人，2008 年；Bakker 等人，2013 年；Salonen 等人., 2018)。由于上次冰期期间 LIG 沉积物的侵蚀（Helmens，2014 年），来自北高纬度地区的代理数据缺乏，使建模挑战更加复杂。为了阐明 LIG 期间欧洲冬季气候的演变，我们提出了一个数据建模综合沿大西洋沿岸从伊比利亚南部到北极的高分辨率气候代理档案。我们的数据揭示了北欧和南欧冬季温度和降水的强烈对比趋势，与在早期和晚期 LIG 气候模型实验中看到的中纬度西风带加强的大气环流状态的转变一致。我们使用来自 R/V Marion Dufresne 在1995、1999 和 2004（图 1；有关海洋岩心的详细信息，请参见 Sánchez Goñi 等人 [2018]）。来自加利西亚边缘（42.15°N，9.68°W）的 MD99-2331 和来自比斯开湾（45.35°N，5.22°W）的 MD04-2845 核心捕获了来自邻近大陆温带森林的花粉雨，具有冬季温度作为主要的生态控制。我们将冬季温度定性推断为温带树木花粉的总和，定量推断为基于花粉的一月平均温度 (TJan) 重建。岩心 MD95-2042 在伊比利亚西南部边缘（37.80°N，10.17°W）和地中海气候区的更南处钻探，花粉数据主要受冬季降水的影响。为了表示冬季降水变化，我们使用了地中海森林类群和半干旱类群的花粉总量。Shackleton 等人确定并确定了几个同位素地层事件。(2003) 在核心 MD95-2042 中，并用于开发线性插值年龄模型。核心 MD99-2331 和 MD04-2845 的年代学基于将森林增加与在核心 MD95-2042 中确定和测年的等效花粉地层特征相关联（Sánchez Goñi 等人，2005 年，2012 年）。对于北欧，我们使用了 Sokli来自芬兰北部的序列（67.80°N，29.30°E），基于花粉重建 1 月和 7 月 (TJul) 平均温度（来自 Salonen 等人，2018 年）和支持冬季条件的水生代理数据（Plikk 等人，2016 年，2019 年；Kylander 等人，2018 年）。在 Sokli 年龄模型中，花粉数据中确定的间冰期开始和结束与北欧和中欧 U/Th 日期洞穴的间冰期条件的开始和结束一致（Salonen 等，2018）。我们从 TJan 重建核心 MD99-2331 和 MD04-2845 使用与之前用于 Sokli 的方法相同的方法（Salonen 等，2018）。该方法使用六个花粉气候校准模型的集合，但此处的校准数据集向南扩展以包括地中海气候区和总共 1295 个地表花粉样本（参见补充材料 1）。每次重建都被总结为六个模型的中值，并使用 LOESS（局部估计的散点图平滑）平滑器（跨度 = 0.1）拟合中值。校准模型的性能通过 10 倍交叉验证进行估计，表明决定系数 (R2) 范围为 0.72 至 0.91，重建误差（预测的均方根误差）范围为 2.08 °C 至 3.67 °C（有关更多详细信息，请参见补充材料中的图 S1 和 S2 以及表 S1。我们在这里讨论了由 Otto-Bliesner 等人进行的 130 ka 和 120 ka 的两个全球气候模型模拟的结果。(2013) 使用完全耦合的社区气候系统模型，第 3 版（CCSM3，https://www.cesm.ucar.edu/models/ccsm3.0/），具有大气-陆地表面-海洋-海洋的动态分量冰系统（柯林斯等人，2006 年）。这两个模拟之间的主要区别涉及轨道参数的变化，导致大气顶部在 50°N 的入射太阳日照出现明显的季节性差异，从 7 月份的 130 到 120 ka，减少了 20 W/m2，但增加了1 月份降低了 10 W/m2 (Berger, 1978)。此外，温室气体 CO2 和 CH4 的大气水平下降，因为 CO2 从 130 ka 时的 300 ppmv（百万分率）降至 120 ka 时的 272 ppmv，而 CH4 从 130 ka 时的 720 ppbv 降至 120 时的 570 ppbv K a。所有其他边界条件（包括陆海分布、植被、冰盖、太阳常数）都保持固定在现代值。应用的模型版本在大气中具有 T85 光谱分辨率，对应约 1.4° 的经纬度网格间距。海洋模型的水平分辨率约为 1° 纬度-经度。在北欧，Sokli 的 TJan 重建显示出跨越 LIG 的强烈变暖趋势（图 2B）。在基础花粉数据中，海洋气候的不同植物指标如 Corylus（榛树）、Quercus（橡树）和 Osmunda（皇家蕨类植物）（Birks 和 Paus，1991 年；Salonen 等人，2012 年；Seppä 等人，2015 年）在夏季降温趋势期间忍受 LIG 后期（图 2A）。来自相同序列的硅藻数据支持温和的 LIG 晚期冬季，表明季节性减少（Plikk 等，2016）。这些来自 Sokli 的多代理数据得到了来自芬兰西部的 Mertuanoja 序列的进一步证实，虽然分辨率低，但在 Osmunda 显示了一个明显的后期 LIG 峰（Eriksson 等，1999）。可靠的降水或水平衡代理指标在北欧很少见。在 Sokli，摇蚊、植物大化石、硅藻和地球化学数据表明 LIG 晚期化学风化和溶解有机碳输入增加，这可能反映了降水增加，但这也可能与云杉（云杉）在流域（Plikk 等人，2016 年，2019 年；Kylander 等人，2018 年）。相比之下，在欧洲南部，温带森林花粉百分比下降和核心 MD04-2845（图 2D 和 2E）和 MD99-2331（图 2F 和 2G）中重建的 TJan 下降表明冬季降温趋势，而地中海森林花粉的减少（图 2H）和核心 MD95-2042 中半干旱花粉的增加（图 2I）表明冬季干燥趋势。在介入的中欧部分，代理数据更加模糊，Zagwijn (1996) 综合了来自 45°N 到 56°N 的 31 个核心序列的花粉和植物大化石数据，并使用重建的古等温线提取一个地点（阿姆斯特丹）的古温度曲线，显示 TJan 的中 LIG 最大值（图 2C）。然而，来自该纬度带个别地点的其他研究表明，LIG TJan 早期出现最大值，然后是中度（Kühl 和 Litt，2003 年）或强（Field 等人，1994 年；Brewer 等人，2008 年）降温。因此，我们的数据显示了 LIG 上北欧（变暖）和南欧（变冷和干燥）冬季气候的强烈和相反的趋势。重要的是，重建的早期 LIG（约 130–125 ka) 南欧 TJan 最大值的时间与 Bakker 等人的建模相矛盾。(2013)，这表明西班牙北部和法国西部的 TJan 最大值为 120-119 ka，但具有很大的不确定性（标准偏差~4 ky）。Bakker 等人没有确定时间。(2013) 对于我们的北欧数据区域，因为大多数集合模型无法在 LIG 均值误差之外找到 TJan 峰值。我们数据中看到的温度和降水趋势的纬度模式与空间特征大致一致北大西洋涛动 (NAO) 型环流的转变，从早期 LIG 的更负 (NAO-) 状态到晚期 LIG 的更正 (NAO+) 状态（Wanner 等，2001）。130 ka（图3A）和120 ka（图3A）的模拟海平面压力（SLP）模式。图 3B) 显示了更广泛的冰岛低压，以及在 120 ka 处表现得更明显的亚速尔群岛高压，而从南到北的 SLP 梯度在 120 ka 处变得更陡峭（图 3C）。NAO 指数值根据里斯本（葡萄牙）和 Stykishholmur/Reykjavik 之间标准化 SLP 的差异计算，显示从 130 ka 时的轻微负值 (-0.245) 转变为 120 ka 时的中度正值 (+0.829)。与欧洲西北部大部分地区的 130 ka 相比，120 ka 的纬向风变得更强（图 3D），这与 120 ka 时更陡的 SLP 梯度一致。对于冬季降水，模拟趋势（图 3E）具有 NAO+ 特征的一些大尺度特征（Wanner 等人，2001 年；Gouveia 等人，2008 年），在北欧有降水增加的趋势，在大约 50°N 时转变为干燥趋势并且降水减少也模拟了格陵兰岛南部。我们的降水代理数据站点 MD95-2042 位于以干燥趋势为标志的宽阔纬度带内，这里的建模趋势与代理数据相匹配（图 2H 和 2I）。然而，LIG TJan 模式（图 3F）偏离了经典的平均 NAO+ 特征，其特征是欧洲大部分地区 40°N 以北的冬季温度异常（Wanner 等人，2001 年；Gouveia 等人，2001 年）。 , 2008)。相反，在欧洲西北部模拟了负 TJan 趋势，类似于东西向的 NAO+ 配置，但缺少地中海沿岸的冷却带（Rousi 等，2020）。结果，我们的北欧（Sokli）和南欧（核心 MD99-2331 和 MD04-2845）温度代理站点都位于模拟趋势符号发生变化的区域。中欧地区也是如此，该地区拥有高密度的 LIG 站点 (Zagwijn, 1996)，显示出相反的温度趋势，模拟横跨法国和德国的西南-东北-趋势线。虽然已发表的冬季温度重建中的一些差异归因于不同的重建算法（Kühl 和 Litt，2003 年），但这些差异也可能是由于中欧和西欧温度趋势的实际变化造成的。正如我们的数据和模型所表明的那样，具有增强的经向压力梯度和增强的中纬度西风，可能是由于夏季日照减少（图 2L），导致冰盖扩大，海冰覆盖和积雪更广泛，以及从针叶林到苔原植被的转变，在大约 20 年的气候建模中开始。122 ka（Crucifix 和 Loutre，2002 年）并得到欧洲代理数据集的支持，显示落叶树线中的主要并发向南位移（Sánchez Goñi 等人，2005 年）。日照强迫和反照率反馈显着放大了大约 10 小时内的南北热梯度。122–120 ka (Sánchez Goñi et al., 2005)，导致更活跃的大气环流，这意味着在约 130–125 ka（林德，2000 年）。我们的数据与 LIG 中期过渡的建模 122 ka 时间大致一致（Crucifix 和 Loutre，2002），在所有七个冬季代理系列的间冰期剖面中具有最佳断点（Gurarie，2014），时间为 122.2 ± 1.6 ka（图 2B，2D-2I）。在北欧代理数据中看到的强烈冬季变暖趋势可能是受到大西洋经向翻转环流 (AMOC) 的影响，该环流基于瞬态模拟实验（Bakker 等人，2013 年）和代理数据，继续保持强劲甚至加强（Sánchez Goñi 等人，2012 年；Galaasen 等人， 2014 年）通过 LIG 晚期（图 2J），近端由温暖的 LIG 晚期北欧海（图 2K）（Zhuravleva 等人，2017 年）。然而，欧洲西南部冬季强烈降温和干燥的反向趋势与这些强迫机制以及中纬度地区冬季日照增加（图 2M）背道而驰，因此需要额外的解释因素。还，与 LIG 上重建的 TJan 降低高达 10°C（图 2D 和 2F）相比，模拟的 TJan 变化在 ±0.5°C 以内（图 3F）。温和的模拟温度变化可能是由于格陵兰岛和南极洲的固定植被和固定冰盖而导致反馈缺失的结果。此外，CCSM3 模型的工业化前控制运行显示北大西洋地区存在冷偏差（Lunt 等人，2013 年），由于 130 ka 的升温能力受限，可能导致冷却响应减弱。相比之下，建模和数据与欧洲西南部的降水趋势一致，模拟的 1 月降水量大幅减少（图 3E）可能足以解释伊比利亚南部植被的响应（图 2H 和 2I）。在现代西南部伊比利亚，植被动态直接受西风带区域影响变化的控制，在现代 NAO+ 状态下看到的西风带向北移动，导致植被活动减少（Gouveia 等，2008），这种响应与LIG.Salonen 晚期伊比利亚地中海西南部森林感谢芬兰科学院的资助（项目 310649）。我们感谢 Karin Helmens（瑞典自然历史博物馆）和三位匿名审稿人的评论，感谢 Anastasia Zhuravleva（GEOMAR Helmholtz 海洋研究中心，德国基尔）帮助获取数据，以及 Bette Otto-Bliesner（美国国家大气研究中心，美国科罗拉多州），用于使 CCSM3 模型输出可用于进一步分析。在现代 NAO+ 状态下看到的西风带更向北的轨迹导致植被活动减少（Gouveia 等人，2008 年），这一反应与 LIG 晚期伊比利亚地中海西南部森林的减少一致。Salonen 承认来自美国科学院的资助芬兰（项目 310649）。我们感谢 Karin Helmens（瑞典自然历史博物馆）和三位匿名审稿人的评论，感谢 Anastasia Zhuravleva（GEOMAR Helmholtz 海洋研究中心，德国基尔）帮助获取数据，以及 Bette Otto-Bliesner（美国国家大气研究中心，美国科罗拉多州），用于使 CCSM3 模型输出可用于进一步分析。在现代 NAO+ 状态下看到的西风带更向北的轨迹导致植被活动减少（Gouveia 等人，2008 年），这一反应与 LIG 晚期伊比利亚地中海西南部森林的减少一致。Salonen 承认来自美国科学院的资助芬兰（项目 310649）。我们感谢 Karin Helmens（瑞典自然历史博物馆）和三位匿名审稿人的评论，感谢 Anastasia Zhuravleva（GEOMAR Helmholtz 海洋研究中心，德国基尔）帮助获取数据，以及 Bette Otto-Bliesner（美国国家大气研究中心，美国科罗拉多州），用于使 CCSM3 模型输出可用于进一步分析。与 LIG 晚期伊比利亚地中海西南部森林减少一致的回应。Salonen 承认来自芬兰学院的资助（项目 310649）。我们感谢 Karin Helmens（瑞典自然历史博物馆）和三位匿名审稿人的评论，感谢 Anastasia Zhuravleva（GEOMAR Helmholtz 海洋研究中心，德国基尔）帮助获取数据，以及 Bette Otto-Bliesner（美国国家大气研究中心，美国科罗拉多州）用于使 CCSM3 模型输出可用于进一步分析。与 LIG 晚期伊比利亚地中海西南部森林减少一致的回应。Salonen 承认来自芬兰学院的资助（项目 310649）。我们感谢 Karin Helmens（瑞典自然历史博物馆）和三位匿名审稿人的评论，感谢 Anastasia Zhuravleva（GEOMAR Helmholtz 海洋研究中心，德国基尔）帮助获取数据，以及 Bette Otto-Bliesner（美国国家大气研究中心，美国科罗拉多州），用于使 CCSM3 模型输出可用于进一步分析。