末次间冰期全球气候瞬变模拟与平衡态模拟的对比研究

Transient and time-slice simulations of global climate change during the Last Interglacial:Model-model and model-data comparisons

采用水平分辨率约1°的地球系统模式CESM,本研究开展了末次间冰期129~124 ka的瞬变模拟试验(TGCS-LIG)和127 ka的平衡态切片模拟试验(127ka-LIG),量化了数值试验方案对模拟的末次间冰期全球气候的影响,以期加深理解造成模拟结果与重建数据之间分歧的可能原因。TGCS-LIG试验和127ka-LIG试验均表明:末次间冰期全球年平均温度较工业革命前低,降温幅度分别为0.4℃和0.2℃;在夏季,北半球升温(1.2/1.5℃),而南半球降温(0.9/0.7℃)。就年降水而言,末次间冰期北半球季风区降水增加,南半球季风区降水减少,但存在显著的区域差异。与重建数据相比,TGCS-LIG试验和127ka-LIG试验的模拟-数据符号一致性与均方根误差均相近,但127ka-LIG试验模拟效果略好。TGCS-LIG试验与重建的全球年平均(夏季)海温符号一致性为34.9%~44.7%(36.8%~42.5%),较127ka-LIG偏差约2%~4%(约5%~7%);均方根误差为2.9~3.2℃(2.9~3.4℃),较127ka-LIG试验偏差小于约0.1℃。对于全球年平均降水,TGCS-LIG试验与重建数据的符号一致性为63.8%,略高于127ka-LIG试验(63.1%)。相较于模拟和重建之间的绝对偏差而言,数值试验方案对于改进两者分歧的作用十分有限。需要注意的是,TGCS-LIG试验表明:末次间冰期全球关键气候模态的年际变率在129~124 ka间存在显著的内部变化,其中ENSO年际变率随时间逐渐增强,北半球环状模与南半球环状模的年际变率呈现千年尺度波动的特征;而127ka-LIG试验无法刻画气候模态的时间演变特征。总之,本文研究结果表明,采用平衡模拟试验或是瞬变模拟试验并不是造成末次间冰期模拟-数据分歧的主要原因,但开展高时空分辨率的瞬变模拟试验对研究末次间冰期气候变率具有重要的科学价值。

Performing transient simulation during 129124 ka(TGCS-LIG)at high temporal and spatial resolutions(~1°×1°,without orbital acceleration)and time-slice experiment at 127 ka(127ka-LIG)can advance our knowledge on the role of experimental setup in the simulation of climate change during the Last Interglacial(LIG).Here,we quantitatively estimate the impact of the experimental setup on the simulated global climate change during the LIG,which may shed light on the potential reasons for model-data discrepancies during the LIG.According to TGCS-LIG/127ka-LIG experiments,the global annual mean surface temperatures decreased by 0.4℃/0.2℃during LIG relative to the preindustrial era,with the summer surface temperature over the Northern(Southern)Hemisphere increased(decreased)by 1.2℃/1.5℃(0.9℃/0.7℃).During LIG,the annual mean monsoonal precipitation increased(decreased)over the Northern(Southern)Hemisphere,with large discrepancies across sub-regions.The hit rates(i.e.,percentage agreement between model and reconstructed data)and root mean square errors were comparable between the two experiments.For global annual(summer)sea surface temperatures,the hit rates between TGCS-LIG and reconstructions were 34.9%44.7%(36.8%42.5%),biased by~2%4%(~5%7%)compared with that of 127ka-LIG;root mean square errors between TGCS-LIG and reconstructions were 293.2℃(2.93.4℃),which were~0.1℃biased compared with 127ka-LIG(generally consistent with 127ka-LIG).For global annual precipitation,the hit rate between TGCS-LIG and multi-proxies was 63.8%,higher by 0.7%relative to 127ka-LIG.The above results indicated that model-model differences were relatively small compared with model-data discrepancies.It is worth noting that TGCS-LIG suggested that the interannual variability of the ENSO showed an increasing trend,whereas the interannual variability of the Northern Hemisphere Annular Mode and Southern Hemisphere Annular Mode fluctuated.These characteristics of varied interannual variability revealed by TGCS-LIG were hardly deciphered in 127ka-LIG.Overall,we suggest that the differences in experimental setup for LIG climate simulations(i.e.using the time-slice experiment or the transient simulation)may have limited impact on reconciling model-data discrepancies during the LIG,nevertheless,we still highlight the importance of performing transient simulation at high resolution in the investigation of climate variability during the LIG.