氣候回饋與海洋：全球暖化下兩者的不確定性與交互作用

Abstract

此篇論文的最終目標是建立一個對於氣候敏感度之不確定性更全面的理解。具體而言，此論文闡述海洋在影響氣候敏感度所扮演的角色。透過改變海表溫度的結構，高緯度的海洋熱吸收對全球的對流層穩定度與輻射收支都相當關鍵。同時，本論文將多個完整耦合的氣候模式裡的氣候敏感度之不確定性部分追溯至北大西洋深海環流模擬之不確定性，並且提供物理過程來解釋北大西洋深海環流模擬的不確定性與其和氣候平均場的連結。 在大多數完整耦合的氣候模式中，淨氣候回饋在二氧化碳變為兩倍或四倍的模擬下，於數十年內會越趨敏感，這代表使用數十年的暖化資料來推估平衡態的氣候敏感度會造成低估。過去的研究將此低估歸因至熱帶太平洋的海表溫度之結構變化。在此研究中，我更深入地去理解海表溫度之結構的形成機制。我將簡化的淺水模式裡所建構之格林函數與完整耦合模式裡所診斷之海洋熱吸收兩者進行整合，由此將氣候回饋隨時間的改變定量地歸因至不同區域的海洋熱吸收。研究結果顯示南大洋的熱吸收隨著時間有不同的減弱幅度，此為造成模式 CESM1 裡氣候回饋發生改變的主要因素。在二氧化碳增加為四倍的數十年過後，南大洋的熱吸收改變能透過遙相關過程使熱帶東南太平洋的海表溫度有增強的暖化作用，此將使對流層的穩定度降低，導致雲的輻射回饋更加敏感。 當模式們大致上都呈現南大洋熱吸收隨時間減弱與東南太平洋的暖化隨時間增強，模式們對於北大西洋的中高緯海洋熱吸收與海表溫度的隨時間反應卻有著很大的不確定性。無論是CMIP5或CMIP6 模式，研究結果均顯示模式間之所以在氣候回饋改變之推估上有所差異，部分可以歸因至各個模式的北大西洋深海環流在暖化下有不同的隨時間反應。那些預測北大西洋深海環流在暖化後期有強度回復的模式，會伴隨北半球更顯著的升溫，此結果與北大西洋深海環流將南半球的能量傳往北半球之特性一致。更顯著的北半球升溫將透過降低低對流層的穩定度，導致穩定度所貢獻的輻射回饋與短波雲輻射回饋隨時間更加敏感。此二者貢獻至淨氣候回饋隨時間越趨敏感的變化。 鑒於北大西洋深海環流的強度改變能夠對具重要輻射意義的海表溫度產生影響，此研究就北大西洋深海環流的預測不確定性之時間尺度與物理過程進行探討。我發現若模式在氣候平均態有較強的北大西洋深海環流，將會對應層化較弱的拉不拉多海，亦即當地較強的混合作用。因此，當二氧化碳增加時，這樣的模式可以將海表的暖水更有效率地混合到拉不拉多海的次表層，尤其是在每年的冬春季。無論是在完整的耦合模式裡進行診斷或是在純海洋模式裡進行溫度強迫之模擬，都顯示拉不拉多海的次表層相對暖化會在數年後導致副熱帶的北大西洋深海環流強度減弱。此結果強調了全球暖化下海洋反應的預測很大程度被海洋的氣候平均場所控制。 綜上所述，本篇論文分別闡述了兩個看似獨立的問題：（一）氣候回饋與氣候敏感度在中長期預測的不確定性（二）海洋環流的模擬不確定性。同時，本論文也連結了上述兩項不確定性。海洋不僅可以透過影響當地的熱吸收來影響氣候回饋的發展，也能透過遙相關過程影響全球的海溫結構與對流層穩定度，進而改變氣候回饋隨時間的變化。

The overarching goal of this thesis is to construct a more comprehensive understanding regarding the uncertainty of climate sensitivity. Specifically, the thesis demonstrates the role of ocean in altering the climate sensitivity. Through shaping the sea surface temperature (SST) pattern, the ocean heat uptake in high latitudes is key to variations in global tropospheric stability and radiative budget. Part of the climate sensitivity uncertainty in multiple fully-coupled climate models are traced to the uncertainty of the Atlantic Meridional Overturning Circulation (AMOC) projections. Physical processes to explain the uncertain AMOC and their mean state dependence are also provided. In most fully-coupled climate models, the net radiative feedback becomes more amplifying few decades after CO2 doubling or quadrupling, indicating an underestimation of equilibrium climate sensitivity inferred from decadal warming. Previous studies have attributed the underestimation to variations in the sea surface temperature patterns over the tropical Pacific. In this study, I take a step further to understand the forming mechanisms of the surface temperature patterns. I quantify the dependence of time-evolving radiative feedbacks on regional ocean heat uptake (OHU) by convolving the Green's Function derived from a simplified, slab-ocean model with the diagnosed OHU from a fully-coupled model. The results suggest that the time-dependent weakening of OHU over the Southern Ocean is the main contribution to the net radiative feedback change in the model CESM1. The remote impact from OHU over the Southern Ocean gives rise to increasingly enhanced surface warming in the Southeastern Pacific, which leads to decreasing tropospheric stability and more sensitive cloud feedback decades after quadrupling CO2. While models generally show robust weakening of the Southern Ocean heat uptake and enhanced Southeastern Pacific warming with time, their projections of OHU and surface warming over North Atlantic are highly uncertain. In both CMIP5 and CMIP6 models, results show that the inter-model spread of changes in net radiative feedback can be partially traced to the time evolution of the AMOC in each model. Models with stronger AMOC recovery tend to project more amplified warming in the Northern Hemisphere, consistent with stronger northward heat transport by the AMOC. The relatively amplified warming in the Northern Hemisphere leads to larger increases in both lapse‐rate and shortwave cloud feedbacks through decreasing the low-level tropospheric stability, accounting for a more amplifying net radiative feedback with time. Since the changes in the AMOC strength could modulate the radiatively-important surface warming pattern, timescales and processes for the inter-model spread of the projected AMOC weakening are investigated. I report that the models with stronger AMOC strength in the mean state climate are associated with less stratified upper Labrador Sea, allowing for stronger mixing. Hence, in response to CO2 increase, the surface warming is mixed to the subsurface Labrador Sea more efficiently, especially in winter and spring. Both diagnostics from fully-coupled models and temperature-perturbing simulations within an ocean-only model suggest that the relatively enhanced subsurface warming in the Labrador Sea would further lead to stronger AMOC weakening in subtropics in several years, highlighting the mean state control of ocean circulation in projected oceanic responses under global warming. Overall, the thesis elaborates on two seemingly independent questions: (1) the uncertainty of long-term projections of climate feedback and sensitivity (2) the uncertainty of ocean circulation projections. At the same time, the thesis fills the gap by linking the above two uncertainties — the ocean modifies the time evolution of climate feedback not only through the local ocean heat uptake, but also via the teleconnections that shape the global SST pattern and tropospheric stability.