A Least-Squares Fit of an Ocean Model to Deglacial Radiocarbon Records

当高纬度地区的表层海水变得足够冷且密度很大时，深海就会得到更新或“通风”。这些水随后横向扩散，填满各种深海盆地，并最终返回海面。因此，通风是深海与大气交流的过程，因此被认为在气候系统中起着特别重要的作用。该项目将估算最后一次冰川消融期间深海通风的变化——这是自然气候变化的最大表现，在地质记录中保存得相对较好。它将(i)通过阐明海洋通风变化对冰芯中观测到的大气二氧化碳浓度的冰川消融上升的贡献，促进科学进步；（ii）通过免费提供为本项目开发的计算机代码，有益于更广泛的科学界；（iii）通过让三名学生参与并起草一本关于本项目使用的方法的袖珍书来支持教育，特别是针对本科生和研究生；（iv）通过进一步了解海洋在大气中二氧化碳的自然变化中的作用来造福社会。二氧化碳是大气中仅次于水蒸气的最重要的温室气体，也是当前主要的环境和社会问题。用更专业的术语来说，深海盆地过去2万年的通气历史将通过海洋放射性碳（14C）记录与使用递归最小二乘技术（卡尔曼滤波和相关平滑器）的海洋环流模型的定量组合来估计。在地质时间尺度上应用这些强大但计算密集型的技术所带来的挑战将通过使用具有简化动力学的粗分辨率模型以及有效的滤波和平滑算法来解决。在考虑到放射性碳记录的不确定性和稀疏分布的情况下，将对海洋通风变化导致末次冰消期大气CO2浓度变化的假设进行检验。本研究的工作计划将包括四项任务。(1)采用基于粗分辨率和行星地圈的中等复杂程度的海洋环流模式。它将有一个网格，容纳14C记录的空间分布，并包括一个简化的海冰模型和14C的传输方程。(2)现代和古数据，包括1000多个来自底栖有孔虫和深海珊瑚的14C年龄估计，将与不确定性估计结合在一起。(3)在末次盛冰期（LGM）至今的前向积分过程中，利用卡尔曼滤波将模型与数据结合起来。(4)将使用平滑器在时间上向后传播数据（包括现代数据）提供的信息，从而进一步约束过去的海洋状态。从（3-4）中，将得出从LGM到今天的海盆尺度海洋通气率的估计值，该估计值与数据和模型相一致，给出了它们各自的误差统计估计值。

The deep ocean is renewed, or "ventilated", when surface waters present at high latitudes become sufficiently cold and very dense. These waters subsequently spread laterally, filling the various deep oceanic basins, and eventually return to the sea surface. Ventilation is therefore the very process by which the deep ocean communicates with our atmosphere and as such, is thought to play a particularly important role in the climate system. This project will estimate the changes in deep ocean ventilation which took place during the last deglaciation -- the largest manifestation of natural climate change that remains relatively well preserved in the geologic record. It will (i) promote the progress of science by elucidating the contribution of ocean ventilation changes to the deglacial rise in atmospheric concentration of carbon dioxide (CO2) observed in ice cores, (ii) be profitable to the broader scientific community by making freely available the computer codes developed for this project, (iii) support education by engaging three students and by drafting a little pocket book on the methods used for this project and aimed particularly at undergraduate and graduate students, and (iv) benefit society by furthering understanding of the ocean's role in the natural changes of atmospheric CO2 -- the most important greenhouse gas in our atmosphere after water vapor and the subject of major current environmental and societal concerns.In more technical terms, the history of the ventilation of deep oceanic basins over the past 20,000 yr will be estimated from the quantitative combination of ocean radiocarbon (14C) records with an ocean circulation model using recursive least-squares techniques (a Kalman filter and a related smoother). The challenge posed by the application of these powerful but computationally intensive techniques over geologic time scales will be addressed by using a coarse-resolution model with simplified dynamics in concert with efficient filtering and smoothing algorithms. The hypothesis that ocean ventilation changes contributed to the changes in atmospheric CO2 concentration of the last deglaciation will be tested with due regard for the uncertainties and sparse distribution of the radiocarbon records. The work plan of this research will comprise four tasks. (1) An ocean circulation model of intermediate complexity, based on coarse resolution and planetary geostrophy, will be applied. It will have a grid that accommodates the spatial distribution of 14C records and include a simplified sea-ice model and transport equation for 14C. (2) Modern and paleo-data, including more than 1000 14C age estimates from benthic foraminifera and deep-sea corals, will be assembled together with uncertainty estimates. (3) A Kalman filter will be used to combine the model with the data during a forward integration of the model from the Last Glacial Maximum (LGM) to today. (4) A smoother will be used to propagate backwards in time the information provided by the data (including modern data) and hence further constrain past ocean states. From (3-4), an estimate of basin-scale ocean ventilation rates from LGM to today will be derived, one that is consistent with the data and the model, given estimates of their respective error statistics.