The contemporary global terrestrial carbon cycle - a systemic model-data fusion analysis

The terrestrial carbon (C) cycle is critical to climate regulation, provisioning ecosystem services and biogeochemical cycles. To develop effective climate mitigation strategies and understand climate risks requires rigorous, uncertainty bounded, systemic information on C-dynamics and their underpinning ecological processes. Significant effort has been invested to quantify contemporary C-cycle states (e.g., LAI, woody biomass) and exchanges of C (e.g., GPP) through use of global observations of land surface properties. These state and flux products are often incomplete in space and time, contain poorly characterised errors, and lack internal consistency checks, making the construction of a reliable global C budget a significant challenge. We present a rigorous, global (0.5×0.5 deg), multi-decadal (2003-2024), data-informed analysis of the terrestrial C-cycle at a monthly time step. We use a state-of-the-art Bayesian model-data fusion (MDF) framework (CARDAMOM) to calibrate an intermediate complexity model of terrestrial ecosystems (DALEC) using ecologically relevant spatio-temporal observations (leaf area, absorbed photosynthetically active radiation, gross primary production, woody biomass, soil C), and forcing (meteorology, atmospheric CO2, burned area and forest loss). The resulting systemic analysis (i.e. consistent between states and fluxes) is a pixel-scale calibration at the finest typical global spatial resolution used by global land surface modelling activities (e.g., Global Carbon Project). CARDAMOM uniquely propagates observational uncertainties through to the retrieved DALEC parameters (i.e. ecosystem properties) and simulates C pools and fluxes for each of 55,246 pixels across the vegetated land surface. This paper's key scientific result is that current global multi-decadal datasets are inadequate to provide >95 \% confidence on the sign of net exchange across 91% of the vegetated land surface. Thus, we are able to confidently determine that 0.23% is a net source, 0.75% a net sink and 7.3% as neutral. However, our main objective here is to enhance accessibility of this open-access dataset through a thorough description of its key features: calibration and evaluation metrics, photosynthate allocation fractions and tissue residence times, and their uncertainties. These outputs can inform the evaluation of land surface models that are too computationally expensive to directly constrain with MDF approaches, and identify key uncertainties in C cycle understanding as a first step to targeting them.

陆地碳循环（terrestrial carbon cycle）对气候调节、生态系统服务供给与生物地球化学循环至关重要。制定有效的气候减缓策略、理解气候风险，需要严谨且经过不确定性界定的系统信息，以解析碳动态及其支撑的生态过程。学界已投入大量精力，通过全球陆地表面属性观测，量化当代碳循环状态变量（如叶面积指数（LAI）、木本生物量）与碳交换过程（如总初级生产力（Gross Primary Production, GPP））。这类状态变量与通量产品往往存在时空覆盖不全、误差特征描述不足、缺乏内部一致性检验等问题，使得构建可靠的全球碳预算成为一项重大挑战。

本文提出一项严谨的全球（0.5°×0.5°）、多年代际（2003-2024年）月尺度陆地碳循环数据驱动分析。我们采用当前领先的贝叶斯模型数据融合（Model-Data Fusion, MDF）框架（CARDAMOM），利用生态相关的时空观测数据（叶面积、吸收光合有效辐射、总初级生产力、木本生物量、土壤碳）与外强迫数据（气象数据、大气CO₂浓度、火烧面积与森林丧失数据），校准陆地生态系统中等复杂度模型（DALEC）。最终得到的系统分析结果（即状态变量与通量之间保持一致），是在全球陆地表面建模活动（如全球碳计划（Global Carbon Project））所采用的最精细典型全球空间分辨率下开展的像元尺度校准。

CARDAMOM可独特地将观测不确定性传递至反演得到的DALEC模型参数（即生态系统属性），并对植被覆盖陆地表面上55246个像元的碳库与碳通量进行模拟。本文的核心科学结论为：当前全球多年代际数据集不足以在91%的植被覆盖陆地表面上，为净交换的符号提供≥95%的置信度。据此我们可以确定，0.23%的区域为净碳源、0.75%为净碳汇、7.3%为中性区域。

然而本文的主要目标是通过详细阐述该开放获取数据集的关键特征——包括校准与评估指标、光合产物分配比例与组织停留时间及其不确定性——来提升该数据集的可访问性。这些产出可用于评估那些因计算成本过高而无法通过MDF方法直接约束的陆地表面模型，并可作为识别碳循环理解中关键不确定性的第一步，进而针对性开展相关研究。