Computationally inexpensive carbon cycle models serve as critical and efficient tools for illuminating the complex dynamics of the carbon cycle and its interplay with the climate system, offering insights into how our planet has responded to climate perturbations throughout its history. During geologic hyperthermal events, carbon cycle models are employed to trace the trajectory of carbon emissions and establish a connection between the emission trajectory and changes in the Earth's surface environment. To date, the prevalent method to estimate the carbon emission rate relies on coupling the carbon cycle modeling and proxy reconstructions. Most previous studies employ a forward methodology, i.e., they force the model with an array of predefined carbon injection scenarios and select the one that produces the best fit to one or more specific proxy-derived records (e.g., atmospheric pCO, sea surface pH, and calcite compensation depth) as the optimal solution. However, this forward method has two potential disadvantages. First, it can be computationally expensive, particularly when tens of thousands of scenarios need to be conducted to find the best solution. Second, it might not yield the best injection trajectory if none of the predefined carbon emission curves represents the realistic emission curve. Hence, an inverse model that can reconstruct the carbon emission trajectory directly from the record/records is urgently needed. In this study, building upon the Long-term Ocean-atmosphere-Sediment CArbon cycle Reservoir (LOSCAR) Model (Zeebe, 2012), we develop an interactive carbon cycle model (named iLOSCAR) using the open-source Python language and include two options, a forward model and an inverse model. The forward model replicates the original LOSCAR model, while the inverse model calculates the emission trajectory constrained by the proxy records in a single run. Both models are accessible via a web-based interface, which allows users to interactively tune model parameters and conduct experiments. In this paper, we present the details of iLOSCAR, including model structure and derivation of key equations. We then validate iLOSCAR's performance through an identical twin test and model intercomparison. We also apply it to a climatic perturbation event to diagnose features of the emission pattern that were overlooked in previous studies. Finally, we discuss the possible directions for model's future development.

中文翻译：

---

iLOSCAR：交互式长期海洋-大气-沉积物碳循环储层模型 v1.0

计算成本低廉的碳循环模型是阐明碳循环复杂动态及其与气候系统相互作用的重要而有效的工具，为我们的星球在整个历史中如何应对气候扰动提供了见解。在地质高温事件期间，碳循环模型用于追踪碳排放轨迹，并在排放轨迹与地球表面环境变化之间建立联系。迄今为止，估计碳排放率的流行方法依赖于耦合碳循环模型和代理重建。大多数先前的研究都采用前向方法，即，它们使用一系列预定义的碳注入情景来强制模型，并选择最适合一个或多个特定代理导出记录（例如，大气 pCO2、海面 pH 值、和方解石补偿深度）作为最优解。然而，这种前向方法有两个潜在的缺点。首先，它的计算成本可能很高，特别是当需要执行数万个场景才能找到最佳解决方案时。其次，如果没有任何预定义的碳排放曲线代表实际的排放曲线，它可能不会产生最佳的注入轨迹。因此，迫切需要一种可以直接从记录重建碳排放轨迹的逆模型。在这项研究中，基于长期海洋-大气-沉积物碳循环库（LOSCAR）模型（Zeebe，2012），我们使用开源Python语言开发了一个交互式碳循环模型（名为iLOSCAR），并包括两个选项，正向模型和逆向模型。正向模型复制了原始的 LOSCAR 模型，而逆向模型则计算单次运行中受代理记录约束的排放轨迹。这两种模型都可以通过基于网络的界面访问，该界面允许用户交互式地调整模型参数并进行实验。在本文中，我们介绍了 iLOSCAR 的详细信息，包括模型结构和关键方程的推导。然后，我们通过同卵双胞胎测试和模型比较来验证 iLOSCAR 的性能。我们还将其应用于气候扰动事件，以诊断先前研究中忽视的排放模式特征。最后，我们讨论了模型未来发展的可能方向。