Understanding past eruption dynamics at a volcano is crucial for forecasting the range of possible future eruptions and their associated hazards and risk. In this work we use numerical models to recreate the footprints of pyroclastic density currents (PDCs) and tephra fall from three eruptions at Gede volcano, Indonesia, with the aim of gaining further insight into these past eruptions and identifying suitable eruption source parameters for future hazard and risk assessment. Gede has the largest number of people living within 100 km of any volcano worldwide, and has exhibited recent unrest activity, yet little is known about its eruptive history. For PDCs, we used Titan2D to recreate geological deposits dated at 1.2 and c. 1 kyrs BP. An objective and quantitative multi-criteria method was developed to evaluate the fit of 342 model simulations with field observations. In recreating the field deposits we were able to identify the best fitting values to reconstruct these eruptions. We found that the 1.2 kyrs BP geological deposits could be reproduced with Titan2D using either a dome-collapse or a column-collapse as the triggering mechanism, although a relatively low basal friction angle of 6° would suggest that the PDCs were highly mobile. For the 1 kyrs BP PDC, a column-collapse mechanism and a higher basal friction angle were required to fit the geological deposits. In agreement with previous studies, we found that Titan2D simulations were most sensitive to the basal friction angle parameter. We used Tephra2 to recreate historic observations of tephra dispersed to Jakarta and Gunung Patuha during the last known magmatic eruption of Gede in 1948. In the absence of observable field deposits, or detailed information from the published literature, we stochastically sampled eruption source parameters from wide ranges informed by analogous volcanic systems, allowing us to constrain the eruption dynamics capable of dispersing tephra to the most populous city in Indonesia, Jakarta. Our modelling suggests that the deposition of tephra fall in Jakarta during the November 1948 eruption was a very low probability event, with a < 1% chance of occurrence. Through this work, we show how the reconstruction of past eruptions with numerical models can improve our understanding of past eruption dynamics, when faced with epistemic uncertainty. At Gede volcano, this provides a crucial step towards the reduction of risk to nearby populations through volcanic hazard assessment.

了解火山过去的喷发动态对于预测未来可能喷发的范围及其相关的危害和风险至关重要。在这项工作中，我们使用数值模型来重建印度尼西亚格德火山三次喷发的火山碎屑密度流 (PDC) 和火山灰的足迹，目的是进一步了解这些过去的喷发并确定合适的喷发源参数以应对未来的灾害和风险评估。格德是世界上居住在任何火山 100 公里范围内的人口最多的地区，并且最近有过动荡活动，但对其喷发历史知之甚少。对于 PDC，我们使用 Titan2D 重建了年代为 1.2 和 c 的地质沉积物。1 基尔 BP。开发了一种客观和定量的多标准方法来评估 342 个模型模拟与现场观察的拟合。在重建现场沉积物时，我们能够确定重建这些喷发的最佳拟合值。我们发现 1.2 kyrs BP 地质沉积物可以使用 Titan2D 使用圆顶坍塌或柱坍塌作为触发机制再现，尽管 6° 的相对较低的基础摩擦角表明 PDC 具有高度移动性。对于 1 kyrs BP PDC，需要柱塌陷机制和更高的基底摩擦角来适应地质沉积物。与之前的研究一致，我们发现 Titan2D 模拟对基础摩擦角参数最敏感。我们使用 Tephra2 重新创建了在 1948 年最后一次已知的格德岩浆喷发期间散布到雅加达和 Gunung Patuha 的火山灰的历史观测结果。由类似的火山系统告知的范围，使我们能够限制能够将火山灰扩散到印度尼西亚人口最多的城市雅加达的喷发动态。我们的模型表明，在 1948 年 11 月火山喷发期间，雅加达火山灰的沉积是一个非常低概率的事件，发生的几率小于 1%。通过这项工作，我们展示了在面对认知不确定性时，如何用数值模型重建过去的喷发可以提高我们对过去喷发动力学的理解。