The First Environmental Digital Twin Dedicated to Understanding Tropical Wetland Methane Emissions for Improved Predictions of Climate Change

甲烷是一种主要的温室气体。它的大气寿命很短(~9年)，这意味着我们可以减轻它的排放和变暖影响。在COP26上，各国签署了甲烷承诺，并在COP27上得到加强，承诺在2030年减少30%的排放，到2050年消除超过0.2摄氏度的全球变暖。挑战是甲烷有许多来源，人工的和天然的。人为排放包括化石燃料(111 Tg CH4年-1)和农业/废物(217 Tg CH4年-1)的重大贡献，自然信号以湿地排放为主(181 Tg CH4年-1，占总排放量的30%)。据估计，热带湿地贡献了湿地总排放量的65%，占全球甲烷预算总额的20%以上。然而，这些估计存在极大的不确定性。为了充分了解甲烷的预算，我们必须监测这些自然排放，并了解它们是如何、何时何地产生的，以及它们在未来气候情景下可能会发生什么变化。如果做不到这一点，将限制向政策提供信息和采取缓解行动的能力。这个问题正变得更加紧迫。近年来，大气中甲烷的含量出现了令人惊讶的快速增长。全球价值在2020年增加了15 ppb，2021年增加了18 ppb，而近年来增加了5-12 ppb。这种加速是令人担忧的，并表明了重大的气候反馈，这些反馈既没有得到充分的理解，也没有预料到。利用我的工作产生的卫星数据(例如Qu等人，2022年，冯等人，2022年)进行的研究得出结论，热带湿地可能是这种新的、迄今尚未解释的甲烷增长率增加的来源。我们知道沼气是由微生物在湿地中产生的，但温度、水位和土壤类型等因素的影响仍然存在疑问。最先进的基于过程的陆面模型可以产生湿地甲烷排放，但模型估计之间的巨大差异限制了它们的实用，根据观测对这些模型进行评估是关键。重要的是，我们也不知道这些产生甲烷的湿地有多大，因为它们的大小会随着降雨和河流流量的变化而不断变化。因此，即使模型捕捉到正确的湿地甲烷气候响应，湿地范围本身也会限制准确估计排放量的能力。因此，问题是双重的：1)我们能否调和我们对湿地甲烷排放对气候反馈的反应建模能力上的巨大差异？2)我们能否极大地改进我们对湿地范围的估计，以限制甲烷排放的时空变化？该奖学金将利用卫星观测和陆地表面模型来建立一个创新的专门的湿地数字双胞胎；一个能够估计湿地产生甲烷的机器学习系统，改变我们对甲烷排放原因的理解和对气候变化的反应。同时，我们需要更好地了解湿地的位置以及它们如何随着时间的变化而变化。通过将新的机器学习方法应用于极高分辨率的卫星图像，并结合先进的水文模型，我将更好地绘制这些湿地地区的地图，并了解它们的动态。为了实现这一目标，我将与项目合作伙伴密切合作，专门从事陆地表面建模(GCP、UKCEH、英国气象局)、机器学习和人工智能(ESA Phi-Lab、NEODAAS)、IT基础设施(NEODAAS、Jasmin、CGI)、高分辨率遥感(Planet)和气候建模(英国气象局)，同时还与从湿地生态系统专家到政策制定者(例如COP/IPCC、UNEP、RAMSAR、CIFOR、CEOS/GCOS)的一系列利益攸关方接触。这一新的湿地数字孪生能力，由地球观测数据驱动，由机器学习提供动力，这将使我们能够开发能够为政策制定者提供决策支持的气候服务，并使我们能够更好地了解这些关键生态系统的气候反应。

Methane (CH4) is a major greenhouse gas. Its short atmospheric lifetime (~9 years) means we can mitigate its emissions and warming effects. At COP26, countries signed up to the Methane Pledge, strengthened at COP27, committing to reduce emissions in 2030 by 30%, eliminating over 0.2C of global warming by 2050.The challenge is that methane has many sources, man-made and natural. Man-made emissions include significant contributions from fossil fuels (111 Tg CH4 yr-1) and agriculture/waste (217 Tg CH4 yr-1), with natural signals dominated by wetland emissions (181 Tg CH4 yr-1, >30% of total emissions). Estimates suggest tropical wetlands contribute >65% of all wetland emissions, over 20% of the total global methane budget. However, these estimates are hugely uncertain. To fully understand the methane budget, we must monitor these natural emissions and understand how, when and where they are produced and how they might change under future climate scenarios. Failure to do so would restrict capability to inform policy and take mitigation action.The problem is becoming more urgent. Recent years have seen a rapid and surprising increase in atmospheric methane. Global values increased by 15 ppb in 2020 and 18 ppb in 2021, compared to 5-12 ppb in recent years. This acceleration is alarming and points to significant climate-feedbacks that are not fully understood nor expected. Studies using satellite data generated by my work (e.g. Qu et al., 2022, Feng et al., 2022) have reached the conclusion that tropical wetlands are the likely source of this new, and as of yet unexplained, increase in the methane growth rate. We know methane is produced in wetlands by microbes but questions remain on the effect of factors such as temperature, water level and soil type. State-of-the-art process-based land surface models can produce wetland methane emissions but huge discrepancies between model estimates limit their utility and assessing these models against observations is key. Importantly, we also do not know how large these methane-producing wetland areas are, as they continually change in size in response to rainfall and riverflow. Therefore, even if models capture the correct wetland methane climate-response, the wetland extent itself will limit ability to accurately estimate emissions. The problem therefore is two-fold: 1) Can we reconcile large discrepancies in our ability to model the wetland methane emission response to climate feedbacks? 2) Can we dramatically improve our estimates of wetland extent to constrain the spatial/temporal changes in methane emissions?This fellowship will use satellite observations and land surface models to build an innovative and dedicated Wetland Digital Twin; a machine-learning system capable of estimating methane produced by wetlands, transforming our understanding of the causes of methane emissions and responses to the changing climate.In parallel, we need much better knowledge of wetland locations and how they change over time. By applying new machine-learning methods to very-high-resolution satellite imagery and combining with advanced hydrological modelling, I will better map these wetland areas and understand their dynamics. To achieve this, I will work closely with Project Partners, specialising in land surface modelling (GCP, UKCEH, UK Met Office), machine learning and artificial intelligence (ESA Phi-Lab, NEODAAS), IT infrastructure (NEODAAS, JASMIN, CGI), high-resolution remote sensing (Planet) and climate modelling (UK Met Office) while also engaging with a range of Stakeholders from wetland ecosystem specialists to policymakers (e.g. COP/IPCC, UNEP, RAMSAR, CIFOR, CEOS/GCOS).This new Wetland Digital Twin capability, driven by Earth Observation data and powered by machine learning, will allow us to develop climate services that are capable of providing decision-support for policymakers and enable better understanding of the climate response of these critical ecosystems.