Bridging theory to reality in projections of the Asian and West African monsoons (Bridge)

在亚洲和西非，大部分降雨都落在夏季季风季。季风雨对农业至关重要，而较晚或较弱的季风可能会对农作物造成灾难，以至于印度财政部长曾将季风描述为该国“真正的财政部长”。然而，尽管我们对受热力学(如温度、海平面)控制的气候变化有很强的概念性理解，但受风型(如区域降水)控制的变化却远没有那么直观。最先进的模型很难正确模拟当今季风降雨的模式，并预测未来的一系列变化。如果没有对控制季风的风循环的基本了解，我们就不可能判断哪些预测是可信的，无论是季节性的还是在全球变暖的情况下。最近，通过使用非常抽象的模型配置：水行星(只覆盖水的行星)和包括简单大陆的模拟，我们对控制季风的机制的理解取得了重大进展。通过剥离现实世界的复杂性，这些模型终于为我们提供了控制纬向平均热带降雨时间和地点的基本理论。然而，这些成功的理论总体上还没有适应区域规模，这为我们对区域季风及其变异性和对气候变化的反应的基本理解提供了一个巨大的机会。为了解决将理论与现实联系起来的挑战，我们确定了一种新的方法，将机器学习方法与模型模拟和数据的层次结构相结合。模型层次结构将使我们能够研究季风环流的行为和理论随着复杂性的增加而表现如何。特别是，我们将利用一种新的、高度可配置的理想化气候模型，ISCA，它允许我们在一个单一、一致的框架内运行从非常简单的水行星到简化的地球模型的各种模拟。理解区域季风的一个主要挑战是，支撑理论的数学在局部尺度上变得高度复杂。然而，机器学习最近也被应用于海洋科学中的类似问题，以识别由不同关键过程管理的区域。我们将使用这些技术来简化数学和发展季风雨的区域理论。适用于每个区域的理论将被用来解释最新的最新气候模型的行为。我们将使用对过去和未来气候的模拟，以及更理想的模拟，目的是在模拟导致气候变化的过程(例如海冰、植物生理学)时确定模型之间的差异。这应该有助于理解历史气候模拟中的偏差，并限制全球变暖下气候预测的模型间差异。通过确定哪些模型可以可靠地模拟季风和未来变化的驱动因素，应该可以对这些关键地区做出更可靠和更有用的预测。该项目的最后阶段将测试是否需要不同的理论来理解不同时间尺度上的季风行为。气候降雨理论是否也能解释降雨量的周与周、或十年与十年之间的变化？这些过程是否有一个提前期，可以为从亚季节性到季节性或年代际预测提供信息？理论洞察力能帮助我们解开海洋表面温度的年代际变化如何调节季风雨年际变化的过程吗？通过从一个新的角度处理这些复杂的问题，布里奇的目标是最终建立起我们对环流控制的季风降雨预测的信心水平，就像我们对热力学控制的气候特征的预测一样。

In Asia and West Africa the majority of rain falls during the summer monsoon season. Monsoon rain is vital for agriculture, and a late or weak monsoon can mean disaster for crops, to the point where the Indian finance minister once described the monsoon as the country's 'real finance minister'. However, while we have a strong conceptual understanding of climatic changes controlled by thermodynamics (e.g. temperature, sea level), changes controlled by wind patterns (e.g. regional precipitation) are far less intuitive. State-of-the-art models struggle to correctly simulate patterns of monsoon rainfall in the present day, and predict a range of future changes. Without basic understanding of the wind circulations controlling the monsoons it is impossible to judge which predictions we can trust, both seasonally and under global warming.Recently, major advances have been made in our understanding of the mechanisms controlling the monsoons by using very abstract model configurations: aquaplanets (planets covered only in water) and simulations including simple continents. By stripping back the complexity of the real world, these models have at last given us basic theories for the controls on when and where zonal-mean tropical rain falls. However, these successful theories have in general not yet been adapted to the regional scale, and this presents an enormous opportunity for a step-change in our fundamental understanding of regional monsoons, their variability and response to climate change. To address the challenge of connecting theory to reality, we have identified a novel approach combining machine learning methods with a hierarchy of model simulations and data.The model hierarchy will allow us to study how monsoon circulations behave and theory performs as complexity increases. In particular, we will make use of a new, highly-configurable idealised climate model, Isca, which allows us to run simulations ranging from very simple aquaplanets up to a simplified model of Earth within a single, consistent framework. A major challenge in understanding regional monsoons is that the mathematics underpinning theory becomes highly complex at a local scale. However, machine learning has recently been applied to similar problems in oceanic science to identify regions governed by different key processes. We will use these techniques to simplify the mathematics and develop regional theories for monsoon rainfall.Theories appropriate to each region will be used to interpret the behaviour of the latest state-of-the-art climate models. We will use both simulations of past and future climate, and more idealised simulations targeted at identifying differences between models in simulating processes contributing to climate change (e.g. sea ice, plant physiology). This should help in understanding biases in simulations of historical climate and constraining intermodel differences in projections of climate under global warming. By identifying which models can be trusted to simulate the monsoons and the drivers of future changes, it should be possible to produce more robust and useable projections for these key regions.The final phase of the project will test whether different theories are needed to understand monsoon behaviour on different timescales. Do theories for climatological rainfall also explain rainfall variations week-to-week, or decade-to-decade? Do these processes have a lead time which could provide information for subseasonal-to-seasonal or decadal forecasting? Can theoretical insight help us untangle how decadal variations in sea surface temperature modulate the processes governing interannual variability in monsoon rain?By approaching these complex problems from a new perspective, Bridge aims to at last build the same level of confidence in our predictions of circulation-governed monsoon rainfall as we have in thermodynamically controlled climate features.