DEPARTMENT FOR SCIENCE, INNOVATION AND TECHNOLOGY

AI for Climate - calls - tender - UNIVERSITY OF LEEDS

Project disclaimer

Disclaimer: The data for this page has been produced from IATI data published by DEPARTMENT FOR SCIENCE, INNOVATION AND TECHNOLOGY. Please contact them (Show Email Address) if you have any questions about their data.

Description

Kilometre-Scale Simulations for Al Training - High-impact weather (HIW) events, such as heavy rain and consequent flooding and landslides, or drought, can be devastating to the livelihoods of local people and the economy of many countries in the tropics. The societal and economic impacts of HIW include loss of human life, damage to property, destruction of crops, loss of livestock, poor health, displacement of populations, loss of infrastructure, severe disruption to transportation from heavy rainfall, and the suspension of many economic activities (UNDRR, 2019, 2020a, 2020b, 2020c). Almost all HIW is expected to increase across the tropics and sub-tropics with ongoing climate change, affecting the poorest and most vulnerable. Improved projections of climate change in HIW aid adaptation, and motivate mitigation. However, in many tropical regions it is unclear whether regions will become wetter or drier (IPCC), limiting adaptation. UPSCALE will focus on the tropics, where moist convection dominates rainfall and is a primary source of heating to the tropical atmosphere, and where we can use the full model hierarchy including the cyclic-tropical-channel. The UPSCALE project will conduct research into (1) the evaluation of the newly developed Met Office Convection-Permitting Models (CPM) hierarchy of simulations, and (2) the development and application of novel process-based diagnostics and propose sensitivity experiments to understand the mechanisms of up and down scale interactions in the CPMs vs. current simulations with parametrised convection, focusing on the value of large pan-tropical domains. These activities will benefit weather forecasting and climate prediction, especially for the tropics/sub-tropics, including the development of machine learning-based predictions. The K-Scale simulations work package would exploit UK and international research in K-scale modelling with both developed (Australia, U.S.) and developing countries (India, S. Africa) to derive additional value from these high-resolution simulations as training data for AI data driven prediction systems that could then be exploited by partners. The resource would accelerate development and evaluation of the K-scale predictions and work with dataset curators/developers to ensure efficient workflows for ML applications. Initially the work will be on using research we are currently collaborating with partners on and deployment with in-country partners in subsequent years.

Objectives

UPSCALE has two main goals: (1) Evaluation of the newly developed Met Office CPM hierarchy of simulations. This will focus on the impact of resolving convective scales on the larger scale circulation (100 km to planetary scales), including time-mean basic state, aspects of weather-to-climate variability including extremes and teleconnections, and key features of the general circulation such as diabatic heating. It will, however, also include evaluation of aspects of the convective scales themselves, particularly with reference to potential upscale impacts (e.g. convective organisation). (2) Building on the knowledge gained from the evaluation studies, we will develop and apply novel process-based diagnostics and propose sensitivity experiments to understand the mechanisms of up and down scale interactions in the CPMs vs. current 10km+ simulations with parameterised convection, focusing in particular on the value of large pan-tropical domains for CP simulations. This analysis will include feedbacks across the earth system between atmosphere, land, and oceans as future fully coupled CPM simulations become available. Achieving each of these goals will yield new insights both into the value of CPMs for training better ML models, and into the metrics and methodologies needed for physical-science-based evaluation of ML-based models and parameterisations.

---

Location

The country, countries or regions that benefit from this Programme.

Developing countries, unspecified

Disclaimer: Country borders do not necessarily reflect the UK Government's official position.