Introduction
Funded by the Ministry of Business, Innovation and Employment (MBIE)’s Endeavour fund, NIWA researchers are developing a physics-informed and artificial intelligence (AI)-driven method to vastly reduce the computer power needed to provide accurate climate change projections. These projections will be based on multiple models (known as ensembles), which is critical to characterising uncertainty associated with climate-related risks.
The work will support Aotearoa’s transition to a low-emissions, climate-resilient economy. It will help New Zealanders better understand their climate-related risks across immediate, medium and long-term timescales This will improve decision-making for climate adaptation and support resilience to extreme weather events.
For instance, more actionable climate projections will help:
- farmers plan for future irrigation needs
- horticulturalists to understand future heat stress and plan crops
- winemakers understand how temperatures might change and affect grape varieties
- councils to plan infrastructure and new developments based on future weather hazards.
Climate models require vast computing power
Our understanding of how global climate will respond to future increases in greenhouse gases depends on sophisticated climate models that require vast computing power to run. For New Zealand, we can use these climate models to tell us how weather patterns may change in a future with increased greenhouse gas emissions.
Even with massive supercomputers, the complexity of these global models means they can only be run at 50-150km resolutions – so places that are close together but with dramatically different weather and climate (think Mt Cook vs Twizel, for instance) have the same projections. The coarse resolution of these climate models is a barrier to societal decision making, as it cannot resolve of more localised changes across New Zealand’s terrain.
Regional Climate Models (RCMs) are computer models which further enhance the resolution of these global climate models and provide higher resolution climate projections that resolve New Zealand’s complex terrain. However, there remains an enormous computational cost of generating high-resolution climate projections.
The animation below illustrates preliminary results from this project: the top image shows a simulation from a global climate model at 100km resolution across Aotearoa New Zealand. Bottom left compares downscaling to 12km resolution by a dynamical (physics-based) model compared with bottom right, downscaling to 12km resolution using machine learning (with much lower computational cost and much quicker to run).
About the project
The computational expense of current regional climate models (RCMs) directly limits the number of ensemble members that can be simulated at high-resolution. Without a 10-fold or more increase in RCM compute speed, downscaling climate projections of a large model ensemble at high spatial resolution is not likely feasible in the next decade.
The NIWA-led team is developing the first physics-informed artificial intelligence (AI)-driven RCM emulator for climate projections, enabling large ensembles of very high-resolution (2-12km) projections for the entire country. This will allow them to generate high-resolution climate projections that are at least 1000 times computationally faster and cheaper than current methods.
The team will engage with key stakeholders throughout the project, including iwi/hapū, regional councils, central government, and industry leaders to ensure research outputs will enable higher-value products and services. The provision of improved climate projections will support a variety of industries which span finance, insurance, weather forecasting, and primary industries.
Research aims
RA1.1 Improving boundary conditions in GCMs with Artificial Intelligence
Previous work by the team has shown that many global climate models (GCMs) misplace the position and strength of the jet and storm tracks over New Zealand, causing issues for RCMs. While prior research has explored GCM bias correction in the context of downscaling, there remains a major scientific challenge in capturing physical consistency between multiple climate variables. Instead, bias correction schemes implement simple statistical approaches, which only correct the climatological mean and variance on an individual variable and grid cell basis.
Because GCMs are free-running simulations, traditional regression-type approaches are not suitable, requiring a one-to-one mapping with observations. However, Generative Adversarial Networks (GANs) are well-suited to learning complex spatial relationships between multivariate distributions, extending biases correction far beyond an individual variable and grid cell basis. The team will also incorporate physical constraints (e.g. the position and magnitude of the jet and storm track) to nudge the AI-based algorithms towards physical consistency.
RA1.2 AI-based Regional Climate Model emulator
The International Panel on Climate Change (IPCC)’s projections, known as the CMIP6 archive, include over 50 GCMs (with 500+ ensemble members). However, it is only computationally feasible to downscale approximately six GCMs for New Zealand at intermediate resolutions (12-25km). This work will address the major issue of RCM computational speed by creating an AI-driven RCM emulator capable of running orders of magnitude faster at higher spatial resolution.
The team will first train a Convolutional Neural Network (CNN) model on an intermediate resolution (12km) RCM driven by a single GCM. To ensure the model generalizes across historical and future climates, they will compare the RCM emulator against benchmark RCM output provided MBIE-funded NIWA Projections Project.
RA1.3 Application of AI-based RCM ensemble climate projections
To provide a more complete picture of possible climate futures for New Zealand, the researchers will apply our final high-resolution RCM emulator to an extensive range of GCMs. Including a much wider range of GCMs to downscale through the RCM emulator will enable probabilistic climate projections for New Zealand for the first time. They will work alongside industry leaders in a Project Advisory Group, and Māori Advisory Group to construct case studies which explore the added value of large ensemble high-resolution climate projections.
