Infrastructure & Investment

Gaussian process regression-based Bayesian optimisation (G-BO) of model parameters – a WRF model case study of southeast Australia heat extremes

Key Points

• Our study optimises WRF model parameters for Southeast Australia heat extremes, enhancing the accuracy of the model simulation. 

• G-BO method finds optimal parameter ranges, substantially improving the simulation of temperature and humidity. 

• Results suggest updating WRF model's default settings for better extreme heat event simulations.

Abstract 

In Numerical Weather Prediction (NWP) models, such as the Weather Research and Forecasting (WRF) model, parameter uncertainty in physics parameterization schemes significantly impacts model output. Our study adopts a Bayesian probabilistic approach, building on prior research that identified temperature (T) and relative humidity (Rh) as sensitive to three key WRF parameters during southeast Australia’s extreme heat events. Using Gaussian process regression-based Bayesian Optimisation (G-BO), we accurately estimated the optimal distributions for these parameters. Results show that the default values are outside their corresponding optimal distribution bounds for two of the three parameters, suggesting the need to reconsider these default values. Additionally, the robustness of the optimal parameter distributions is validated by their application to an independent extreme heat event, not included in the optimisation process. In this test, the optimised parameters substantially improved the simulation of T and Rh, highlighting their effectiveness in enhancing simulation accuracy during extreme heat conditions. 

Plain Language Summary 

This study aims to enhance the accuracy of a numerical weather model called the Weather Research and Forecasting (WRF) model, especially for simulating extreme heat events in Southeast Australia. Typically, the accuracy of such models depends on specific settings, which are often set to default values. Our research used a method known as Gaussian process regression-based Bayesian Optimisation (G-BO) to determine the best range of values for these settings. We found that the default settings were not optimal. By applying G-BO, we identified more effective values that substantially improved the model’s simulations of temperature and humidity during heat extremes. This improvement was consistent even when tested on an independent extreme heat event. These advancements are vital for more accurate weather forecasting, which is essential for emergency services, electricity management, and agriculture planning during extreme heat conditions.

Moving to General Equilibrium: The Role of CGEs for Economic Analysis of Water Infrastructure Projects

Authors' Summary: The desire of policymakers and public finance institutions to understand the contribution of water infrastructure to the wider economy, rather than the value of project-level outputs in isolation, has spawned a multidisciplinary branch of water resource planning that integrates traditional biophysical modeling of water resource systems with economy-wide models, including computable general equilibrium models. Economy-wide models include several distinct approaches, including input–output models, macro-econometric models, hybrid input–output macro-econometric models, and general equilibrium models—the term “economy-wide” usually refers to a national level analysis, but could also apply to a sub-national region, multi-nation regions, or the world. A key common characteristic of these models is that they disaggregate the overall economy of a country or region into a number of smaller units, or optimizing agents, who in turn interact with other agents in the economy in determining the use of inputs for production, and the outcomes of markets for goods. These economic agents include industries, service providers, households, governments, and many more. Such a holistic general equilibrium modeling approach is particularly useful for understanding and measuring social costs, a key aim in most cost–benefit analyses (CBAs) of water infrastructure investments when the project or program will have non-marginal impacts and current market prices will be impacted and an appropriately detailed social accounting matrix is available.

This article draws on examples from recent work on low- and middle-income countries (LMICs) and provides an outline of available resources that are necessary to conduct an economy-wide modeling analysis. LMICs are the focus of larger water resource investment potential in the 21st century, including large-scale hydropower, irrigation, and drinking water supply. A step-by-step approach is illustrated and supports the conclusion that conditions exist to apply these models much more broadly in LMICs to enhance CBAs.

Zero hunger. Affordable and clean energy. Reduced inequalities. These are among the  sustainable development goals that the United Nations has established in pursuit of the long-term well-being of the Earth and its inhabitants. But achieving goals like these—whether by the UN’s 2030 deadline or beyond—requires a detailed understanding of the many complex, interconnected, co-evolving natural, social and technological systems upon which all life depends.

Upgrading computational tools for sustainable development
Spatially Resolved Temperature Response Functions to CO2 Emissions

Abstract

Carbon dioxide (CO2) emissions affect local temperature; quantifying that local response is important for learning about the earth system, the impacts of mitigation, and adaptation needs. We assume the climate system can be represented as a time-dependent linear system, diagnosing Green's Functions for the spatial temperature response to CO2 emissions based on CMIP6 earth system models. This allows us to emulate the linear component of the temperature response to CO2. This approach is sufficient to capture the spatial temperature response of CMIP6 experiments within one standard deviation of the multimodel spread across most regions, though accuracy is lower in the Southern Ocean and the Arctic. Our approach reveals where nonlinear feedbacks are important in current CMIP6 models, and where the local system response is well represented by a time-dependent linear differential operator. It incorporates emissions path dependency and may be useful for evaluating large ensembles of emission scenarios.

Key Points

 

  • With a Green's Function approach, we emulate the linear component of the spatially resolved temperature response to CO2 emissions

  • We reproduce the temperature response well within multi-model uncertainty except in the Arctic and Southern Ocean

  • This approach allows expedient quantification of the spatial and temporal temperature response to varying CO2 emissions pathways

 

Plain Language Summary

Carbon dioxide (CO2) emissions impact surface temperature. It is well established that the global mean temperature change is proportional to the cumulative emissions of CO2. This has led to the creation of carbon budgets to reach temperature goals. We test this relationship at the spatio-temporal scale, quantifying a simple approach that estimates the local temperature response to CO2 emissions alone. We use an approach built from the Climate Model Intercomparison Project Phase 6 (CMIP6) Earth System Models, based on the concept that an additional unit of CO2 can be scaled for larger emissions and summed over time to estimate cumulative impacts. We evaluate this with additional CMIP6 experiments, showing that this approach captures the temperature response in most locations with lower accuracy in the Arctic and Southern Ocean. This type of approach may be useful to evaluate many policy scenarios and to better understand earth system processes that are represented in the models, as it takes around one second to quantify 90 years' worth of temperature change on a local computer, while Earth System Models can require weeks of runtime on supercomputers.

Showcasing advances and building community in modeling for sustainability (Intro to SPECIAL FEATURE: Modeling Dynamic Systems for Sustainable Development)

Abstract: We organized this Special Feature on “Modeling Dynamic Systems for Sustainable Development” to showcase the field’s recent advances. Much recent research in sustainability science has mobilized data and theory to better understand systems that include interacting people, technologies, institutions, ecosystems, and both social and environmental processes. A recent National Academies workshop and an Annual Review paper identified several challenges and open questions for the field, stressing the importance of developing and testing new theories to advance knowledge and guide action. However, there has been less attention in sustainability science towards integrating modeling with theory and data-focused approaches. Modeling is necessary for making projections about the dynamical implications of our present understanding of nature-society systems – which is essential to determining whether long term trends in nature-society interactions are consistent with sustainable development goals, and to analyze whether particular interventions (e.g. technologies, policies, behavior) are likely to change those interactions in ways that promote such goals.

The papers in this Special Feature highlight advances in simulating coupled nature-society systems. We believe that these techniques, if they were more widely adopted, could significantly improve the capacity of sustainability science researchers to test
theory, mobilize data, and inform action. Each contribution to the Special Feature addresses a specific area in which novel modeling approaches have demonstrated the capacity to advance theory and insight more broadly. The contributions were selected to be illustrative rather than comprehensive, and to facilitate connections across the communities they represent.

 

Pages

Subscribe to Infrastructure & Investment