earth 2 nvidias state of technology in weather forecasts

link between climate change and extreme weather events has strengthened in recent years, supported by a growing body of observational evidence and modeling studies.. Moreover, climate ch

VIETNAM NATIONAL UNIVERSITY, HANOI UNIVERSITY OF ENGINEERING AND TECHNOLOGY

Ngô Minh Đức - 21020620

Major: Computer Science Course/Code: 2324II_INT3011E_20

HANOI - 04/2024

Table of Contents

I Problem Statement & Introduction 3

1 Climate change and the extreme weather 3

2. How AI models affect the climate change 7

3 The Solution & Prevention 7

4 The EARTH-2 Platform & Services 9

4.1 Creator: NVIDIA NVIDIA CUDA-X Microservices– 9

4.2. Earth-2 Overview 11

II Detail Workflows of Earth-2 13

1 The process of weather forecasting 13

2 The Models 16

2.1 ICON Model 16

2.2. FourCastNet Model 18

2.3 CorrDiff Model 21

III Final Steps of Platform & Conclusion 23

1 Combining all the technologies 23

2. Conclusion 24

IV References 25

I Problem Statement & Introduction

1 Climate change and the extreme weather

Climate change, spurred by human activities, stands as a monumental challenge of our era, exacerbated in significant part by the unintended consequences of technological advancement The symbiotic relationship between technology and climate change reveals a darker side where innovation, often heralded for progress, concurrently fosters environmental degradation and exacerbates the climate crisis The combustion of fossil fuels, the cornerstone of modern industrialization and technological progress, remains a primary driver of climate change

According to the Global Carbon Project, fossil fuel combustion accounted for a staggering 76% of global greenhouse gas emissions in 2019 The proliferation of carbon-intensive technologies across various sectors, including energy production, transportation, and manufacturing, has led to unprecedented levels of greenhouse gas emissions, primarily carbon dioxide (CO2), into the atmosphere The widespread use

of internal combustion engines in automobiles and aircraft, coupled with reliance on coal-fired power plants and other fossil fuel-based energy sources, underscores the detrimental impact of technology on climate stability Additionally, industrial processes such as cement production, responsible for approximately 7% of global CO2 emissions, further compound the problem The relentless pursuit of economic growth and technological innovation, often at the expense of environmental sustainability, has accelerated deforestation, habitat destruction, and resource depletion, exacerbating climate change through loss of carbon sinks and biodiversity

Moreover, the proliferation of electronic waste, driven by rapid technological obsolescence and consumption patterns, contributes to environmental pollution and resource depletion, further exacerbating the ecological footprint of technology Thus, while technology has undeniably propelled human civilization forward, its detrimental environmental impacts underscore the urgent need for a paradigm shift towards sustainable innovation and responsible consumption to mitigate the adverse effects of climate change

Figure 1: In 2018, Algerian temperature hit a peak of 51.3 Celsius

Climate change intensifies and exacerbates extreme weather events through a combination of factors, including increased temperatures, altered precipitation patterns, and changes in atmospheric circulation patterns These changes amplify the frequency, intensity, and duration of extreme weather phenomena such as hurricanes, typhoons, heatwaves, droughts, floods, and wildfires The scientific consensus on the

link between climate change and extreme weather events has strengthened in recent years, supported by a growing body of observational evidence and modeling studies

One prominent example is the increase in the intensity and frequency of tropical cyclones, including hurricanes and typhoons Warmer ocean temperatures provide more energy and moisture to fuel these storms, leading to stronger winds and heavier rainfall According to the Intergovernmental Panel

on Climate Change (IPCC), there has been a discernible increase

in the intensity of the strongest tropical cyclones in the North Atlantic since the 1970s, with a notable uptick in the proportion

of Category 4 and 5 hurricanes Similarly, in the Western North Pacific, where typhoons occur, studies have observed a trend towards more intense storms and a poleward shift in their tracks, potentially increasing the risk of landfall in highly populated coastal areas

Moreover, climate change contributes to the exacerbation of other extreme weather events, such as heatwaves and droughts, which can have devastating impacts on ecosystems, agriculture, and human health The World Meteorological Organization (WMO) reports that heatwaves have become more frequent and prolonged in many regions, with record-breaking temperatures occurring with increasing frequency For instance, the European heatwave of 2019 shattered temperature records across the continent, with France, Germany, Belgium, the Netherlands, and the United Kingdom experiencing their highest temperatures on record These extreme heat events pose significant risks to human health, particularly among vulnerable populations, and can also exacerbate drought conditions, further straining water resources and agricultural productivity

Furthermore, climate change contributes to changes in precipitation patterns, leading to more intense rainfall events and an increased risk of flooding Warmer air can hold more moisture, resulting in heavier downpours and an elevated risk

of flash floods and urban flooding The IPCC projects an increase in the frequency and intensity of heavy precipitation events in many regions as global temperatures continue to rise One notable example is the catastrophic flooding caused by Hurricane Harvey in 2017, which dumped unprecedented amounts of rainfall over Houston, Texas, leading to widespread inundation, infrastructure damage, and loss of life

In addition to tropical cyclones and flooding, climate change also influences other extreme events such as wildfires and storm surges Rising temperatures and prolonged periods of drought create conducive conditions for wildfires to ignite and spread, as seen in the devastating wildfires that ravaged Australia, California, and the Amazon rainforest in recent years Meanwhile, sea-level rise, driven by thermal expansion

of ocean waters and melting ice caps, exacerbates storm surges associated with tropical cyclones, posing increased risks to coastal communities and infrastructure

Overall, the evidence linking climate change to the increasing frequency and intensity of extreme weather events is robust and compelling As global temperatures continue to rise, the imperative to mitigate greenhouse gas emissions and adapt to the impacts of climate change becomes ever more urgent, to reduce the risks and vulnerabilities associated with extreme weather phenomena and safeguard the well-being of communities and ecosystems worldwide

2 How AI models affect the climate change

The training and inference phases of large language models (LLM) such as multi-modal chatbots are intensive processes Large-scale algorithms impose a substantial burden on energy and natural resources, often fueled by nonrenewable sources They require ballooning computational resources, just as the energy and infrastructure needs of AI systems are growing themselves thanks to advancements such as generative AI and mounting industry competition

Generative models with broad functionality such as Bard or ChatGPT employ far more power per query than conventional applications and require hardware at levels beyond task-specific computing systems The more we ask models to perform wide-ranging tasks, the greater the amount of energy and carbon necessary For instance, supporting AI’s pervasive need for real-time application requires a power increase for data centers

And LLMs grow much bigger year after year Training earlier chatbots models such as GPT-3 led to the production of 500 metric tons of greenhouse gas emissions equivalent to about —

1 million miles driven by a conventional gasoline-powered vehicle This same model required more than 1,200 MWh during the training phase roughly the amount of energy used —

in a million American homes in one hour Future iterations may perpetually increase these metrics Updated versions such as GPT-4 have much greater needs and generate higher carbon emissions, though a lack of accessible input and output information renders analyses difficult

3 The Solution & Prevention

Artificial Intelligence (AI) has the potential to be a powerful tool in combating climate change by enhancing resource

efficiency, enabling more informed decision-making, and fostering innovations in energy systems and carbon management Here are several key areas where AI can contribute significantly:

Climate Modeling and Prediction AI, especially machine : learning techniques, can be used to improve the accuracy and granularity of climate predictions It can analyze vast amounts

of environmental data much faster than traditional methods, helping scientists and policymakers understand climate dynamics, predict future scenarios, and assess the impact of different policies

Energy Sector Optimization: AI can optimize the generation, distribution, and consumption of renewable energy It can predict energy demand and supply fluctuations, improve grid management, and enhance energy storage efficiency This supports the shift towards more sustainable energy sources by making them more reliable and cost-effective

And other aspects for example: Smart Infrastructure Carbon , Capture and Storage Precision Agriculture Environment , , Monitoring and Enforcement Climate Finance and Risk , Management, Public Awareness and Behavioral Change …

By integrating AI into these areas, not only can efficiencies be improved, costs reduced, but new solutions can emerge to tackle the complex challenges associated with climate change However, it's crucial to ensure that the deployment of AI in environmental contexts is done ethically and sustainably, considering broader societal impacts and minimizing potential negative outcomes such as biases or increased inequality

This article will mainly focus on EARTH-2, a Platform&Service created by NVIDIA that helps monitoring and predicting weather, climate condition

4 The EARTH-2 Platform & Services

4.1 Creator: NVIDIA NVIDIA CUDA-X Microservices –NVIDIA Corporation is a prominent American technology company incorporated in Delaware and based in Santa Clara, California It was founded in 1993 by Jensen Huang, Chris Malachowsky, and Curtis Priem NVIDIA is best known for its Graphics Processing Unit (GPU) products, which are high-performance hardware used in gaming, Figure 2: AI can have several good impacts on fighting climate change

professional visualization, data centers, and automotive markets

NVIDIA's GPUs are particularly well-known in the gaming industry, where they are used to create complex graphics and realistic visual effects The company's GeForce line is one of the most popular choices among gamers for its performance and efficiency

Beyond gaming, NVIDIA has expanded into AI and deep learning markets Their CUDA technology, a parallel computing platform and application programming interface, enables dramatic increases in computing performance by harnessing the power of the GPU NVIDIA also ventures into the automotive industry, providing AI solutions for autonomous vehicles and infotainment systems Their Tegra mobile processors are found in self-driving cars, where they process visual data and make navigational decisions

Overall, NVIDIA stands as a leader in the creation of GPUs for a wide range of applications, pushing forward innovation in gaming, professional visualization, and AI technologies.)

NVIDIA CUDA is a parallel computing platform, stands for Compute Unified Device Architecture

CUDA-X Microservices are developer tools, accelerated libraries, and technologies packaged as cloud APIs They are easy to integrate, customize, and deploy in data processing, AI, and HPC applications

GPU-CUDA-X microservices include NVIDIA® Riva for customizable speech and translation AI, NVIDIA Earth-2 for high-resolution climate and weather simulations, NVIDIA cuOpt for routing optimization and ™

NVIDIA NeMo™ Retriever for responsive augmented generation (RAG) capabilities for enterprises 4.2 Earth-2 Overview

retrieval-Earth-2, an AI-powered Earth climate digital twin designed to revolutionize weather and climate simulation

This initiative aims to address the growing economic losses, estimated at $143 billion, caused by extreme weather events due to climate change (2000 2019) –

It is a full-stack, open platform that accelerates climate and weather predictions with interactive, AI-augmented, high-resolution simulation It includes physical simulation of numerical models like ICON; machine learning models such as FourCastNet, GraphCast, and

through NVIDIA Modulus; and data federation and visualization with NVIDIA Omniverse ™ Running

and OVX supercomputers, Earth-2 will provide a path ™

to simulate and visualize the global atmosphere at unprecedented speed and scale

Features of Earth- 2:

Higher Resolution and Large-Scale AI Training: The Earth-2 accelerated systems will let climate scientists produce kilometer (km)-scale climate simulations, conduct large-scale AI training and inference, and achieve low-latency interactivity NVIDIA Modulus integrates support for numerous neural network models for climate and weather simulation

GPU-Optimized and Accelerated Climate Simulation The : Earth-2 development platform is optimized for GPU-

accelerated numerical climate simulations at km-scale to maximize simulated days per day (SDPD)

Data Federation and Interactive Weather Visualization: NVIDIA Omniverse enables ultra-large-scale, high-fidelity, interactive visualizations that depict weather conditions across the globe Omniverse Nucleus includes

a data federation engine that offers transparent data access across external databases and live feeds

High-level overview of how Earth-2 works and its capabilities:

Earth-2, integrated into NVIDIA's CUDA-X microservices ecosystem, introduces cloud-based APIs accessible through NVIDIA DGX Cloud, allowing users to create AI-driven simulations of weather phenomena with remarkable detail and efficiency These simulations range from global atmospheric patterns to localized weather conditions such as cloud cover, typhoons, and turbulence Earth-2's also has integration with NVIDIA Omniverse aims to provide a powerful platform for developing 3D workflows and applications based on Universal Scene Description (OpenUSD)

By harnessing the computational power of NVIDIA DGX Cloud, Earth-2 facilitates full-stack acceleration for climate and weather solutions This includes optimizing

AI pipelines for models like FourCastNet and Deep Learning Weather Prediction, as well as GPU acceleration for numerical weather prediction models such as ICON

II Detail Workflows of Earth-2

To simulate such an enormous object like the Earth, it requires numerous techniques and technologies, from the collecting data step to processing it, and lastly, which also an important step is how

to visualize it Fortunately, NVIDIA is such a big company that they can do all these things with the products and humans The ir collecting data step is obvious, data about current weather conditions has been collected for a long time, as far back to 1920s with various techniques Then NVIDIA introduced several models working together to perform predictions and visualize the conditions namedly GraphCast, FourCastNet and CorrDiff Then all the data is visualized by NVIDIA Omniverse

1 The process of weather forecasting

Observing Current Weather Conditions

Weather forecasts start with collecting data about the current state of the atmosphere This is done through various observation systems, including:

Automated Surface Observing Systems (ASOS): A network of over 900 stations across the U.S that constantly monitor

Figure 3: Create a digital twin of the Earth is a long process yet need to be done