Welcome to the AI and Machine Learning in Atmospheric Sciences Course#
My name is Carlos D. Hoyos, a professor at the Universidad Nacional de Colombia, Facultad de Minas, Sede Medellín. I’m here to guide you through the transformative intersection of atmospheric sciences and AI/ML. This course blends theory and hands-on applications, equipping you with skills to tackle real-world challenges. Course materials are hosted as Jupyter Books on GitHub, where you can explore models and datasets, and engage directly with advanced AI/ML concepts tailored to atmospheric science. This course also emphasizes recent advancements, including the work by NVIDIA, DeepMind, Microsoft, and Huawei, along with foundational ML theory and practical applications in atmospheric science.
The Role of AI and ML in Atmospheric Sciences#
AI and ML are transforming atmospheric sciences by enabling faster, more accurate analysis of complex systems. However, fully leveraging AI and ML requires a strong foundation in atmospheric sciences and the underlying physics governing these phenomena. Without an understanding of atmospheric physics, data scientists and engineers may overlook critical constraints or make inaccurate predictions, especially when interpreting or applying AI models to real-world atmospheric systems.
1. Weather Forecasting Improvements#
AI enhances weather forecasting by working alongside traditional numerical weather prediction models. For instance, deep learning models such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs) help analyze vast datasets, providing quicker and potentially more accurate forecasts, especially for extreme weather events like hurricanes. Major weather agencies, including the Met Office and ECMWF, are actively developing AI applications to improve forecast accuracy at a fraction of the computational cost required by traditional methods [ZLWC21].
2. Air Quality Monitoring and Prediction#
Machine learning models predict air quality by assessing pollutants like PM2.5, ozone, and nitrogen dioxide. These models, trained on satellite and ground sensor data, predict spatial and temporal patterns in pollution, aiding public health and environmental policies. For example, ML algorithms have been used to measure the effectiveness of air quality policies by identifying pollution trends and detecting hotspots [KSAA21, ZLWC21].
3. Renewable Energy Optimization#
AI plays a critical role in forecasting renewable energy availability, particularly for wind and solar sources. By analyzing atmospheric data on wind patterns and solar radiation, machine learning models can predict energy production, supporting grid stability and efficient energy supply forecasting. These models often surpass traditional approaches in handling the complex dependencies within this data, thus optimizing energy generation systems [SS19, ZLWC21].
4. Climate Modeling and Downscaling#
AI, especially physics-informed neural networks (PINNs), is transforming large-scale climate modeling by enabling high-resolution local predictions through downscaling. PINNs incorporate physical principles into neural networks, creating detailed climate data more affordably and efficiently than traditional methods. This high-resolution data is crucial for local planning and adaptation to climate impacts [SS19, ZLWC21].
5. Tracking and Predicting Extreme Weather Events#
By applying historical data and real-time satellite imagery, AI models detect and forecast extreme weather events, such as flash floods and hurricanes. Using anomaly detection and deep learning techniques, these models provide early warnings, which are invaluable for emergency response and disaster preparedness [KSAA21].
6. Health Impacts of Airborne Pollution#
AI also supports studies linking air quality with health outcomes, especially in areas with limited air quality monitoring. ML models trained on satellite and sensor data estimate pollution exposure levels, correlating trends with respiratory or cardiovascular disease rates. This approach has proven effective for assessing health risks in urban areas and regions with high pollution, informing healthcare and environmental policies [KSAA21, ZLWC21].
7. Risk Management and Early Warning Systems#
AI and ML are integral to enhancing risk management and early warning systems for natural disasters. By analyzing historical and real-time data, these models can forecast extreme events such as floods, droughts, and hurricanes. For instance, ML models use satellite and sensor data to identify patterns and precursors to extreme events, generating early alerts that enable communities and agencies to prepare effectively. These AI-driven systems are crucial for reducing the impacts of disasters on vulnerable populations, optimizing resource allocation, and supporting timely decision-making during emergencies [SS19, ZLWC21].
8. Agricultural Applications from a Climate Perspective#
In agriculture, AI and ML models help monitor climate conditions, predict crop yields, and assess climate-related risks. For example, ML techniques can forecast drought conditions by analyzing soil moisture, rainfall, and temperature data, allowing farmers to optimize irrigation and water management. Additionally, AI-powered models are used to predict pest and disease outbreaks by detecting climate conditions that favor their spread. These applications help improve food security, adapt agricultural practices to changing climates, and minimize losses due to extreme weather events [KSAA21, SS19, ZLWC21].
These applications showcase the essential role AI and ML play in advancing atmospheric science, providing tools to handle complex, data-intensive challenges in climate resilience, public health, and disaster preparedness.
Practical Relevance: Climate Resilience, Renewable Energy, and More#
Our study goes beyond theory, showing how AI/ML enables advancements in climate resilience, environmental protection, energy, and agriculture. For instance, machine learning helps improve wind and solar energy forecasting, optimize climate models, and predict extreme weather events. Additionally, AI-driven insights are crucial for managing water resources and informing climate adaptation policies. By mastering these techniques, you’ll gain a toolkit for addressing both global and local challenges.
Why Jupyter Books?#
Jupyter Books provide a unique platform for integrating live code, dynamic visualizations, and interactive examples directly into our course materials. This allows us to break down complex concepts and equations in real-time, giving you the opportunity to not only read but also engage with the material. Imagine being able to run a simulation of a fluid system while learning the theory behind it, or exploring interactive visualizations that help you better grasp the forces at play in different fluid scenarios.
The flexibility of Jupyter Books makes them an ideal medium for the topics of this course, a subject that often requires hands-on analysis and computational tools to fully understand. As we progress through the course, I encourage you to experiment with the code and simulations provided in these materials to deepen your learning. You can learn about Jupyter Books at jupyterbook.org.
Why GitHub?#
By hosting these materials on GitHub, I’m inviting you into a collaborative learning environment. Version control on GitHub ensures that the materials are always up to date, and it offers an opportunity for you to suggest improvements, raise issues, and even contribute to new content. This platform promotes transparency, community-driven learning, and continuous improvement—values that are essential in both academia and engineering practice.
If you don’t already have a GitHub account, I recommend creating one, as it will be useful not only for this course but also for future projects and collaborations in your career.
Why in English?#
While classes are in Spanish, course materials are in English to familiarize you with the technical terminology in AI and atmospheric sciences. Engaging with the content in English will also prepare you for global collaboration in research and industry, opening doors to global research efforts, international conferences, publications, and partnerships. It’s an investment that will expand your opportunities both academically and professionally. Many of the key resources—such as academic papers, research articles, and advanced textbooks—are published in English, and mastering this language will help you stay connected with the global scientific community.
A Bit About Me#
I have a Ph.D. in Earth and Atmospheric Sciences from Georgia Institute of Technology, along with a Masters in Oceanic and Atmospheric Sciences, a Masters in Water Resources, and a Civil Engineering degree. My work has taken me through diverse fields—studying the behavior of atmospheric currents, understanding water resources, and applying data science to solve problems related to agriculture, energy, risk management, and climate. I’ve had the privilege of using ML to address atmospheric challenges across diverse fields. This course reflects my commitment to teaching not just the fundamentals of AI and ML but also how to apply them in meaningful ways within atmospheric sciences, contributing to sustainable solutions and innovative approaches to pressing global issues.