Tutorials
Question Answering over Sustainability Reports: Information Richness and Answer Quality
In this tutorial, we learn about an advanced strategy for information retrieval for question answering with Large Language Models (LLMs) in knowledge-intensive domains like sustainability reporting. We analyze the quality and quantity of sources before answering a question and quantify uncertainty around answering a question after LLM generation.
Climate Policy Radar's Open Knowledge Graph
Climate Policy Radar (CPR) helps people access and understand vast amounts of climate documents: laws, policies, NDCs, corporate transition plans, litigation documents, reports by statutory advisory bodies and industry bodies, and more. By linking this expert knowledge of climate change to the extensive curated database of climate documents, we show you how to create a climate policy knowledge graph. This can then be used in turn to analyse the global policy landscape.
PiggyCast - Improving Weather Prediction Accuracy through a Stacking-Based Ensemble AI Approach.
Accurate weather prediction is fundamental to understanding and adapting to the impacts of climate change. As our climate shifts, the frequency and intensity of extreme weather events are changing, making reliable forecasts more critical than ever. This tutorial introduces PiggyCast, an ensemble machine learning model designed to improve weather prediction accuracy by stacking forecasts from various numerical, AI-based, and hybrid weather prediction models. We demonstrate how a combined approach, using gradient-boosted decision trees (XGBoost), can surpass the predictive performance of individual base models.
Flood mapping with optical and microwave satellite data: From indices to machine learning
Flood mapping is a critical component of disaster response and climate adaptation, but it is often hindered by cloud cover, ambiguous spectral signals, and trade-offs in image resolution. This tutorial introduces participants to practical methods for flood detection using freely available multi-sensor satellite data. We begin with optical imagery from MODIS and Sentinel-2, showing how spatial resolution affects the level of detail captured in flood maps. We then move to microwave data from Sentinel-1 SAR, which can penetrate clouds and provide more reliable monitoring during flood events. Participants will implement rule-based approaches such as spectral indices, band ratios, and Z-score anomalies, followed by machine learning models including Logistic Regression and Random Forests. Finally, we demonstrate how trained models can be applied to flood events in different parts of the world, highlighting both the opportunities and limitations of generalizing across geographies. By the end of the tutorial, participants will understand how different satellite data sources and modeling approaches influence flood mapping outcomes, and how these methods can support disaster response and long-term climate resilience.
Agricultural Monitoring with Fields of The World (FTW)
This tutorial demonstrates how to generate field boundaries globally using the Fields of The World dataset, pretrained models, and command line interface (CLI). We then show how to use those boundaries in agricultural monitoring tasks under climate change, including crop type classification and forest loss monitoring. By the end, users will be able to perform the following tasks to support climate change-related decision-making: (1) Extract agricultural field boundaries for any location, (2) Build machine learning models for crop type classification, and (3) Analyze forest loss within agricultural landscapes. By equipping users with the ability to generate field boundaries and link them to climate-relevant monitoring tasks, this tutorial lowers the barrier for researchers, practitioners, and policymakers to access and deploy advanced geospatial AI.
AI‑Powered Measurement & Verification: Building Interpretable Counterfactual Models to Verify Energy Savings in Buildings
Buildings use a lot of energy, but proving that retrofits actually save it is hard. This tutorial is a hands-on guide to make energy efficiency measurable, investable, and a reliable climate asset, using Machine Learning. The goal is to construct interpretable counterfactual baselines from smart-meter and weather data, quantify savings with defensible uncertainty, and communicate results that meet IPMVP and ASHRAE Guideline 14.
Understanding Drivers of Climate Extremes Using Regime-specific Causal Graphs
The goal of this tutorial is to show how we can use methods from constraint-based causal discovery to uncover the causal relationships that are present in different moisture regimes. In doing that, we aim to improve our general understanding of the dynamics of extreme events, with application to understanding drivers of soil-moisture under different, more extreme, regimes.
Empowering Safe Reinforcement Learning for Power System Control with CommonPower
This tutorial introduces the CommonPower library, designed to benchmark safe reinforcement learning (RL) algorithms on control problems for power systems. We highlight two crucial issues in RL for power system control: safeguarding RL decision-making and assessing the impact of forecast quality on control performance. Participants will learn how to use CommonPower to further invesitigate these topics.
Planning for Floods & Droughts: Intro to AI-Driven Hydrological Modeling
A guide to model hydrological system using the real-world CAMELS dataset, which contains weather drivers for 531 basins across the continental United States. Through this modeling process, we will demonstrate various methods to predict streamflow, aiding in flood and drought planning.
Agile Modeling for Bioacoustic Monitoring
Bioacoustic monitoring promises to help unlock the ability to monitor biodiversity, ecosystem health, and endangered species cost effectively. This tutorial presents an "agile modeling" approach that enables users to build custom classifier systems efficiently for species of interest using transfer learning, audio search, and human-in-the-loop active learning.
Aquaculture Mapping: Detecting and Classifying Aquaculture Ponds using Deep Learning
Mananging aquaculture ponds is vital for environmental monitoring and conservation. This tutorial presents how to leverage satellite imagery and semantic segmentation models to detect and map aquaculture ponds based on production intensity.
Zero-Emission Vehicle Intelligence (ZEVi): Effectively Charging Electric Vehicles at Scale Without Breaking Power Systems (or the Bank)
This tutorial surveys different methods to formulate electric vehicle (EV) charging and energy dispatch as an optimization problem, using tools such as convex optimization, Markov decision process (MDP), and reinforcement learning (RL).
Sea Water Flood Risk Assessment in Egypt using Deep Learning, Sentinel-1 & 2, and Copernicus DEM: Part I
Floods in coastal areas can be extremely destructive natural hazards resulting in societal and economical damage. In this tutorial, explore how to predict building and population density to understand the potential impact from a flooding event.
Sea Water Flood Risk Assessment in Egypt using Deep Learning, Sentinel-1 & 2, and Copernicus DEM: Part II
Part II of this tutorial series demonstrates how train a CNN on satellite data to detect the extent of flooding in a potential disaster.
Reducing your Climate Impact when Training ML Models
Learn how to measure a machine learning model's carbon footprint and practice strategies that can help shrink the energy involved in training these models.
Introduction to Camera Trap Recognition with Deep Learning
The greatest logistical barrier to long-term wildlife monitoring with camera traps is the overwhelming amount of human labor needed to annotate thousands or millions of images for ecological analysis. In this tutorial, learn how to train a deep learning model to automatically identify animals in camera trap images.
Predicting Mobility Demand from Urban Features
Ensuring a shared bike is readily available can reduce demand for less climate-friendly transportation options. Explore how to model the relationship between bike usage and points of interest (POIs) to identify the best locations for new shared-bike stands.
AI for Optimal Power Flow
AC Optimal Power Flow (OPF) attempts to determine the setpoints of generators that would minimize the operating cost of a power system while meeting other operational constraints. In this tutorial, learn how to leverage PyTorch to train a neural network to approximate the optimal solutions.
NLP Models for Climate Policy Analysis: Part I
Explore how Natural Language Processing (NLP) can be used to assist in identifying and mapping climate-relevant literature using a supervised learning approach.
NLP Models for Climate Policy Analysis: Part II
Build on what you learned in Part I, and leverage a state of the art Large Language Model (LLM) to classify climate policy documents.
Estimating Coal Power Plant Operation From Satellite Images with Computer Vision
Explore how to monitor coal power plant activity by leveraging satellite imagery and computer vision models.
Smart Meter Data Analytics: Practical Use-Cases and Best Practices of Machine Learning Applications for Energy Data in the Residential Sector
A practical guide to current trends in smart meter data analytics with a focus on feature engineering and machine learning scenarios for energy data at 15-minute resolution. Gain insights into current trends and use cases in the energy field and get a sense of typical and atypical energy consumption in a residential building.
Quantus x Climate: Applying Explainable AI Evaluation in Climate Science
In climate science, explainable artificial intelligence (XAI) can be used to improve and validate deep learning methods, but evaluation and selection of XAI methods is challenging. Learn how to use the explainable AI evaluation package Quantus to compare and select an appropriate XAI for your climate AI research task.
CityLearn: Reinforcement Learning Control for Grid-Interactive Efficient Buildings and Communities
Learn how to design simple and advanced control algorithms to provide energy flexibility, and acquire familiarity with the CityLearn environment and its datasets for extended use in projects. The tutorial provides a walk-through on how to set up and interact with the environment using a real-world dataset in three hands-on control experiments.
Disaster Risk Monitoring Using Satellite Imagery
Learn how to build and deploy a deep learning model to automate the detection of flood events using satellite imagery. This workflow can be applied to lower the cost, improve the efficiency, and significantly enhance the effectiveness of various natural disaster management use cases.
ClimateLearn: Machine Learning for Predicting Weather and Climate
Apply machine learning to predict climate variables into the future and transform low-resolution outputs of climate models into high-resolution regional forecasts.
FourCastNet: A practical introduction to a state-of-the-art deep learning global weather emulator
Learn how to use FourCastNet, a weather model based on deep learning, to obtain short to medium-range forecasts of crucial atmospheric variables such as surface wind velocities.
Automating the Creation of LULC Datasets for Semantic Segmentation
Learn how to create a custom LULC dataset by combining multispectral satellite and vector data.
Forecasting the El Nino/ Southern Oscillation with Machine Learning
Existing El Niño forecasts use dynamical models that rely on the physics of the atmosphere and ocean. Learn how to create El Niño forecasts using machine learning instead, which uses statistical optimization to issue forecasts.
Check out the introductory slides for more information.
Building Load Forecasting with Machine Learning
Accurate forecasts of energy demand and supply are essential to mitigate climate change. Discover how to train and evaluate building load forecasts using off-the-shelf ML models.
Building Control with RL using BOPTEST
Apply reinforcement learning to a building emulator to intelligently control HVAC systems.
Check out the recorded talk for more information.
Land Use and Land Cover (LULC) Classification using Deep Learning: Part I
Mapping the extent of land use and land cover categories over time is essential for better environmental monitoring, urban planning and nature protection. Train and fine-tune a deep learning model to classify satellite images into 10 LULC categories.
Check out the introductory slides for more information.
Land Use and Land Cover (LULC) Classification using Deep Learning: Part II
Use the neural network trained in LULC Classification Part I to generate land use and land cover maps with Python GIS.
Check out the introductory slides for more information.
Open Catalyst Project: An Introduction to Machine Learning for Material Discovery
Chemical reactions are one approach to convert intermittently available renewable energy, such as solar or wind, to other fuels, like hydrogen. Learn how ML can discover low-cost catalysts to drive these reactions at high rates.'
Check out the recorded talk for more information.