Resource Repository - CCAI Virtual Summer School 2026
The Climate Change AI Virtual Summer School aims to democratize expertise in responsible AI and climate, by providing essential grounding in AI, climate, responsible design, and their intersection, as well as exposure to a wide range of AI-for-climate applications. The 2026 program featured two types of content modules:
- Foundation modules provide important technical and socio-technical foundations for doing responsible, impactful, and scientifically-grounded AI-for-climate work. Modules include introductions to AI and climate change; responsible, ethical, and sustainable AI; and considerations for shaping AI-for-climate projects in practice.
- Sectoral modules contextualize where both basic and advanced AI methods play a role across different climate-relevant sectors, including opportunities, limitations, and important considerations. These modules are meant to provide both broad overviews and concrete examples of AI-for-climate work within a given sector. The 2026 program included 17 sectoral modules.
Materials from the program are below, organized by content module. Each module includes lecture recordings, slides (in English, Spanish, Portuguese, French, and Italian) and, for many modules, hands-on coding tutorials with walkthrough videos. Recordings are also available via this YouTube playlist.
Click a module to see its materials.
Foundation Modules
Tackling Climate Change with Machine Learning
About this lecture
This lecture introduces how AI can help address climate change, with applications ranging from environmental monitoring to improving energy systems. It provides a foundation in climate change and AI, before exploring where AI might be useful and where other approaches may be more appropriate. The lecture explores considerations for conducting impactful AI-for-climate projects, including assessing the available data and working with people who understand the problem. It emphasizes planning for deployment early and considering who will benefit from a proposed solution. Participants also holistically examine AI’s environmental impacts and learn ways to get involved, with a central message: climate needs should guide how and when to use technology.
Introduction to AI
About this lecture
This lecture covers what artificial intelligence (AI) and machine learning (ML) methods actually are and how they work. This includes what an ML model is, the main types of ML models and methods (including several paradigms particularly relevant for climate applications), and what the ML modeling pipeline looks like (including considerations such as data quality and preparation; training, validation, and test splits; and overfitting). The lecture highlights the importance of understanding that ML is a tool (not a silver bullet), and includes discussion of when ML tools are/aren’t a good fit, as well as discussion of important ethical considerations.
About this tutorial
Build and evaluate an introductory machine-learning model using USGS earthquake observations to identify earthquakes likely to generate tsunami alerts.
Introduction to Climate Change
About this lecture
This lecture explains what climate change is, what causes it, and where things stand now. The lecture presents key terms and taxonomies, including mitigation, adaptation, climate science, and the IPCC sector categories, as well as the main levers for cutting emissions across sectors, using frameworks like the Kaya Identity and rough order-of-magnitude estimates of what each lever can deliver. It also covers who acts and how, from policy and regulation to markets, technology, corporate action, and individual choices, as well as important frameworks and considerations such as climate justice, equity, co-benefits, and maladaptation (and why they matter for climate action that holds up in practice).
About this tutorial
Use the En-ROADS climate simulation model to construct emissions-reduction policy scenarios targeting warming below 2°C (ideally 1.5°C), and evaluate implementation trade-offs.
Responsible & Ethical AI
About this lecture
This lecture explores how human choices, values, and power shape AI and its applications to climate challenges. The lecture examines both systemic and geopolitical impacts and project-level responsibilities, beginning with whether AI should be used at all for a given task (“Question Zero”) and whose needs it would serve. The lecture considers what fairness and accountability mean in practice, recognizing that these concepts involve value judgments that technical metrics alone cannot resolve. It introduces frameworks for responsible AI and data governance, following a project from its initial design through its eventual retirement. Particular attention is given to meaningful stakeholder engagement, including community ownership and agency. The central message is that responsible AI requires ongoing participation and human judgment, extending beyond technical fixes to address the broader social consequences of technology.
Sustainability Impacts of AI
About this lecture
This lecture explores how AI affects the environment and what it would take to align its development and use with climate goals. It begins with the environmental costs of computing, examining the resources needed to produce hardware and train and run models. It explains how these demands vary by task and how smaller, purpose-built models can reduce them. The lecture then considers application impacts, showing how AI can support climate solutions but also expand emissions-intensive activities such as oil and gas production. At the system level, it explores how increased consumption can offset efficiency gains and how AI can influence misinformation and the distribution of power. These discussions inform a broader assessment of claims about AI’s “net” climate benefits, including why potential emissions savings do not justify unchecked growth in computing. The lecture closes with approaches to regulating AI’s energy and resource use, emphasizing transparency and the need to consider local communities when planning data centers.
About this tutorial
Measure the carbon footprint of training machine learning models and practise strategies to reduce energy use and associated emissions.
Introduction to AI Policy & Regulation
About this lecture
This lecture explores how AI is governed and the implications for climate, innovation, deployment, and decision-making. The lecture introduces who shapes AI policy and how potential harms inform decisions about oversight. Drawing parallels with climate risk, it examines how the likelihood and consequences of harm can guide appropriate safeguards. Examples within and outside the U.S. illustrate how approaches centered on managing risks and protecting rights translate into policy. Particular attention is given to the connections between AI expansion and energy infrastructure, including how growing electricity demand affects communities and climate goals. The lecture closes by exploring principles for supporting clean and affordable power as AI develops. Its central message is that AI policy is also climate and infrastructure policy, with decisions made today shaping long-term environmental and social outcomes.
Introduction to Climate Policy & Regulation
About this lecture
This lecture explores why climate policy is needed and how it shapes the development and use of AI for climate action. The lecture introduces mitigation and adaptation as policy goals and examines how countries’ different circumstances influence their approaches. It explains the international climate policy landscape and considers questions of equity, including who bears responsibility for emissions and who has the resources to act. Drawing on examples from India and other countries, the lecture then explores what it takes to turn AI predictions into useful decisions. Case studies show why representative data and testing under real-world conditions matter, and why institutions need clear responsibility for acting on model outputs. A detailed example of dengue forecasting illustrates these challenges in practice. The central message is that effective climate AI requires both useful models and the capacity to translate their results into sustained action.
Shaping Your AI-for-Climate Project in Practice
About this lecture
This lecture discusses what it takes to turn an AI-for-climate idea into something that actually works and has actual on-the-ground impact. It starts from the perspective that the model is only one layer of the system, and that data infrastructure, governance, and how the tool fits into people's workflows matter just as much. The lecture provides a set of questions for deciding whether AI is even the right fit – such as whether the data is good enough, whether operational constraints allow it, whether stakeholders will trust the outputs, and whether a simpler method would do the job – and gives a clear view of what is deployable today vs. what still needs real innovation. The lecture emphasizes that deployment is a “loop” that involves piloting, getting feedback, and iterating before scaling, instead of a one-off rollout. The lecture also emphasizes the importance of measuring what matters, working backward from the desired outcome rather than simply measuring whatever is easiest. It ends on power and data ownership, and why communities should be co-developers and not just data sources. Overall, this lecture provides a helpful set of tools for those aiming to scope impactful AI-for-climate projects or for identifying why one may have stalled.
Sectoral Modules
AI for Accelerated Materials Science
About this lecture
This lecture presents a detailed overview of how machine learning is speeding up the search for materials that matter for climate, such as next-generation photovoltaics, batteries, alternative fuels, and CO2 sorbents. It covers where AI fits across the discovery pipeline, from predicting properties and screening candidates to characterization, optimization, and experimental validation. The lecture deep-dives into how ML makes huge chemical spaces searchable, using material representations and learned latent spaces to surface candidates that conventional approaches would miss. It also covers generative AI and inverse design, where one starts from the properties they want and then works toward the material, instead of only predicting what known materials do. Autonomous labs are also briefly discussed, where models propose candidates, experiments produce new data, and that data feeds back into better predictions. The lecture is clear that faster discovery does not automatically mean climate impact; it looks at what gets in the way, including lifecycle impacts, toxicity, critical-mineral use, access to data and compute, lab safety, and getting materials out of the lab and into deployment.
About this tutorial
Use machine-learning interatomic potentials to investigate catalytic materials and accelerate computational catalyst discovery.
AI for Agriculture and Food Security
About this lecture
This lecture explores how AI can turn Earth observations into useful information for agricultural decisions, from individual farms to global food security monitoring. It begins with agriculture’s dual role as a contributor to climate change and a sector vulnerable to its impacts. The lecture explains how satellite imagery complements field observations and other agricultural data, emphasizing that data and AI methods should match the scale and purpose of the decision. Examples include guiding fertilizer application within fields, mapping crops and land use, and monitoring risks to food production. It also introduces geospatial foundation models and their potential to support multiple agricultural tasks. The lecture examines why models may struggle when applied to new regions or when changing climate conditions make past patterns less reliable. Throughout, it emphasizes combining AI outputs with human expertise and communicating uncertainty clearly so that results can support informed action.
About this tutorial
Use Fields of The World data and pretrained geospatial models to delineate agricultural field boundaries, classify crop types, and monitor forest loss.
AI for Biodiversity & Ecosystems
About this lecture
This lecture discusses how machine learning helps us track life on Earth, a system we will never fully observe. It compares the main ways of collecting data: camera traps (analyzed by models like MegaDetector), acoustic recorders (analyzed by models like BirdNET), eDNA, and drone, aerial, and satellite imagery. The lecture highlights these data collection methods’ trade-offs in cost, coverage, and resolution, to provide insight into what data collection methods to use when. The lecture also covers species distribution models, both mechanistic and correlative, and why they are hard to get right: absence of data, environmental layers that may map the habitat instead of the species, and correlations that look convincing but mean nothing. It also discusses ecosystem questions like plant traits, phenology, invasive species, and how distributions might shift under climate change, and provides a candid overview about what is still unresolved. The lecture also covers multimodal foundation models and when to be careful with them, biodiversity indicators like the Living Planet Index, and area prioritization with CAPTAIN, along with FAIR and CARE data principles. It ends with an actionable list of datasets, models, and communities to help get started in the AI for biodiversity space.
About this tutorial
Develop efficient wildlife bioacoustic classifiers using transfer learning, audio search, and human-in-the-loop active learning.
AI for Buildings & Cities
About this lecture
This lecture covers how machine learning can cut energy use in a sector that accounts for roughly 30% of global final energy. The talk presents a “sense, model, decide, verify” framework for thinking about machine learning’s role, and applies it to four problems: (1) pulling appliance-level usage out of a single smart meter (NILM), (2) forecasting demand and shifting it away from peaks, (3) coordinating HVAC, blinds, and lighting in a way that is seamless for building occupants, and (4) searching for low-carbon building designs. A recurring question throughout the lecture is whether a result survives a new building, the real grid, and real occupants. The lecture provides a frank perspective about what goes wrong, such as models with low average error that still miss the events that matter, sensors that fail in real homes, and the privacy cost of fine-grained data. It also emphasizes the importance of simple baselines, which are often effective in practice.
About this lecture
This lecture provides an overview of how AI can be used to help planners understand cities. It discusses the role of AI as pattern recognition that supports human judgement, with planning values coming first. The work is framed as a loop from city to data to model to insight to policy, and back again. AI's role is taxonomized into categories including prediction, simulation, interpretation, and intervention, with examples from mobility, urban heat, urban form, and housing. The lecture provides a detailed overview of spatial data (points, lines, polygons, and rasters), why scale and neighbourhood effects matter, and how bias in who gets measured and who can access the data limits what a model can say. It then covers planning-related methods and applications including computer vision on street imagery, graph learning, explainable AI, and simulation for testing scenarios that planners can actually debate.
About this tutorial
Use the CityLearn environment and real-world building data to design and test energy-flexibility control experiments, including reinforcement learning.
AI for Carbon Accounting (MRV)
About this lecture
This lecture talks about how satellites and machine learning are changing the way we check who is emitting what, particularly for methane. It sets bottom-up accounting, built on self-reported activity data and emission factors, against top-down estimates from atmospheric observations, and asks which one you should trust, why, and when. A large portion of the lecture follows the methane plume pipeline, from raw satellite radiance to plume detection and emission rates. It shows where AI helps, such as in finding emitting infrastructure, estimating emissions for plants that do not report, and spotting plumes faster, as well as where it struggles. The lecture observes that labelled data is scarce (so synthetic plumes are often used for training), and that in the tests shown, general Earth foundation models did worse than CNNs. The lecture is candid that ground truth is missing at several stages, and that a false positive can trigger a regulatory fee, as in the US super-emitter reporting rules. It notes that pairing ML with a physics check is one practical approach to these challenges.
About this tutorial
Use satellite imagery and computer vision to monitor activity at coal-fired power plants for emissions-related monitoring and reporting.
AI for Climate Finance
About this lecture
This lecture explores how AI can inform financial policy and investment decisions that support climate mitigation and adaptation. The lecture begins with why finance matters for climate action and how AI can help fill information gaps. It first examines international climate finance, connecting countries’ historical responsibility and ability to pay with the needs of vulnerable countries. A case study shows how machine learning can analyze project descriptions to independently assess reported climate funding. The lecture then shifts to financial markets, examining physical and transition risks and using company disclosures to identify the financial effects of extreme weather. Finally, it explores financing for climate innovation, showing how AI can track investment flows and help evaluate whether public funding attracts additional private investment. It closes with datasets and tools for getting started, emphasizing the importance of understanding data limitations and checking findings against independent evidence and expert judgment.
About this tutorial
Estimate hurricane wind value-at-risk using an ensemble of AI weather forecasts and financial exposure information.
AI for Climate Science
About this lecture
This lecture explores how machine learning is advancing weather forecasting and the analysis of climate data. The lecture examines how generative modeling can be used to produce ensembles of plausible forecasts to represent uncertainty, while self-supervised learning can be used to extract useful patterns from datasets with limited labels. It also introduces applications in downscaling and bias correction, helping align climate simulations with observations. Throughout, the lecture emphasizes adapting methods to the physical properties of geophysical data and evaluating predictions beyond average error alone. It closes by considering climate data equity and how transfer learning can support regions where observations are scarce.
About this tutorial
Train and evaluate a convolutional neural network on New Zealand CORDEX ML-Bench precipitation data to downscale coarse GCM projections into higher-resolution local climate information.
AI for Forestry
About this lecture
Given that forests are too big to survey on foot, this lecture shows how AI can help fill the gap via analysis of satellite data. The lecture touches on methods from Random Forests and SVMs to CNNs, vision transformers, and now foundation models, with a clear view of where each one fits. Regarding foundation models, the lecture observes that while today’s foundation models show promise, they do not always transfer to tropical forests, are costly to run, and are hard to interpret. The lecture then discusses AI-for-forestry applications including tree species classification from airborne LiDAR, wildfire prediction and real-time fire segmentation from drone video, and biomass and carbon estimation. The lecture observes that combining optical, radar, and LiDAR data often turns out to matter more than which machine learning algorithm one picks. The lecture also looks at systems already in use, like INPE's monitoring programs in Brazil, Global Forest Watch, and initiatives across Africa. It finishes on explainable AI, and why forest managers need models they can trust before they act on them.
About this tutorial
Detect tropical deforestation from Sentinel-2 NDVI time series with Brazilian Amazon PRODES labels, extending to AlphaEarth embeddings and a time-series Transformer.
AI for Oceans & Marine Systems
About this lecture
The ocean absorbs about 90% of the excess heat from global warming and roughly a quarter of our CO₂. This lecture therefore posits that the ocean is the centre of the climate system, and identifies where machine learning can help with ocean monitoring and resource management. The lecture starts from the simple physics of temperature and salinity that drives global ocean circulation, then looks at what warming, acidification, sea level rise, and marine heatwaves do to plankton and fisheries. Since no single data source is enough to monitor oceans, it compares in-situ observations, satellites, and numerical models, and where each one falls short, including who gets left out. Machine learning is framed as a bridge between theory and data, not the destination. The lecture covers example applications of machine learning for oceans, including estimating surface pCO₂ from satellite data, tracking Antarctic plankton, attributing marine heatwaves, and using the THOR framework to see how Atlantic overturning circulation changes across climate models. It also covers illegal fishing detection with Global Fishing Watch and why detection without enforcement capacity can shift the burden onto the wrong people.
About this tutorial
Use satellite observations and machine learning to assess coastal water quality and marine ecosystem changes.
AI for Policy, Economics, and Social Sciences
About this lecture
AI can touch every stage of the climate policy cycle, and this lecture discusses where it helps and also where it does not. The lecture starts by discussing the use of AI for measuring the policy world, covering applications such as using natural language processing to harmonize fragmented policy databases and to map hundreds of thousands of papers; with that said, the lecture also cautions that high-resolution data can give false precision where data is thinnest. The lecture then covers the use of AI for modelling futures, including AI-based emulators that make integrated assessment models fast enough to explore, as well as reinforcement learning (including multi-agent versions) for electricity market design. On evaluation, the lecture covers causal methods such as synthetic control, with Sweden's carbon tax as the example, as well as AI for verifying carbon credits and the limits of LLMs in policy workflows. The lecture also discusses AI's own environmental footprint.
About this lecture
Climate social science examines what people say, write, and do, and this lecture shows how AI is changing how researchers study it. Examples include classifying climate salience in political manifestos across 20+ languages, tracking contrarian claims in US Congressional speeches, mapping polarization around the COPs on social media, and linking firms' earnings calls to their lobbying. The lecture also covers AI for faster evidence synthesis and AI-assisted qualitative analysis. The lecture also discusses the cautionary side of using AI for these applications: for instance, misclassification bias and why validation can't be skipped; why LLMs don't reliably simulate survey respondents; the risk of deskilling research and early-career work; and who controls the technology, given the concentration of AI compute in a small number of countries. The throughline of the lecture is the importance of keeping focus on the climate-related research question, rather than having the focus be the AI tool itself.
About this tutorial
Use supervised NLP models to identify climate-related literature (Part I), then large language models to classify climate-policy documents (Part II).
About this tutorial
Use supervised NLP models to identify climate-related literature (Part I), then large language models to classify climate-policy documents (Part II).
AI for Power & Energy Systems
About this lecture
As more wind, solar, electric vehicles, and distributed resources are added to the power grid to enable decarbonization efforts, this makes balancing electricity supply and demand harder. This lecture provides an overview of where AI helps manage the increasing complexity of power grids, covering applications across planning, operations, maintenance, markets, and demand response. The lecture includes in-depth discussion of solar, wind, and demand forecasting, including differences in approach between big-data vs. small-data settings, temporal and spatial forecasting settings, the importance of uncertainty-aware forecasts, and link between forecasts and action. The lecture also includes in-depth discussion of physics-constrained AI for grid optimization, using the case study of optimal power flow to walk through PINNs, constraint-aware architectures, graph neural networks, and other approaches to align neural network outputs with the laws of physics. The lecture also highlights risks and challenges associated with AI model deployment and data scarcity, including perspectives on the level of autonomy vs. oversight that is realistic for AI in power grids today, and pointers to an open grid dataset built from low-cost sensors.
About this tutorial
Train a PyTorch neural network to approximate solutions to alternating-current optimal power flow subject to power-system operational constraints.
AI for Public Health
About this lecture
This lecture looks at how AI can be used to link climate and environmental data to health outcomes. The lecture covers early warning systems for heat, floods, and air pollution, as well as a pipeline that connects satellite and weather data to disease burden. Malaria, acute respiratory infection, child wasting, and musculoskeletal injury serve as case studies. Grounded in these case studies, the lecture provides important observations for impactful research and deployment – for instance, as exemplified by Ethiopia's EPIDEMIA malaria forecasting system that has provided effective forecasts for over a decade, the best model is not always the one with the top benchmark score, but one that is actually deployable, due to factors such as usability, ready access to data, and thoughtful co-design with health authorities. The lecture also discusses topics such as inconsistent definitions of heat, missing and conflicting patient records, data and algorithm bias that can lead to unequal care, and the need for common data models. It also covers climate-health co-benefits and AI's own footprint, and it ends on the point that how this technology gets used is a human choice.
About this lecture
This lecture looks at how AI can be used to link climate and environmental data to health outcomes. The lecture covers early warning systems for heat, floods, and air pollution, as well as a pipeline that connects satellite and weather data to disease burden. Malaria, acute respiratory infection, child wasting, and musculoskeletal injury serve as case studies. Grounded in these case studies, the lecture provides important observations for impactful research and deployment – for instance, as exemplified by Ethiopia's EPIDEMIA malaria forecasting system that has provided effective forecasts for over a decade, the best model is not always the one with the top benchmark score, but one that is actually deployable, due to factors such as usability, ready access to data, and thoughtful co-design with health authorities. The lecture also discusses topics such as inconsistent definitions of heat, missing and conflicting patient records, data and algorithm bias that can lead to unequal care, and the need for common data models. It also covers climate-health co-benefits and AI's own footprint, and it ends on the point that how this technology gets used is a human choice.
About this tutorial
Create a dengue outbreak analytics workflow that integrates prediction, retrieval-augmented generation and evaluation of LLM-generated responses.
AI for Risk Assessment, Disaster Management & Relief
About this lecture
This lecture explores how AI can support disaster preparedness and response by improving our understanding of risk. The lecture begins with hazard forecasting and mapping, then examines methods for identifying the people and infrastructure in harm’s way as exposure changes over time. It next considers vulnerability, including how building conditions and access to resources shape the harm communities experience and their ability to recover. These components come together in risk assessments that estimate potential losses and disruptions to essential services. The lecture then follows decision-making across the disaster cycle, from early warnings and forecast-based financing to emergency response and post-event recovery. It closes with responsible implementation, addressing uncertainty and data gaps that can leave vulnerable communities overlooked. Throughout, the emphasis is on tools that support clear decisions and fit institutional capacity, with local knowledge and human oversight guiding their use.
About this tutorial
Assess coastal flooding exposure in Egypt using Sentinel-1/2 and Copernicus elevation data, including mapping population or building density and CNN-based flood detection.
About this tutorial
Assess coastal flooding exposure in Egypt using Sentinel-1/2 and Copernicus elevation data, including mapping population or building density and CNN-based flood detection.
AI for Transportation
About this lecture
This lecture explores transportation emissions and climate targets before examining why the sector is difficult to decarbonize. It provides an overview of AI applications and considers where they can support climate change mitigation. The lecture then takes a closer look at how AI can support electrification and help shift travel toward lower-emission modes. Examples include smart charging and vehicle energy management. It also examines emerging uses of large language models to simulate travel behavior and inform transportation planning. A further section explores how AI can improve transport efficiency through applications such as traffic signal optimization and vehicle routing. Throughout, the lecture considers equity and the risk that efficiency gains could encourage more travel. It closes by connecting technical opportunities to broader transportation priorities and offering resources and next steps for getting involved.
About this tutorial
Model associations between bike-share usage and nearby points of interest to propose new shared-bike station locations.
AI for Water Resources and Hydrology
About this lecture
This lecture explores how AI and optimization methods can support reservoir management under changing water availability. The lecture introduces genetic algorithms and explains how they search for solutions by selecting and modifying candidate solutions over successive generations. Reservoir case studies demonstrate how these methods can maximize hydropower generation while meeting operational constraints. The lecture then examines how multi-objective optimization reveals trade-offs between hydropower production and irrigation needs. It also introduces neural network approaches for forecasting reservoir inflows while considering climate variability. A forecasting case study compares methods across different time scales and shows how forecasts can inform reservoir operating decisions. The lecture closes with opportunities to integrate these approaches into decision-support systems for climate-resilient reservoir management.
About this tutorial
Use the CAMELS dataset of 531 continental US watersheds to build streamflow prediction models for flood and drought planning.
AI for Weather
About this lecture
This lecture explores how machine learning is changing forecasts, spanning horizons from a few days ahead to the coming season. The lecture begins with short-range prediction, showing how models such as GraphCast, GenCast, and AIFS achieve accuracy comparable to physics-based numerical weather prediction for many variables, but also how their accuracy can degrade faster than the IFS numerical weather prediction model as lead time grows. The lecture also explores the use of graph transformers for high-resolution refinement of predictions in specific regions, and discusses why models trained on ERA5 need correcting toward satellite estimates, particularly for rainfall. Probabilistic skill is emphasized throughout, including calibration, discrimination, and uncertainty. The lecture then turns to seasonal forecasting, where raw model bias must be corrected with methods such as quantile mapping and Season-Net before forecasts can support decisions. It considers how correlation and error metrics can hide poor skill on impact-based events like heavy rain days and warm spells. A farmer-focused example, ImvulaNet, illustrates the importance of co-production with local communities, and the lecture closes with the effects of sparse observational data in Africa, emphasizing that forecast skill only creates value when it reaches the people who act on it.
About this tutorial
Stack IFS HRES, GraphCast, Pangu-Weather and NeuralGCM weather forecasts with XGBoost; use SHAP to explain the ensemble’s predictions.
Combining AI and Indigenous Knowledge for Climate Action
About this lecture
This lecture starts with what Indigenous and local knowledge actually is: a living system tied to land, language, community, and governance, where facts come with responsibilities. It then looks at why that matters for climate, including Indigenous lands covering more than a quarter of the world's land surface, thousands of local observations of climate change that largely match modelled trends, and studies of Indigenous fire stewardship. Much of the lecture is about what goes wrong when knowledge gets extracted, such as a review of 81 adaptation studies that mostly treated it as data to fill scientific gaps. The lecture further covers data sovereignty through the CARE principles, the right to refuse or withdraw, and a governance lifecycle for AI projects, from community mandate to co-design to returning benefit.
About this tutorial
Build a bird-bioacoustics monitoring pipeline for the Lkotkote case study, integrating conservation-oriented AI with locally grounded knowledge.