Centram Centre for Applied Artificial Intelligence

AI ANALYTIC AND AUTOMATION FOR CLIMATE CHANGE MANAGEMENT

Course Overview

This course is designed for a professional in climate change, it balances theory, hands‑on labs, automation/deployment, ethics and policy, and a practical capstone.

Course title

AI Analytics and Automation for Climate Change Management

Course description

This course introduces methods and tools that combine AI, data analytics and automation to monitor, predict and manage climate risks and mitigation/adaptation interventions. Covers climate data sources, machine learning for time series and geospatial analysis, remote sensing, decision support and automated operational pipelines (MLOps), with case studies (flooding, wildfire, carbon accounting, agriculture, urban heat). Emphasis on reproducible, ethical and policy‑aware solutions.
Learning objectives

By course end students will be able to:

  1. Identify and access key climate and environmental datasets (satellite, reanalysis, observational).
  2. Build ML models for climate-relevant tasks: forecasting, classification, segmentation, anomaly detection and causal inference.
  3. Apply geospatial and remote-sensing techniques (GEE, raster/vector handling).
  4. Design automation pipelines for data ingestion, model training, monitoring and deployment (MLOps).
  5. Evaluate model fairness, uncertainty and robustness; understand ethical and policy implications.
  6. Deliver an end‑to‑end, production-ready climate analytics solution and communicate results to stakeholders.

Course Duration and delivery method

The course duration is 2 weeks’ intensive psychical boot camp training PLUS 3 weeks of online study OR 4 weeks of intensive physical boot camp training whichever is convenient to the participant

Course Content

Foundations

  1. Overview of climate risk management and decision contexts (mitigation, adaptation, monitoring).
  2. Climate data types: in-situ, reanalysis (ERA5), satellite (MODIS, Sentinel, Landsat), socio-economic data, emissions inventories.
  3. Tools & environment: Python, Jupyter, Git, cloud basics.

Data engineering for climate analytics

  1. Data ingestion, cleaning, gridding/resampling, gap-filling for time series and spatial data.
  2.  APIs and platforms: Copernicus, NASA Earthdata, GEE, Google Cloud Public Datasets.
  3. Lab: Download and pre-process ERA5 or Sentinel-2 sample datasets; build reproducible ETL.

Time series forecasting & anomaly detection

  1.  Classical & ML methods: ARIMA, Prophet, LSTMs, Temporal CNNs, Transformers for time series.
  2. Anomaly/outlier detection for early warning (isolation forest, autoencoders).
  3. Lab: Short-term temperature/precipitation forecasting and anomaly detection for early flood/wildfire alerts.

Remote sensing & geospatial ML

  1. Image preprocessing, indices (NDVI, NBR), change detection, segmentation (U-Net), object detection.
  2. Pixel vs object-based analysis; classification accuracy metrics.
  3. Lab: Land cover mapping or burned-area detection with Sentinel/Landsat.

Climate attribution, causal inference and scenario analytics

  1. Basics of causal inference for climate impacts (counterfactuals, difference-in-differences).
  2.  Emissions scenario modeling, linking physical models with statistical models.
  3. Lab: Simple causal analysis of heatwave impacts on hospital admissions (synthetic or public data).

Emissions monitoring & carbon accounting with AI

  1. Satellite-based emissions detection (methane plumes, thermal anomalies), top‑down vs bottom-up accounting.
  2. Data fusion for inventory improvements.
  3. Lab: Detecting methane plumes or thermal hotspots; estimating emissions magnitudes.

Optimization & decision support for mitigation/adaptation

  1. Optimization models for energy systems, land-use, evacuation planning; reinforcement learning basics for resource allocation.
  2. Multi-criteria decision analysis and uncertainty quantification.
  3. Lab: Optimize PV + battery sizing under climate-informed irradiance scenarios.

Automation, reproducibility & MLOps

  1. Pipelines for data ingestion, training, CI/CD, model registry, monitoring and retraining.
  2. Tools: Airflow/Prefect, Docker/Kubernetes, Kubeflow, MLflow, GitOps.
  3. Lab: Build CI pipeline for scheduled model retraining and deployment of a small forecasting model.

Model interpretability, uncertainty & risk communication

  1. Explainable AI (SHAP, LIME), probabilistic forecasting and communicating uncertainty to stakeholders.
  2. Visualizations and dashboards for decision-makers.
  3. Lab: Create dashboard for flood risk forecasts with uncertainty bounds.

Ethics, governance, policy and deployment considerations

  1.  Bias, privacy, equity, false alarms, cost of errors, regulatory frameworks.
  2.  Data governance, open science and stakeholder engagement.
  3. Guest lecture / panel (policy makers, climate data NGO).

Case studies & domain applications

  1. Deep dives: wildfire detection & response automation, crop yield forecasting & advisories, heat-health early warnings, coastal flood planning.
  2. Participants project checkpoints and feedback.

Capstone presentations and course wrap-up

  1. Final project presentations: end-to-end solution (data pipeline → model → deployment/automation + evaluation + stakeholder communication).
  2.  Course synthesis, next steps and resources.

Practical labs & project ideas

  1. Flood forecasting pipeline using precipitation forecasts + hydrologic model + ML residual correction; automated alerts.
  2. Satellite-based deforestation monitoring with automated change detection and alert generation.
  3. Methane plume detection and emissions estimation from hyperspectral/thermal imagery and automation of reporting.
  4.  Crop yield forecasting combining remote sensing and weather forecasts, with recommender for adaptation measures.
  5. Urban heat island mapping, forecasting and automated scheduling for cooling centres.

Datasets and platforms to use

  1. ERA5 (ECMWF reanalysis), CMIP6 scenario outputs for climate projections
  2. Sentinel-1/2, Landsat, MODIS (Copernicus, USGS)
  3.  NASA Earthdata, GHSL (population), EDGAR (emissions), OpenWeather/NOAA observations
  4. Google Earth Engine, Google Cloud Public Datasets, AWS Open Data
  5.  Local/national agency datasets (where available)

Recommended tools & Libraries

  1.  Python stack: xarray, rasterio, rioxarray, geopandas, pyproj, netCDF4
  2.  ML: scikit-learn, PyTorch/TensorFlow, statsmodels, Prophet, tsai/tslearn
  3. Remote sensing & imagery: rasterio, GDAL, EO-Learn, segmentation-models
  4. MLOps & automation: Airflow/Prefect, Docker, Kubernetes, MLflow, Great Expectations
  5. Visualization/dashboards: Plotly/Dash, Streamlit, Kepler.gl

Core readings & resources

  1. IPCC Assessment Reports (summaries and relevant chapters)
  2. “Deep Learning for Time Series Forecasting” — Brownlee (practical resource)
  3. Relevant papers/case studies from journals (e.g., Nature Climate Change, Remote Sensing)
  4. Google Earth Engine tutorials, Copernicus/Earthdata documentation
  5.  Selected ethics & governance readings (IEEE, World Economic Forum, OECD guidance on AI)