Awesome-Earth-Artificial-Intelligence

A curated list of tutorials, notebooks, software, datasets, courses, books, video lectures and papers specifically for Artificial Intelligence (AI) use cases in Earth Science — with emphasis on open-source tools, freely accessible papers, and reproducible benchmarks (including geospatial and weather/climate foundation models).

Maintained by ESIP Machine Learning Cluster. Free and open to inspire AI for Good.

Contributions are most welcome. Please refer to our contributing guidelines, what is awesome?, and Code of Conduct.

Contents

ML-enthusiastic Earth Scientific Questions

Courses

• :sunglasses::sparkling_heart: GeoSMART Machine Learning Curriculum

• :sunglasses::sparkling_heart: Introduction to Machine Learning for Earth Observation (EO College MOOC) - Free MOOC from TUM/DLR covering classification, object detection, change detection, SAR, and self-supervised learning for EO

• :sunglasses::sparkling_heart: GeoAI with Python: A Practical Guide to Open-Source Geospatial AI Zenodo - Open-access book with 23 chapters of executable code for segmentation, detection, change detection, and foundation models

• ICESat-2 Hackweek

• ML Seminar: Physics-informed Machine learning for weather and climate science (57:35) by Dr. Karthik Kashinath from Lawrence Berkeley National Lab, Mar 19, 2021

• ML Seminar: Scalable Geospatial Analysis (53:23) by Tom Augspurger from Microsoft AI for Earth, May 20, 2021

• Fundamentals of ML and DL in Python - A series of Jupyter notebooks that walk you through the fundamentals of Machine Learning and Deep Learning in python using Scikit-Learn and TensorFlow.

• Trustworthy Artificial Intelligence for Environmental Science (TAI4ES) Summer School will be virtually the week of July 26-30, 2021.

• Artificial Intelligence for Earth System Science (AI4ESS) Summer School repo readinglist

• American Meterological Survey AI Webinar Series

• USGS Artificial Intelligence/Machine Learning Workshop

• Stanford CS 229 ML Cheatsheets

• RadiantEarth ML4EO Bootcamp 2021

• Summer School on High-Performance and Disruptive Computing in Remote Sensing - Scaling Machine Learning for Remote Sensing using Cloud Computing

Books

• :sunglasses: :sparkling_heart: Artificial Intelligence in Earth Science

• :sunglasses: :sparkling_heart: Artificial Intelligence Methods in the Environmental Sciences

• Deep Learning for the Earth Sciences

• How to achieve AI maturity and why it matters? (PDF)

• 70-Years-of-Machine-Learning-in-Geoscience-in-Review

Tools

Earth observation, geospatial, weather, and climate software only. Entries are sorted alphabetically by name. :sunglasses: marks maintainer picks for this list (not a quality tier). For pretrained model weights, see Foundation Models. For general ML infrastructure, see RelatedAwesome.

• AI Segmentation by TerraLab - QGIS plugin for point-and-click segmentation of buildings, trees and any object in satellite and drone imagery into vector polygons, with a free CPU-only local mode

• ai-models - Open-source CLI to run AI weather models (GraphCast, FourCastNet, Pangu-Weather) with ECMWF data pipelines

• ClimateLearn paper - PyTorch library for weather forecasting and climate downscaling benchmarks (ERA5, CMIP6)

• EarthML website - Tools for working with machine learning in earth science

• eo-learn - Earth observation processing framework for machine learning in Python

• GEO-Bench-2 leaderboard paper - Reproducible benchmark for geospatial foundation models across 19 permissively licensed datasets

• :sunglasses: GeoAI docs - Unified Python framework for EO deep learning: segmentation, detection, change detection, and foundation model workflows

• GRIME AI website - Ecohydrological workflow suite for ground-based time-lapse imagery, from acquisition through ML applications

• GRIME2 website - Camera-based water level measurement from ground-based time-lapse imagery

• Makani - Scalable training framework for ML weather models (FourCastNet 3); Apache 2.0

• Microsoft AI for Earth API Platform - Distributed API hosting for long-running geospatial ML model inference on Azure/Kubernetes

• :sunglasses: pygeoweaver - Python library for AI & geospatial workflow management, FAIRness, tangibility and productivity improvement

• segment-geospatial (samgeo) docs - Segment Anything Model (SAM) and HQ-SAM for geospatial imagery segmentation

• SeisBench docs - Open toolbox for earthquake ML: phase picking, event detection, pretrained models, and benchmark datasets

• :sunglasses: TerraTorch paper - Fine-tuning and benchmarking toolkit for geospatial foundation models; integrates with GEO-Bench-2 and Hugging Face weights

• torch-harmonics - Differentiable signal processing on the sphere for geometric weather ML; BSD-3-Clause

• :sunglasses: TorchGeo docs - PyTorch domain library with 100+ geospatial datasets, spatial samplers, multispectral transforms, and pretrained backbones

• WeatherBench 2 docs - Open evaluation framework and leaderboard for data-driven global weather models

• Xarray-Beam - Python library for building Apache Beam pipelines with Xarray datasets

Foundation Models

Pretrained model weights and primary model repositories for Earth observation, weather, and climate. Sorted alphabetically within each group. For fine-tuning toolkits and benchmarks, see Tools. For task-specific application code and vision-language models, see Code.

Earth Observation

• AlphaEarth Foundations embeddings - Global 10 m embedding field layers (2017–2024) for sparse-label mapping; annual embeddings on Google Earth Engine and GCS

• :sunglasses::sparkling_heart: Clay docs weights - Sensor-agnostic MAE foundation model (v1.5) for EO embeddings across Sentinel-2, Landsat, Sentinel-1, and custom sensors; Apache 2.0

• Copernicus-FM paper - Unified Copernicus foundation model across Sentinel missions with Copernicus-Pretrain and Copernicus-Bench

• DOFA paper - Dynamic One-For-All multimodal foundation model with wavelength-conditioned hypernetworks for cross-sensor generalization

• :sunglasses::sparkling_heart: Prithvi-EO-2.0 weights paper - Multi-temporal ViT foundation model (300M/600M) trained on 4.2M global HLS time series at 30 m; fine-tune via TerraTorch

• TerraMind weights paper - Any-to-any generative multimodal EO foundation model (IBM/ESA Φ-lab); fine-tune via TerraTorch

Weather and Climate

• Aurora docs paper - 1.3B-parameter atmospheric foundation model for weather, air pollution, and ocean waves

• FourCastNet 3 - Probabilistic spherical-convolution weather ensemble forecasting at 0.25°; training via Makani

• GraphCast / GenCast GraphCast paper GenCast paper - GNN-based medium-range global weather forecasting and diffusion ensemble forecasting; Apache 2.0

• NeuralGCM dycore paper - Differentiable hybrid general circulation model combining physics-based dynamics with learned components; Apache 2.0 code, CC BY-SA 4.0 weights

• :sunglasses::sparkling_heart: Prithvi-WxC weights paper - 2.3B-parameter weather/climate foundation model on MERRA-2 for forecasting, downscaling, and parameterization

Tutorials

• :sunglasses::sparkling_heart: GeoSMART Machine Learning Curriculum & Use Cases

• :sunglasses::sparkling_heart: GeoAI with Python Book Code - Executable notebooks for seven core GeoAI tasks and foundation model workflows

• :sunglasses::sparkling_heart: GeoAI Video Tutorials docs - Step-by-step GeoAI package tutorials from Open Geospatial Solutions

• :sunglasses::sparkling_heart: TerraTorch Documentation - Fine-tuning guides for Prithvi, TerraMind, Clay, and GEO-Bench-2 benchmarking

• :sunglasses::sparkling_heart: NeuralGCM Inference Quickstart - Run pretrained hybrid GCM weather forecasts with open checkpoints on GCS

• :sunglasses::sparkling_heart: NASA Openscapes Earthdata Cloud Cookbook

• :sunglasses::sparkling_heart: Artificial Intelligence in Earth science Book Materials

• :sunglasses::sparkling_heart: RadiantEarth MLhub Tutorials

• Machine Learning Tutorials (general, not Earth science specific)

• Pixel-level land cover classification

• EO-learn-workshop - EO-learn-workshop: Bridging Earth Observation data and Machine Learning in Python,

• Machine Learning for Development Machine Learning for Development: A method to Learn and Identify Earth Features from Satellite Images,

• ELSI-DL-Bootcamp - Intro to Machine Learning and Deep Learning for Earth-Life Sciences,

• UW WaterhackerWeek - Introduction to Machine Learning on Landslide Data and Scikit-learn from UW WaterhackerWeek,

• Planet Snow Mapping - Introduction to using Planet imagery to map snow cover

• Machine Learning Pipeline for Climate Science - an end-to-end pipeline for the creation, intercomparison and evaluation of machine learning methods in climate science

• AI Cheatsheets - Essential Cheat Sheets for deep learning and machine learning engineers. It contains a lot of useful tutorials to learn awesome tricks on AI engineering

Training Data

• Kaggle Earth Science Training Dataset

• Radiant MLHub - an open library for geospatial training data

• Google Earth Engine Data Catalog

• AlphaEarth Satellite Embeddings paper - Global annual 10 m embedding fields (2017–2024) from AlphaEarth Foundations; also on GCS

• GEO-Bench-2 Datasets - 19 permissively licensed benchmark datasets for geospatial foundation model evaluation on Hugging Face

• Copernicus-Embed-025deg - Global 0.25° embedding map integrating multi-source Sentinel observations (released with Copernicus-FM)

• WeatherBench 2 ERA5 Zarr - Open cloud-optimized ERA5 and baseline forecast data for ML weather model training and evaluation

• University of California Irvine Machine Learning Repository

• EuroSAT Dataset - EuroSAT Dataset: Land Use and Land Cover Classification with Sentinel-2,

• Awesome Satellite Imagery Datasets - Awesome Satellite Imagery Datasets: A curated list of deep learning training datasets,

• STanford EArthquake Dataset (STEAD) - A Global Data Set of Seismic Signals for AI

• ZipCheckup - Free ZIP-level environmental safety dataset for 42,000+ US ZIP codes covering water quality, air quality, PFAS contamination, radon, lead, flood risk, and 11 additional verticals. Public REST API and npm/PyPI packages for ML pipelines. CC BY 4.0.

Code

Task-specific implementations and Earth-facing applications. Foundation model weights live under Foundation Models; fine-tuning toolkits under Tools.

• :sunglasses::sparkling_heart: Earth System Emulator (ESEm) - A tool for emulating geophysical datasets including (but not limited to) Earth System Models

• :sunglasses::sparkling_heart: EmissionAI - Microsoft AI for Earth Project: AI Monitoring Coal-fired Power Plant Emission from Space

• EarthDial paper - Multi-spectral, multi-temporal vision-language model for EO dialogue across 44 downstream datasets

• GeoChat paper - Grounded large vision-language model for remote sensing QA, captioning, and referring detection

• Global Forest Watch - ML-powered deforestation and forest cover change monitoring from satellite imagery

• iNaturalist Computer Vision - Species identification from community-contributed observations (76,000+ taxa)

• TEOChat paper - Temporal vision-language assistant for change detection, damage assessment, and EO dialogue

• Wildlife Insights - Automated species identification from camera trap images; integrates with GBIF

• BassNet,paper-preprint - Deep Learning for Land-cover Classification in Hyperspectral Images,

• MTLCC - Multitemporal Land Cover Classification Network (ConvLSTM, ConvGRU),

• Landsat Time Series Analysis for Multi-Temporal Land Cover Classification

• EarthEngine-Deep-Learning - Deep Learning on Google Earth Engine,

• Continuous Change Detection and Classification - Continuous Change Detection and Classification (CCDC) of land cover using all available Landsat data,

• Object-based Classification on Earth Engine - Object-based land cover classification with Feature Extraction and Feature Selection for Google Earth Engine (GEE),

• Earth Lens - Earth Lens, a Microsoft Garage project is an iOS iPad application that helps people and organizations quickly identify and classify objects in aerial imagery through the power of machine learning.

• Image Classification Neural Network Ranking with source code - paperswithcode has put together a list of cutting-edge papers and ranked them with the claimed accuracy.

• EQTransformer - An AI-Based Earthquake Signal Detector and Phase Picker.

• Tropical Cyclone Windspeed Estimator - Winning solutions for Tropical Cyclone Wind Speed Prediction Competition

• Infernis - Open-source ML-powered wildfire risk prediction engine for British Columbia. XGBoost + CNN trained on 10 fire seasons (2015-2024) from 21 open government and scientific data sources. Provides a free REST API with daily predictions at 5km resolution.

Videos

• GeoAI Tutorials Playlist - Open Geospatial Solutions tutorials on segmentation, detection, and QGIS GeoAI plugin workflows

• Tutorial on Microsoft Azure Machine Learning Studio (AutoML-Regression), created by Microsoft AI for Earth Project: AI Monitoring Coal-fired Power Plant Emission from Space.

Papers

• :sunglasses: :sparkling_heart: A Review of Earth Artificial Intelligence

• :sunglasses: :sparkling_heart: Foundation Models for Remote Sensing and Earth Observation: A Survey - Taxonomy of visual, vision-language, and LLM-based RSFMs with benchmarking across public datasets

• Towards practical artificial intelligence in Earth sciences

• A Review of Practical AI for Remote Sensing in Earth Sciences

• Prithvi-EO-2.0: A Versatile Multi-Temporal Foundation Model for Earth Observation Applications

• Prithvi WxC: Foundation Model for Weather and Climate

• TerraMind: Large-Scale Generative Multimodality for Earth Observation

• TerraTorch: The Geospatial Foundation Models Toolkit

• AlphaEarth Foundations: An embedding field model for accurate and efficient global mapping from sparse label data

• GEO-Bench-2: From Performance to Capability, Rethinking Evaluation in Geospatial AI

• Neural Plasticity-Inspired Foundation Model for Observing the Earth Crossing Modalities (DOFA)

• FourCastNet 3: A geometric approach to probabilistic machine-learning weather forecasting at scale

• GraphCast: Learning skillful medium-range global weather forecasting

• GenCast: Diffusion-based ensemble forecasting for medium-range weather

• Aurora: A Foundation Model of the Atmosphere

• Neural General Circulation Models for Weather and Climate

• TEOChat: A Large Vision-Language Assistant for Temporal Earth Observation Data

• GeoChat: Grounded Large Vision-Language Model for Remote Sensing

• EarthDial: Turning Multi-sensory Earth Observations to Interactive Dialogues

• Towards a Unified Copernicus Foundation Model for Earth Vision arxiv

• Advances on Multimodal Remote Sensing Foundation Models for Earth Observation Downstream Tasks: A Survey - Open-access review of vision-X multimodal RSFMs

• Big Earth data analytics: A survey

• Adoption of machine learning techniques in ecology and earth science

• CIRA Guide To Custom Loss Functions For Neural Networks In Environmental Sciences - Version 1

• Zero-Shot Learning of Aerosol Optical Properties with Graph NeuralNetworks

• NeuralHydrology - a collection of papers on using neural networks in hydrology

• Ten Ways to Apply Machine Learning in Earth and Space Sciences

• Advancing AI for Earth Science: A Data Systems Perspective

• Google Earth Engine: Planetary-scale geospatial analysis for everyone

• WeatherBench 2: A benchmark for the next generation of data-driven global weather models

• ClimateLearn: Benchmarking Machine Learning for Weather and Climate Modeling

• PCA-OS: A Planetary Climate Adaptation Operating System (KDD 2026 Blue Sky Ideas Track) - Frames climate adaptation as a continual learning and decision loop over an intervention-aware global causal knowledge graph, fusing Earth-observation signals and operational traces into versioned, auditable adaptation interventions and robust decision portfolios.

Reports

• Workshop Report: Advancing Application of Machine Learning Tools for NASA’s Earth Observation Data

Thoughts

• :sunglasses: :sparkling_heart: Learning earth system models from observations: machine learning or data assimilation?

• Artificial intelligence: A powerful paradigm for scientific research

• Why 90% of machine learning models never hit the market

• 'Farewell Convolutions' – ML Community Applauds Anonymous ICLR 2021 Paper That Uses Transformers for Image Recognition at Scale

• 37 reasons why your neural network is not working

Competitions

• :sunglasses::sparkling_heart: GeoAI Challenge - aimed at providing solutions for collaboratively addressing real-world geospatial problems by applying artificial intelligence (AI)/machine learning (ML)

• 2025 GeoAI Challenge: Cropland Mapping in Dry Environments - ITU/FAO challenge on distinguishing cropland from pasture in Fergana and Orenburg using time-series satellite imagery

• 2026 GeoAI Challenge: Reaching new heights with GeoFM - ITU/ESA Φ-lab challenge on global surface height and land-cover mapping with open satellite imagery and GeoFM embeddings

• GPU Hackthons - designed to help scientists, researchers and developers to accelerate and optimize their applications on GPUs.

• LANL Earthquake Prediction

• HackerEarth

Communities

• ESIP Machine Learning Cluster

• ESIP Agriculture and Climate Cluster

• AI Alliance Climate & Sustainability Group - Community behind GEO-Bench-2 and open geospatial foundation model evaluation

• TorchGeo Community - OSGeo community project for geospatial deep learning in PyTorch

RelatedAwesome

• Awesome-Open-Geoscience – A list is curated from repositories that make our lives as geoscientists, hackers and data wranglers easier or just more awesome. In accordance with the awesome manifesto, we add awesome repositories.
• Awesome-Spatial – Awesome list for geospatial, not specific to geoscience but significant overlap
• Awesome Open Climate Science – Awesome list for atmospheric, ocean, climate, and hydrologic science
• awesome-weather-models – Catalogue of AI-based weather forecasting models with open-source and open-weights status
• awesome-WeatherAI – Papers, datasets, and open model implementations for AI weather and climate
• Awesome_AI4Earth – Deep learning for Earth system science, especially data-driven weather prediction
• Awesome-AI-for-Atmosphere-and-Ocean – Research papers on AI for atmospheric science and oceanography
• Awesome Coastal – Awesome list for coastal engineers and scientists
• Awesome Satellite Imagery Datasets - List of aerial and satellite imagery datasets with annotations for computer vision and deep learning
• Awesome Workflow Engines - A curated list of awesome open source workflow engines
• Awesome Pipeline - A curated list of awesome pipeline toolkits inspired by Awesome Sysadmin
• Awesome Machine Learning - A curated list of awesome Machine Learning frameworks, libraries and software

General ML infrastructure (companion tools)

These are useful in Earth AI workflows but are not Earth-specific; we list them here rather than in Tools.

• BentoML – Open-source framework for high-performance ML model serving
• Dopamine – Research framework for reinforcement learning prototyping
• flashlight – C++ standalone library for machine learning
• MindsDB – Explainable AutoML framework on PyTorch
• Netron – Neural network and ONNX/Keras/TFLite model visualizer
• ml.js – Machine learning tools in JavaScript
• MLflow – Machine learning lifecycle platform
• OneFlow – Performance-centered open-source deep learning framework
• Polyaxon – ML platform for Kubernetes training and monitoring
• Snips NLU – Natural language understanding for structured extraction from text
• SynapseML – Scalable ML pipelines on Apache Spark
• TensorFlow Hub – Repository of reusable TensorFlow SavedModels
• TransmogrifAI – AutoML library on Apache Spark (Scala)