Geospatial Artificial Intelligence For Animals

GAIA is a groundbreaking initiative developing an advanced cloud-application for detecting whales in very high resolution satellite imagery.

Geospatial Artificial Intelligence for Animals is a pioneering initiative that plans to use cutting-edge satellite technology and artificial intelligence to detect whales and other marine mammals from space. Developed by NOAA’s Northeast and Alaska Fisheries Science Centers in partnership with public and private collaborators—including the U.S. Geological Survey, Microsoft AI for Good, and the Naval Research Laboratory—GAIA is revolutionizing how we monitor marine species across vast and remote ocean regions

GAIA is integrating very high resolution satellite imagery with advanced cloud computing, geospatial analysis, and machine learning to build a scalable, automated detection system. The platform will help scientists locate endangered species like North Atlantic right whales and Cook Inlet belugas, contributing to conservation and management efforts under the Marine Mammal Protection Act and the Endangered Species Act.

At the core of GAIA is a secure, cloud-based application that enables expert annotation and validation of satellite imagery. The system streamlines data collection, processing, and dissemination. It supports collaboration across NOAA’s regional science centers and helps address key conservation priorities. With future expansions planned—including broader satellite sensor support and multi-species annotation—GAIA is laying the groundwork for a new era of satellite-enabled wildlife monitoring.

GAIA’s journey began with a bold question: Can we consistently detect whales from space to better understand where they are, much like we already do with planes and boats? This idea sparked Biologist’s Guide to the Galaxy—a landmark paper written by marine biologists turned satellite imagery experts. The guide outlines the potential and challenges with using very high resolution (VHR) satellite imagery to detect marine animals, and calls for a shared, global effort to build an automated whale detection system.

GAIA is working toward a reliable, real-world system that delivers whale detections with a large spatial-temporal footprint—turning satellite data into an everyday tool for marine conservation, monitoring, and protection. The paper’s insights into sensor capabilities, data standards, and large-scale coordination have helped shape GAIA’s mission: to transform satellite whale detection into a survey tool as routine and dependable as traditional platforms.

GAIA is uniquely positioned to advance this work because of access to U.S. government contracts through the National Geospatial-Intelligence Agency and the National Reconnaissance Office. The contracts are distributed by the U.S. Geological Survey with support from the Civil Applications Committee. These partnerships have enabled targeted image collection over seasonal whale aggregations, focusing on endangered species like the North Atlantic right whale and Cook Inlet beluga whale. Access to large volumes of satellite imagery is essential for building robust AI training datasets.

GAIA partnered with the British Antarctic Survey, who led the development of a standardized workflow for manually annotating VHR optical satellite imagery to find whales and create an expert-validated training dataset. This workflow, implemented in ESRI ArcMap and ArcGIS Pro, is detailed in the MethodsX publication Annotating very high-resolution satellite imagery: A whale case study. The paper outlines a step-by-step protocol for reviewing, annotating, bounding, and clipping satellite imagery into AI-ready image chips, towards the creation of a robust “ground truth” dataset to support machine learning.

To accelerate the process of manual annotation, GAIA partnered with Microsoft’s AI for Good Lab, which created a prototype tool, WHALE — a semi-automated, cloud-based application designed to streamline the annotation of satellite imagery. Built on Microsoft Azure and available as open-source software, this tool allowed researchers to annotate imagery using a web-based interface that presents a curated catalog of “interesting points” likely to contain whales. Users could quickly view image metadata and label interesting points, assign species and confidence levels. This system boosted annotation efficiency and represented a powerful step toward scalable, AI-assisted wildlife monitoring.

Building on the Microsoft AI for Good prototype, GAIA has deployed and customized the application to meet NOAA security requirements. GAIA expanded the system’s capabilities by integrating a backend database, enabling efficient storage and spatial querying of annotation data. We added additional functionality to better support data ingestion, annotation workflows, and validation of results. This secure version GAIA version 0.1, provides a powerful system for generating high-quality training data.

To ensure imagery is optimal for species identification, GAIA created a specialized preprocessing workflow. This includes geometric corrections, noise reduction, and contrast enhancements to improve visual clarity. These best practices support accurate annotation and species detection, helping build a reliable inventory of whale observations. This work will be published soon.

To support large-scale data collection, GAIA has developed a flexible codebase that automates the retrieval of satellite imagery through the USGS EarthExplorer API. This system streamlines access to extensive archives of very high-resolution (VHR) imagery, enabling the team to efficiently search, filter, and download scenes based on key parameters including location, date, and sensor type. By automating this process within a unified workflow, GAIA ensures rapid and consistent access to imagery needed for annotation and analysis.

As part of expanding efforts under the Strategic Initiative on Remote Sensing, GAIA is also supporting image collection in areas and for species that align with the regional priorities of NOAA’s science centers:

These targeted efforts allow GAIA to better support regional monitoring needs.

GAIA version 0.1 is our recently deployed cloud application designed to support the collaborative annotation and validation of VHR satellite imagery for whale detection. The platform features an intuitive annotation interface that allows 3 independent subject matter experts to review each image. Once all 3 annotations are complete, a final validation step ensures consensus and data quality. The system displays cloud optimized GeoTIFFs (COGs) of preprocessed Maxar level 1B imagery, which users can adjust for brightness and contrast directly in the interface to enhance image clarity.

The first case study for the GAIA application will use imagery collected in Cape Cod Bay, Massachusetts, from 2020-2024, that spatially and temporally overlap with Center for Coastal Studies aerial surveys that detected whales. All annotation and validation data are stored in a SpatiaLite database, supporting efficient and structured data management.

The GAIA team is actively exploring a wide range of new features and enhancements that will build upon our current capabilities and respond to the real-world needs of scientists. We’re committed to making the platform more powerful, intuitive, and collaborative. Below is a preview of some exciting improvements on the horizon in our beta version GAIA 0.2 - designed to help researchers work more efficiently, engage in smoother teamwork, and uncover even deeper insights from satellite imagery.

GAIA version 0.1 is currently designed to support annotation of North Atlantic right whale imagery from Cape Cod Bay, Massachusetts. In version 0.2, we’re expanding the platform to support multiple annotation projects simultaneously, enabling us to process imagery for several research efforts at once. This upgrade will help us better address animal detection needs across all NOAA science centers, in alignment with priorities set by the Strategic Initiative on Remote Sensing.

We’re streamlining the process for scientists to add satellite imagery to their projects by building a user-friendly interface for accessing the image repositories via Application Programming Interface. This means scientists will be able to easily load new imagery into their projects from the Earth Explorer repository —no coding required. Once added, each image will be automatically ingested with its relevant metadata, triggering our full preprocessing pipeline to produce high-quality visualizations that are ready for annotators to review.

Currently, GAIA version 0.1 only supports imagery from the Maxar WorldView-3 satellite. We're actively working to expand compatibility to include additional high-resolution sensors such as WorldView-2 and GeoEye. This enhancement will significantly increase the volume of available imagery for scientists to analyze and will extend our ability to detect whales across a wider range of locations and time periods.

To support collaborators, we’re adding a new feature that will allow users to export validated whale detections as CSV files. This straightforward yet powerful capability will make it easier for researchers to share results, perform additional analyses, and incorporate the annotated data into their own workflows—such as training and evaluating machine learning models. By enabling seamless data export, we’re amplifying the impact of GAIA’s outputs and supporting the community.

Last updated by Northeast Fisheries Science Center on 04/10/2026