Agenda
Tracks
- Track 1: On-ramp to Cloud-Native Geospatial Data: Capacity building for professionals who are new to working with cloud-native formats and workflows.
- Track 2: Cloud-Native Geospatial in Practice: Best practices, cost savings, and innovation enabled by the adoption of cloud-native technologies.
- Track 3: Building Resilient Data Infrastructure: Creating the infrastructure and systems we need to make trustworthy data reliably available to solve global challenges.
Day 1
Breakfast
Welcome and Keynote
Welcome: Jed Sundwall | Keynote: Chris Holmes
Track Introductions: Why this is important?
Julia Wagemann (Track 1), Aimee Barciauskas (Track 2), Brianna Pagán (Track 3)
Morning Break
SLOT 1
Track 1: STAC (Wasatch A)
Speaker | Talk |
---|---|
Matthew Hanson [Element 84] |
Introduction to STAC Discover the power of STAC (SpatioTemporal Asset Catalog) in this presentation, which simplifies its core concepts and provides a clear overview of the current state of the STAC ecosystem. Learn about best practices, the latest advancements, available public STAC catalogs, and how to navigate the myriad extensions available |
Benjamin Clark [Overture Maps Foundation & Meta] |
STAC-ing GeoParquet Overture maps wants the CNG community to be able to find our data easily. We also want to enable developers to write awesome tools to use our data! We think STAC can help us accomplish both of these things. I’ll outline how Overture went from hand-rolling our own release manifests to discovering the STAC format, and how we’re working towards publishing STAC representations of our monthly data releases. |
David Raleigh [Umbra] |
Effortless STAC Queries: CQLAlchemy and the Power of Autocompletion The intricate syntax of OGC’s Common Query Language (CQL2-JSON) introduces a steep learning curve, hindering widespread adoption within the SpatioTemporal Asset Catalog (STAC) community. Its verbose JSON structure makes programmatic query generation challenging, impeding the STAC goal of seamless geospatial data discovery. To address this, I developed CQLAlchemy, a Python Object Document Mapper library that bridges the gap between CQL2-JSON’s power and STAC’s vision of user-friendly access. |
Track 2: Cloud-Native Innovations 1: Cloud-Native Formats (Magpie A)
Speaker | Talk |
---|---|
Lindsey Neild [Earthmover] |
Zarr for Cloud-Native Geospatial. When and Why? At Earthmover, we’re seeing a growing number of teams building planetary-scale earth-observation data cubes with Zarr. In the CNG community, Zarr has traditionally been recommended for weather and climate data, while Cloud-Optimized GeoTIFF (CoG) is recommended for geospatial rasters. So why are people choosing Zarr over the highly successful CoG/STAC architecture? This talk explains this trend and clarifies where Zarr is the right choice. The key takeaway is that Zarr brings significant advantages for level-3 data (harmonized spatiotemporal grid), while CoG/STAC remains the optimal choice for level-2 data (many individual scenes with different footprints). |
Sai Cheemalapati [Google] |
Cloud-Native Geospatial in Earth Engine: COGs and Beyond To enhance cloud interoperability, Google Earth Engine now supports Cloud-Optimized GeoTIFFs (COGs) hosted on Google Cloud Storage. Integrating COGs and achieving performance comparable to Earth Engine’s traditional internally-stored assets presented a substantial technical challenge. This presentation will describe our COG support in detail, examine the technical challenges of integrating COGs with Earth Engine’s infrastructure, and provide an analysis of the suitability of COGs for various Earth Engine applications. We will also briefly discuss our exploration of Zarr support, highlighting the format’s perceived strengths and weaknesses for Earth Engine applications. |
Matthias Mohr [Independent Software Engineer and Consultant] |
The Future of ARD: Composable Building Blocks CEOS-ARD is likely the largest collection of specifications for geospatial analysis-ready data (ARD) that is adopted by a wide range of data providers. Currently, each specification - usually one for each product type - consists of a hand-crafted specification document with similar “boilerplate” and the actual requirements. With the addition of more product types the maintenance effort of these documents is growing significantly and the consistency across documents is decreasing. |
Track 2: GeoAI: Tools (Magpie B)
Speaker | Talk |
---|---|
Chris Stoner [Amazon Web Services] |
Using AWS Open Data with Amazon Bedrock Looking for datasets to fine tune your generativeAI applications? Come to this talk where we will cover how to use datasets in the Registry of Open Data on AWS with the Amazon Bedrock service. |
Kwin Keuter & Brad Andrick [Earth Genome] |
Scaling EarthIndex: AI Meets Cloud-Native Geospatial Dive into the technical and financial challenges and solutions behind building a scalable geospatial AI tool. From processing global imagery mosaics to publishing, cataloging and searching billions of embeddings. We’ll cover how defaulting to cloud-native technologies and geospatial standards allows us to build global scale solutions on a non-profit budget. |
Simon Ilyushchenko [Google] |
Google Earth Agents: Building and Benchmarking AI Assistants Google Earth Engine (EE) hosts nearly 100PB of public remote sensing data. Together with its powerful processing language, this presents an ideal platform for deploying Generative AI agents. We have developed several agents to assist users in the often arduous journey of using EE. These agents help users find the most appropriate datasets, learn how to use them, and even write and debug EE code. Additionally, agents can autonomously solve EE tasks—but how well? One of our ongoing projects involves benchmarking agent performance on a carefully selected variety of real-world tasks. |
Track 3: Disaster Response Discussion (Wasatch B)
Speaker | Talk |
---|---|
Cristiano Giovando [Humanitarian OpenStreetMap Team] |
OpenAerialMap: Streamlining Access to Open Disaster Imagery OpenAerialMap (OAM) addresses the need for openly licensed, high-resolution imagery in disaster response, where current access is often fragmented and slow. OAM is moving to a STAC-based architecture to become the main access point for all post-disaster imagery made available in cloud-native format by providers. |
Duk-jin Kim [Seoul National University] |
Satellite-Based Disaster Monitoring System Utilizing Cloud Computing With the increasing frequency and severity of natural disasters, the need for efficient and automated disaster monitoring systems has become critical. This study presents a satellite-based disaster monitoring system utilizing cloud computing, developed to enhance the rapid detection and analysis of various disaster events such as flooding, earthquakes, oil spills, and illegal maritime activities. |
Brianna Pagán [Development Seed] |
Discussion After Cristiano Giovando and Duk-jin Kim give their talks, we will continue a conversation around disaster response and its relation to geospatial. |
Lunch
SLOT 2
Track 1: Cloud-Native Data Formats (Wasatch A)
Speaker | Talk |
---|---|
Howard Butler [Hobu, Inc] |
Data Gravity Shapes the Architecture of Cloud-Native Geospatial Data doesn’t weigh but has gravity just the same. Gravity can be an apt metaphor to describe architectural choices driving the development cloud-native geospatial tools, formats, and specifications. This was especially true for the development of Cloud Optimized Point Cloud, whose niche is filled with very large, information-sparse datasets that require complex selectivity, backward compatibility, and high efficiency. |
Emma Marshall [University of Utah] |
Cloud-native geospatial datacube workflows with Xarray and Zarr The increasing availability and co-location of compute, storage, and cloud-optimized data is transforming the pace and nature of scientific discovery, and has the potential to democratize computationally-intensive scientific inquiry. However, the nature of these changes can place a significant burden on users looking to adopt new skills across a range of domains. |
Tom Nicholas [Earthmover] |
VirtualiZarr and Icechunk: How to build a cloud-optimised datacube of archival files in 3 lines Many level-3 datasets are distributed as collections of thousands of individual files or granules, making it difficult to address the data as a coherent datacube. Worse, the data is often stuck in pre-cloud archival file formats, precluding efficient access from object storage. |
Julia Wagemann [thriveGEO] |
Sentinel's EOPF Toolkit: Driving Community Adoption of the Zarr data format for Copernicus Sentinel Data An overview of cloud-native data formats for geospatial data, including COG, Zarr, and other emerging formats. Learn how these formats enable efficient data access and processing in the cloud. |
Track 2: Cloud-Native Innovations 2: New Ideas (Magpie A)
Speaker | Talk |
---|---|
Shane Loeffler & Raphael Hagen [CarbonPlan] |
Exploring questions around visualization and data distribution of vector CNG formats In a new project centered around a CONUS-wide, high-resolution, vector dataset, we’re exploring using CNG vector formats for our data visualization and distribution strategies. Ideally, our analysis outputs, visualization layer, and distribution strategies could all use the same underlying cloud-native data. GeoParquet has emerged as an exciting potential format for this, but several open questions remain. |
Pete Gadomski [Development Seed] |
Right-sizing STAC: An exploration of STAC API backends, including stac-geoparquet Since its release in May 2021, the SpatioTemporal Asset Catalog (STAC) specification has been widely adopted by the public and private sectors as a core technology for indexing and searching geospatial assets. Part of its success has stemmed from its wide ecosystem of open source tooling in a variety of languages. |
Julia Signell [Element 84] |
STAC <> Zarr STAC (SpatioTemporal Asset Catalogs) is a specification for defining and searching any type of data that has spatial and temporal dimensions. STAC has seen significant adoption in the earth observation community. Zarr is a specification for storing groups of cloud-optimized arrays. Zarr has been adopted by the earth modeling community (led by Pangeo). |
Florent Gravin [Camptocamp] |
Leverage cloud-native vector data in your webmapping application, simply In the rapidly evolving landscape of geospatial data, cloud-native vector data has emerged as a critical component, supported by cloud-native databases and warehouses such as BigQuery, Redshift, and Snowflake, as well as formats like GeoParquet. While SaaS providers like CARTO offer robust tools for large-scale geospatial analysis, there is a growing interest in building custom web mapping applications that leverage these technologies without the associated fees. |
Track 2: Applications: Cloud-Native Geospatial in Practice (Magpie B)
Speaker | Talk |
---|---|
Frederic Leclercq [Flanders Marine Institute] |
Cloud-Optimized Formats for Scalable and Efficient Data Management in the EDITO Datalake The EDITO (European Digital Twin of the Ocean) datalake is designed to support the management and delivery of vast marine and environmental datasets, enabling efficient access and analysis for a wide range of applications and stakeholders. This presentation focuses on the strategies and technologies implemented to ensure interoperability, scalability, and performance in a cloud-native environment. |
Jeffrey Gillan [Data Science Institute - University of Arizona] |
Making the World's Drone Data Open & Cloud-Native The view from above provided by drone imagery has opened up new insights and opportunities across agriculture, geology, construction, mining, forest and wildfire management, ecological research, and disaster response. But as groups across the globe take to the skies, they each encounter similar big data management challenges around processing, analysis, storing, and sharing. |
Charlie Savage [Privateer/Orbital Insight] |
Building a Geospatial Platform with CNG Technologies I’d like to present how Orbital uses CNG technologies in our Terrascope platform. Real world examples include dark ship detection, global monitoring of GPS spoofing/jamming events and observations of ports/airports. These use cases require using both imagery (EO, SAR, Lidar, etc.) and geolocation data (AIS, ADSB, RF, etc.). |
Naomi Provost [CTrees] |
Moving from Science to Product: Making Cloud-Native Geospatial Work for Us At CTrees, we create machine learning models that integrate multiple data sources to produce high-resolution, time-series datasets on forest carbon and activity. Our outputs—raster-based estimates of carbon stocks, emissions, removals, and forest change—support projects ranging from jurisdictional analysis to deforestation monitoring. |
Track 3: The Where of it All Discussion (Wasatch B)
Speaker | Talk |
---|---|
Katie Baynes [NASA], Lena Trudeau [Inclined Analytics], & Jen Marcus [Taylor Geospatial Engine] [Discussion Leaders] |
The Where of it All Where will geospatial projects live in the future? Given shifts in research funding, what organizations will house open-source projects and other resources needed by the geospatial community? How do we reliably create and maintain data needed for planetary-scale challenges? What are sustainable funding models? |
Workshop: Interfacing with Cloud-Native Overture Data and the GERS Ecosystem (Ballroom 2)
Interfacing with Cloud-Native Overture Data and the GERS Ecosystem
Workshop Leaders:
- Dana Bauer & Jennings Anderson [Overture Maps Foundation]
Transition
SLOT 3
Track 1: Cloud-Native Geospatial and ArcGIS (Wasatch A)
Speaker | Talk |
---|---|
David Wright [Esri] |
Cloud-Native Geospatial and ArcGIS Cloud-Native Geospatial data and ArcGIS are a long-standing duo for efficient desktop, web and mobile GIS collaboration through 2D, 3D and multidimensional data experiences. It starts with providing a comprehensive set of tools for imagery and remote sensing data management, analysis and visualization, making it easy to work with cloud-optimized formats such as COG, MRF, CRF, Zarr, and I3S. More recently a STAC user experience was introduced in ArcGIS Pro, along with STAC methods in ArcGIS Python APIs, further opening access to public and private data at the asset level. We’ll showcase these capabilities, and how Esri continues to empower professionals by enhancing their ability to securely manage, publish, share and analyze geospatial data effectively using cloud-native deployment patterns that support small to massive scale operations. |
Track 2: Cloud-Native Innovations 3: Performance and Scale (Magpie A)
Speaker | Talk |
---|---|
Zachariah Dicus [KBR, contractor to USGS] |
Zarr: Landsat Trade Study at Scale The USGS EROS is exploring the transition to the Zarr format. This trade study examines the effects of data processing methods, product archiving, and metadata management at scale as the USGS plans for the future storage of over 200 Petabytes of data in the cloud over the next 15 years. The study will discuss the Zarr structure, compression, processing benefits, visualization, and geospatial readiness. |
Jeffrey Albrecht [Regrow] |
Are COGs Actually Scaleable? During this talk we’ll set out to answer a simple question;are Cloud-Optimized GeoTiffs actually scaleable? We’ll examine the math and theory behind COG access patterns, look at benchmarks of current software, and discuss potential improvements! You can expect to learn about the COG data format, blob stores, caching, and distributed system design! |
Kyle Barron [Development Seed] |
Super-fast cloud storage operations with Obstore Cloud-native geospatial is predicated on efficient access to cloud storage. Instead of downloading files wholesale, clients need to be able to quickly retrieve portions of files via range requests. But interfacing with multiple clouds is a pain. Using the official Python libraries for each cloud — such as boto3 for AWS and similar counterparts for Google Cloud Storage and Azure Storage — requires creating your own abstraction layer on top. The most common Python library to abstract access to object storage, fsspec , doesn’t naturally fit cloud storage providers because it attempts to abstract files, not the stateless HTTP requests that cloud operations are much more similar to. fsspec can also be hard to use. For example, it’s missing static typing support; both a necessity for today’s production-caliber Python applications and a big productivity boost for typing-enabled IDEs. This talk will introduce Obstore, a new Python library to interface with Amazon S3, Google Cloud Storage, and Azure Storage. Powered by a Rust core, it’s the simplest, highest-throughput Python interface to cloud object storage. It provides significantly higher throughput than alternatives like fsspec or aioboto3, while being simpler to install and use. |
Track 2: GeoAI: Strategies (Magpie B)
Speaker | Talk |
---|---|
Hugo Lee [Jet Propulsion Laboratory, NASA] |
Multiscale Geospatial and AI-Driven Approaches for Climate Model Evaluation As Earth System Models (ESMs) continue to produce massive datasets at increasingly fine resolutions, climate scientists face challenges in efficiently evaluating and interpreting these large-scale, multi-resolution outputs against observational data. This presentation introduces an advanced analytical framework, combining Hierarchical Data Analysis (HDA), Topological Data Analysis (TDA), and Retrieval-Augmented Generation (RAG) integrated with Large Language Models (LLMs), to streamline the evaluation of climate model performance across spatial scales. |
Erin Trochim [University of Alaska Fairbanks] |
Questioning Candor: AI-Driven Frameworks for Geospatial Data As we continue developing geospatial access for the 21st century, maps alone will not provide sufficient context. We need to re-examine and address the usability and trustworthiness of data. Geospatial data is inherently complex and diverse, where robust infrastructure and intelligent systems have long been the solution to unlock its full potential. |
Aashish Panta [University of Utah] |
Scalable Web-Based Exploration and RAG-enhanced Insights for NASA's Downscaled Climate Data NASA generates petabytes of geospatial and climate data daily to uncover environmental patterns and predict future scenarios. Understanding long-term climate trends is essential for identifying risks, assessing mitigation strategies, and making informed decisions at planetary scales. |
Afternoon Break
Plenary Talk
Lynne Schneider
Plenary Lightning Talks
Ballroom 2 + 3: Plenary Lightning Talks
Speaker | Talk |
---|---|
Tyler Erickson [VorGeo] |
Mapping the CNG Ecosystem |
Martha Morrissey [Pachama] |
Helping Restore Nature with Cloud-Native Geospatial |
Jessie Pechmann [Humanitarian OpenStreetMap Team] |
Cloud-Native Geo for Good: Empowering Humanitarian Action with Accessible Data |
Arvind Mohan [Center for Nonlinear Studies] |
Bridging "Nerdville" and "Fieldville": A Vision for Resilient Data Infrastructure for Real-World AI in Crisis Management |
Natalie Evans Harris [State of Maryland] |
Maryland State Chief Data Officer Lightning Talk |
Benjamin P. Stewart [The World Bank] |
The World Bank Lightning Talk |
Michal Migurski [CARTO] |
CARTO Lightning Talk |
Tim Wallace [The New York Times] |
The New York Times Lightning Talk |
Day 2
Breakfast
Keynote
Julia Stewart Lowndes
Transition
SLOT 1
Track 1: Making Geospatial Workflows More Accessible (Wasatch A)
Speaker | Talk |
---|---|
Matt Fisher [Schmidt DSE @ UC Berkeley] |
Exploring more approachable geospatial data workflows as a community In this talk, I will present the ongoing work of the GeoJupyter community, an open and collaborative effort to reimagine geospatial interactive computing experiences for education, research, and industry. We aim to combine the approachability and playfulness of desktop GIS tools, the flexibility and reproducibility of coding-driven GIS methods, and the collaborative and storytelling power of Jupyter to enable more researchers, educators, and learners to confidently engage with geospatial data. |
Eric Charles [Datalayer] |
AI Agent for Jupyter to simplify Earthdata Discovery and Analysis In this demo, we will showcase AI tools such as NASA Earthdata search and download, as well as Python code generation in Jupyter Notebooks. These tools are designed to help geospatial analysts, especially beginners, quickly locate relevant datasets and streamline their analysis. |
Nathan Swain [Aquaveo | Tethys Geoscience Foundation] |
Tethys Platform: Bridging Earth Science and Cloud-Native Geospatial Technology Tethys Platform is an open-source framework that enables scientists to build and deploy GIS web applications for Earth science research, ensuring that findings reach the end users who can benefit from them. It integrates proven open-source components—OpenLayers, CesiumJS, Bokeh, Plotly, GeoServer, THREDDS, and PostGIS—through an SDK that simplifies common web GIS development workflows such as data visualization, exploration, and analysis. |
Greg van Gaans [Department for Housing and Urban Development (DHUD), South Australia Government] |
Delivery of Enterprise Geospatial Services using a Cloud-Native Geospatial Approach DHUD is responsible for delivering the land and built environment geospatial data (some 250 datasets) for South Australia. Some key datasets include cadastre, planning zones, address and suburbs. This presentation explores the delivery of this foundational geospatial data through a serverless geospatial data lake architecture built on AWS. |
Track 2: Systems: Innovations in System Design (Magpie A)
Speaker | Talk |
---|---|
Lukas Bindreiter [Tilebox] |
Beyond Points: Efficient Spatio-Temporal Polygon Indexing for Scalable Catalogs The first obstacle when developing geospatial data pipelines is finding relevant input data and organizing output data in the form of a catalog. Current storage engines for spatio-temporal metadata are often limited to indices either on the space dimension or on the time dimension, but not on both. |
Alexander Merose [Open Athena AI Foundation] |
Why machine learning people should think about databases: lessons from AI weather models From when the word “computers” referred to buildings of people to the design of the Von Neumann architecture (i.e. the CPU), the history of computing is marked by inventions in service of predicting the weather. Machine learning has recently added to this legacy with a revolution in weather forecasting. |
Tom Augspurger [NVIDIA] |
GPU-accelerated Cloud-Native Geospatial Cloud-native geospatial promises the ability to scale your workloads to massive datasets with ease. As the datasets become larger and your workload becomes more computationally intensive, accelerated computing can save us from ever increasing runtimes. |
Matthew Powers [Wherobots] |
Introducing geospatial support in Apache Iceberg This talk explores Apache Iceberg’s new native geospatial support (Iceberg Geo), tackling challenges in large-scale spatial data management. It covers Iceberg Geo’s development, design, and goals, highlighting its impact on both geospatial and Iceberg communities. |
Track 3: Beyond Open Data Discussion (Wasatch B)
Speaker | Talk |
---|---|
Marc Prioleau [Overture Maps Foundation], Vikram Gundeti [Foursquare], & Tom Lee [Mapbox] [Discussion Leaders] |
Beyond Open Data How do we create data products that can be relied upon by the research and commercial sectors? Discussion points include licensing, provenance, data ownership, ongoing maintenance, and the ability to determine the value of data products (in terms of economic benefit and societal good). |
Track 1 Workshop: On-ramp to CNG: Part 1 (Ballroom 2)
Cloud-Native Geospatial for Earth Observation
Workshop Leaders:
- Alex Leith [CNG Workshop]
Earthmover Workshop: Earthmover Workshop (Ballroom 3)
Zarr, Icechunk, & Xarray for cloud-native geospatial data-cube analysis
This 3-hour long workshop is designed as an on-ramp to using the Zarr data format for cloud-native geospatial datacube analysis. Guided by experience running similar workshops for the earth, ocean, atmosphere, & climate sciences, we focus on the conceptual underpinnings of array data analytics using Zarr & Xarray.
Our learning goals are:
- Understand the Zarr chunked n-dimensional array format.
- Understand how the Zarr data model complements the Raster data model underlying the GDAL/GeoTIFF/STAC ecosystem.
- Learn how to navigate the two data models by building cloud-native datacubes from an input GeoTIFF dataset.
- Understand how to use Xarray and the surrounding ecosystem for geospatial analytics
Each hour will contain 30 minutes of instruction, 20 minutes of semi-structured hands-on coding time with support from instructors, and 10 minutes break. Below we list areas of focus for each hour:
Hour 1: Introduction to Zarr & Xarray. Compare and contrast with the raster data model.
Hour 2: Building your own Zarr data cube from an input GeoTIFF dataset. We will have examples of a small number of data ingestion workflows (approx. 3) that attendees can choose from to build their own data cube. This hour will be mostly hands-on coding with 20 minutes of discussion to address attendee questions and any misconceptions.
Hour 3: Analytics with Zarr data cubes using Xarray including raster-vector joins and vector data cubes using a zonal statistics example. End with one or two demos of cloud-native data cube workflows using Zarr. Options include:
- AI/ML training using a Zarr datacube
- Building a production-grade serverless continental-scale datacube pipeline with Zarr
These proposed demos are also the subject of a parallel talk submission. The focus here will be pedagogical, the focus there is demonstrative.
Workshop Leaders:
- Deepak Cherian, Joe Hamman, Emma Marshall, Tom Nicholas, Lindsey Nield [Earthmover]
Coworking: Birds of a Feather & More (Magpie B)
Birds of a Feather and Coworking Space
Morning Break
SLOT 2
Track 1: Scaling Geospatial Intelligence (Wasatch A)
Speaker | Talk |
---|---|
Noah Slocum [Esri] |
Spatial analysis at scale with ArcGIS GeoAnalytics Engine and Apache Spark A key aspect to capacity building for cloud-native geospatial data is considering professionals that are experienced with cloud-native formats but new to geospatial data. As a community, it is important to think about how we can make it easier for such professionals to feel confident in the accuracy, suitability, and stability of the geospatial methods supporting their workflows. |
Damian Wylie [Wherobots] |
Extract insights from satellite imagery at scale with WherobotsAI Inferring objects and detecting change in satellite imagery was once reserved for companies with the talent, money, and time to build, manage, and run sophisticated, self-managed machine learning (ML) inference solutions against satellite data. |
Brandon Liu [Protomaps] |
Minimum Viable CNG What is the essence of cloud-native geospatial? It might not be a SaaS, but instead simple primitives that are simple to publish anywhere. This is a hands-on workshop to making global raster and vector data publishable and accessible, through the lens of the Protomaps open source project. |
Track 2: Systems: Building a Cloud-Native Geospatial Ecosystem (Magpie A)
Speaker | Talk |
---|---|
Samapriya Roy [Desert Research Institute | Spatial Bytes] |
Cloud-Native Commons: Lessons from Building the Earth Engine Community Catalog Democratization requires cost absorption, whether by vendors or communities. This talk reframes the cloud-native conversation through the lens of the Google Earth Engine (GEE) Community Catalog, an initiative that has grown since 2020 to host 3,800+ datasets, over half a petabyte of data, and serve millions of users worldwide. |
Jose Macchi [Geotekne] |
Unleashing Massive Cloud Power: Scalable & Enriched Access to OvertureMaps with GeoServerCloud This presentation showcases a powerful architecture for dynamically consuming Overture Maps data at scale, without requiring local storage. By placing a robust mapping solution—GeoServer Cloud—at the forefront, we enable not only efficient and parallelized access to Overture Maps through distributed WFS services but also the ability to enrich data access with advanced business logic. |
Lukas Bindreiter [Tilebox] |
Tooling Designed for the Complexities of Space Data: A Hyperspectral Data Workflow The growth of space-derived geospatial data is exponential. And the rise of hyperspectral satellite imaging is not only expanding Earth Observation (EO) applications, it is drastically increasing the volume of data you have to downlink, transfer, process, and store. |
Sarah Johnson [Coiled] |
Scaling Cloud-Native Geospatial Workflows with Dask, Xarray, and Coiled Geospatial datasets are growing rapidly, often exceeding 100TB and reaching petabyte scale. While many of these datasets are publicly available, unlocking their full potential requires scalable infrastructure, efficient computation, and a diverse set of tools. |
Track 3: Workforce Development Discussion (Wasatch B)
Speaker | Talk |
---|---|
Community Discussion [Community Discussion] |
Workforce Development How do we grow our community on purpose, particularly helping poor people get paying jobs? Funding models. |
Track 1 Workshop: On-ramp to CNG: Part 2 (Ballroom 2)
Deep Dive into Cloud-Native Geospatial Raster Formats
Workshop Leaders:
- Jarret Keifer [CNG Workshop]
Earthmover Workshop: Earthmover Workshop (Ballroom 3)
Zarr, Icechunk, & Xarray for cloud-native geospatial data-cube analysis
This 3-hour long workshop is designed as an on-ramp to using the Zarr data format for cloud-native geospatial datacube analysis. Guided by experience running similar workshops for the earth, ocean, atmosphere, & climate sciences, we focus on the conceptual underpinnings of array data analytics using Zarr & Xarray.
Our learning goals are:
- Understand the Zarr chunked n-dimensional array format.
- Understand how the Zarr data model complements the Raster data model underlying the GDAL/GeoTIFF/STAC ecosystem.
- Learn how to navigate the two data models by building cloud-native datacubes from an input GeoTIFF dataset.
- Understand how to use Xarray and the surrounding ecosystem for geospatial analytics
Each hour will contain 30 minutes of instruction, 20 minutes of semi-structured hands-on coding time with support from instructors, and 10 minutes break. Below we list areas of focus for each hour:
Hour 1: Introduction to Zarr & Xarray. Compare and contrast with the raster data model.
Hour 2: Building your own Zarr data cube from an input GeoTIFF dataset. We will have examples of a small number of data ingestion workflows (approx. 3) that attendees can choose from to build their own data cube. This hour will be mostly hands-on coding with 20 minutes of discussion to address attendee questions and any misconceptions.
Hour 3: Analytics with Zarr data cubes using Xarray including raster-vector joins and vector data cubes using a zonal statistics example. End with one or two demos of cloud-native data cube workflows using Zarr. Options include:
- AI/ML training using a Zarr datacube
- Building a production-grade serverless continental-scale datacube pipeline with Zarr
These proposed demos are also the subject of a parallel talk submission. The focus here will be pedagogical, the focus there is demonstrative.
Workshop Leaders:
- Deepak Cherian, Joe Hamman, Emma Marshall, Tom Nicholas, Lindsey Nield [Earthmover]
Coworking: Birds of a Feather & More (Magpie B)
Birds of a Feather and Coworking Space
Lunch
SLOT 3
Track 1: Geospatial Workflows (Wasatch A)
Speaker | Talk |
---|---|
Tonian Robinson [USGS] |
Accessing and Processing Landsat Data in the Cloud Since December 2020, the United States Geological Survey (USGS) Landsat Collection 2 data archive has been available in an uncompressed Cloud Optimized GeoTIFFs (COG) format on Amazon Web Services (AWS) Simple Storage Services (S3). |
Guyu Ye [Amazon Web Services] |
Modeling Sustainable Urban Spaces on AWS Heat risk poses major threats to human and environmental health and safety. According to an Atlantic Council study, there are currently more than 8,500 deaths annually associated with daily average temperatures above 90 degrees Fahrenheit (32 degrees Celsius). |
Dean Hintz [Safe Software] |
Model-based data transformation workflows to support loading and automation pipelines for cloud-native with FME |
Track 2: Applications: Weather (Magpie A)
Speaker | Talk |
---|---|
Sarah Zwiep [Floodbase] |
Why We Don't Use Zarr (Yet) Floodbase delivers near-real time (NRT) flood monitoring and expertise for parametric flood insurance and disaster relief programs worldwide. Cloud-native geospatial (CNG) technologies like COG, STAC, and geoparquet have enabled this work to scale effectively, despite complexities of diverse data sources and the constraints of real-time monitoring and customer-facing products. |
Hans Mohrmann [Brightband] |
A Cloud-Native Dataset of Atmospheric Observations for ML Applications A growing research community is turning to work directly with observational datasets to advance data-driven weather modeling and evaluation. These observations span diverse sensing modalities, from in situ surface stations, radiosondes, and aircraft data, to remotely sensed data across the electromagnetic spectrum. |
Antoine Dolant [University of Illinois at Urbana-Champaign] |
Agentic LLM Framework for Adaptive Decision Discourse Effective decision-making in complex systems requires synthesizing diverse perspectives to address multifaceted challenges under uncertainty. This study introduces a real-world inspired agentic Large Language Models (LLMs) framework, to simulate and enhance decision discourse—the deliberative process through which actionable strategies are collaboratively developed. |
Alfonso Ladino [The University of Illinois at Urbana-Champaign] |
Efficient Weather Radar Data Management in Practice: FAIR Principles and Cloud-Native Solutions Radars play a critical role in meteorology due to their high spatio-temporal measurement resolution, enabling early detection and tracking of severe weather phenomena. These capabilities empower meteorologists to issue timely alerts and warnings, thereby safeguarding lives and reducing property damage. |
Track 1 Workshop: On-ramp to CNG: Part 3 (Ballroom 2)
Deep Dive into Cloud-Native Geospatial Raster Formats
Workshop Leaders:
- Jarret Keifer [CNG Workshop]
Coworking: Birds of a Feather & More (Magpie B)
Birds of a Feather and Coworking Space
Afternoon Break
Recap from Track Leaders
Julia Wagemann (Track 1), Aimee Barciauskas (Track 2), Brianna Pagán (Track 3)
Plenary Panel
Builders Panel: Mo Sarwat, Amy Rose, Sean Gorman
Day 3
Breakfast
Keynote & Discussion
Drew Breunig