Project Geospatial

In this session from FOSS4G NA 2025, Adeel Hassan, a senior machine learning engineer at 84.2, presents Bringing it all Together: Knowledge Graphs, LLMs, and Cloud-Native Geospatial Data. Adeel discusses the Noah Knowledge Mesh project, an ambitious initiative funded by NOAA to bridge the gap between complex satellite datasets and real-world user questions through natural language processing.The presentation details a sophisticated architecture that combines Knowledge Graphs for data discovery with Large Language Models (LLMs) for data analysis. By translating natural language queries into a structured, custom formal language, the system can autonomously navigate thousands of NetCDF files, perform complex spatial-temporal computations using Xarray and Dask, and deliver high-level insights directly to the user.Highlights:🛰️ The Missing Link: How AI and LLMs are finally connecting massive data silos with natural language user needs🔄 Knowledge Graphs and Ontologies: Leveraging RDF and schema-based metadata to make NOAA's vast data catalogs discoverable🧩 Virtual Zarr and Kerchunk: Transforming decades of legacy NetCDF files into high-performance, cloud-native virtual stores⚙️ Structured Reasoning: Why 84.2 uses a custom formal language and Pydantic models instead of raw Python code generation for AI safety🌐 Data Analysis Agents: A look at the multi-agent system that delegates tasks between data search, analysis, and user orchestration🏢 Performance Optimization: Achieving a 40x speedup by storing Zarr metadata in Parquet format to handle massive time-series data🏗️ Real-World Demos: Watch the system calculate sea surface temperature anomalies and compare temperatures across geographical regions in real-timeFor more content like this check out www.projectgeospatial.com#Geospatial #FOSS4G #LLM #KnowledgeGraph #NOAA #AI #CloudNative #Xarray #Dask #NetCDF #Zarr #DataScience #MachineLearning #ProjectGeospatial #DataDiscovery

In this session from FOSS4G NA 2025, Pete Gadomski of Element 84 presents Cloud Native Geospatial Metadata with STAC GeoParquet. Pete explores the practical application of the SpatioTemporal Asset Catalog (STAC) specification and how the emerging GeoParquet format is addressing the scalability limitations of traditional static and API-based metadata catalogs.

The presentation compares traditional "static" (JSON blobs on blob storage) and "dynamic" (API/database-backed) access patterns, highlighting where each falls short when handling millions of items. Pete demonstrates how STAC GeoParquet provides a middle ground—enabling bulk metadata access, efficient full-scans, and direct-to-file searching without the overhead of a database. He also showcases "stac-map," a browser-based visualizer that uses DuckDB to filter and display GeoParquet files directly in the browser, and discusses how organized metadata snapshots can be used to feed and train Large Language Models (LLMs).

Highlights:

🛰️ The STAC Journey: A brief history of the SpatioTemporal Asset Catalog and its recent adoption as an OGC Community Standard

🔄 Static vs. API Patterns: A critical look at the trade-offs between cheap, simple blob storage and powerful but complex database-backed APIs

🧩 Enter STAC GeoParquet: How converting JSON metadata to a columnar format results in massive storage savings—from 21MB down to 200KB for 10,000 items

⚙️ Performance at Scale: Real-world results from a 10-million-item glacier dataset, showing how sorted GeoParquet files can reduce search times from 187 seconds to just 3 seconds

🌐 Universal Compatibility: Why using Parquet allows non-geospatial tools like DuckDB to analyze STAC metadata without specialized GIS libraries

🏢 The "One STAC" Vision: Moving toward a world where every database provides a GeoParquet snapshot for easier crawling and cross-catalog discovery

🏗️ STAC-Map Demo: A look at a new public tool for snapshotting STAC searches and uploading bundles for interactive browser-based filtering

For more content like this check out www.projectgeospatial.com

#Geospatial #FOSS4G #STAC #GeoParquet #Metadata #CloudNative #BigData #GIS #OpenSource #DuckDB #DataEngineering #Element84 #ProjectGeospatial #DataDiscovery #Xarray

In this session from FOSS4G NA 2025, Bruce Momjian, a co-founder and Core Team member of the PostgreSQL Global Development Group, shares his decades of experience on Building Open Source Teams. Bruce explores the unique challenges of leading a global community of volunteers and how "servant leadership" is the key to sustainable open-source success.

Moving beyond typical corporate management, Bruce explains why traditional organizational structures often fail in open-source environments. He dives into the deep-seated motivations that drive developers to contribute—ranging from intellectual stimulation to a belief in helping others—and provides a roadmap for how leaders can foster a culture of ownership and pride in code.

Highlights:

🛰️ Transitioning to Leadership: Bruce’s journey from software engineering to leading one of the world's most successful open-source projects

🔄 Corporate vs. Open Source: Why managing volunteers requires a complete departure from enterprise-level leadership tactics

🧩 Understanding Motivation: A look at the six core forces that drive open-source participation, including career advancement and mental stimulation

⚙️ The Power of Physical Presence: Why getting on a plane to meet contributors in person is the ultimate tool for motivation

🌐 The Servant Leader Model: Why leadership in open source must be about serving the team rather than exercising authority

🏢 The 5% Rule: Understanding the small percentage of "true" open-source contributors and how to nurture them within a company

🏗️ Building Global Culture: Overcoming challenges in communication, language barriers, and 24-hour development cycles

For more content like this check out www.projectgeospatial.com

#PostgreSQL #OpenSource #FOSS4G #Leadership #SoftwareDevelopment #TechCommunity #ServantLeadership #OpenSourceManagement #ProjectGeospatial #TechLeadership #BruceMomjian #GIS #CommunityBuilding #GlobalTeams #SoftwareEngineering

In this session from FOSS4G NA 2025, Sina Kashuk, co-founder of Fused, presents DuckDB + Rasters: High-Performance Geospatial Analytics with H3 Hexagons. Sina explores how to overcome the traditional challenges of joining raster and vector data by converting raster datasets into a uniform H3 hierarchical grid system, enabling high-speed analytics directly in the browser via DuckDB-Wasm.

The presentation features a live demonstration of "Fused Canvas," a serverless Python runtime that processes massive datasets—including CDL (Cropland Data Layer), DEM (Digital Elevation Model), and ERA5 climate data—on the fly. Sina showcases how "hexagonifying" data allows for complex spatial joins, real-time histogram updates, and petabyte-scale time-series analysis while bypassing the limitations of traditional cloud-optimized geotiffs. He also conducts a browser-based stress test, pulling 67 million records and nearly a gigabyte of data in seconds to demonstrate the power of parallelized, cloud-native parquet queries.

Highlights:

🛰️ Why H3? The benefits of a hierarchical hexagonal grid for uniform adjacency, efficient partitioning, and analytics near the poles

🔄 Raster-to-Hex Conversion: Transforming standard XYZ raster data into DuckDB-compatible rows where geometry is represented as a hex ID

🧩 Browser as Database: Using DuckDB-Wasm to perform real-time SQL queries on millions of spatial records directly within the user interface

⚙️ Fused Canvas: A look at the new serverless Python platform that allows for arbitrary code execution and dynamic tiling in the cloud

🌐 ERA5 at Scale: Navigating 75 years of climate data to visualize global warming patterns and temperature shifts using spatial-temporal partitioning

🏢 Overcoming WASM Limitations: How Fused manages range requests and virtual files to prevent unnecessary full-file downloads in the browser

🏗️ Stress Test Demo: Watch the system saturate network bandwidth by fetching 20 million records in 10 seconds using 30 parallel browser threads

For more content like this check out www.projectgeospatial.com

#Geospatial #FOSS4G #DuckDB #H3 #RasterData #BigData #CloudNative #DataScience #OpenSource #Python #ERA5 #Fused #ProjectGeospatial #Wasm #SpatialAnalytics

In this session from FOSS4G NA 2025, Randal Hale of North River Geographic Systems presents Dragging Some FOSS4G into your Closed Source Life. Randal shares a decade of experience transitioning public and private organizations from proprietary GIS stacks to open-source solutions like QGIS and PostGIS.

Through a series of candid—and often humorous—case studies, Randal explores the "messy middle" of open-source adoption. He details the reality of replacing expensive commercial software in small-town 911 centers and state agencies, offering practical advice on managing IT hurdles, securing leadership buy-in, and avoiding the "easy button" trap. The talk emphasizes that moving to FOSS4G is a journey rather than a weekend switch, requiring careful planning, community engagement, and a healthy dose of patience.

Highlights:

🛰️ The Caribbean Leap: How a consulting gig in the islands forced a sudden, total immersion into the world of open-source mapping

🔄 Sneaker-Nets and Lenovo Servers: A look at the real-world hardware and data management challenges facing small county GIS departments

🧩 The "Infection" Strategy: Why starting with low-impact datasets like bus routes is the best way to introduce PostGIS to an organization

⚙️ Beyond the Price Tag: Debunking the myth that open source means "spending no money" and the importance of budgeting for support

🌐 Community-Driven Solutions: Navigating the unique support structure of the FOSS4G community—from email lists to bug reporting

🏢 Hybrid Realities: Why a "burn the licenses" approach isn't always practical and how a hybrid stack can provide the best of both worlds

🏗️ Failure and Recovery: Lessons learned from 10,000 lines of dead ArcPy code and the ongoing effort to modernize state-level standards with SQL

For more content like this check out www.projectgeospatial.com

#Geospatial #FOSS4G #OpenSource #QGIS #PostGIS #GIS #DigitalTransformation #PublicSector #SmallBiz #TechTransition #OpenData #ProjectGeospatial #SoftwareDevelopment #Governance #CommunityBuilding

In this session from FOSS4G NA 2025, Calvin Metcalf of the Sanborn Map Company takes a deep dive into an unlikely hero of the modern web: **Cloud Optimized Shapefiles**. While shapefiles are often seen as a relic of the past, Calvin demonstrates how these multi-file structures can be modernized to work natively on the web through clever indexing and byte-range requests.

Calvin breaks down the anatomy of shapefiles—from the .shp and .dbf to the often-overlooked .shx, .prj, and .cpg files—and explains the engineering required to make them "cloud optimized." By implementing a QIX (spatial index) file and leveraging Hilbert curves for spatial ordering, he shows how to structure geospatial data so users only download the specific parts they need, significantly reducing data overhead and server requests.

Highlights:

🛰️ What is a Shapefile? A breakdown of the three to five files that make up the "yesterday's format of tomorrow"

🔄 Anatomy of Cloud Optimization: How byte-range requests allow for downloading specific parts of a file without the whole package

🧩 The Power of the QIX Index: Using tree-based nodes to skip irrelevant data and jump directly to spatial bounding boxes

⚙️ Big Endian vs. Little Endian: Navigating the "split the baby" compromise in shapefile data structures

🌐 The Hilbert Curve Advantage: Why space-filling curves are the superior method for keeping spatially close shapes adjacent in the file

🏢 Practical Implementation: A look at the open-source library and command-line tool for building and reordering cloud-optimized shapefiles

🏗️ Browser Hurdles: Addressing the specific Firefox bug that impacts byte-range requests on compressed files

For more content like this check out www.projectgeospatial.com

#Geospatial #FOSS4G #Shapefile #CloudOptimized #GIS #WebMapping #OpenSource #DataEngineering #HilbertCurve #SanbornMapCompany #SoftwareDevelopment #ProjectGeospatial #VectorData #WebDev #SpatialAnalysis

In this session from FOSS4G NA 2025, Annie Cartas from the University of Maryland’s CATT Lab presents Evolving the Trip Analytics Platform. Annie shares the technical journey of modernizing a transportation analysis web application—transitioning from a legacy Leaflet-based system plagued by "fog of war" loading screens to a high-performance stack powered by MapLibre, PostGIS, and vector tiles.

The presentation provides a candid look at how the CATT Lab optimized their geospatial architecture by moving heavy processing from the front-end (Turf.js) to indexed spatial databases, implementing robust three-pronged caching strategies, and leveraging modern open-source tools like deck.gl and PGTILESERVE. Annie also discusses the practical use of AI helpers for PostGIS syntax and MapLibre styling, and the performance benefits of adopting STAC and cloud-native visualization techniques for historical vehicle trip data.

Highlights:

🛰️ Modernizing Transportation Data: How trip analytics helps DOTs explore historical vehicle data to make infrastructure decisions like bridge closures

🔄 Moving Beyond GeoJSON: The transition from sending massive coordinate arrays via websockets to efficient ID-based vector tile serving

🧩 PostGIS Power: Overcoming "SQL fear" to utilize spatial indexes and server-side processing for complex study area intersections

⚙️ PGTILESERVE and Martin: A look at faster developer iteration with automated tile serving and the trade-offs between custom and out-of-the-box servers

🌐 Visualizing Big Data: Using deck.gl and globe views for GPU-based rendering of millions of traffic arcs and H3 global grids

🏢 Caching Strategies: A deep dive into Tanstack Query, NGINX, and Redis for minimizing costly database re-executions

🏗️ AI in GIS: How ChatGPT assisted in writing complex octet-string calculations and navigating the MapLibre style specification

For more content like this check out www.projectgeospatial.com

#TransportationAnalytics #FOSS4G #PostGIS #MapLibre #Geospatial #GIS #VectorTiles #OpenSource #DataVisualization #ProjectGeospatial #TechMigration #BigData #SmartCities #SQL #UrbanPlanning

In this session from FOSS4G NA 2025, Dan Baston of I-Sciences presents GDAL's New CLI: A Ground-Up Redesign for Consistency and Power. Dan explores the significant overhaul of the GDAL (Geospatial Data Abstraction Library) command-line interface, moving away from the "swiss-army knife" approach of massive, inconsistent tools like gdalwarp and ogr2ogr toward a suite of single-purpose, highly predictable commands.

The presentation introduces the concept of "pipelines," a powerful new way to chain geospatial operations—such as clipping, reprojecting, and simplifying—without the need for intermediate files on disk. Dan also details the new Python API mapping, enhanced tab-completion for complex options, and over 10 brand-new native C++ utilities for raster calculation, zonal stats, and geometry cleaning that were previously only available as Python scripts or separate libraries.

Highlights:

🛰️ The "Burn the Village" Moment: Why the GDAL team decided to build a completely new CLI to fix decades of inconsistent command arguments

🔄 Single-Purpose Tools: Breaking down complex utilities into specialized commands like gdal_raster_clip, gdal_raster_mosaic, and gdal_raster_repro

🧩 Introducing Pipelines: How to use the exclamation mark (!) character to chain multiple vector and raster steps into a single readable command

⚙️ Automated Python Mapping: A look at the new gdal.alg module that provides direct, native access to all CLI tools without using subprocesses

🌐 Enhanced Discoverability: Using tab-completion to find available formats, EPSG codes, and even files stored in cloud-based virtual file systems (S3/GCS)

🏢 Serialized Workflows: Using the new .gdal_alg.json format to record and share complex processing chains as virtual datasets

🏗️ New Native Capabilities: A showcase of new C++ tools for raster reclassification, focal stats, moving window averages, and polygon coverage repair

For more content like this check out www.projectgeospatial.com

#GDAL #FOSS4G #Geospatial #CLI #GIS #Python #Cplusplus #RasterData #VectorData #OpenSource #DataEngineering #ProjectGeospatial #SoftwareDevelopment #Pipelines #ZonalStats

In this session from FOSS4G NA 2025, Sophia Parafina, a cloud advocate for the EarthScope Consortium, presents Geophysics in the Cloud. Sophia introduces the mission of EarthScope, a nonprofit organization funded by the National Science Foundation (NSF) that provides the global scientific community with access to high-quality geophysical data—including seismic, GNSS, SAR, and magnetotelluric observations.

The presentation details EarthScope's massive "lift and shift" migration from legacy on-premise infrastructure to a cloud-native environment in AWS, managing over a petabyte of open data. Sophia showcases "GeoLab," a specialized JupyterHub environment that allows researchers to process large-scale geophysical datasets directly in the cloud using tools like Dask, PyTorch, and Apache Arrow. She also demonstrates real-world applications of geophysics, such as using GNSS reflectometry to measure snowpack and soil moisture with millimeter precision.

Highlights:

🛰️ What is EarthScope? An overview of the consortium’s role in providing open access to seismic and geodetic instrumentation and data

🔄 Migration to AWS: Breaking down decades of Fortran-based processing into modern AWS Lambdas and S3-based workflows

🧩 GeoLab & JupyterHub: A deep dive into the cloud-based research environment offering specialized containers for seismic and GNSS analysis

⚙️ Modern Data Formats: How EarthScope is moving toward Apache Arrow and TileDB to enable high-speed, zero-copy data engineering

🌐 Global Seismic Picking: A look at how researchers used EarthScope’s petabyte-scale data to build a global earthquake catalog using machine learning

🏢 Hazard Monitoring: Using seismometers and GNSS stations to track everything from hurricane-induced ocean vibrations to ionospheric disturbances caused by tsunamis

🏗️ GNSS Reflectometry: Explaining how satellite signal noise ratios are used to measure surface deformation, snowpack, and vegetation moisture

For more content like this check out www.projectgeospatial.com

#Geophysics #FOSS4G #CloudComputing #AWS #EarthScope #Seismology #GNSS #JupyterHub #DataScience #OpenSource #OpenData #ProjectGeospatial #Dask #MachineLearning #NaturalHazards

In this session from FOSS4G NA 2025, Julia Signell and Jarrett Keifer of Element 84 present Is Zarr the New COG? The presenters tackle the growing debate around whether the Zarr data format is poised to replace Cloud Optimized GeoTIFFs (COGs) by providing a technical comparison of how each format handles multi-dimensional arrays, chunking, and compression.

Julia and Jarrett break down the "apples to oranges" comparison, explaining that while COG is a specialized specification built on top of the TIFF format, Zarr is a generic format for storing hierarchies of n-dimensional arrays. They explore the critical roles of chunking and compression codecs in data performance and demonstrate how, with identical settings, the bytes on disk for a Zarr chunk and a COG tile can be virtually the same. The talk also looks ahead to emerging Zarr conventions, such as GeoZarr and multi-scales, that aim to bring COG-like features to the Zarr ecosystem.

Highlights:

🛰️ Defining the Formats: A look at the structural differences between Cloud Optimized GeoTIFFs and the hierarchical array format of Zarr

🔄 Linearization and Chunking: How data is projected from multi-dimensional space to a single-dimension string of bytes on disk

🧩 Compression Codecs: A deep dive into the five types of codecs—entropy, predictive, structural, quantization, and mapping—used to optimize data storage

⚙️ The Faster Myth: Why performance differences are often due to default compression and chunking settings rather than the format itself

🌐 Interoperability vs. Multi-dimensionality: When to stick with the portability of TIFF and when to leverage the n-dimensional power of Zarr

🏢 Zarr Conventions: Understanding how new, community-driven specifications are breaking the deadlock for Zarr's geospatial adoption

🏗️ Virtual Zarr Stores: Using references to access legacy COG data through a modern Zarr interface without duplicating storage

For more content like this check out www.projectgeospatial.com

#Geospatial #FOSS4G #Zarr #COG #GeoTIFF #CloudNative #DataEngineering #Compression #BigData #GIS #OpenSource #Element84 #Xarray #Kerchunk #ProjectGeospatial #DataAnalysis