Datacubes: Enabling Space/Time Analytics and AI
summary
Argues that storing gridded Earth observations as multidimensional data cubes (space, depth, time, spectral) makes analytics and AI simpler and faster than juggling files or MapReduce. Describes an array-database implementation (rasdaman) running dozens of petabyte deployments, a WCPS server-side analytics language for pixel-wise expressions, WCS/OAPI coverages and ISO/OGC coverage models, interactive demos at standards.rasdaman.com (MODIS, Sentinel, Cesium/WebGL), and a federated Earth Server that proves the approach at scale.
AI-written summary · about
chapters (14)
- 0:00Introduction and welcome
- 2:57What are data cubes
- 6:50User-centric vs sensor-centric views
- 10:44Approaches to data cubes
- 13:35Rasterman array database
- 15:00Why standards matter
- 17:51OGC coverage standards
- 23:54ISO and INSPIRE standards
- 26:29Web Coverage Service explained
- 28:14WCPS language for analytics
- 30:42Live demo and examples
- 34:14Q&A on validation and tools
- 36:42Q&A on array databases vs alternatives
- 38:47Closing and summary
related talks
31mOpenDataCube - Petabyte Scale Satellite Imagery Processing in Python
20mCloud Native Patterns for Geospatial Analytics and AI
22mTake your Business to a New Geography
48mAccelerate science, research, and innovation with AWS
26mWho Said Wrangling Geospatial Data at Scale was Easy?
10mPaco Nathan (Liber 118) Interview - OSCON 2014
36mArcticDB, the OLAP antidote
40mData-centric exploration using intake, dask, hvplot, and more