Support and encourage the use of geospatial data & technology in research and teaching at Stanford

(whatever that means at any given time)

Earth Observation Today

From Observation

to

Continuous Monitoring

From Observation to Continuous Monitoring

The number of EO satellites is expected to increase from over 900 today to more than 2,300 by 2032

Daily data collection could reach around 230 Petabytes (PB) 

Drone Imaging

• Sub-centimeter spatial resolution

• Ultra-high cadence (on-demand, event-driven)

• Ideal for:

  • Cultural heritage monitoring

  • Disaster damage assessment

  • Illegal logging & mining detection

  • Agricultural & ecological surveys

• Produces massive volumes of imagery

• Drives need for real-time automated analysis

The Other Massive Data Stream

Earth Observation Data Infrastructures

"Stanford used more Planet imagery than NASA last year..."

Joe Mascaro - Director of E&R Planet.com

HOW DO YOU MAKE SENSE OF THESE QUANTITIES OF SENSOR DATA?

The Data Deluge

• Millions of satellite scenes daily

• Billions of drone pixels per mission

• Hyperspectral & multispectral streams expanding

• Impossible for human-only analysis

• Massive Parallelization required simply for processing

• AI required to extract insight at operational scale

Until 2008, only 4% of the Landsat archive had even been examined

"Often it turns out to be more efficient to move the questions than to move the data."

-Jim Gray (1944-2007)

The Fourth Paradigm: Data-Intensive Scientific Discovery

https://code.earthengine.google.com/8de743027454b85f04223a811d85f863

Foundation Models/Embeddings for EO

• Numerical vectors representing imagery "meaning"

• Enable cross-sensor comparison

• Used for:

  • Similarity Search

  • Classification

  • Change detection

CLAY Foundation Model

• Foundation EO model

• Accepts imagery characteristics via config

• Works with:

  • Optical

  • Multispectral

  • Hyperspectral

  • SAR

• Sensor-agnostic embedding generation

CALCULATE YOUR OWN EMBEDDINGS

Historic Maps Present a Scaling Problem, too

MapReader (Turing Institute)

Authoritarian regimes often try to control the narrative by manipulating geography: altering maps, renaming places, or deleting settlements.
 

The Human Capital Behind Map Resilience

• Archivists

• Cartographers

• Historians

• Local knowledge keepers

• Stewards of cultural memory

• Maps preserve truth of memory.

• EO preserves reality.

• AI provides understanding at scale.