Cropland Mapping @30 using Google Earth Engine Jun Xiong 29 st Oct 2015.
Tyler Erickson Google Google Earth Engine Aspen …...Google Earth Engine a cloud-based geospatial...
Transcript of Tyler Erickson Google Google Earth Engine Aspen …...Google Earth Engine a cloud-based geospatial...
Water Perspectives from Google Earth Engine
Tyler EricksonGoogle
Aspen-Nicholas Water ForumMay 29, 2015
source: http://www.lternet.edu/node/49495Photo: John Marr
Tyler EricksonDeveloper AdvocateEarth Engine ProjectGoogle
USGS Landsat Program● 30m resolution● ~10 spectral channels● ~500MB per scene● 45 seconds to acquire a scene● >40 years of observations
Satellite Remote Sensing Data Example
USGS Earth Resources Observation & Science (EROS) Center
European Commission Joint Research CentreGlobal Environmental Monitoring Unit
Global Surface Water Map
What is Earth Engine?
Google Earth Enginea cloud-basedgeospatial processing platform
Goals● Make substantive progress on global challenges
that involve large geospatial datasets.Approach● Build a geospatial analysis platform that allows
both highly-interactive algorithm development and global-scale analysis.
What is Earth Engine?
GeospatialDatasets
AlgorithmicPrimitives
add
focal_min
filter
reduce
join
distancemosaic
convolve
Results
Storage and Compute
Requests
a.k.a. Tyler's Provocative Statements
The Earth Engine View of the World
Too much data for a single machine.Big Data
Medium Data
Small Data An amount of data that humans can use to make a decision.
Fits on a single machine.
Google Works with Big Data
Value is in the Usage of Data
Disk CPU
Key lessons:● Bandwidth is (relatively) expensive,
so co-locate CPU and disk. Bring the algorithm to the data!
● Disk is cheap,so bring everything online.
● CPU is even cheaper,so don’t pre-process needlessly.
Bandwidth
Data Transfer is the Limiting Factor
> >
Exploration: only feasible if results are FAST
Federated Systems Look Goodin Presentations
(but that's about it)
USGS EROS
NASA/JPL PO.DAAC
NASA NEX
NASA/USGS LP DAAC
NASA LDAS
ORNL DAAC
NASA GES DISC
Earth Engine Public Data
Archive
NOAA NCEP
Evapotranspiration Modeling
Input datasets:▪ Satellite Imagery (Landsat)▪ Elevation data (National Elevation Dataset)▪ Land use (National Land Cover Dataset - NLCD)▪ Weather data (NLDAS / gridMET)
Christmas Valley, ORJuly 15th, 2014
source: Baburao Kamble, Ayse Kilic (UNL), Rick Allen (Univ. Idaho) & Justin Huntington (DRI)
Remote Sensing Data is Messy!
● clouds● haze● view angle● sun angle● properties change● sensor calibration ● lions, tigers, bears….
Our "Big Data" is just Sparse Sampling
● spatial resolution● spectral resolution● revisit frequency
You think this is the Big Data era?This is just the beginning...
Transparency Can Force Change