cloud computing for drought monitoring with google earth engine
TRANSCRIPT
CLOUD COMPUTING FOR DROUGHT
MONITORING WITH GOOGLE EARTH ENGINE
Landsat 8 John Abatzoglou
Katherine Hegewisch
Alex Peterson
Donny VanSant
Rick Allen
Ayse Kilic
Tyler Erikson
David Thau
Noel Gorelick
Rebecca Moore
Mike Hobbins
Jim Verdin
Justin Huntington
Britta Daudert
Charles Morton
Dan McEvoy
Andy Joros
Landsat 8
Introduction• Collaboration with Google Earth Engine Team
• DRI and U-Idaho received two Google Earth Engine Faculty Research grants in 2014 to develop
software and provide guidance for monitoring of drought and evapotranspiration (ET)
• One of many results – a web application so anyone can process and visualize map and time series
and users can download results
MODIS April – October 2014 Median NDVI made with Earth Engine in about 7 seconds
Landsat for Vegetation Water Use and
Drought Monitoring• Remote sensing using Landsat is arguably the only way to detect vegetation stress and
ET at field scales over large areas
• Landsat pixel size (30m x 30m) is optimal for evaluating individual fields, riparian zones, and meadows (1985-pres)
• MODIS pixel size (250m x 250m) is optimal for regional analysis (2000-pres)
• To better understand if vegetation changes are natural or anthropogenic we need ~30+ years of satellite data, and paired with climate archives
• Better understanding vegetation and ET varies with climate at field and regional scales will increase the effectiveness of biological and hydrological monitoring plans, and drought monitoring
Landsat MODIS
Cloud Computing with Climate and Remote Sensing Data
• Develop a tool to better understand the long term spatial and temporal variability of ET from irrigated agriculture and groundwater dependent ecosystems (riparian areas, wetlands, springs)
• Rely on Landsat satellite imagery (16 day return intervals) to compute vegetation indices and energy balanced based ET
• Rely on gridded weather data to estimate PPT and ETo
• Problem – lots and lots of data and processing..
• 21 scenes for NV
• 1000+ Landsat images per path/row since 1985
• Equates to >20,000 images to process..
Google Earth Engine Cloud Computing• Google has the entire archive of Landsat and MODIS imagery and CFSR, NLDAS, and downscaled NLDAS
gridded weather data available for massive parallel processing in the cloud
• This technology has changed the paradigm of how we process and analyze satellite imagery and gridded
weather data
• https://earthengine.google.org/#timelapse/v=40.86687,-117.50682,8.988,latLng&t=2.85
Java Script and
Python
Application
Programming
Interface (API)
Max 30m Landsat 8 NDVI 5/2014 to 10/2014
Landsat and Drought Monitoring
Lovelock, Nevada – Humboldt River Basin• No groundwater pumping for irrigation (too salty)
• Very little storage upstream
• Extremely sensitive to persistent hydrologic drought
-Growing Season Max NDVI (30m Pixels) – Computed using Google Earth Engine
-Google hosts the entire 40yr+ Landsat archive and provides parallel cloud computing
2011 2013 2014
Wet Dry Drier
Landsat
~0%
water
delivery
Monitoring Spring declines – Needle Point Spring, UT• GW modeling is a necessary tool for assessing and predicting capture of natural
SW/GW discharge• https://earthengine.google.org/#timelapse/v=38.74288,-114.04747,10.812,latLng&t=0.61
Needle Point Spring stopped flowing
in 2001
Pumpers are point fingers at each
other as to who is responsible
Hearing just held at NV State
Engineer’s Office
Hydrologic Modeling & Remote Sensing• Groundwater modeling and remote sensing
tools are mutually supportive tools• Modeling supports the remote sensing
• Remote sensing supports the GW modeling
• Newly developed front end web application
to mine the Landsat and other remote
sensing and gridded weather data archives
in the cloud to evaluate change, and better
understand causality of change
Google Earth Engine Landsat Time
Series Tool – Climate Engine
Demo of Climate EngineHopefully a tool relevant at field and regional scales
www.climateengine.org – still in development..
Demo of CLIM Engine and some other tools that can be
useful for field level inventories and drought assessments
• Show case studies on Climate Engine
• Show PDSI for last year (case study link)
• Maggie Creek Restoration (Landsat summer NDVI time series)
• Range allotment by South Fork (MODIS summer time series, summer PPT)
• -115.92 E , 40.56 N
• Indian Valley dry vs drier year map and time series
• Impacts of pumping on Needle Point Spring vegetation (daily Landsat time
series)
Landsat 8, Launched Feb 11, 2013
Contact Information:
775-673-7670
Many thanks to:
You
Collaborators
BLM
USGS/NASA
Landsat Science Team
NV Division of Water Resources
University of Idaho
Drought and Past Precipitation & Temperature
• Precipitation in NW NV is extremely variable from year to year
• Only a dozen years are near the10 inch average
• Recent droughts are relatively short compared to the 30s and 50s droughts
• Current temperatures are higher than the 30s drought; high temperatures increase severity of
droughts ( i.e. drought feedbacks)
*
Figures modified from Mike Dettinger, USGS
Precipitation 5yr avg.
Temperature 5yr avg.
Temperature Trend
Precipitation
Extended Droughts
Drought Evolution and Monitoring
Drought evolution is complex: it has multiple drivers and develops and
recovers at different time scales
Long Term
vs.
Short Term
Drought
Hydrologic
vs.
Soil Moisture/
Rangeland
Drought
Blue = Wet : Red = Dry
Jan – Dec 2014 Precipitation Anomaly July – October 2014 Precipitation Anomaly
• Fish Lake Valley example of pairing Landsat NDVI with PPT and pumping
• Groundwater is primary source of water for irrigation in the valley
• Test – can we see changes in greasewood NDVI due to pumping?
• https://earthengine.google.org/#timelapse/v=37.82067,-118.03078,10.812,latLng&t=2.86
Fish Lake Pumping and Wetland / Greasewood Flat
Vegetation
Result – Fish Lake Pumping &
Greasewood Phreatophytes
• Arlemont Ranch well (117 S01 E35 35CC 1) measured by NDWR
• Digitized polygon around well, ~ 0.25 miles across; is largerly comprised of greasewood
• Evaluated spatial average Aug-Sept NDVI, PRISM PPT, and water levels
• NDVI declining; GW levels declining
Avg. PPT = 5in/yr
Pre
cip
ita
tio
n F
rac
tio
n o
f N
orm
al
2013 July-Aug Max NDVI 2014 July-Aug Max NDVI
Indian Valley supports the largest
Sage-Grouse lek in NV (i.e. aggregation
of males dancing for females)
Indian Valley, NV
Remote Sensing for Sage-Grouse Sensitive Areas
Capture of Groundwater Discharge
• Appropriation of the full perennial yield assumes capture all the natural groundwater discharge
• By design, long-term groundwater pumping causes a lowering of the water table and reduces groundwater ET (ETg)• Capture of ETg is put to beneficial use (for humans)
• Capture of ETg reduces vegetation vigor and biological diversity
• In most cases, groundwater appropriation is based on the ETg from phreatophyte vegetation
Sources of Water to a Pumped Well
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.00 10 20 30 40 50 60
TIME, IN YEARS
FR
AC
TIO
N O
F P
UM
PIN
G R
AT
E
GW storage
“Capture”
- capture of SW and ETg
Theis (1940) “All water discharged by wells is balanced by a loss of
water somewhere else”
“the idea of safe yield…in which the size
of a development if it is less than or equal
to the recharge is considered to be ‘safe’
is fallacious”
“Often streams are depleted long before
the pumping reaches the magnitude of
recharge.”
“…if pumping equals recharge (or discharge),
eventually streams, marshes, and springs dry up”
“Despite being discredited repeatedly in the literature,
safe yield continues to be used as the basis of water-
management policies, leading to continued ground-
water depletion, stream dewatering, and loss of
wetland and riparian ecosystems.”
Monitoring, Management, Mitigation (3M) Plans
• Baseline and future hydrologic and biological monitoring (~7yrs)
• Establishment of groundwater management actions• Staged development
• Trigger levels
• Pumping schedules
• Assess response of ecosystems to withdraw
• Refinement of unreasonable adverse effects
• Mitigation measures• Operational adjustment
• Change in pumping location
• Reduction in pumping / curtailment
• Provide alternative water source
Shoshone Ponds
Stipulation Requirements for Hydrologic Monitoring
• Monitor stream / spring discharge
• Monitor vegetation vigor
• Separate impacts of climate from
pumping effects
• Remote sensing, including both aerial
photography and satellite imagery
“However, currently available technology does
not provide sufficient precision to detect short-
term changes in vegetation that may be
induced by groundwater withdrawal at the fine
scales necessary to meet the monitoring
requirements of the Plan. Instead, permanent
line transect data will be used to detect these
fine-scale vegetation changes.”
Shoshone Ponds