Integrating Satellite Data with Ecosystem Models Progress

1
Integrating Satellite Data with Ecosystem Models
Progress ProspectsRamakrishna NemaniPetr
VotavaAndy MichaelisForrest MeltonHirofumi
HashimotoWeile WangCristina MilesiSam
HiattNASA Ames Research CenterMoffett Field,
CAGlobal Vegetation Workshop, Missoula, MT,
June 24, 2009
2
outline
Need for integrating satellite data into
ecosystem models Retrospective to near-realtime
integration Drowning in data From data-poor to
a data-rich environment Building community
focused data - modeling centers
3
Integrated, Multiple Constraints on the Biosphere
1000 km
Upscaling Prediction
10 km
1km
1ha
1m
Downscaling Verification
4
Original FOREST-BGC flow diagram, emphasizing
dual time steps, critical role of LAI, C-H2O-N
interactions, and remote sensing applications,
1988
5

• no LAI spoken here

6
First? spatial ecosystem model driven by remote
sensingTwo weather stationsOne AVHRR
sceneTwo years after acquisition24 hours to
run the modelLess than 1500 km2

• Running et al. Ecology 1989

7
Retrospective to Real time
Key elements Monitoring Modeling Forecasting
Local to Global
Nemani et al., RSE, 2008
8
Daily Weather Surfaces 1KM
Raw data
Gridded Products
9
Satellite data
1/05
2/05
10
California Ecological Daily Nowcast at 1km
Feb/01/2006
Climate Satellite
Carbon and water cycles
T
P
ET
GPP
0
2.5
5
GPP (gC/m2/d) ET (mm/d)
Biome-BGC Simulation models
RAD
Outputs include plant growth, irrigation demand,
streamflow
11
Daily Nowcasts California
Ecosystem
Vegetation
Hydrology
Meteorology
12
Ensemble Model Experiments
TOPS
BGC
CASA
LPJ
BEAMS
SimCYCLE

More
Data sets are available at the TOPS website
(http//ecocast.arc.nasa.gov)
Wang, Michaelis, Dungan
13
Near realtime monitoring of global NPP anomalies
Running et al., 2004, Bioscience, 54547-560
14
Irrigation Forecasts
Irrigation Forecast for week of July 19-26,
2005 Tokalon Vineyard, Oakville, CA CIMIS
Measured Weather Data through July 18, 2005 NWS
Forecast Weather Data July 19-26, 2005
0
30
Forecast Irrigation (mm)
0
1000
meters
N
Fully automated web delivery to growers
Seasonal
15
Drowning in Data
Less than 10 years ago, modelers spent nearly
90 of their time gathering appropriate
data Now the DAACs distribute about 4TB a
day Global data at 8km, 1km, 500m, 250m and 30m
16
From data to knowledge/decisions
17
Limitations of Current Standard Practice

• Data volume Long term analyses of climate and
  ecosystem change increasingly require retrieval
  and processing of multiple terrabytes of data
  from a diversity of models and sensors.
• Bandwidth As the growth in bandwidth begins to
  flatten, transmission of data from centralized
  data archives presents an increasing challenge.
• Data management costs and redundant storage and
  processing Costs associated with local storage
  and management of data and compute resources add
  significant costs to individual research and
  application development efforts.
• gt100k/year

18
Proposed solution
Community-focused data-modeling centers (bring
your code to the data) Proposals solicited for
50 of 400M for NASA Earth Sciences Prototyping
exercise, funded as part of the stimulus
activity Possibly matching funds from CISCO
19
Proposed Data Modeling Center
What will it have 2 x 512 core CPUs 2 PB
disk storage Data MODIS, AVHRR,
LANDSAT Models publicly available
codes Software Utilities public domain Web
Portal
20
How will it work?Generalized Disturbance Index
Mildrexler et al 2006
Mildrexler et al 2006
21
Gaining access
Users register Only registered users can
request project resources Project Request
Template A one page summary of proposed
work Requested Datasets Spatial
subset Temporal subset Requested
models Requested utilities Estimated CPU
hours Length of time the resources are
requested for We're working on licensing issues
for commercial packages
22
Create virtual environment
Allocate computing resources Allocate disk
space Make available all requested data
sets Create a project page on the Social
Network
23
User gains access
Shell access Can bring their own models Can
bring their own data Run and modify their
algorithms Publish results and share them with
other researchers Save their environment (data,
algorithms etc.) for future re-use and
reproducibility of results This environment can
be shared with other people
24
Clean-up
Resources get recycled If desired, captured
knowledge is archived Environment Results Data
Models
25
Social networking component
Search for related work Who's doing what
where Map interface Categorized by
discipline Categorized by geographic
region Create your own project Control the
visibility of your project Control the access to
your project Control level of sharing Link to
other projects Add publications
26
Phasing access
A prototype will be ready in Spring
2010 Available first to NASA Centers NASA
funded researchers then to the rest of the
community
27
It is not just an idea. We are funded to build
this. It is supposed to serve the
community. Please get involved, send comments,
suggestions to ecocast_at_gmail.com