terrestrial observation and prediction system development of a biospheric nowcast and forecast...
TRANSCRIPT
Terrestrial Observation and Prediction System Development of a Biospheric Nowcast and Forecast Capability
Ramakrishna NemaniNASA/Ames Research Center
Collaborators:
Keith Golden, Petr Votava, Michael White, Andy Michaelis, Forrest Melton, Matt Jolly, Kazuhito Itchii, Hirofumi Hashimoto, Clark Glymour, Steve Running, Ranga Myneni and Patricia Andrews
NASA Biodiversity and Ecological Forecasting Team MeetingAugust 30, 2005
Turning Observations into Knowledge Products
With the Launch of Aura, the 1st Series of EOS is Now Complete
Goal
Specific goal for this project is to develop a biospheric nowcast and forecast system useful for monitoring and predicting key ecosystem variables relevant in natural resources management
Key elements:
Monitoring
Modeling
Forecasting
Scale flexibility
Terrestrial Observation and Prediction System
Technology focusDistributed Agent Architecture
UWF,Tetrad IV
CMU
Nat’l. Data Centers
UWPRECISE
NASA ARCTOPS/IMAGEbot
UMTTOPS Appl
Scripps Inst. OceanographyCO2/Climate Forecasts
Evaluation criteria
Time and resources needed to implement over a new geographic region add a new sensor/new data source add a new model adapt to a new domain
Ability to quantify improvements
gridding climate data
RAWS
Modular
Unattended
Tmax / TminVPD, precipitationSolar radiationDaylength
Any userDefined grid
Jolly, nemani, Running…. 2004. Envi. Modeling and Software
Global Vegetation Production Anomaly May 2005
Potential Climate Limits for Plant Growth
Temperature
WaterSunlight
Each month, our analysis identifies climate-related
causes behind the predicted NPP anomalies
Brian Bonnlander/Clark Glymour/Votava, IHMC/ARC
Train the algorithms on all the non-arson fires during 2000-2002
Methods include:Support Vector MachinesArtificial Neural NetworksLogistic Regression
Data-driven modelsMODIS data in mapping wildland fire risk
Predicting fire risk
Brian Bonnlander/Clark Glymour/Votava, IHMC/ARC
CAL-SYNERGY1km Daily weather, satellite and model data
MaximumAir Temperature
Vegetation density Vegetation Growth Soil Moisture
Most downloaded data setUsed by USGS, CDW, NPS, BLM andWine industry
MODIS MODEL
Monitoring snow conditionsMonitoring snow conditions Columbia river basinColumbia river basin
Interannual variability in snow conditions
Sn
ow C
ove
r A
rea
(1
05
km
2)
Collaboration with the National Park Service
TOPS Irrigation Scheduling
LAI from NDVIImagery
Limited Farm-scale Soils Data
0
5
10
15
20
25
30
J F M A M J J A S O N D
0
10
20
30
40
50
60
70
80
90
Tavg, C
ETo, mm
Ppt, mmMet Data from CIMIS
CropParamsfrom Variety
Irrigation ForecastsCrop Monitoring
Inputs Modeling Outputs
Forecast from NWS
Maintaining optimal water stress for better vintages
Vineyard Water ManagementIrrigation forecasts
Used to maintain vines at specificwater stress level to maximize
fruit quality
Forecasts integrate high-resolutionsatellite/aircraft data, weather, soils
and NWS short-term forecasts
Irrigation Forecast for week of July 27, 2005
Partners include Constellation/Mondavi,Hess collection, Kendall Jackson and
several other small wineries1000meters N
interannual climate-wine quality
Nemani et al., 2001 Climate Research
Interannual variability
Change in Spring (March-April-May) Temperature, oC
[1998-2004] - [1991-1998]
Decadal climate changes and U.S wine industry
Cooler springs after 1998
Late budbreak Slow ripening
Delayed harvest Increasing risk from frost
Predicted Changes in phenology in response to climatic changes
Later bloom over the west after 1998
Changes in start of growing season derived from satellite data
Planning/Execution Agent technologies beyond TOPS
FutureCurrent
Ecological Forecastinghttp://ecocast.arc.nasa.gov
Summary
Willem de Kooning (1904-1997)A Tree in Naples (1960)Museum of Modern Art
the end
more information at: http://ecocast.arc.nasa.gov
Unprecedented data volumes
Working with large data sets requires robust automation
Planning/Execution technologies allow integration of distributed & heterogenous data sets
TOPS is not model-centric, allowing rapid adaptation to new domains
Potential for mimicking the weather service with ecological forecasts of various lead times
Characterizing and communicating the uncertainty inecological forecasts remains a challenge
Summary