u.s. department of the interior u.s. geological survey michael budde, james verdin, and james...
TRANSCRIPT
U.S. Department of the InteriorU.S. Geological Survey
Michael Budde, James Verdin, and James RowlandUSGS – FEWS NET
Sentinel-2 Agriculture User Consultation25 April, 2012 – Frascati, Italy
Remote Sensing Support to Famine Early Warning:
Prospects for Uptake of Sentinel-2 Observations and Products
An activity of the Office of Food for Peace at USAID, which directly supports its goal:
“to ensure that appropriate… emergency food aid is provided to the right people in the right places at the right time and in the right way”
FEWS NET is a food security decision support system with its own climate services
Famine Early Warning Systems Network
FEWS NET Implementing Partners
US Geological Survey - EROS
University of California, Santa Barbara (UCSB)
NOAA CPC, NOHRSC, ESRL
NASA GSFC
USDA FAS
Chemonics International (Prime Contractor)
• Evidence for Development - Livelihoods
• Overseas Strategic Consulting - Web Dev.
• Action Against Hunger - Nutrition
CORE PARTNERS
Kimetrica
NETWORK MEMBERS
World Food Program (WFP)
Food and Agriculture Organization (FAO)
National Government Ministries
Price/Market Information Systems
Meteorological Centers
Non-governmental Organizations (NGOs)
Regional Institutions
(SADC, CILSS, COMESA, IGAD, ASEAN)
Other UN agencies (UNICEF, UNHCR)
FEWS NET – Where we work
The Purpose of FEWS NET
• To prevent famine and mitigate food insecurity by providing decision makers with information that is accurate, credible, timely, and actionable.
• To strengthen the ability of FEWS NET countries and regional organizations to provide timely early warning and vulnerability analysis.
FEWS NET Early
Warning Analysis
Convergence of
evidence
Livelihoods based
analysis
Scenario building
Comparable classification
of food security
outcomes
Effective decision support
Convergence of Evidence
Remote sensing
Rainfall Forecasts
Market/Price monitoring
Field Assessments
Nutrition Surveys
Conflict
The FEWS NET Approach
FEWS NET Early
Warning Analysis
Convergence of
evidence
Livelihoods based
analysis
Scenario building
Comparable classification
of food security
outcomes
Effective decision support
The FEWS NET ApproachConsider evidence in the context of local livelihoods
FEWS NET Early
Warning Analysis
Convergence of
evidence
Livelihoods based
analysis
Scenario building
Comparable classification
of food security
outcomes
Effective decision support
The FEWS NET ApproachScenario Building
Current FS status
Other Info: Trade, Conflict, Health…
Climate Forecasts
Livelihoods
Seasonal Calendars
National/Regional Food Security Outcomes
FEWS NET Early
Warning Analysis
Convergence of
evidence
Livelihoods based
analysis
Scenario building
Comparable classification
of food security
outcomes
Effective decision support
The FEWS NET ApproachComparable Classification
East Africa Estimated food security Outcomes, Sept 2011
FEWS NET Early
Warning Analysis
Convergence of
evidence
Livelihoods based
analysis
Scenario building
Comparable classification
of food security
outcomes
Effective decision support
The FEWS NET ApproachEffective Decision Support
• Livelihood systems are based on subsistence agriculture and/or pastoralism, and are highly climate-sensitive
• Conventional climate station networks are sparse and often late reporting or don’t report at all
• Satellite remote sensing and modeling fill the gap, and provide the basis for early detection of agricultural drought
• Rainfall, vegetation, snow pack, actual ET, crop water requirements are monitored for both rainfed and irrigated crop lands as well as pastoral areas
• Observations are focused by region and seasonality based on crop calendars
Remote Sensing / Modeling
Current real-time Monitoring Capabilities
Operational Remote Sensing
Modeled Derivatives
NDVI
RFE
LST
Current real-time Monitoring Capabilities
http://earlywarning.usgs.gov/fews
• Africa - 60+ products
• Afghanistan - 25+ products
• S. Central Asia - 8 products
• S. Asia - 5 products
• L. America / Caribbean - 12+ products
• Global - 5 products
USGS FEWS NET Data
Regional interactive map viewersEarly Warning Explorer (EWX)
RFE Anomaly MODIS LST eMODIS NDVI eMODIS Anom
Time series of eMODIS NDVI, cumulative RFE, and Dekadal RFE
User interface to dynamically compare standardized anomalies
Framework to incorporate NASA data into routine analysis
Multi-level time series analysis Ability to analyze analog years Time series data download
FEWS NET Online Analysis Tools
Operational Monitoring – eMODIS Example
eMODIS (EROS MODIS) background• Developed by the LP DAAC (USGS-EOS) at EROS in 2007• Funded by vegetation monitoring stakeholders• Expedited and Historical products for Terra and Aqua MODIS• CONUS, Alaska, C. America/Caribbean, Africa, and Central Asia
eMODIS Provides• Ability to customize compositing periods• Consistent and timely processing (Expedited within 12 hours
and Historical 3-5 days)• Minimized Re-sampling• Improved spatial resolution (i.e. 250m)
LANCE is an Earth Observing System Data and Information System (EOSDIS) and provides near real-time capabilities for applications, operational agencies, and researchers. The LANCE system supports eMODIS expedited processing through the provision of timely surface reflectance data.
Expedited• LANCE Level-2 Surface Reflectance• Final eMODIS product delivered within 12 hours of acquisition
Historical• LAADS Level-2 Surface Reflectance (ephemeris &
atmosphere)• Final eMODIS product delivered within 3-5 days of acquisition
eMODIS NDVI - LANCE
eMODIS SystemUSGS/EROS
Terra MODIS
LANCE
USGS FEWS NET
Processing System (stack, smooth, create products, develop time series, package data for distribution)
EDOS
MODIS L0 Data T+3hrs
T+9hrs
Composite +18 hrs
FEWS NET Decision Support System (USAID FFP)
MODISL2 Data
FEWS NET eMODIS NDVI and Anomaly Maps
Input to Impact Assessment s and Food Security
Decision Making
eMODIS NDVI – Expedited Data Flow
Potential Applications of Sentinel-2 DataExtension of Existing Waterhole Monitoring
• Active for 41 waterholes in Kenya / Ethiopia• Website provides daily information on the status of a
waterhole relative to the median and previous year• High resolution multi-spectral imagery needed for
waterhole identification/delineation Validation with field data captures trends and seasonal variation successfully
Potential Applications of Sentinel-2 DataTargeted Rangeland / Cropland Monitoring
• Delineation of known dry-season grazing areas and water-points.
• Identification of robust indicators and analysis• Refined mapping of migratory patterns.• Identification of key livestock markets & trade flows.• Potential constraints i.e human/wildlife conflicts,
animal diseases and Govt. policies (X-border).
Continuous monitoring of rangeland resources; focusing on critical period(s) of livestock migration to “dry-season” grazing areas.
Focused Cropland Assessments
Potential Applications of Sentinel-2 DataIrrigated Area Mapping / Delineation
False Color Landsat Image : May 27, 2004
Irrigated Areas 2004Using peak NDVI threshold
Helmand Province, Afghanistan
• Utilizes high resolution imagery in combination with 250m MODIS NDVI to annually map irrigated areas in Afghanistan.
• Provides knowledge of expanding or abandoned areas under irrigation.
• Allows for more accurate monitoring of irrigated agriculture conditions.
Potential Applications of Sentinel-2 DataCropped Area Estimation
• High-resolution imagery allows us to “visit” a large number of locations with no access restriction
• Imagery is interpreted as crop, no-crop, and other using a dot grid approach
• Dots are attributed with external data – geophysical and demographic
• Statistical models are built - generalized additive mixed model
• Models predict cropped area for locations where we do not have imagery
• Produce country-level estimate of cropped area
• Cropped area is a critical component of estimating food production in the developing world and is often inaccurately reported .
• Independent cropped area estimates provide an unbiased assessment.
The Need
The Approach
High resolution image (~1 M) for Haiti with a 500 M dot grid overlain (yellow dots)
Requirements for use of Sentinel-2 Data• Near real-time processing capabilities (e.g. LANCE system for
MODIS surface reflectance) - expedited data delivery is key to being applicable as real-time monitoring tools
• Ability to streamline data delivery process - similar to the Web-enabled Landsat data (WELD) Project
• Sentinel-2 data available at no cost – even low cost creates a bottleneck in the ability to utilize the frequent revisit time for early warning applications
• Interoperability with Landsat 8 – the ability to leverage observations from both Sentinel-2 and Landsat 8 to create comparable products at even better frequency than either sensor can offer is a true benefit to the earth observing community
Benefits for Famine Early WarningThe need for a “closer look” – as FEWS NET, we need to assess broad-scale food security issues, but are increasingly requested to focus our attention at finer scales - an example from S.Somalia
Worst Drought in 60 Years • The worst drought in 60 years, on the heels of two failed rainy seasons, had devastating impacts on the eastern Horn of Africa in 2011 and led to a declaration of famine in southern Somalia.
• The drought conditions coupled with conflict in Somalia created uncertainty in the state of the agricultural system going into the short rainy season (Oct – Dec, 2011)
2010/11 rainfall compared to historical totals since 1950/51 in select pastoral areas of Kenya and Ethiopia
Benefits for Famine Early Warning• Together with the Food Security and Nutrition Analysis Unit (FSNAU), FEWS NET monitored the
short rainy season and potential crop conditions for specific irrigated and rainfed areas on a frequent basis, updating time series every 5 days based on eMODIS NDVI.
• Extremely good rains throughout the season led to improvements in pastoral conditions and cropped areas appeared to perform remarkably well (time series below).
• However, there was still concern over whether traditionally cropped areas were actually planted due to the likelihood of outmigration resulting from conflict.
• Sentinel-2 observations with their frequent revisit schedule, fine resolution, and applicable spectral characteristics would have been suitable to provide that “closer look”.
Thank you