U.S. Geological Survey Earth Resources Operation Systems (EROS) Data Center World Data Center for Remotely Sensed Land Data
USGS EROS DATA CENTER Land Remote Sensing from Space: Acquisition to Applications Earth Observation Satellites USGS National Archive Challenge Data Applications Declassified Systems Preserve Landsat 1-5,7 Provide Access NOAA - POES Process Shuttle Radar Reproduce TERRA (1999) Distribute NASA-EOS (1999) Hold in Trust High Resolution Systems Expanding to over 18 million images of the earth! Land Cover Environmental Monitoring Emergency Response Fire Danger Rating DOI Land Management Natural Hazards Coastal Zones
USGS EDC Data Holdings ¾ Aerial Photographs ¾ 1940-present ¾ U.S. coverage ¾ > 9 million frames ¾ Scale: 1-2 meter Natl. Aerial Photography Program (NAPP), Dallas/Fort Worth Airport
USGS EDC Data Holdings ¾ Landsat Satellite Images ¾ 1972-present ¾ > 18 million frames ¾ Global coverage ¾ 15-80 meter Landsat 5 MSS
USGS EDC Data Holdings ¾ AVHRR Satellite Images ¾ 1987-present ¾ Global coverage ¾ 1 km resolution AVHRR Time Series
Using Landsat satellite imagery to estimate agricultural chemical exposure in an epidemiological study Susan Maxwell, PhD (USGS EROS Data Center) Interface 2002, Montreal, Canada Collaborators: Dr. Jay Nuckols, EHASL, Colorado State University Dr. Mary Ward, National Cancer Institute Eric Smith, EHASL, Colorado State University Leanne Small, EHASL, Colorado State University Fort Collins, Colorado - Landsat 7 - July 26, 1999
Why use satellite imagery? ¾ Traditional methods of collecting chemical exposure data don t work well (environmental/biological sampling, questionnaires) Spray drift Dust Agriculture Chemicals ¾Fertilizers ¾Pesticides Drinking water
Why use satellite imagery? ¾ Cancers generally take several years to develop, therefore need to reconstruct historical exposure ¾ Our approach: use Landsat imagery to create historical land use/crop type maps integrate with other data (chemical use, soils, wind, etc.) to estimate exposure
Metric Development Transport Modeling # Residence with 500 Meter Buffer U.S. Census Bureau Place Areas Cultivated with Sorghum # # 0.22-0.24 0.18-0.22 0.14-0.18 0.1-0.14 0.06-0.1 0.04-0.06 0.02-0.04 0.01-0.02 0.005-0.01 0.003-0.005 0.001-0.003 No Data N 0 1 Mile (Ward et al. Environmental Health Perspectives, 2000)
Why Landsat? ¾ Longest running satellite sensor (1972-current) ¾ Successful crop type mapping applications (AGRISTARS, etc.) ¾ Appropriate spectral bands (visible, near infrared, middle infrared) ¾ Appropriate spatial resolution (30-80 meter) ¾ Inexpensive (compared to higher resolution data sets)
Crop Type Classification - Sheldon, NE
Case Study Mapping Corn ¾ Chemicals used on corn (nitrogen, atrazine) have been associated with several cancers and birth defects Ground-water contamination risk From: USGS 1225, The quality of our nation s waters
Traditional classification methods are not appropriate ¾ Only want CORN ¾ BIG Data Sets Large geographical regions File size ~500 Mb/image Multi-year 30 years
Traditional classification methods are not appropriate (cont.) ¾ Usually need ground reference data expensive, difficult to get for historical data ¾ Time-consuming process
Crop characteristics ¾ Corn dominates corn soybeans sorghum dry beans sugarbeets corn soybeans sorghum dry beans sugarbeets 0.8 100 Hectares (m illion) 0.6 0.4 0.2 0.0 33 32 31 30 29 28 Landsat Path Number Proportion (% ) 80 60 40 20 0 33 32 31 30 29 28 Landsat Path Number
Crop characteristics ¾ Large, homogeneous fields ¾ Spectral characteristics differ from other major crops (soybeans, alfalfa, winter wheat, etc.) ¾ Spectrally similar to deciduous trees, riparian area
Case Study Mapping Corn ¾ Initial method software was developed to. ¾ Use existing land cover maps (NLCD) to eliminate non-row crop classes (spring grains, hay/pasture, trees, urban, wetland, etc.) ¾ Use existing USDA acreage estimates to target specific geographic region (i.e., county) to collect training statistics ¾ Use maximum likelihood algorithm to classify the entire image ¾ Use the Mahalanobis distance image in combination with USDA acreage estimates to identify cut-off for highly likely corn, likely corn and unlikely corn
Method cont. ¾ Use existing land cover maps (NLCD) to eliminate non-row crop classes (spring grains, hay/pasture, trees, urban, wetland, etc.)
Method cont. ¾ Use USDA acreage estimates to target specific geographic region (i.e., county) to collect training signature 80 Hall 1000's of Hect 60 40 20 0 Corn Sorghum Soybeans All Hay Winter Wheat
Method cont. ¾ Use the Mahalanobis distance image in combination with USDA acreage estimates to identify cut-off for highly likely corn, likely corn and unlikely corn Mahalanobis distance image Highly Likely Corn Likely Corn
Mahalanobis Distance Threshold Mahalanobis Distance Value Land Area (Hectares) Cumulative Total (Hectares) Cumulative Total (% of NASS) Classification Code 1 1206.4 1206.6 2.1 1 2 4413.2 5619.6 9.6 1 3 1364.4 6984.0 11.9 1............ 55 581.0 44107.2 75.2 1 56 517.7 44624.9 76.0 2 57 741.2 45366.1 77.3 2 58 141.8 45507.9 77.5 2............... 131 1066.3 59082.1 100.7 2 132 417.2 59499.3 3............ 1787 0.4 82893.2 3
Results ¾ >80% average accuracy ¾ Higher errors occur when Spectrally similar cover types in same area (millet, sorghum) Image date is too early in growing season Non-parametric signature (clouds/haze, irrigated/nonirrigated corn)
Thank You Susan Maxwell maxwell@usgs.gov