corrie hannah mentor: dr. stuart e. marsh, office of arid lands studies nasa space grant symposium

Corrie HannahMentor: Dr. Stuart E. Marsh, Office of Arid Lands Studies

NASA Space Grant Symposium

April 17, 2009

Arizona State University

Using Remote Sensing to Map and Manage Using Remote Sensing to Map and Manage Buffelgrass Infestations in Tucson and the Buffelgrass Infestations in Tucson and the

Santa Catalina MountainsSanta Catalina Mountains

What is Buffelgrass?What is Buffelgrass?

• Invasive species originally brought from Africa to the Americas for cattlegrazing

• Threatens to out-compete native species• Potential to transform the fireproof Arizona

Sonoran Desert into a burning savannah

National Fire Plan http://www.forestsandrangelands.gov

((Pennisetum ciliare)Pennisetum ciliare)

What is Remote Sensing?What is Remote Sensing?A science that uses satellite imagery and aerial

photography to identify features and natural processes on the Earth’s surface

ObjectivesObjectives• Identify a practical method of mapping and

managing buffelgrass in Southern Arizona • Determine benefits of different spectral

and spatial satellite and aerial imagery to locate buffelgrass

Identifying Remotely Sensed Identifying Remotely Sensed Imagery to Map BuffelgrassImagery to Map Buffelgrass

TypeType Spectral Resolution Spectral Resolution Spatial Spatial ResolutionResolution

PAGPAG Aerial Photography

3 color bands 1ft

DOQQDOQQ Aerial Photography

3 color bands plus near infrared

1m

LandsatLandsat Satellite Imagery

6 reflective bands 30m

ASTERASTER Satellite Imagery

3 color & near infrared/6 shortwave infrared bands

15m/30m

MethodsMethods• Identified remotely sensed imagery• Manually digitized high resolution aerial imagery• Located buffelgrass locations: Presence & Absense• Fieldwork! Validated digitized ground control points• Classified buffelgrass locations in images• Accuracy assessments

Area of Study: Santa Catalina Mountains North of Tucson

Heads-up DigitizingHeads-up Digitizing

Buffelgrass!!!

Supervised ClassificationsSupervised Classifications

1ft Aerial Image

30m ASTER Satellite Imagery

1m Aerial Image

30m Landsat Satellite Imagery

Red polygon areas indicate buffelgrass presence

Texture Filters and TopographyTexture Filters and Topography

1ft Aerial Image + Texture + Topography 1m Aerial Image + Texture + Topography

15m ASTER Satellite Image + Topography 30m Landsat Satellite Image + Topography

Red polygon areas indicate buffelgrass presence

Multi-temporal StacksMulti-temporal Stacks

15m ASTER scene composite: 4/22/2006 and 7/18/2006

15m ASTER scene composite: 7/18/2006 and 11/27/2008

15m ASTER scene composite + Topography: 4/22/2006 and 7/18/2006

15m ASTER scene composite + Topography: 7/18/2006 and 11/27/2008

Red polygon areas indicate buffelgrass presence

Binary Ensemble Binary Ensemble

Yellow = 5 or more models predict buffelgrass

Red = 10 or more models predict buffelgrass

Green/White = less than 5 models predict buffelgrass

A composite stack of the twenty classifications: buffelgrass presence or absence

Place where the most models were Place where the most models were successful at locating buffelgrasssuccessful at locating buffelgrass

Photo from the Field

1ft Aerial photography Binary Ensemble: Close-up

Place where most models were not as Place where most models were not as successful at locating buffelgrasssuccessful at locating buffelgrass

Photo from the Field

1ft Aerial photography Binary Ensemble: Close-up

Photo from the Field

1ft Aerial photography

Binary Ensemble: Close-up

Place where the most models were Place where the most models were successful at locating buffelgrasssuccessful at locating buffelgrass

ResultsResults

• High-resolution imagery was more reliable than moderate-resolution imagery at locating buffelgrass

• Manual digitization of the1ft aerial photography was the most accurate method of classifying buffelgrass

• Of the automated classified images, the 1ft aerial photography with topography and texture was the most accurate

ConclusionConclusionWe suggest that 1ft resolution or higher aerial

photography, which covers much of the infested areas surrounding Tucson, should be used to most

effectively map and manage buffelgrass infestations in the future.

AcknowledgementsAcknowledgements

• Aaryn D. OlssonAaryn D. Olsson• Kyle A. HartfieldKyle A. Hartfield• Stuart E. MarshStuart E. Marsh• Barron J. OrrBarron J. Orr• Grant CasadyGrant Casady• Arizona Remote Sensing CenterArizona Remote Sensing Center• Office of Arid Lands StudiesOffice of Arid Lands Studies• NASA Space Grant ConsortiumNASA Space Grant Consortium

Thank YouThank You