Detecting and Mapping Invasive Phragmites australis in the Coastal Great Lakes with ALOS PALSAR Imagery

Transcription

1 Detecting and Mapping Invasive Phragmites australis in the Coastal Great Lakes with ALOS PALSAR Imagery Brian Huberty U.S Fish & Wildlife Service Region 3 Ecological Services Laura L. Bourgeau-Chavez, Kirk Scarbrough, Anthony Landon Michigan Technological University Michigan Tech Research Institute (MTRI) Ann Arbor, MI Martha Carlson Mazur, Kurt Kowalski, and David M. Galbraith USGS Great Lakes Science Center October 30, 2012

2

3 Invasive Phragmites australis PROBLEM: Aggressive invader that Impacts ecosystem services: Displaces native vegetation in wetlands Reduces biological diversity Dries wetland soils and traps sediment Restricts shoreline views Forms monospecific stands Changes structural complexity Difficult to control

4 Project Objectives Map current invasive Phragmites extent for Great Lakes coastal zone Remote Sensing Field Work Validation Identify major environmental drivers of Phragmites australis distribution Provide decision support tool Assess vulnerable areas to new invasion

5 Project Overview Project goal: Develop methods for creating a distribution map of invasive Phragmites for management and control decision support Solution: Use satellite remote sensing, synthetic aperture RADAR (SAR) at 20 m resolution to obtain mmu of ½ acre Methods: Extensive field surveys and Classification of Satellite SAR imagery, Unsupervised-Supervised Classification Pilot Study - Lake St. Clair Results: Maps of distribution of invasive Phragmites (field data and PALSAR maps) and accuracy assessments Decision Support: USGS is working on the development of publicly-accessible tool to assess vulnerability and aid land managers in allocating limited resources

6 Mapping Approach for U.S. Coastal Great Lakes Basin Target: Monotypic stands Phragmites australis, ½ acre minimum mapping unit Area of interest: 10 km inland from the coastal zone Approach use Radar Sensor: Japanese satellite ALOS PALSAR 23 cm wavelength, m resolution 2 bands L-HH and L-HV Multi-season (spring summer and fall) datasets Requires ~ 87 (70x70km swath) 3- date image stacks 6

7 Why Radar? Landsat ETM (Aug 01) 5,4,3 False Color Composite Radarsat (Oct 98), JERS (Aug 98), JERS (March 95) False Color Composite LANDSAT can be used to identify a broad spectrum of land cover types Radiant energy reflectance from vegetation varies depending on features at the cellular level (e.g. chlorophyll, leaf moisture), as well as variations in surface or background reflectance (e.g., soil type, water). SAR can differentiate wetland types based on: Inundation/water level patterns Vertical structure Soil moisture Biomass

8 Why use Synthetic Aperture RADAR (SAR)? Phragmites SAR can be used to differentiate wetland species based on: Inundation/water level patterns (HH) Vertical Structure (HH) Soil moisture (HH) Biomass (HV) Seasonal (spring, summer, fall data) Phenological variation Water level cycles Typha

9 DATA:Three-season PALSAR Mosaics JAXA JAXA All PALSAR data were processed, terrain corrected and georectified by Don Atwood and staff of Alaska

10 Field Measurements Collected in ½ acre plots GPS locations Center of ½ acre plots Photos with GPS tag 4 cardinal directions (over 3000 photos in archive) Dominant covertype- Vegetative composition Wetland Ecosystem type Average Veg. height (3) Density of Phrag and Typha only Phragmites presence Recent changes/ herbicide/burn treatments

11 Field Training and Validation Data 1145 unique field site visits. 782 validation, 363 training Phragmites observed at 30% of sites. 28% Validation sites 36% training sites Only NWI "Palustrine Emergent" polygons used to generate random points for validation sites of these, only 51% were documented as emergent in the field observations

12 Example Site Map 12

13 Example Field Data Sheet

14 Web-based Data Entry Used to manage spatial, attribute, and image data collected by field teams Web Browser Web Server Output Products (shapefiles, KML, etc.) Spatial Database

15 Field Photos with Geotags Over 3,000 GPS-encoded photos Created kml (for use in GoogleEarth) of photos for distribution From the Field Data GIS: Validation point and Field of View (FOV) for Digital Photos for a Phrag site. Corresponding GPS-encoded photo shown above for highlighted FOV.

16 Field Data Results Coastal Lake Basin Validation 0.2 ha Sites Validation Sites with Phragmites present Training 0.2 ha sites Training sites with Phragmites present Erie (46%) (61%) Ontario (10%) 31 6 (19%) Huron (42%) (53%) Michigan (29%) (29%) Superior (0%) 80 3 (4%) Total (28%) (36%) Validation locations were randomly selected within NWI classed Emergent wetlands Need 377 validation locations per Lake basin for 95% confidence level ( we have 109 to 255 per basin) Phragmites was often targeted in the Training data collection

17

18 Area Mapped as Invasive Phragmites Coastal Lake Basin Coastal area in 10 km buffer (ha) Dominant Area of wetland and select ecosystem types in the filter (ha) Hectares of Phragmites mapped in the filtered areas Erie 778,447 96,862 8,233 Michigan 1,724, ,320 6,002 Ontario 442, , Superior 1,270,484 N/A N/A Huron 650,715 75,402 10,395 Total 4,866, ,640 24,643 Phragmites is more prevalent in the more southern coastal areas where human development and populations are greater. Lakes St. Clair and Erie which are 2 of the shallower of the Great Lakes are heavily infested. Lake Ontario has the least amount of infestation of the southern 4 lakes, and it also has less coastal wetland and less water level fluctuations.

19 Final Potential Phragmites Maps,

20 Great Lakes Basin Accuracy Assessment Potential Phragmites Dominant with greater than 90% cover Entire Basin SAR Classes Phrag other Total Producer s accuracy (Omission Error) Field Validation Phrag % other % Total User s accuracy (Commission Error) 42% 96% 85%

21 Project Objectives Map current invasive Phragmites extent for Great Lakes coastal zone Remote Sensing Field Work Validation Identify major environmental drivers of Phragmites australis distribution Provide decision support tool Assess vulnerable areas to new invasion

22 Photo: M. Carlson Mazur Martha. L. Carlson Mazur 1,2, Kurt P. Kowalski 2, David M. Galbraith 2, Laura L. Bourgeau-Chavez 3, Liza Jenkins 3, Colin Brooks 3 1 Boston College, Chestnut Hill, MA, USA 2 U.S. Geological Survey, Ann Arbor, MI, USA 3 Michigan Tech Research Institute, Ann Arbor, MI, USA

23 Existing Phragmites stands

24 and habitat suitability index

25 Existing Phragmites stands

26 and habitat suitability index

27 URL: Coming soon!

28 Factors that improve habitat suitability for Phragmites: Flat terrain Close proximity to development and agriculture Greater road density Poorly drained soils Phragmites may respond differently to habitat characteristics in the upper and lower Great Lakes Decision support tool and webinars coming soon! Photo: M. Carlson Mazur

29 Outreach/ Product Sharing MTRI project website Jpegs of 3-season radar image mosaics for Lakes Huron, Ontario, Michigan, and Erie 2010 field data in GoogleEarth as KMLs 2010 site visit, geotagged field photos in GoogleEarth Join the Website list for updates and distribution of final map products Decision support tool from USGS

30 THANKS - QUESTIONS? brian_huberty@fws.gov