Loading…
US-IALE 2018 has ended
Back To Schedule
Tuesday, April 10 • 11:45am - 12:00pm
INVASIVE SPECIES: Integration of Radar and Optical Satellite Data for Great Lakes Coastal Wetland Monitoring

Sign up or log in to save this to your schedule, view media, leave feedback and see who's attending!

AUTHORS: Matt Unitis*, Jodi Brandt – Boise State University

ABSTRACT: Great Lakes coastal wetlands have an ecological importance disproportionate to their size, yet are extremely vulnerable to invasion by introduced plants. Plant invasions can be effectively controlled in early stages of invasion, but early detection of invasive wetland plants using traditional ground survey methods is time-prohibitive, and remote detection is challenging because of the heterogeneity of plant composition during early stages of invasion. New satellite data sources and analysis techniques offer promise for wetland and invasive plant remote sensing, and in this study we used Sentinel-1 C-band SAR from three different dates with RapidEye multispectral optical imagery to detect phenological and spectral differences between wetland plant communities in the Great Lakes. Of particular interest was detecting hybrid cattail (Typha x. glauca), which imperils the high quality coastal wetlands of the St. Mary’s River. We compared wetland classifications using Randomforest and Maximum Likelihood algorithms, and subsequently determined the best Randomforest model using various data combinations of optical and radar inputs. Both techniques produced classifications of similar accuracy (Random Forest achieved a 94.0% average overall accuracy and Maximum Likehood achieved 90.1% overall accuracy). The best-performing classification was a Randomforest model with multi-temporal SAR and optical bands as input variables, achieving a 97.4% producer’s accuracy for hybrid cattail. Randomforest variable importance demonstrated that the Near Infrared band of RapidEye and the August 22nd SAR were the two most important variables to discriminate between wetland plant communities. Our results indicate that by combining moderate resolution SAR and optical datasets, we can map detailed wetland communities at a scale and resolution to guide early detection and management of wetland plant invasions. Furthermore, our methods can be used for repeat monitoring of hybrid cattail, and are likely adaptable for remote monitoring in other wetland environments due to our exclusive use of spaceborne platforms with global coverage.

Tuesday April 10, 2018 11:45am - 12:00pm CDT
Spire Parlor

Attendees (6)