Dr Colin Polsky, Dr. Robert Gilmore Pontius Jr, Albert Decatur, Dan Runfola, Nick Giner, Rahul Rakshit, Matt Salem, Nick Perdue, Tom Hamill

Title
Looking for Lawns: How We Know What’s In Your Yard

Abstract
Holmes is a half-meter, large extent land cover mapping project with the objective of building a dataset to explore the patterns of suburbanization occurring at various geographies in the coupled human-environment system of suburban Boston. Using an object based methodology on 4-band aerial imagery; the Holmes project has created land cover maps for 14 of the 26 towns within the Plum Island Ecosystem in northern Massachusetts. The object-based methodology works by grouping continuous spectrally similar pixels into objects that can be understood in an intuitive and contextual manner. The landscape has been divided into seven classes- coniferous, deciduous, fine green, bare soil, impervious, water, and wetlands- and the data put into a analysis model which allows for the calculation of percent land cover at geographic scales ranging from the extent of a town to sub-parcel. Validation on the classifications is done with a virtual fieldwork methodology which allows for a random sample to be created over the full extent of a town using virtual. Virtual fieldwork allows for aerial, oblique, and street views as a means of validation of all the sample points and eliminates the ambiguity of determining the land cover of a validation point which falls within a shadow. With this methodology, the land cover classifications for the town of Lynnfield Massachusetts are 93% correct.