Using Urban Landscape Trajectories to Develop a Multi-Temporal Land Cover Database to Support Ecological Modeling

Abstract

Urbanization and the resulting changes in land cover have myriad impacts on ecological systems. Monitoring these changes across large spatial extents and long time spans requires synoptic remotely sensed data with an appropriate temporal sequence. We developed a multi-temporal land cover dataset for a six-county area surrounding the Seattle, Washington State, USA, metropolitan region. Land cover maps for 1986, 1991, 1995, 1999, and 2002 were developed from Landsat TM images through a combination of spectral unmixing, image segmentation, multi-season imagery, and supervised classification approaches to differentiate an initial nine land cover classes. We then used ancillary GIS layers and temporal information to define trajectories of land cover change through multiple updating and backdating rules and refined our land cover classification for each date into 14 classes. We compared the accuracy of the initial approach with the landscape trajectory modifications and determined that the use of landscape trajectory rules increased our ability to differentiate several classes including bare soil (separated into cleared for development, agriculture, and clearcut forest) and three intensities of urban. Using the temporal dataset, we found that between 1986 and 2002, urban land cover increased from 8 to 18% of our study area, while lowland deciduous and mixed forests decreased from 21 to 14%, and grass and agriculture decreased from 11 to 8%. systems that are present in this region were likely significantly altered by these changes in land cover. Our results suggest that multi-temporal (i.e., multiple years and multiple seasons within years) Landsat data are an economical means to quantify land cover and land cover change across large and highly heterogeneous urbanizing landscapes. Our data, and similar temporal land cover change products, have been used in ecological modeling of past, present, and likely future changes in ecological systems (e.g., avian biodiversity, water quality). Such data are important inputs for ecological modelers, policy makers, and urban planners to manage and plan for future landscape change.

Carbon consequences of land cover change and expansion of urban lands: A case study in the Seattle metropolitan region

“…To assess trends in population, land cover, and terrestrial carbon changes in the Seattle MSA, we integrated US census data, a time series of Landsat-based land cover maps (Alberti, Weeks, & Coe, 2004;Hepinstall-Cymerman, Coe, & Alberti, 2009), a digital elevation model, and field measurements of aboveground live biomass (Hutyra et al, 2011). Most of the historical population data were obtained from the US Census for the Seattle MSA region, with population estimates for 1880 and 1890 originating from Sale (1978).…”

“…Most of the historical population data were obtained from the US Census for the Seattle MSA region, with population estimates for 1880 and 1890 originating from Sale (1978). Land cover data were obtained from the University of Washington, Urban Ecology Research Laboratory (Hepinstall-Cymerman et al, 2009). Remote sensing images from the Landsat Thematic Mapper and the Enhanced Thematic Mapper were used to classify the land cover at a 30 m resolution across the Seattle region for the years 1986years , 1991years , 1995years , 1999years , 2002years , and 2007years (see HepinstallCymerman et al, 2009 for full methodological details and data processing methods).…”

Monitoring gradual ecosystem change using Landsat time series analyses: Case studies in selected forest and rangeland ecosystems

“…In general, these types of events radically alter the spectral properties of the land surface, and are readily discernible in Landsat imagery. The literature is replete with examples of mapping and monitoring major disruptive changes using remotely sensed data (Clark and Bobbe, 2006;Coppin et al, 2004;Hansen et al, 2008;Hepinstall-Cymerman et al, 2009;Masek et al, 2008;Wulder et al, 2009). Challenges to mapping and monitoring these events often revolve around developing consistent, efficient and operational approaches to enable accurate characterization of these events (Brink and Eva, 2009;Huang et al, 2010a;Kennedy et al, 2010).…”

Urban growth patterns and growth management boundaries in the Central Puget Sound, Washington, 1986–2007

“…We used 14-class land cover data for 1986, 1991, 1995, 1999, 2002, and 2007 developed from a combination of Landsat Thematic Mapper (TM) and Enhanced TM (ETM+) imagery (Hepinstall-Cymerman et al 2009). Multiple methods were used to differentiate 14 land cover classes in each image.…”