ABSTRACT. Ecologists have developed terminology to distinguish ecosystems based on the degree of human alteration. To this end, ecosystems can be characterized as "novel ecosystems," "impacted ecosystems," or "designed ecosystems," depending on the role of human management in ecosystem development and effects on ecosystem properties. Properly classifying an ecosystem as novel, impacted, or designed has critical implications for its conservation and management, but a broadly applicable definition for a "novel ecosystem" does not exist. We have provided a formal definition of "novel ecosystem" that facilitates its use in practical applications and have described four characteristics of such an ecosystem. A novel ecosystem can be identified by its origins rooted in human agency, the ecological thresholds it has crossed, a significantly altered species composition, and a capacity to sustain itself. Ecosystem classification in the literature has been inconsistent. We have illustrated the application of our definition using multiple case studies representing impacted, designed, and novel ecosystems.

a b s t r a c tFoliar nitrogen (N) in plant canopies is central to a number of important ecosystem processes and continues to be an active subject in the field of remote sensing. Previous estimates of foliar N at the landscape scale have primarily focused on intact forests and grasslands using aircraft imaging spectrometry and various techniques of statistical calibration and modeling. The present study extends this work by examining the potential to estimate the foliar N concentration (%N) of residential, agricultural and other cultivated grassland areas within a suburbanizing watershed in southeastern New Hampshire. These grasslands occupy a relatively small fraction (17.5%) of total land area within the study watershed, but are important to regional biogeochemistry and are highly valued by humans. In conjunction with ground-based vegetation sampling (n = 20 sites with 54 sample plots), we developed partial least squares regression (PLSR) models for predicting mass-based canopy %N across management types using input from airborne and field-based imaging spectrometers. Models yielded strong relationships for predicting canopy %N from both ground-and aircraft-based sensors (r 2 = 0.76 and 0.67, respectively) across sites that included turf grass, grazed pasture, hayfields and fallow fields. Similarities in spectral resolution between the sensors used in this study and the proposed HyspIRI mission suggest promise for detecting canopy %N across multiple forms of managed grasslands, with the possible exception of areas containing lawns too small to be captured with HyspIRI's planned 60 m spatial resolution.