Using Landsat imagery to detect, monitor, and project net landscape change

Large-area land change traditionally has been monitored through comparison of maps derived from remotely sensed data for two or more snapshots in time. The products are generally informative about types, rates, and patterns of land change, though the time steps may be too coarse to capture some important changes. The National Land Cover Database, as one such example, has provided the foundation for several USGS efforts for forward and backward modeling of land-cover.

USGS-EROS has produced unprecedented national-scale land-cover projections for four IPCC emissions scenarios from 1992 to 2100, and is currently completing associated historical projections from 1992 to 1938 to create a >160 year dataset of annual land-cover maps. Although a valuable dataset, the characteristics of these projections may not meet the needs of all users. Therefore, USGS-EROS is leading the development of a stand-alone modeling tool that will enable stakeholders to quickly develop customized, climate-based land-cover projections to satisfy input needs for their own conservation applications.

Also on the near-term horizon is a radical new paradigm for characterizing and detecting changes in land-cover that will revolutionize our abilities to monitor and project land change. This new approach uses the depth of the Landsat archive to mathematically characterize the temporal behavior of pixels through time and detect when pixel responses diverge from past trajectories.

This presentation will highlight 1) existing land-cover modeling capabilities and datasets 2) development of a public land-cover modeling tool, and 3) upcoming products from USGS-EROS for monitoring, assessing, and projecting land-cover change.