| Abstract: |
The USGS land cover Trends project conducts detailed classification of 10- and 20-km^2 blocks based on fieldwork, 60-m satellite imagery, and ancillary data. The research provides highly accurate land cover estimates for 5 temporal “windows” in the period 1973 to 2000. I have applied stochastic methods to estimate empirical Markov chain matrices for each of 62 blocks in the 160,000-km^2 Chesapeake Bay Watershed. The matrices describe shifts among 3 land cover classes during each period and are used to estimate parameters for point forecasts of decadal 2000-2050 regional land cover. I then interpolate land cover fields for the entire region based on the estimates and error measurements. Results provide animations based on regional maps of where the landscape is changing fastest and toward what kinds of land covers. I intend to use the stochastic model both to relate simple, robust measures of land cover status and dynamics to descriptions of the drivers of landscape transformation (population growth, economic activity, climate change) and to forecast the consequences of landscape change for Chesapeake Bay water quality, habitat, and living resources. |
