providing information necessary for building and assessing quantitative models that may
be used for forecasting in the context of sustainability. However, as for all monitoring
systems, the design of process-driven measurement systems such as NEON must be
discovered in the context of known constraints (e.g., logistical, financial).
Global Forest Network– The Global Forest Resources Assessment 2010 (FAO
2010) examines the current status and recent trends for the extent, condition, uses and
values of forests and other wooded land. Information has been collated from 233
countries and territories for four points in time: 1990, 2000, 2005 and 2010. The results
are presented according to seven themes associated with sustainable forest
management. A systematic sampling design based on intersections of whole degrees of
longitude and latitude was used with a reduced intensity above 60 degrees North/South
latitude. At each sample site, a 10-km × 10-km area is accessed via interpretation and
classification of four Landsat satellite images dating from circa 1975, 1990, 2000 and
2005.
3.3. Data fusion methods
Development of data fusion methods will be necessary for combining multiple
sources of information. Data fusion could come in the form of fusion of one set of
observations with another set, obtained under different sampling paradigms, or fusion of
observations with model-based output. In some cases, we may consider model outputs
as a form of data observation (e.g., climate model outputs may be used as “data” for a
biogeochemical cycle model or an economics model), and hence they involve the same
mathematical problem. The general problem is to come up with an estimate of the
underlying latent (or unobservable) process(es) of interest based on the two (or more)
sets of observations. Approaches for fusing different datasets are just beginning to
emerge. For example, Nguyen, Cressie, and Braverman (2010) combined data from two
different instruments on the same satellite to find optimal spatial predictors of aerosol
optical depth. The idea is to expand upon such data fusion methods so that they can
accommodate a wide range of datasets that may inform similar or overlapping
processes.
Data fusion examples– The combination of ground data obtained from sample
plots and remotely sensed data such as satellite imagery are often combined to estimate
baseline conditions and to assess change. Combining such independently acquired
data produces multiple sources of uncertainty: (1) rectification of the imagery to the plot
coordinate system is not without error, (2) ground plot coordinates have varying degrees
of error depending on the quality of global positioning system (GPS) receivers, (3) plots
and image pixels are nearly always of different sizes, and (4) plot measurement and
image acquisition dates are seldom the same. Although reasonably accurate maps and
estimates of land cover parameters may be obtained, the effects of the sources of
uncertainty on the uncertainty of land cover parameter estimates is generally unknown.
Forests on the Edge (Stein et al., 2009), a project of the U.S. Forest Service,
uses geographic information systems (GIS) techniques to assess the contributions of
private forest land and threats to those contributions. The analyses are based on the
fusion of independently constructed data layers of different spatial resolution obtained
from a diverse set of underlying data sources. The primary layer from which a private
forest land map is constructed is based on a 30-m x 30-m, satellite image-based
forest/non-forest layer (Homer et al., 2007) and a layer depicting land protection status