Geoscience Reference
In-Depth Information
As satellite sensors and related algorithms continue to improve, many of the technical obstacles
addressed above may be overcome. However, it is essential that groups producing global LC
products have a thorough awareness of technology improvements across the range of satellite
sensors, including optical, microwave LIDAR, and SAR. Such awareness can be supported through
LPV's interaction with the other WGCV subgroups, including discussions at the semiannual WGCV
plenary meetings and utilizing the projects and publications available through the other subgroups
of WGCV. Further, coordination of various LC products can help determine the most suitable
approach to using multiple products. For example, the MODIS product has been operationally
produced since 2001. Careful examination of this product as well as the GLC 2000 product could
lend insight into the best way to use both in a complementary fashion.
3.4.1
Fine-Resolution Image Quality and Availability
Data sharing of high-resolution imagery may be one of the most immediate and concrete ways
in with LPV can support global land-product validation. Using the NASA Earth Observing System
Land Validation Core Sites (Morisette et al., 2002) as an example, the LPV and WGISS subgroups
are establishing an infrastructure for a set of “CEOS Land Product Validation Core Sites.” The initial
sites being considered for this project are shown in Plate 3.1, which represents an agglomeration
of three entities: the EOS Land Validation Core Sites, the VAlidation of Land European Remote
sensing Instruments (VALERI) project, and the CEOS “LAI Inter-comparison” activity (Table 3.1).
The concept is to establish a set of sites where high-resolution data will be archived and proved
free or at minimal cost over locations where field and/or tower measurements are continuously or
periodically collected (Plate 3.1). These core sites are intended to serve as validation sites for
multiple satellite products. Specific products appropriate for validation depend on the individual
field tower measurement parameters (Morisette et al., 2002). Practically, the limited number of sites
(approximately 50), which are not based on a random sample, cannot be used for statistical inferences
on a global product. However, in terms of LC validation, the high-resolution data from these sites
would allow a set of common “confidence-building sites” that could be shared by GLC 2000 and
MODIS as well as future global LC mapping efforts. LC product comparisons with high-resolution
data and cross-comparison with other global LC products over the core sites would provide sub-
stantive information for initial quality control. Additionally, within a given site a random sample
could be collected and design-based inference carried out for that particular “subpopulation.” So,
while the core site concept has limitations with respect to statistical inference, the opportunities for
data sharing and initial cross-comparison at a set of core sites seems worthwhile.
3.4.2
Local Knowledge Requirements
The LPV was strategically designed to complement the objectives of the Global Observation of
Forest Cover/Land Dynamics (GOFC/GOLD) program (http://www.fao.org/gtos/gofc-gold/). This
partnership provides a context for validation activities (through LPV) within the specific user group
(GOFC/GOLD). GOFC/GOLD is broken down into three implementation teams that include: (1)
LC characteristics and change, (2) fire-related products, and (3) biophysical processes. Initial activ-
ities of LPV have also focused on these three areas through topical workshops and initial projects.
A major component of GOFC/GOLD is to build on “regional networks.” These networks involve
local and regional partners who are interested in using the global products and serve to provide
feedback to the data producers. This regional network concept has proven to be a significant resource
to support validation efforts. The IGBP experience indicates that the knowledge gained through
regional collaborators is critical. LPV can use the regional networks as an infrastructure to gain
local expertise for product validation. This infrastructure can provide assistance with the difficult
and labor-intensive task of design-based inference planned for both MODIS and GLC 2000.
