continuous satellite data incorporate mixed signals from all plants as well as
background substrates. Therefore, the exact biophysical meanings of LSP, in
particular SOS, remain vague when ground data that lack landscape representativeness are employed for validation.
To meet this challenge, phenology data collection in a spatially concentrated
manner which allows detailed characterization of landscape heterogeneity became
crucial (Liang and Schwartz 2009; Liang et al. 2011; Schwartz et al. 2013).
Beginning from 2006, high density phenology observations were carried out in a
mixed forest located in northern Wisconsin, near an AmeriFlux tower site (Park
Falls/WLEF). A high resolution spatial sampling of major canopy species was
deployed at the study sites, initially for a 625 9 275 m area (2006 and 2007), and
later for two 625 9 625 m expanded areas (since 2008). At least three of the
largest trees at 288 plots were sampled. The coverage of sampling grids matched
the size of MODIS pixels, and the density of sampling allowed capturing spatial
variations within the pixels. As shown in Fig. 4.3, the spatial setting of in situ data
collection allowed better compatibility with coarse resolution satellite pixels in
both extent and representativeness. In addition to high spatial density, phenology
observation was conducted with high temporal frequency. The field crew took
measurements for all sampled trees every other day (modified to a 4-day interval
since 2010) throughout the early spring time period (about a month). A detailed
field protocol was used to describe phenology from buds to leaves and with
Fig. 4.3 High resolution
sampling design for satellite
phenology validation
showing: (1) in situ
phenological observation
transects/plots (according to a
cyclic sampling design, see
Burrows et al. 2002); (2)
MODIS pixel grids
(250 9 250 m squares,
sinusoidal projection); and
(3) 2.4 m resolution NDVI
image derived from a May
18, 2007 QuickBird image
underlying the plots and
MODIS pixel grids. This
figure is reprinted from Liang
et al. (2011) with permission
from Elsevier
112
J. M. Hanes et al.
background substrates. Therefore, the exact biophysical meanings of LSP, in
particular SOS, remain vague when ground data that lack landscape representativeness are employed for validation.
To meet this challenge, phenology data collection in a spatially concentrated
manner which allows detailed characterization of landscape heterogeneity became
crucial (Liang and Schwartz 2009; Liang et al. 2011; Schwartz et al. 2013).
Beginning from 2006, high density phenology observations were carried out in a
mixed forest located in northern Wisconsin, near an AmeriFlux tower site (Park
Falls/WLEF). A high resolution spatial sampling of major canopy species was
deployed at the study sites, initially for a 625 9 275 m area (2006 and 2007), and
later for two 625 9 625 m expanded areas (since 2008). At least three of the
largest trees at 288 plots were sampled. The coverage of sampling grids matched
the size of MODIS pixels, and the density of sampling allowed capturing spatial
variations within the pixels. As shown in Fig. 4.3, the spatial setting of in situ data
collection allowed better compatibility with coarse resolution satellite pixels in
both extent and representativeness. In addition to high spatial density, phenology
observation was conducted with high temporal frequency. The field crew took
measurements for all sampled trees every other day (modified to a 4-day interval
since 2010) throughout the early spring time period (about a month). A detailed
field protocol was used to describe phenology from buds to leaves and with
Fig. 4.3 High resolution
sampling design for satellite
phenology validation
showing: (1) in situ
phenological observation
transects/plots (according to a
cyclic sampling design, see
Burrows et al. 2002); (2)
MODIS pixel grids
(250 9 250 m squares,
sinusoidal projection); and
(3) 2.4 m resolution NDVI
image derived from a May
18, 2007 QuickBird image
underlying the plots and
MODIS pixel grids. This
figure is reprinted from Liang
et al. (2011) with permission
from Elsevier
112
J. M. Hanes et al.