Least-squared linear regression results and mean differences
among site measurements and model simulation and
radar estimates of spring initiation, fall cessation and associated
growing season length for the 2-year study period are
summarized in Table 4. Scatterplots of the relationships
between site measurements and radar estimates of growing
season initiation and total length are presented in Fig. 7.
Radar estimates of growing season parameters were significantly
( p < 0.025) associated with site measurements, accounting
for 75.5%, 22.6% and 68.2% of the variation in
growing season initiation, cessation and length, respectively.
Correspondences between radar and BIOME-BGC estimates
of these parameters were generally lower, with
radar-based results accounting for 61.1% and 65.3%
( p < 0.001) of the variation in model estimates of growing
season initiation and length, respectively. Correspondence
between model and measured results was generally good,
with model results accounting for more than 78%
( p < 0.001) of the variability in site measurements of growing
season parameters. Radar estimates of growing season
initiation generally occurred prior to, but within 10 and 16
days of site-based measurements and model results, respectively.
The high correspondence between radar results and
site measurements are due to the relatively high sensitivity
of the Ku-band radar to freeze –thaw processes in snow and
the strong link between spring initiation of the growing
season in boreal and subalpine evergreen coniferous forests
and the general onset of seasonal snowmelt in the spring
(e.g., Jarvis & Linder, 2000; Monson et al., 2002). The earlier
onset observed by the radar may be due to its sensitivity to the early stages of snowmelt, while net CO2 uptake and sap
flow are delayed until there is sufficient available soil water.
Reduced correspondence between radar remote sensing and
BIOME-BGC estimates of growing season initiation is due
to the dependence of model computations on simple air
temperature and estimated surface soil layer temperature
thresholds, independent of snowmelt dynamics.